Deterrence, punishment severity and drink-driving

Author:
Briscoe, Suzanne Marie
Publication Date:
2005
DOI:
https://doi.org/10.26190/unsworks/23230
License:
https://creativecommons.org/licenses/by-nc-nd/3.0/au/
Link to license to see what you are allowed to do with this resource.

Downloaded from http://hdl.handle.net/1959.4/23442 in https://

unsworks.unsw.edu.au on 2024-05-14
 DETERRENCE, PUNISHMENT SEVERITY AND
DRINK-DRIVING

SUZANNE BRISCOE

A thesis submitted in fulfilment

Of the requirements for the degree of
Doctor of Philosophy

School of Social Science and Policy University of New South Wales

August, 2005
ABSTRACT
This thesis tests one of the major propositions of deterrence theory: that increases in
the severity of criminal punishment can reduce the likelihood of offending. To this
end, a case study in which the statutory penalties were doubled for almost all drink-
driving offences in New South Wales, Australia, is examined. This case study was
considered a good test of deterrence hypotheses regarding the efficacy of punishment
severity because it satisfied several preconditions necessary for optimising deterrent
effects.

Two quasi-experimental studies were undertaken to assess the impact of these

legislative changes on drink-driving offending levels: an interrupted time-series
analysis of road crash rates (Study 1) and an analysis of drink-driving reoffending
rates before and after the penalty changes were implemented (Study 2).

Study 1 found a significant increase in monthly single-vehicle night-time crash rates

(a surrogate measure of alcohol-related road crashes) after the tougher drink-driving
penalties were introduced in New South Wales. The analyses also revealed a
significant increase, around the same time, in non alcohol-related crashes (as
measured by multiple-vehicle day-time crashes). Further exploratory analyses
indicated that the rate of increase observed after the penalty changes, was greater for
multiple-vehicle day-time crashes than for single-vehicle night-time crashes. This
suggests that the significant increase in the surrogate measure of alcohol-related
crashes may have been driven primarily by a rise in non alcohol-related crashes that
coincided with the policy’s implementation. Two possible conclusions regarding the
deterrent effect of the tougher penalties are drawn from these findings: (1) that there
was a reduction in alcohol-related road crashes but this effect was overwhelmed by
the rise in non alcohol-related crashes occurring around the same time or (2) that
there was no reduction in crash rates after the tougher penalties were introduced.

To test deterrence hypotheses further, the reoffending rates of two drink-driver

cohorts were compared: one consisting of all drink-drivers convicted in NSW Local
Courts in the year before the penalty changes and the other consisting of all drink-

ii
drivers convicted in the year after. Logistic regression and survival analyses
demonstrated that those drink-drivers who were convicted under the new penalty
regime were less likely, and took longer, to reappear before the courts for a new
drink-driving offence than drink-drivers convicted before the introduction of the new
penalties. This reduction in reoffending was, however, only apparent for drink-
drivers residing in country and regional areas and was small in magnitude. While
these findings are consistent with the possibility that the penalty changes resulted in a
reduction in alcohol-related road crashes which could not be detected in Study 1,
they suggest that any decrease is likely to have been relatively small.

An examination of the context in which the drink-driving sentencing policy was

implemented identified two potential explanations for the small deterrent effect
observed in the quasi-experimental research. Firstly, NSW Police appeared to have
altered their enforcement practices around the time that the legislative amendments
were introduced. These changes may have reduced the perceived risk of detection
amongst some drink-drivers and therefore undermined the deterrent effect of the
increase in punishment severity. Secondly, the tougher drink-driving penalties
received limited publicity, thus many drivers may have been unaware of the changes
to drink-driving punishment.

To further explore the deterrent effect of increases in punishment severity a third

study using a scenario-based survey methodology was undertaken. This work builds
on Study 1 and Study 2 by examining the relationship between legal sanctions and
willingness to drink-drive, in an experimental setting where it was possible to control
for other contextual influences. The results of this study showed that participants
who made more mistakes in a test of knowledge about drink-driving penalties in
New South Wales stated that they were more likely to offend in the drink-driving
scenario than were participants who made fewer mistakes, controlling for other
factors. This would suggest that were the changes to drink-driving penalties
(examined in the case study) communicated more effectively to the target audience
then there may have been a greater reduction in drink-driving offending levels in
New South Wales. However, these conclusions are tentative given that there was
only weak evidence for a relationship between the number of mistakes on the

iii
knowledge test and participants’ ratings of penalty severity and, further, that no
significant interaction was found between perceptions of punishment severity and
perceptions of punishment certainty.

The sentencing policy examined in this thesis presented a good opportunity to test
the deterrent efficacy of tougher penalties in a real world situation where several
necessary conditions were in place for optimising deterrent effects. However, not all
relevant contextual factors could be controlled for, and therefore definitive
conclusions about the deterrent effect of the penalty changes could not be drawn
from its results. Despite these limitations, the small effect of the sentencing changes
on offending levels observed in this research, in conjunction with the weak evidence
for punishment severity effects found in the scenario-based survey, is consistent with
much of the available evidence on deterrence. The implications of these findings for
deterrence theory, criminal justice policy and future research in this area are
discussed.

iv
Acknowledgements
I would like to acknowledge my thesis supervisors, Professor Janet Chan and Dr Don
Weatherburn, for their continual support and guidance throughout this project. In
particular, I would like to thank Don for encouraging me to embark on this journey,
for allowing me the opportunity to undertake this research in a supportive and
resourceful work environment and for always making time.

I would also like to thank several staff from the NSW Bureau of Crime Statistics and
Research (both past and present) who have helped me at various stages throughout
this research. In particular, thanks go to Neil Donnelly, Marilyn Chilvers, Shuling
Chen and Bronwyn Lind for their invaluable advice on the statistical methods used in
these analyses and also Jiuzhao Hua, whose ingenious work on the Bureau’s
Reoffending Database made the recidivism analysis reported in this thesis possible.

The NSW Roads and Traffic Authority should also be acknowledged for providing
the road crash data for the interrupted time-series analyses and for supplying offence
data to validate the accuracy of the prior offending variable included in the Bureau’s
Reoffending Database.

Finally, I would like to thank my family and friends who endured much of this
journey with me and who were extremely supportive throughout. In particular, I
would like to thank my parents who proofread this thesis in record time and
especially my husband-to-be for his support and continual encouragement, for
listening to my ideas and entertaining my theories, and for all the lost weekends.

v
Previous publications
Some of the analyses appearing in this thesis have previously been published. A
paper describing the results from the interrupted time-series analysis of road accident
data (Study 1) was published in the international, peer-reviewed journal Accident,
Analysis and Prevention in 2004 (Briscoe 2004a). The recidivism analysis,
comparing the reoffending rates of drink-drivers convicted in 1997 with those
convicted in 1999 (Study 2) was also published in 2004 as an Alcohol Studies
Bulletin for the NSW Bureau of Crime Statistics and Research (Briscoe 2004b). All
the work reported in this thesis was completed during my PhD candidature at the
University of New South Wales and received approval from the Human Research
Ethics Committee at this institution.

vi
TABLE OF CONTENTS
Page Number

ABSTRACT ................................................................................................... II

Acknowledgements..................................................................................................... v

Previous publications ................................................................................................ vi

CHAPTER 1. INTRODUCTION TO THESIS AND OVERVIEW .................... 1

History of deterrence theory ..................................................................................... 2

General tests of deterrence theory............................................................................ 3

Deterrence and the certainty of formal punishment................................................. 4
Deterrence and the severity of formal punishment .................................................. 7
Limitations of aggregate-level deterrence research ............................................... 13
Perceptual research in deterrence........................................................................... 15

The current study..................................................................................................... 17

Thesis structure ...................................................................................................... 18

CHAPTER 2. DETERRENCE THEORY ...................................................... 21

Definitions of deterrence.......................................................................................... 22

Modern deterrence theory....................................................................................... 26

Expanding the Subjective Expected Utility Model................................................ 28
Critiques of the Rational-Choice Perspective ........................................................ 29

Deterrence and specific offences ............................................................................. 34

The role of informal social control ......................................................................... 36

Further interaction effects in deterrence ............................................................... 40

Individual characteristics and sanction perceptions............................................... 41
Perceived risk of apprehension and perceptions of punishment severity .............. 47

Drink-driving as a case study.................................................................................. 50

CHAPTER 3. DETERRENCE RESEARCH ON DRINK-DRIVING .............. 53

Aggregate-level research ......................................................................................... 54

Punishment certainty.............................................................................................. 54
Punishment severity ............................................................................................... 63
Punishment celerity................................................................................................ 71

vii
Individual-level research ......................................................................................... 74
Reoffending studies................................................................................................ 74
Perceptual studies................................................................................................... 78

Significance of the current study ............................................................................ 83

Methodological issues in policy impact studies..................................................... 84
Knowledge gaps in the deterrence literature.......................................................... 86

CHAPTER 4. INCREASED DRINK-DRIVING PENALTIES IN NEW SOUTH

WALES: A CASE STUDY ........................................................................... 87

Drink-driving legislation in New South Wales and the 1998 penalty changes ... 89

The case study........................................................................................................... 92

Measuring policy effectiveness - The quasi-experiment ....................................... 95

Alcohol-related road crashes - Study 1 .................................................................. 96
Drink-driving recidivism rates - Study 2 ............................................................. 102

Perceptual research - Study 3 ............................................................................... 108

Hypotheses ........................................................................................................... 109
Participants and methodology .............................................................................. 115

CHAPTER 5. IMPACT OF THE 1998 PENALTY CHANGES ON

OFFENDING RATES IN NEW SOUTH WALES ....................................... 120

Study 1 - Road crash analysis ............................................................................... 120

All road crashes resulting in injury ...................................................................... 121
Alcohol-related road crash rates .......................................................................... 124
Control series ....................................................................................................... 129
Summary .............................................................................................................. 131
Alcohol-related crashes versus control series ...................................................... 133
Preliminary conclusions ....................................................................................... 135

Study 2 - Analysis of recidivism rates .................................................................. 137

Implementation of statutory amendments............................................................ 137
Characteristics of offender cohorts ...................................................................... 141
Probability of reappearing for a drink-driving offence ........................................ 143
Time to first drink-driving offence ...................................................................... 146
Summary and discussion of recidivism analysis.................................................. 151

Conclusions from the quasi-experimental studies............................................... 154

CHAPTER 6. CONTEXT OF POLICY IMPLEMENTATION ...................... 156

Enforcement of drink-driving offences ................................................................ 157

Police breath tests and public perceptions of enforcement activity ..................... 160
Police targeting of high-risk times and locations................................................. 163

Media coverage of statutory penalty increase ..................................................... 165

viii
 Newspaper coverage of penalty changes ............................................................. 165
RTA advertising campaigns and other publicity.................................................. 171

What went wrong? ................................................................................................. 173

CHAPTER 7. KNOWLEDGE OF PENALTIES, PERCEPTIONS OF

SANCTION SEVERITY AND DRINK-DRIVING ........................................ 176

Formation of risk perceptions............................................................................... 177

Legal knowledge as a factor in the deterrence process ........................................ 179

The scenario-based study - Study 3 ...................................................................... 185

Hypotheses ........................................................................................................... 185

Results of Study 3 ................................................................................................... 188

Sample demographics .......................................................................................... 188
Independent variables........................................................................................... 189
Predicting stated likelihood of offending............................................................. 193
Linear regression models including knowledge and certainty variables ............. 196
Testing the assumptions of linear regression ....................................................... 200

Summary and discussion of survey findings........................................................ 204

Limitations of Study 3.......................................................................................... 209

CHAPTER 8. SUMMARY AND DISCUSSION .......................................... 211

The deterrent effect of increased drink-driving penalties in New South Wales

.................................................................................................................................. 212
Changes in drink-driving behaviour after the penalty increase ........................... 212
Implementation of the 1998 sentencing policy .................................................... 214
Knowledge of drink-driving penalties and perceptions of sanction severity....... 216

Implications for theory and practice .................................................................... 219

Translating policy into a credible threat............................................................... 220
The marginal deterrent efficacy of punishment severity...................................... 225

Future directions for deterrence research and drink-driving policy ................ 230
Future research ..................................................................................................... 230
Policy recommendations ...................................................................................... 232

REFERENCES .......................................................................................... 237

APPENDICES............................................................................................ 255

Appendix I – Summary of drink-driving studies reviewed in Chapter 3 ......... 255

Appendix II – Participant information sheet and drink-driving questionnaire

.................................................................................................................................. 258

ix
Appendix III – Supplementary analyses .............................................................. 270

x
CHAPTER 1. INTRODUCTION TO THESIS AND
OVERVIEW

Deterrence is one of the major aims of punishments prescribed by legislatures and is

a prominent feature of many criminal justice interventions designed to curtail
criminal behaviour. It assumes that crime is a rational decision-making process in
which potential offenders weigh the costs and benefits associated with an illegal act.
If the anticipated costs are greater than the anticipated rewards, individuals will
choose not to engage in the prohibited behaviour. It is an appealing penal ideology
because it makes sense that people will want to avoid unpleasant consequences and
therefore continues to inform public opinion on crime control and, at the same time,
play an influential role in public policy. Just how effective deterrence is, however, as
a strategy to reduce the incidence of crime still remains somewhat unclear.

The question “Can punishment deter crime?” is one that has been the focus of much
criminological research conducted over the last three or four decades, but is not one
that can be answered categorically. The term ‘crime’ covers a wide range of
behaviours. The nature of these criminal behaviours, as well as the context in which
they are committed, varies substantially. The threat of formal punishment is unlikely
to be equally effective in deterring all forms of crime, under all conditions, for all
offenders. This differential effectiveness of legal sanctions does not necessarily
imply that deterrence theory is irrelevant to offending behaviour, nor does it suggest
that deterrence is an ineffective policy tool. It does mean, however, that empirical
research on deterrence must consider variations in circumstance that could account
for differences in the effectiveness of legal threats.

This thesis is primarily concerned with the deterrent effect of the severity of legal
punishment. It draws upon theory and research on the effectiveness of formal
sanctions, to identify the conditions under which deterrent effects are most likely to
be observed. A case study, in which several of these conditions exist, is then
presented and the deterrent efficacy of an increase in the severity of statutory
penalties is examined. By examining changes in sanction severity within this setting,

1
the research provides a valid test of deterrence hypotheses and addresses important
limitations in previous research conducted in this area. As a result, this work will
lead to more definitive conclusions regarding deterrence and punishment severity,
which will have relevance to criminal justice policy and important implications for
further theory development.

Before exploring these issues further, this chapter will briefly describe the evolution
of deterrence theory, present an overview of the present state of knowledge in the
deterrence field and discuss the current study in light of previous research findings in
this area.

History of deterrence theory

The notion of deterrence originally appeared in the writings of the 18th Century
Utilitarian philosophers, Cesare Beccaria and Jeremy Bentham. These philosophers
viewed human nature as essentially hedonistic and believed that crime can be
deterred if punishment is sufficiently unpleasant. Gibbs asserts that many of the ideas
of these early deterrence theorists can be reduced to the following statement: “the
rate for a particular type of crime varies inversely with the celerity, certainty, and
severity of punishments for that type of crime” (Gibbs 1975, p. 5). However, Gibbs
(1975) also notes that these early ideas on deterrence were little more than a series of
vague statements or what he terms a “deterrence doctrine”. To develop a more
systematic theory of deterrence, the specific sanctions and relevant circumstances
necessary to maximise deterrent effects need to be stipulated in order to determine
how punishment will affect offender behaviour in any given situation.

The deterrence doctrine proposed by Beccaria and Bentham has, for a long time, held
a central position in the philosophy of criminal law, in penal policy and in the
administration of criminal sanctions. However it wasn’t until the 1960s and 70s that
criminologists, sociologists and economists began to take an interest in deterrence
and attempted to construct a more systematic theory of deterrence. Their interest
arose out of disillusionment with other penal ideologies, in particular rehabilitation.
Rehabilitation was found to be a difficult concept on which to base a criminal justice
system because there was little evidence that there were rehabilitative programs that

2
could substantially affect recidivism (Andenaes 1971). This search for a stronger
foundation on which to base criminal justice policy, as well as a growing emphasis
on deterrence-based penal policies in the United States of America (USA) and the
United Kingdom (UK), led to a large body of literature on deterrence and related
issues (Andenaes 1971).

A seminal paper, by Becker (1968), emerged during this period which became
influential in the development of a more systematic theory of deterrence. It proposed
that the decision to offend is a rational one and thus could be treated like any other
decision made under conditions of uncertainty. Drawing upon economic theory,
Becker asserted that potential offenders calculate the expected costs and benefits of
an array of illegal and legal opportunities and commit an offence only if the expected
benefits outweigh the expected costs. These calculations will be discussed in greater
detail in Chapter 2. It is important to note here, however, that they include an
evaluation by the offender of (1) the anticipated gains from committing the crime, (2)
the subjective probability of obtaining those gains, (3) the anticipated ‘pain’ that will
arise from punishment for the crime (severity of punishment) and (4) the subjective
probability that this punishment will be incurred (certainty of punishment).

This view, that criminal behaviour is the outcome of a set of calculated decisions or
choices, deviated from more deterministic models of criminal conduct of the time,
which saw criminal conduct as the result of psychologically, socially or biologically
determined dispositions to offend. In these deterministic models the offender is
portrayed as a passive figure, “prey to internal or external forces outside personal
control” (Clarke & Cornish 1985, p. 148). In contrast, the rational choice model
depicts the offender as an active decision-maker who, when presented with an
offending opportunity, will process and evaluate relevant information from their
environment (Homel 1986).

General tests of deterrence theory

Since Becker’s work in the late 1960s, a vast number of studies, of varying quality,
have been undertaken to test deterrence hypotheses. These studies fall into one of
three categories: ecological studies, quasi-experimental studies and perceptual

3
studies (Nagin 1998). Ecological and quasi-experimental studies generally use
aggregate data to examine the effect of various deterrence-based policies on rates of
particular crimes. Ecological studies look at natural variations in sanction levels and
crime rates across time or space, while quasi-experimental examine the effect of
specific interventions on measures of offending behaviour. The third type of study
used in deterrence research is concerned with individuals’ perceptions of risk. These
studies rely on survey data in order to investigate the relationship between
perceptions of sanction risk and self-reported prior offending or intentions to offend.

In a recent review of this deterrence research, Nagin concluded, “the collective

actions of the criminal justice system exert a very substantial deterrent effect” (Nagin
1998, p. 3). This concurs with Cook’s assessment of the deterrence evidence almost
20 years earlier, from which he affirmed that the “criminal justice system, ineffective
as it may seem in many areas, has an overall crime deterrent effect of great
magnitude” (Cook 1980, p. 213). However, as Nagin (1998) rightly observes, these
conclusions are of little value in formulating policy. He states that the issue for
policy makers is not whether the criminal justice system as an entity deters potential
offenders but whether a specific policy can add to its preventative effect. For this
reason a large body of deterrence research has focused on the deterrent effect of
changes in the certainty and severity of formal punishment.1

Deterrence and the certainty of formal punishment

Many of the early aggregate-level studies that adopted an ecological approach for
studying deterrence found consistent evidence for a negative association between
crime rates and the probability of formal punishment being imposed by the criminal
justice system (Blumstein, Cohen & Nagin 1978). Perhaps the most infamous of
these was the work undertaken by Isaac Ehrlich in the 1970s. Building on Becker’s
economic theory of crime, Ehrlich (1973) developed a rational model of offending
which aimed to predict the annual aggregate murder rate in the USA from estimates
of the probability of apprehension, conviction and execution for this type of crime.
The dependent variable in this analysis was the number of murders/manslaughters

1
Note that very little work has been conducted thus far on the role of celeritous punishment in
controlling crime and therefore deterrence research investigating this aspect of punishment is not
reviewed here. The efficacy of swift punishment is however discussed in later Chapters.

4
per 1,000 civilian population recorded by police each year from 1933 through 1969.
Estimates of punishment risk over this period came from data on the percentage of
all murders cleared by the arrest of a suspect and the fraction of persons charged with
murder who were subsequently convicted. Ehrlich’s analysis also included measures
of other potentially confounding variables, including unemployment, per capita
income and the fraction of young people in the population.

Ehrlich found a significant independent relationship between the annual murder rate
and both the likelihood of arrest and the likelihood of conviction. Ehrlich (1975)
estimated that a one per cent change in the risk of arrest for homicide resulted in a
1.0 - 1.5 per cent change in the aggregate murder rate, while a one per cent change in
the risk of conviction was associated with a 0.4 - 0.5 per cent reduction in the murder
rate. Including the risk of execution once convicted,2 Ehrlich estimated that an
additional execution each year could save between seven and eight lives.

A number of scholars have since made trenchant criticisms of Ehrlich’s study and of
ecological research more generally (Cook 1980; Nagin 1998; Spelman 2000;
Weatherburn et al. 2000; Chan & Oxley 2004). Most notable of these is the
simultaneity problem – the reciprocal relationship that potentially exists between
sanction risks and crime rates. Tougher crime control policies can affect the
incidence of crime but, so can higher crime rates affect sanction risk by influencing
the severity of penal policy and/or the amount of resources devoted to law
enforcement. This violates the basic assumption of regression analysis that the
dependent variable is ‘caused’ by the independent variables included in the model
and thus constitutes a major problem for researchers attempting to infer deterrent
effects from ecological research designs.

Support for the deterrent effect of increases in the certainty of punishment is,
however, also found in quasi-experiments, which overcome many of the
methodological limitations that hinder ecological research (Cook 1980; Nagin 1998).
Law enforcement agencies often attempt to exert an effect on the incidence of crime

2
Ehrlich (1975) estimated that a one per cent change in the risk of execution reduced the murder rate
by between 0.039 and 0.068%. This measure of punishment risk is correctly subsumed under the
‘severity’ of punishment.

5
by launching “crackdowns” or “blitzes” on particular types of crimes or on crime in
particular geographical areas. The effect of these “crackdowns” or “blitzes” on the
actual probability of detection and apprehension for a crime is probably quite limited.
However, their effect on the perceived risk of apprehension can be substantial,
especially if police increase the ratio of police officers to offenders in specific places
or situations, use coercive police powers to stop, search or arrest offenders, or widely
publicise police crackdowns in the mass media (Sherman 1990).

Sherman (1990) undertook a comprehensive review of empirical research on police

interventions in localities where these sorts of initiatives were employed. Of the 18
cases Sherman reviewed, he found that the majority of the crackdowns (15 from 18)
were associated with an initial reduction in levels of the target crime, a result that
Sherman took as evidence for an initial deterrent effect. Furthermore, the most
successful of these police interventions appeared to be those that were implemented
over a short-term time frame (i.e. less than six months). These interventions were
associated not only with an initial deterrent effect but also with a residual deterrent
effect once police efforts were scaled back. In contrast, the initial deterrent effect
created by many of the long-term crackdowns showed evidence of decay as the
intervention progressed. Sherman proposes that the enhanced effect of the shorter
interventions may be due to the fact that police crackdowns increase the uncertainty
surrounding sanction risks, thereby causing people to overestimate the probability of
detection and apprehension. As the intervention progresses over a longer period,
however, people come to learn that this initial reaction was an overestimation and re-
adjust their risk estimates accordingly.

Interestingly, however, Sherman (1990) found differential effects for police

crackdowns targeting different types of crime. In particular, two case studies of drug
crackdowns found evidence of an increase in crime corresponding with the
augmented police activity, and three drug crackdowns were associated with
displacement of criminal activity to other areas. Sherman suggests that, crackdowns
on market-driven criminal activity may have more complex outcomes than those on
“middle-class offences” such as drink-driving and parking violations (Sherman 1990,
p. 36).

6
More recent work by Braga and his colleagues (Braga et al. 2001) has supported
Sherman’s conclusions regarding deterrence-based policing interventions but has
also shown that the way deterrence initiatives are implemented can be critical to their
success. Braga et al. (2001) evaluated the impact of a problem-oriented approach
(known as “Operation Ceasefire” or the “Pulling Levers” program) to the policing of
youth gang violence, particularly gun violence, in the Boston area. A time-series
analysis of youth homicide rates, calls for service from shots being fired and gun
assaults showed that Operation Ceasefire was associated with a significant reduction
in all three of these incident types – a reduction that was greater, and more specific to
youth gun violence, than that found in other major USA cities over the same period.
The authors of this evaluation indicated that the key to the success of this program
was the fact that the intervention directly targeted ‘high-risk’ offenders and
adequately communicated the increased threat to this group: “the pulling-levers
deterrence strategy controlled violence by focusing on particular groups that were
behaving violently, subjecting them to a range of discretionary criminal justice
system action, and directly communicating cause and effect to a very specific
audience” (Braga et al. 2001, p. 219).

Deterrence and the severity of formal punishment

The deterrent efficacy of increases in punishment severity is much less clear. Earlier
work on deterrence and the severity of formal punishment dealt predominantly with
the impact of capital punishment on homicide rates. State-sanctioned termination of
an offender’s life is a highly emotive and political topic and thus has attracted a great
deal of academic attention. This academic interest was boosted by the fact that many
western countries (with the exception of the USA) had abolished the death penalty by
the 1970s, which meant that deterrence hypotheses could be tested using ‘natural’ or
quasi-experimental research designs.

An example of this approach is the work undertaken by Barber and Wilson (1968) in
Australia. These researchers examined trends in homicide rates in Queensland before
and after the abolition of the death penalty in 1922, and compared these trends with
those occurring in two other Australian States, where capital punishment was still

7
available as a sentencing option (i.e. New South Wales and South Australia).
Contrary to deterrence hypotheses, this analysis showed a general downward trend in
homicide rates in Queensland over the 40-year period examined, which did not alter
to any great extent after the death penalty was removed from State legislation.
Meanwhile, the homicide rate in New South Wales (NSW) and South Australia
generally remained at or around the same level, fluctuating around a mean of 10
murders per million population. Furthermore, Queensland had, historically, recorded
a greater number of executions each year, on average, than other Australian States,
but at the same time experienced a much higher murder rate than either of the two
comparison States. While only a descriptive analysis, and therefore limited in its
ability to draw strong inferences, Barber and Wilson conclude from these results that
capital punishment is no more effective in deterring potential offenders than the
alternative of life imprisonment.

Most other empirical studies investigating the deterrent efficacy of capital

punishment have been conducted in the USA. This focus on the USA primarily stems
from the fact that (1) there is variability across USA States in the uptake and
application of the death penalty which allows for cross-sectional comparisons across
jurisdictions and (2) a moratorium on executions existed in the USA from 1968
through to 1977, after a USA Supreme Court decision ruled capital punishment
unconstitutional, which permits longitudinal, time-series analyses of the
effectiveness of different punishment regimes. The majority of studies employing
these types of analyses have found little evidence that the use of capital punishment
can have a positive impact on murder and/or manslaughter rates (for a review see
Chan & Oxley 2004).

For example, Peterson and Bailey (1988) examined the extent to which the
availability of the death penalty in contiguous USA States was able to predict State-
level murder rates and changes in these rates over the period 1973 through 1984.
They conducted a multivariate analysis in which yearly homicide rates were
regressed against the number of months where capital punishment was available in
each State and measures on seven State-level sociodemographic variables (e.g. %
metropolitan population; % poverty; % unemployment in each State). This analysis

8
found no relationship between the availability of capital punishment in any given
year and the average yearly homicide rate in that State, controlling for other
potentially confounding factors. In fact, a further comparative analysis undertaken by
these authors showed that in some cases, the relationship between the availability of
the death penalty and yearly homicide rates across States was in the opposite
direction to that predicted. That is, several States without statutory provisions for
capital punishment recorded homicide rates that were, on average, lower than their
neighbouring State that permitted capital punishment.

Given these findings on the efficacy of capital punishment, many criminologists,

sociologists and behavioural scientists have been quick to dismiss the severity of
punishment as unimportant in deterring offenders. This is illustrated by a position
paper on punishment and behaviour change issued by the Australian Psychological
Society in the 1990s. In this paper, Sanson and his colleagues concluded, “given that
capital punishment represents an extreme form of punishment, its failure to act as a
general deterrent suggests that milder forms of punishment are also unlikely to have
a strong general deterrence effect” (Sanson et al. 1996, p. 160).

However, the dismissal of a whole field of inquiry on the basis of the results from
capital punishment research alone would seem somewhat premature. Death penalty
research only comments on whether capital punishment is any more effective in
reducing homicides than the usual alternative of life imprisonment. It says nothing
about the deterrent effect of milder forms of punishment or the deterrent efficacy of
other types of legal sanctions. Evidence that the abolition of capital punishment has
little impact on aggregate-level murder rates therefore cannot be interpreted as proof
that reducing the penalty for homicide from life imprisonment to a shorter gaol term
will similarly result in no change in offender behaviour. Moreover, capital
punishment is, in most cases, reserved only for the most serious crimes, such as
homicide, which violate the moral code of society. Most people refrain from
committing these types of crimes, not because the costs associated with the formal
sanctioning process are too great, but because they believe the act to be immoral or
inherently wrong and one that will be harshly judged by other members of society
(Andenaes 1966; Gibbs 1975). There are other criminal behaviours, however, that do

9
not invoke the same moral approbation and it is possible that, for these offences,
changes in the magnitude of criminal penalties can have a significant deterrent effect
on rates of offending.

Much of the research available on the deterrent effect of non-capital sanctions

consists of studies investigating the relationship between imprisonment, either the
average length of prison sentences or the likelihood of being sentenced to custody,
and aggregate-level crime rates. A review of this research by von Hirsch and his
colleagues (1999) found only weak, negative correlations between these measures of
punishment severity and crime levels in different jurisdictions.

One of the earlier studies included in this review, which attracted considerable
attention in the deterrence debate (von Hirsch et al. 1999), was the work by
Farrington and Langan (1992).3 This research examined crime trends in the USA
between 1981 and 1986 and in England between 1981 and 1987, and compared
criminal justice practices in both these countries over the same time period. Unlike
many other aggregate-level studies, this study not only relied on police-recorded
offence data (which can be affected by public reporting and police recording
practices) to measure crime but also examined data from victim surveys, thereby
allowing more valid inferences to be drawn from the trend analysis. These data
revealed that, during the early to mid 1980s, England experienced a marked increase
in property crime but no change in violent crime, while the USA experienced a
reduction in both property and violent crime. However, an analysis of the severity of
punishment imposed for these offences revealed no consistent pattern to suggest any
association between the degree of onerousness of formal sanctions for
violent/property offences and rates of these crimes in each country. The three
criminal justice variables measuring punishment severity in this study were the
probability of custody given conviction, the length of imprisonment and average
prison time served. Only the first of these three measures was negatively correlated
with crime rates in England and the USA, and, while this negative relationship was
3
This study was the first of three studies conducted by Farrington and his colleagues which examined
the flow of offenders through the criminal justice system, from offending to incarceration, in several
different countries. All three studies generally found weak, and statistically non-significant,
associations between measures of punishment severity and crime levels (see von Hirsch et al. 1999 for
a review of this research).

10
apparent for all violence and property offences in the USA, in England it was
observed only for the offences of assault and motor vehicle theft.

Nevertheless, von Hirsch et al. (1999) are cautious in their conclusions regarding the
marginal deterrent effect of imprisonment, given that the majority of studies
conducted in this area suffer from several methodological weaknesses. Firstly,
comparisons across jurisdictions or across time must control for other possible
influences on crime rates, including social and economic factors. Yet often these
controls are missing from econometric analyses of the deterrent effect of punishment
severity or, where they have been included, are inappropriate measures of the factors
they supposedly represent. Secondly, many punishment severity studies have
specified prison population as the independent variable in the research but in doing
so fail to distinguish between the certainty and severity of formal punishment. A
tough jurisdiction, where a large proportion of offenders are sentenced to custody
and to long prison terms, could still have a relatively small prison population if the
risk of detection, apprehension and conviction is low in that locale. Thirdly, the bulk
of the empirical research available on the effectiveness of imprisonment as a
deterrent has adopted an ecological approach and, as noted previously, a major
problem with this type of research is the reciprocal relationship that potentially exists
between sanction levels and crime rates.

Spelman (2000), in his review of the literature on imprisonment and crime, identifies
four recent studies that successfully deal with many of the problems associated with
earlier ecological research on the efficacy of prison. All four of these studies found
that higher rates of imprisonment are associated with reduced crime levels and
estimated that a one per cent increase in the USA prison population could reduce
aggregate-level crime rates from anywhere between 0.16 and 0.31 per cent.

One of these studies, undertaken by Levitt (1996), is particularly noteworthy because

it applied a unique approach to deal with the simultaneity bias associated with
ecological research. In this study, Levitt (1996) examined the impact of prison
overcrowding litigation on crime rates in the USA. This litigation resulted in 12 USA
States having to reduce the size of their prison population after the USA Supreme

11
Court ruled overcrowding in prisons to be unconstitutional. Importantly, Levitt
showed that the overcrowding litigation had the intended effect on the size of prison
populations but crime rates, in the years prior to the courts decision, were not
predictive of litigation being filed in particular States. Thus, changes in prison
populations in these 12 USA States could be considered independent of the level of
crime in these jurisdictions. Examining annual crime trends for the years 1973
through 1993, Levitt then demonstrates that the change in prison populations
resulting from the litigation was strongly and negatively correlated with recorded
crime rates, controlling for several other potentially confounding variables (e.g.
unemployment, per capita income, per cent population in metropolitan areas). From
this data Levitt estimates that reducing the prison population by one offender is
associated with an increase of approximately 15 crimes per year, an estimated
elasticity that is two to three times higher than previous studies.

Prima facie, these results appear to support deterrence hypotheses but one other
plausible explanation still remains: incapacitation. States who reduce their prison
population may subsequently experience higher crime rates because motivated
offenders, who are released early, are no longer physically prevented from
committing crime, not because the broader population are less fearful of the
sanctioning process. This is a limitation shared by all four studies included in
Spelman’s (2000) review and one that is yet to be resolved empirically.

Even if we are unable to disentangle the effect of incapacitation from deterrence at

the aggregate level it is possible to examine the deterrent effect of imprisonment at
the individual level by considering the impact of longer sentence lengths on the
likelihood of an offender being reconvicted once released. Again, however, the
available evidence in this area is far from conclusive. A recent study by Spohn and
Holleran (2002) compared the recidivism rates for a group of 301 drug offenders
sentenced to prison with a group of 776 drug offenders who were sentenced to
probation and found no evidence that imprisonment reduced the likelihood of
recidivism. In fact this research showed that drug offenders who were sentenced to
prison had higher recidivism rates and reoffended sooner compared to drug offenders
sentenced to the non-custodial alternative. In contrast, research by Weinwrath and

12
Gartrell (2001) on 514 incarcerated drink-drivers in Alberta, Canada, found that
offenders sentenced to longer periods in detention were less likely to be reconvicted
for a new drink-driving offence than drink-drivers sentenced to shorter periods of
imprisonment, even after controlling for other factors relevant to sentencing and
offending (e.g. age, prior offences, treatment for alcohol addiction, marital status,
employment status and education). However, the relationship between sentence
length and recidivism found in this study was not linear, with the data suggesting the
existence of a punishment threshold (of approximately six months gaol) above which
no greater impact on reoffending was observed.

Limitations of aggregate-level deterrence research

Sherman’s (1990) research on police “crackdowns” and “blitzes” reviewed earlier in
this Chapter, suggests that variations in the probability of formal punishment can
deter some potential offenders, at least in the short term. However, the evidence
presented above on the deterrent efficacy of alterations to the severity of punishment
is equivocal. One reason for the inconsistency of research findings on the impact of
punishment severity is that much of the deterrence research on severity effects has
been undertaken using an ecological approach and, as previously discussed, this type
of research has several methodological limitations that preclude confident
conclusions regarding deterrent effects.

In addition to these methodological issues, there are two other important contextual
factors that may go some way to explaining the minimal, and at times conflicting,
evidence currently available on the deterrent efficacy of punishment severity. Firstly,
legislation that aims to increase the severity of formal punishment is often
unsuccessfully translated into sentencing practice. Policies that raise the severity of
statutory sanctions, for example, can sometimes result in little, or no, change in the
severity of penalties that offenders actually receive if found guilty. This may be
because players within the criminal justice system (e.g. police, prosecutors, judges,
jurors and defendants), believing the increase excessive, choose to charge fewer
offenders or charge offenders with less serious offences, increase plea bargaining,
reduce convictions or impose alternative sanctions, and in doing so undermine the
intended deterrent effect of the legislative changes. Barber and Wilson (1968)

13
highlight this implementation issue in their research on capital punishment in
Australia by showing that jurors are less likely to convict an offender in States where
the death penalty is more likely to be imposed. In this case, the threat of punishment
cannot be considered equal in each of these jurisdictions.

Secondly, to date, investigations of punishment severity conducted at the aggregate-

level have only considered punishment changes in a context that is generally not
conducive to deterrent outcomes. The deterrence model proposed by Becker (1968),
and indeed the deterrence doctrine proffered by the early utilitarian philosophers,
anticipated that punishment certainty and severity combine multiplicatively, such
that the severity of punishment matters most when punishment is likely to be
incurred. However, much deterrence research has been undertaken in jurisdictions, or
for offences, where the certainty of apprehension is relatively low (see von Hirsch et
al. 1999; Robinson & Darley 2004). In these instances, where offenders perceive it
unlikely that they will be caught, the severity of legal sanctions may matter little in
their decisions to offend.

This highlights the need to move beyond the simple question of whether punishment
deters crime. Gibbs (1975) notes that “all of the contradictory general observations
about deterrence can be reconciled by one empirical assertion: In some situations
some individuals are deterred from some crimes by some punishments” (Gibbs 1975,
p. 11). But he also observes that this statement is unfalsifiable, and therefore of no
theoretical value, unless we can identify (1) the individuals who are most deterred by
legal sanctions, (2) the sanctions that are most effective in deterring these
individuals, (3) the crimes that are most susceptible to deterrence-based interventions
and (4) the context in which sanctions work best. Accordingly, one of the major aims
of this thesis is to explore the conditions under which formal punishment is thought
to be most efficacious. Only when these conditions have been established can we
then undertake valid tests of deterrence hypotheses relating to the probability and
severity of legal sanctions.

14
Perceptual research in deterrence
The aggregate-level research summarised in the previous two sections has, by its
very nature another important limitation: the studies in question infer deterrent
effects; they do not directly measure them.

Suppose a study finds no evidence for a significant association between the

availability of capital punishment in a State and murder rates in that State. One might
conclude that the death penalty is no better in deterring offenders than life
imprisonment. But it is also possible that: (1) people perceive an increase in the cost
of committing murder but are not deterred (2) people perceive life imprisonment to
be a sanction that is as severe as the death penalty and therefore do not change their
offending behaviour or (3) there is a difference in perceived severity of the death
penalty compared to life imprisonment but the difference is only slight and unable to
be detected by aggregate-level measures.

A mass of research emerged during the late 1970s and throughout the 1980s, which
utilised survey data to investigate the relationship between risk perceptions and
behaviour. This research consisted of three types of studies: cross-sectional surveys,
which assess current perceptions of sanction risks and past or future intentions to
offend; longitudinal surveys, which measure current perceptions and then, at a later
point in time, subsequent criminal behaviour; and scenario-based designs, which
experimentally manipulate sanction risks within a hypothetical offending scenario
and then measure the likelihood of offending given these conditions.

The results from these studies vary somewhat depending upon the methodology
employed but are generally consistent with aggregate-level research in finding
positive effects for perceived certainty of detection, apprehension and conviction
(punishment certainty). That is, those persons who believe it likely that formal
punishment will be incurred are less likely to offend (e.g. Waldo & Chiricos 1972;
Grasmick & Bryjak 1980; Paternoster et al. 1982; Paternoster 1989; Bachman,
Paternoster & Ward 1992). Perceptual studies have also highlighted the role of
informal social controls in offender decision-making and the interdependency

15
between these informal sanctions and the risk of formal punishment (e.g. Paternoster
& Iovanni 1986; Williams & Hawkins 1986; Klepper & Nagin 1989a).

On the other hand, very few early perceptual studies have investigated the deterrent
effect of sentence severity and those that have, found only weak correlations between
perceptions of the severity of legal sanctions and past or future intentions to offend
(von Hirsch et al. 1999). One study that did examine the relationship between
perceived severity of formal punishment and criminal behaviour was a cross-
sectional study by Grasmick and Bryjak (1980). In this survey of over 400
respondents, people were asked about their prior involvement in various illegal
activities (e.g. theft, assault, tax fraud), the estimated chance of being arrested for
each of these offences and how much of a problem the resultant punishment would
be for them if caught and found guilty. Both the estimated chance of arrest and
ratings of punishment severity were inversely related to previous offending episodes.
Thus, those who had offended less often were respondents who thought it likely they
would be caught and receive a relatively severe punishment. More recent work by
Nagin and Pogarsky (2001) has also examined the marginal deterrent effect of the
perceived severity of legal sanctions but in this case a scenario-based methodology
was employed. This research found a negative association between perceived
severity of legal costs and the stated likelihood of offending in the circumstances
described but this relationship was weak and much less robust than that found for the
perceived certainty of apprehension and conviction.

While the perceptual deterrence research conducted thus far has made an important
contribution to our knowledge regarding risk perceptions and criminal behaviour,
this type of research has generally failed to consider how perceptions of risk are
formed. This is a limitation not only of studies examining the perceived severity of
formal punishment but throughout the perceptual deterrence literature (Nagin 1998;
von Hirsch et al. 1999). The importance of this neglected area of research becomes
immediately obvious when assessing the role of criminal justice policy in the
deterrence process. As Nagin observes, “even if crime decisions are influenced by
sanction risk perceptions… absence some linkage between policy and perceptions -
behavior is immune to policy manipulations” (Nagin 1998, p. 18). Research

16
investigating the link between experiences with, or knowledge of, the criminal justice
system and perceptions of the formal sanctioning process is therefore needed before
we can conclude that criminal justice practice has the potential to affect offender
behaviour through deterrent-like mechanisms.

The current study

The purpose of this thesis is to contribute to the debate on punishment severity by
assessing the impact of an increase in the severity of prescribed legal sanctions for
drink-driving offences in New South Wales, Australia. Specifically, the research
question to be addressed is whether the higher penalties introduced by the NSW State
Government in 1998 were a marginal deterrent for drink-drivers and, if so, by how
much. Positioning this research within a deterrence framework ensures that a small
number of hypotheses can be generated from previous research and empirically
tested. Deterrence is also generally considered one of the major aims of legal
sanctions prescribed for drink-driving offences (Homel 1980; Ross 1984), has been a
potent force in shaping driver behaviour in Australia over the last two decades or
more (Homel 2004) and is viewed by the Australian public as the most effective
strategy to address drink-driving and related issues (Australian Institute of Health
and Welfare 2002a).

The case study that has been selected for this analysis is also a good test of
deterrence because it is set in a context where several of the preconditions for
deterrence success have been met. First of all, drink-driving is an offence that does
not attract the same moral condemnation as other types of crime, such as homicide or
robbery, and is therefore, potentially more susceptible to changes in the probability
and severity of formal punishment than are other criminal behaviours. Furthermore,
in Australia, very few people are gaoled for a drink-driving offence. Thus, we can be
more confident that any reduction in aggregate-level crime rates emanating from a
change in sentencing policy is due to deterrence rather than incapacitation. In
Australian jurisdictions there also exists a relatively high perceived risk of detection
and apprehension for drink-driving, created by the introduction, and the continued
publicity, of random breath testing by police. This high probability of apprehension
is not true of many other offences (Robinson & Darley 2004) or, as will be seen later

17
in this thesis, the offence of drink-driving in other countries. Finally, the measure
typically used to monitor trends in drink-driving behaviour (i.e. road crash rates) is
less prone to bias as a result of underreporting or enforcement practice. It is therefore
a more valid and reliable measure of offending than other available crime indicators,
such as police-recorded arrest data. These features of the selected case study render it
a more definitive test of the marginal deterrent effect of punishment severity than
preceding work that has been undertaken in this area.

This work also builds on previous deterrence research by addressing several

methodological limitations of punishment severity studies. Firstly, it employs a
quasi-experimental research design to examine the impact of the penalty changes on
offending rates in New South Wales. This type of study is able to overcome many of
the problems associated with ecological research because it incorporates several
features of a true experimental design, including a specific well-defined intervention,
comparison groups to control for other possible influences on crime rates and, in the
case of interrupted time-series analyses, many observations before and after the
intervention to distinguish deterrent effects from other fluctuations in crime levels.
Focusing on one specific, well-defined intervention also allows the researcher to
consider implementation issues associated with criminal justice policies within a
realistic setting. Thus, another important component of this thesis concerns how well
the 1998 sentencing policy was translated into a credible threat. In particular, we
address the question of what impact the penalty changes had on actual sentencing
practice and what potential, if any, the policy had to affect perceptions of sanction
severity. These aspects of the research design will improve our ability to make
confident conclusions regarding deterrent effects.

Thesis structure
This thesis begins with a more detailed description of deterrence theory and a
discussion of the basic assumptions underpinning the theory. Chapter 2 outlines
some of the major criticisms of the rational choice model of offender behaviour and
explains how these criticisms have been addressed in modern deterrence theory. It
also explores the conditions thought to affect the efficacy of formal punishment and

18
explains how the offence of drink-driving, which meets several of these conditions,
can be used as a case study to examine deterrent effects.

Previous deterrence research that has been conducted in the drink-driving field is
reviewed in Chapter 3. The studies described in this chapter present evidence
regarding the deterrent effect of changes in the certainty, severity and celerity of
formal punishment for drink-driving offences. Both aggregate-level and individual-
level research are included in this review, as are several methodological limitations
of the drink-driving studies conducted thus far. This section also highlights some
knowledge gaps identified in the broader deterrence literature.

The focus of the thesis then shifts towards the 1998 sentencing policy that increased
statutory penalties for drink-driving offences in New South Wales and its effect on
offender behaviour. In Chapter 4, the case study is described in greater detail and the
methodology employed to investigate the impact of the penalty increase is discussed.
The results of two major studies undertaken to examine any changes in drink-driving
rates which coincided with the increase in punishment severity are presented in
Chapter 5. The first study involves an interrupted time-series analysis of alcohol-
related road crashes before and after the legislative changes and a comparison of
changes in these crashes with those apparent in rates of non alcohol-related crashes
over the same time period. The second study involves a comparison of the
reoffending rates of two cohorts of drink-drivers sentenced before and after the
penalty changes. The second study also verifies the extent to which the 1998
legislative amendments affected actual sentencing practice and considers any
countervailing effects of the policy on the prosecution and conviction of drink-
drivers in New South Wales.

Chapter 6 provides further information about the context in which the sentencing
policy was implemented. Specifically, it examines data bearing on the level of
enforcement of drink-driving offences at and around the time the penalties were
raised and on the level of publicity the sentencing policy received in the media at the
time of implementation. The issues of enforcement practice and public knowledge of
the penalty changes are further explored in the third study undertaken for this thesis,

19
the results of which are presented in Chapter 7. This study uses a scenario-based
survey approach to examine the relationship between knowledge of applicable drink-
driving penalties, perceptions of formal punishment for drink-driving offences and
the stated likelihood of offending. This study builds on the previous two studies by
examining, at the individual-level, the link between sentencing policy in New South
Wales, perceptions of punishment severity and drink-driving behaviour. It also
presents an opportunity to examine the deterrent effect of punishment severity whilst
controlling for other potentially confounding contextual factors, such as the level of
police enforcement targeting drink-drivers.

Finally, in Chapter 8, the results from each of the three studies assessing the deterrent
effect of the 1998 sentencing policy are summarised and the implications of these
findings for deterrence theory, criminal justice policy and future research in this area
are discussed.

20
CHAPTER 2. DETERRENCE THEORY

Deterrence occurs when a person refrains from committing an offence because they
fear the negative consequences associated with the act. Typically, it is the legal
sanctions prescribed by the criminal justice system which constitute the subject
matter for deterrence research. However, as discussed later in this chapter, informal
sanctions or “extralegal” consequences associated with offending are also potential
costs that can influence an individual’s decision to offend. Without dismissing the
importance of these other types of sanctions, the primary focus of the current thesis is
the deterrent efficacy of legal sanctions. The findings from this analysis will have
direct implications not only for deterrence theory but also for criminal justice policy
more generally.

A methodological problem for researchers attempting to investigate the deterrent

efficacy of legal punishment is that, by definition, deterrence denotes the absence of
a behaviour and is therefore, essentially an unobservable phenomenon. This makes
statements such as “punishment deters crime” difficult to falsify unless they are
situated in the broader context of a systematic theory (Gibbs 1975). Deterrence
theorists have thus sought to identify the necessary conditions under which
deterrence occurs, including the types of offences, sanctions, individuals and
contexts that are conducive to deterrent outcomes. Explication of these conditions, a
priori, allows one to formulate predictions regarding the deterrent efficacy of
particular sanctions and modifications to those sanctions (Beyleveld 1979).

This chapter endeavours to explore the factors that are thought to be influential in the
deterrence process. It begins by presenting a typology of deterrent effects to more
clearly define relevant concepts. The expected utility model of offender decision-
making, upon which modern deterrence theory is based, is then described and the
limitations of this rational choice model of criminal behaviour are discussed. Other
factors that may influence the deterrent effect of state-imposed punishment are also
highlighted, such as the applicability of the deterrence model to different types of

21
crimes and the potential for punishment variables to interact with contextual factors
and offender characteristics.

Definitions of deterrence
General and specific (or special) deterrence are key concepts in the deterrence
literature. General deterrence refers to punishment that discourages members of the
public, who learn of the sanctions, from violating the law. Specific deterrence refers
to the effect that punishment has on the criminal behaviour of those who are
punished. Some deterrence theorists claim that general and specific deterrence are
not separate mechanisms by which legal sanctions act to reduce offending, but
involve the same mechanism operating within different populations (Zimring &
Hawkins 1973; Beyleveld 1979; Stafford & Warr 1993). Zimring and Hawkins
(1973), for example, describe previous experience of punishment as just “one of an
enormous variety of factors which condition threat responsiveness” (Zimring &
Hawkins 1973, p. 225). The distinction between specific and general deterrence is,
nevertheless, one that is pervasive throughout the literature on deterrence.

Gibbs (1975) argues that it is too narrow in scope to classify punishment effects as
simply general versus specific deterrence. He proposes, instead, a typology
consisting of 16 different situations in which an individual may consider committing
a criminal offence. This typology classifies deterrent effects according to an
individual’s previous experience with punishment for the particular crime type under
consideration, their experience with punishment for other types of crime and their
offending history, both in terms of the crime in question and other criminal offences.
Gibbs contends that a theory could be formulated for each of the 16 different types of
deterrence that result from these four levels of legal threat and punishment
experience.

A simpler typology offered by Gibbs (1975), which is sufficient for the present
purposes, distinguishes between three main types of deterrence: absolute, restrictive
and specific deterrence. Absolute deterrence occurs in instances where an individual
refrains completely from a particular criminal behaviour because of the risk of being
sanctioned. Restrictive deterrence occurs when an individual curtails, or reduces the

22
frequency of, criminal behaviour during some period in order to lessen the risk of
being sanctioned. These two types of deterrence can be equated with the broader
concept of general deterrence, described above. The definition of specific deterrence
remains the same as that given previously: the omission, or curtailment of criminal
behaviour by an individual who has previously been punished and fears being
punished again.

The distinction between absolute and restrictive deterrence is based on the frequency
with which an individual offends, given the existence of a specific legal threat.
Zimring and Hawkins (1973), on the other hand, make a distinction between the
deterrent effect of introducing sanctions where none had previously existed (absolute
deterrence) and the deterrent effect of incremental changes to an already existing
threat (marginal deterrence). Gibbs (1975) asserts that this distinction between
absolute and marginal deterrence refers to the size of the threat posed by the criminal
justice system rather than two distinctive types of deterrence and thus, should be
replaced by the terms absolute deterrent efficacy and marginal deterrent efficacy of
punishment.

The absolute and marginal deterrent efficacy of legal sanctions is a crucial

distinction to make when assessing the impact of penal policies on rates of offending.
Much deterrence research has been interested in comparing the rate of crime and the
levels of penalty severity across different jurisdictions, thus focusing on the marginal
deterrent efficacy of punishment. Advocates for decriminalisation (or alternatively
prohibition) have sometimes used findings from this type of research to argue that
the removal of some sanction will have no impact on criminal behaviour. Such
arguments are misguided. In order to assess the absolute deterrent efficacy of
criminal penalties it is necessary to evaluate policies that remove a legal threat
completely, or introduce a threat where none had previously existed. Because these
types of policies are rarely implemented, deterrence research is mainly focused on
the marginal deterrent efficacy of legal sanctions.

The distinction made by Zimring and Hawkins (1973) between absolute and
marginal deterrent efficacy also highlights the need to clearly define the nature of the

23
threat being investigated by deterrence research. As discussed above, state-imposed
punishment is the primary focus of the current research. An example of this research
would be a comparison of the murder rate across States or counties that have capital
punishment for murder with those that do not. However, this type of research does
not adequately deal with variations in the application of statutory punishment across
jurisdictions. In one State, the criminal courts may never impose capital punishment
for murder. In another, several convicted murderers may be on “death row” at any
one time. The threat generated by capital punishment legislation in these two
jurisdictions cannot be considered equal. Deterrence research needs to consider both
the impact of legal punishment appearing in statutes and the translation of this
legislation into actual sentencing practice.

A further consideration is people’s beliefs about legal punishment. It is often said

that legal sanctions operate to deter potential offenders by making them fearful of the
consequences of involvement in crime. If the costs being examined are those
threatened by the criminal justice system, then this statement assumes that the
potential offender has some knowledge, however limited, of the actual or prescribed
punishment for that offence. Yet deterrence could also occur if a potential offender
mistakenly believes that legal sanctions exist for a particular behaviour or
overestimates the magnitude of the penalty attached to an offence. This Beyleveld
(1979) labels deterrence by merely believed sanctions to contrast it with deterrence
by prescribed or actual sanctions. While Beyleveld notes that this distinction is of
theoretical significance, he also concedes that only the deterrent effect of prescribed
or actual sanctions will be of relevance to deterrence policies. It does, however,
illustrate the importance of legal knowledge in maximising the effectiveness of
deterrence-based penal policies, an issue that will be discussed in greater detail
throughout this thesis.

One final problem for research investigating the deterrent effects of legal sanctions is
that deterrence is just one of several mechanisms by which legal sanctions can serve
to prevent offending. Gibbs (1975) identifies at least nine other ‘preventive
consequences’ or ‘preventive mechanisms’ of punishment in addition to deterrence.
Three of these, retribution (exacting vengeance through legal means to prevent acts

24
of revenge), rehabilitation (changing an offender’s motivation to commit crime) and
incapacitation (physically preventing an offender from committing further crimes),
are commonly identified as major aims of criminal punishment. However, there are
several other means by which legal sanctions can serve to reduce crime. For
example, the introduction or presence of a legal sanction may demonstrate that a
particular act is wrong and is condemned by the wider community. If the behaviour
is punished regularly and severely, this shared belief will be reinforced and
potentially lead to a reduction in offending (i.e. Normative Validation).
Alternatively, sanctions could further public knowledge that a certain behaviour is
illegal. This knowledge, combined with an unquestioning acceptance of the law, may
improve the legal compliance of some citizens (i.e. Enculturation). These
preventative mechanisms of punishment resemble Andenaes (1974) concept of
general prevention. Unlike general deterrence, which only refers to compliance
resulting from fear of punishment, general prevention refers to a broader range of
controls that include moral inhibitions and socialisation of preventive habits.

Distinguishing deterrent from other long-term preventative effects of punishment in

research, is complicated by the fact that the short-term threat of general deterrence
could achieve these broader social controls. As an example, compliance could
initially be obtained through fear of punishment but eventually obtained because a
social norm develops that inhibits the proscribed behaviour and promotes voluntary
compliance with the law (Ross 1984). Similarly, the threat of punishment may have a
habituative effect, such that adherence to the law is initially secured through fear but
repetition of the law-abiding conduct eventually results in the behaviour becoming
routine practice (e.g. seat-belt laws; Elliott 2003).

This thesis deals with simple or direct deterrence, that is, behaviour changes resulting
from fear of threatened punishment. However, the contribution of other preventative
mechanisms of punishment, in securing legal compliance, should be kept in mind
when considering the implications of the current research. Deterrence is just one of
numerous reasons why people comply with the law. Many citizens refrain from
offending simply out of a moral obligation to obey the law or a general acceptance of
the legitimacy of authority (Tyler 1990). On the other hand, there is a proportion of

25
the population for whom these broader social controls are ineffective, but who are
responsive to other preventative mechanisms of punishment, including deterrence.
Nagin and Paternoster (1994) categorise these persons as “marginal offenders”,
individuals whose self-control is not high enough to totally eliminate the possibility
of breaking the law but neither low enough to make them insensitive to punishment.
It is this group of potential offenders who would be the most responsive to changes
in the threat of punishment and who are thus the target group for deterrence-based
interventions.

Modern deterrence theory

As discussed in Chapter 1, modern deterrence theory is an application of the rational-
choice paradigm of decision-making. It assumes that human beings are essentially
‘rational actors’ who, in circumstances of uncertainty, evaluate the consequences of
alternative choices and estimate the likelihood of their occurrence. The alternative
selected is the one that maximises the expected utility, that is the option maximising
the benefits and minimising the costs to the individual. An important component of
this calculation is that the utility of each alternative relies on a subjective evaluation
of the value of each outcome. For this reason individuals may evaluate the ‘profit’
and ‘loss’ associated with equivalent alternatives differently and may, therefore, vary
in how they respond to similar rewards and punishments. This notion that people
make subjective evaluations rather than relying on objective values differentiates
expected utility theory from expected value theory, though both theories assume that
the probability of an outcome occurring and the value assigned to that outcome
combine in precisely the same way.

Expected utility maximisation as applied to decision-making in a criminal context

predicts that a person will engage in a criminal act if the subjective expected utility
of the criminal act E(U)c is greater than the subjective expected utility of other
legitimate alternatives E(U)a. The subjective expected utility of a crime is calculated
firstly by evaluating the gains associated with the successful completion of the crime
and the losses associated with legal sanctioning if caught. The criminal gains include
tangible goods (e.g. cash) or avoided inconveniences (e.g. having to catch public
transport home in the case of drink-driving). They also include the anticipated

26
benefits of engaging in the behaviour itself (e.g. the thrill obtained from speeding in
a car). The losses or costs, on the other hand, are those imposed by the criminal
justice system (though crime costs have more recently been extended to include
informal social sanctions that may arise from arrest or criminal conviction). As stated
previously, the value assigned to these gains and losses will depend on numerous
factors, such as current wealth status, and thus utility calculations will vary across
individuals.

To determine the expected utility of a course of action an individual must also

calculate the probability of the outcome occurring, that is, the chance that he or she
will be rewarded or sanctioned if he or she were to engage in the behaviour. In a
criminal context, this estimate of outcome likelihood is generally thought to be
subjective in nature (i.e. the perceived probability of being arrested) rather than
objective (i.e. the actual risk of being arrested for an offence), but is otherwise
thought to adhere to the general rules of probability theory (note that the use of
subjective probabilities distinguishes subjective expected utility theory from
expected utility theory).

These two steps, assessment of expected outcomes and the application of subjective
probabilities, can be represented, more formally, by the following equation:
E(U)c = P(G)cU(G)c + P(L)cU(L)c (1)
where E(U)c is the expected utility of committing the crime, P(G)c the subjective
probability of obtaining the gains associated with the crime, U(G)c is the utility of
those gains, P(L)c is the subjective probability of receiving legal sanctions if caught
and U(L)c is the disutility of legal sanctioning if caught. Deterrence theory focuses
attention on the variables affecting the disutility of crime, which is represented in the
right-hand side of the above equation. This theory states that to discourage an
individual from committing an offence, the perceived severity (U(L)c) and the
perceived certainty (P(L)c) of the punishment associated with that crime must be
sufficiently large to reduce its subjective expected utility below that of the
anticipated gains. Thus, an individual will refrain from offending if: P(G)cU(G)c <
P(L)cU(L)c.

27
Importantly, however, as shown in equation 1, the utility of formal punishment and
the subjective probability of it occurring are thought to combine multiplicatively
rather than additively. This means that the deterrent effect of a change in sanction
severity (U(L)c) depends upon the perceived likelihood of the sanction (P(L)c). If the
risk of apprehension is zero then no level of sanction severity will deter an individual
from committing a crime. Likewise, in the absence of sanctions, the risk of
apprehension will not exert a deterrent effect.

Expanding the Subjective Expected Utility Model

According to classical deterrence theory the efficacy of legal sanctions is a function
of the perceived celerity of punishment. That is, swiftly delivered punishment will
deter an individual more than punishment occurring some time in the distant future.
While a time variable is not included in the traditional subjective expected utility
equation, the deterrence model has recently been extended to include the economic
concept of a “discount rate” to account for the hypothesised celerity effect of
punishment (Kenkel 1993; Nagin & Pogarsky 2001). The concept of a discount rate
being used to devalue future costs is derived from economic reasoning about the
“time value of money”. According to this reasoning, a given sum of money is worth
more now to an individual than it is in the future because it can be used to obtain
immediate benefits. Given the choice of receiving $1000 now or $1000 a year later, a
rational actor will always select the former option. This is because any delay in
receiving the immediate benefits that could be derived from the $1000 will reduce
the current worth of that money (example adapted from Nagin & Pogarsky 2001). In
financial decision-making, the size of an immediate reward that just balances the
value of a delayed reward is called the “present value” of the delayed reward. The
more delayed the reward is, the smaller its present value. Similarly, the more delayed
a cost is, the smaller its present value.

If the process of evaluating future punishment is similar to financial decision-

making, long delays in the imposition of punishment for an offence would reduce the
present value of the perceived costs of committing the crime. The consequent
reduction in the costs associated with the criminal behaviour would diminish the
deterrent effects of the prescribed sanctions for that offence. In economic decision-

28
making the devaluation of future costs is both a function of the delay period and
some “discount rate”, which is typically the current market interest rate. In an
analogous way, Nagin and Pogarsky (2001) and Kenkel (1993) model the magnitude
of the celerity effect of punishment as a function of both the length of delay in
punishment and the extent to which an individual prefers to delay punishment. The
relationship is expressed in the following equation:
Dt = 1/(1+r)t (2)
where r is the degree to which an individual devalues future consequences and t is
the delay before the penalty is imposed.

Nagin and Pogarsky (2001) argue that preference for delay (r in the above equation)
can be equated with an individual’s “impulsivity” or “present-orientation”. A person
who is more concerned with what happens to them in the short run will devalue
future punishment to a greater extent than someone who is more concerned about
long-term consequences. Thus, the disutility of a crime not only decreases as a
function of long delays in the imposition of penalties but also as a function of an
individual’s preference for delay or present-orientedness. In the expanded deterrence
model an individual will offend if the expected gains associated with the crime
outweigh the discounted expected costs associated with legal sanctioning if caught:

EU(G)c > P(L)cU(L)cDt (3)

Thus, ceretis paribus, any increase in the severity, certainty or celerity of punishment
for an offence would increase the disutility of the crime and therefore reduce the
prevalence of the criminal behaviour.

Critiques of the Rational-Choice Perspective

Several authors have criticised the rational choice perspective of offending outlined
above on the basis that people do not always combine information in the manner
predicted by the expected utility equation, an assertion that has been supported by
experimental research on human decision-making under uncertainty.

For example, Tversky and Kahneman (1981) presented people with a public health
dilemma in which 600 lives were at risk and asked whether they would prefer to

29
definitely save 200 lives or take a gamble that offered a one-third chance that all 600
lives would be saved. They found that the majority of participants were risk averse in
this scenario, that is, most favoured the certainty of definitely saving 200 lives over
the gamble that potentially saves all 600. However, when participants were presented
the same scenario, but with a slight change of wording from ‘lives saved’ to ‘lives
lost’, this preference for risk aversion shifted to one of risk seeking. In the second
scenario, the majority of participants favoured the gamble offering a one-third
chance that no one would die rather than the certain loss of 400 lives. The finding
that an option with the same probability and outcome can be accepted in one
situation but rejected in another, depending on how the outcomes are described or
“framed”, is at odds with the subjective expected utility model.

Carroll (1978) also provides experimental evidence on human decision-making

which is inconsistent with subjective expected utility theory. He presented a group of
offenders and a group of nonoffenders with a series of criminal opportunities,
varying on four of the dimensions featured in deterrence theory: the probability of a
successful crime, the amount of money to be gained if successful, the probability of
capture, and the penalty if caught. When deciding whether or not to offend given
these conditions, Carroll discovered that participants tended to focus on only one
dimension of the problem rather than fully integrating information on all four
dimensions. The dimension proving to be the most influential in the decision-making
process varied considerably across participants and did not seem to be affected by
previous criminal experience. In general, however, he found that people tended to
place greater weight on monetary incentives associated with the crime rather than
any potential criminal penalties that may be incurred if caught. Tittle (1977) also
found a tendency for individuals to focus on the utility or benefits of the crime, rather
than evaluating all available information, in his cross-sectional household survey
aiming to predict future deviance of almost 2,000 USA residents.

From these, and other similar studies appearing in the behaviour-decision literature,
Johnson and Payne (1986) identify two important stages in the decision process
where criminal choices predictably deviate from the subjective expected utility

30
model: (1) the way in which an individual constructs a representation of the problem
and (2) the way in which an individual evaluates different alternatives.

In discussing problem representation, Johnson and Payne draw on principles of

prospect theory, a descriptive model of decisions under uncertainty formulated by
Tversky and Kahneman (1981). Prospect theory proposes that the subjective value
assigned to a particular outcome will vary depending on whether the choice is framed
as one between possible gains or possible losses. For gains, the utility or value
function is thought to be concave, but for losses, the utility or value function is
convex. Thus, in contrast to the expected utility model which proposes that people
will always tend to be risk averse, prospect theory suggests that decision makers will
tend to be risk averse for gains but risk seeking for losses. Furthermore, in Prospect
Theory the value function for losses is assumed to be steeper than the value function
for gains, thus the loss of a sum of money will be considered more painful than the
pleasure of winning that same amount. Tversky and Kahneman’s public health
dilemma, presented previously, exemplifies these deviations from expected utility
predictions. When the choice presented to participants was phrased in terms of the
number of lives that could be saved, people opted for the certainty of saving 200
lives over the one-third gamble of saving all 600 lives (risk averse). On the other
hand, when the problem was presented in terms of lives lost, people were much more
willing to accept the gamble in order to prevent any loss of life (risk seeking). For
criminal decisions, this could mean that an individual, who chooses not to commit a
crime for gain, might still commit the offence if it means avoiding a possible loss
(e.g. underreporting your taxable income to reduce your tax liability but not to
increase your tax refund, Johnson & Payne 1986).

Prospect theory also proposes that people do not adhere to the rules of probability
theory when constructing a representation of the problem. Instead, individuals will
tend to apply a subjective evaluation or weighting function to outcome likelihood
estimates just as they apply a subjective value function to gains and losses. For
example, this weighting function has individuals discounting altogether near zero
probabilities and overweighting moderately low probabilities. Thus, if there is very
little chance of being apprehended for an offence, this will be perceived as equivalent

31
to zero and enforcement will have no impact on offending behaviour. But if police
are able to alter perceptions to such an extent that the probability of apprehension is
believed to be ‘moderately’ low, then people will tend to exaggerate (out of
proportion) apprehension risk and the success of the campaign in reducing offending
will exceed subjective expected utility predictions.

The second theme in Johnson and Payne’s (1986) discussion of decision-making

concerns the processes that operate on these problem representations. Several factors
in everyday life, such as time pressures, can impinge on available cognitive
resources, thereby reducing the capacity of an individual to evaluate all available
choice alternatives. As a consequence, individuals will tend to apply various
strategies to simplify the task of outcome selection. One common heuristic described
by Johnson and Payne is the noncompensatory strategy. This involves screening
various alternatives on one or two attributes, choosing or rejecting each alternative
on the basis of these attributes alone and then evaluating the selected alternatives
more carefully. This screening approach to decision-making would be most effective
when cognitive demands are high, such as occurs when an individual is presented
with a complex problem or is constrained by time. This is consistent with Carroll’s
(1978) finding that many people are unidimensional in their approach to criminal
decisions, focusing on only one salient feature of the problem presented to them.

Johnson and Payne’s (1986) descriptive model of decision-making is far from an all-
encompassing, systematic theory of offender decision-making but it does serve to
highlight several ways in which criminal decisions may deviate from subjective
expected utility predictions. Cognitive psychology has shown that information is
processed in a memory that has limited capacity, duration and ability to permanently
store relevant information. This results in heuristics or short-cuts being employed,
where necessary, to cope with large amounts of complex information in an
economical fashion. Recognising these limitations, most deterrence theorists have
long since abandoned the idea that the analysis of criminal opportunities involves a
complex mathematical process of utility calculations and probability estimations.
What they do propose, however, is that “within the limits of their inclination and
ability to make such calculations (an individual) will act so as to pursue whatever is

32
at the moment, in their own mind, the most important” (Stover & Brown 1975, p.
369). This feature of modern deterrence theory is referred to as “limited” or
“bounded rationality” (Simon 1983), in which the “incredibly economic man of
decision-making theory (is replaced) with a choosing organism of only limited
knowledge and ability” (Douglas & Wildavsky 1982, p. 77). In this way, although
individuals may not calculate the utility of all available alternatives in precisely the
same way as economists describe, they can be regarded as active decision-makers
who are affected by information that is supplied to them.

Bounded rationality recognises a person’s limited capacity to process and acquire

information. An individual will often try to avoid exhaustive, complex calculations
and instead focus on a few simple and concrete aspects of a problem with the aim to
‘satisfy’, rather than optimise, their needs or wants (Johnson & Payne 1986). Their
failure to seek out all relevant information also means that their knowledge of risks
and rewards associated with various alternatives may be incomplete or biased by the
channels through which the information is sourced. This results in behaviour
decisions that are far from optimal. Carroll (1978) and Cook (1980) maintain,
however, that legal punishment can still modify behaviour through deterrent-like
mechanisms by influencing the “standing decisions”, or rules of thumb, that people
adopt in relation to offending. A person may decide, for example, that they will not
drive above the speed limit under any circumstances because they fear losing their
driver’s licence if caught. Adopting this “standing decision” would eliminate the
need for the driver to undertake unnecessary, complex utility calculations whenever
presented thereafter with an opportunity to offend. An increase in the severity of a
legal threat would aim to persuade more people to adopt such a position in relation to
their driving behaviour. Even in instances where information-processing abilities are
reduced, Cook (1980) suggests that people can still be guided by these standing
decisions because they have been derived through prior contemplation, presumably
in a more rational state.

Cook’s (1980) description of “standing decisions” is indicative of theorists’ attempts

to apply the rational choice perspective, and by extension, the deterrence model not
only to criminal events (crimes) but also to criminal involvement (or criminality).

33
This distinguishes deterrence from other criminological theories that also place
importance on situational determinants of criminal behaviour, such as routine activity
theory (Clarke & Felson 1993). Both routine activity and the rational choice
perspective of offender behaviour view crime as purposive behaviour in which an
offender makes various decisions or choices about the relative attractiveness of
criminal acts. The latter, however, is interested not only in decisions surrounding the
commission of an offence but also decisions to adopt crime as a ‘way of life’. These
criminal involvement decisions entail weighing up of risks and rewards over an
extended period of time, rather than simply assessing information that relates to the
immediate situation, and are likely to be influenced by different factors than those
utilised in event decisions, including any anticipated informal social sanctions
associated with offending (Clarke & Felson 1993).

While it would be wrong to dismiss deterrence theory on the basis that human
decision-making often departs from the mathematical utility equations, it would be
equally misleading to claim that deterrence theory is a complete explanation of
criminal behaviour. Like most other criminological theories, deterrence fails to
account for each and every decision to offend. Individual differences exist in
willingness to take risks, in objective circumstances and in the assessment of costs
and benefits. Rationality in offender decision-making can also sometimes be
compromised by limited knowledge and ability to process relevant information.
However, not all potential offenders need to act predictably all of the time, for
deterrence theory to be useful in guiding policy and research goals. As Cook (1980)
argues, deterrence is concerned with aggregate behaviour. Thus all that is necessary
for deterrence assumptions to hold true is that a proportion of criminal acts be the
result of a rational decision process (or the result of a prior rational decision to offend
in certain circumstances) in which potential offenders consider the possible risks and
rewards associated with the behaviour.

Deterrence and specific offences

Crime covers a wide range of behaviours that have in common only the fact that they
are prohibited by the law. Any theory that attempts to model offender decision-
making must therefore take into account differences in the situational context in

34
which crime is committed, as well as the purpose or motivation of offenders at the
time a crime is committed. Recognising these important differences across various
offence types, deterrence scholars have proposed that the threat of punishment will
necessarily be more important for some types of criminal behaviour than for others.
As Zimring and Hawkins (1973) note “the jealous husband bent on homicide is less
likely to be deterred by legal threat than the …bank clerk who views embezzlement
as a means of securing funds to indulge in speculation” (Zimring & Hawkins 1973,
p. 130).

A distinction appearing in criminal law, which reflects this differential

responsiveness to formal punishment, is that which classifies criminal acts as either
mala per se or mala quia prohibita. The former are actions that are immoral in their
own right (e.g. murder, incest) while the latter are actions that are illegal merely
because they are prohibited by the law (e.g. tax evasion, marijuana use). Andenaes
(1966) contends that since mala per se crimes are supported by the moral code of
society, moral feelings and the fear of social judgement, rather than fear of formal
punishment, act to prevent an individual from committing the crime. For mala quia
prohibita, on the other hand, the law stands alone as a preventative force: so
conformity is only secured by effective legal sanctions. Rates of tax evasion, on this
argument, should be more sensitive to changes in the certainty and severity of legal
punishment than rates of incest.

Other scholars have made the distinction between instrumental crime, which is
committed in order to attain some other goal, and expressive crime, which is
committed because it is pleasurable in and of itself (Chambliss 1967; Zimring &
Hawkins 1973). This distinction focuses attention on offender motivation or purpose
for engaging in the criminal behaviour, as well as the emotional state of the offender
at the time the decision is made. Instrumental crimes, such as parking violations, are
thought to be more normally motivated and rational and are, therefore, more likely to
be susceptible to the threat of punishment. On the other hand, expressive crimes,
such as assault or murder, are often committed on impulse and in contexts of high
emotional arousal. In this situation, where there is little contemplation of the
consequences of the behaviour, punishment is likely to be a weak deterrent.

35
It is easy, however, to imagine crimes which do not fall neatly into the two categories
described above. For example, a domestic violence victim may execute a plan to
murder his or her partner, taking into consideration the risks associated with different
aspects of the crime location and modus operandi. Alternatively, a dependent drug
user who steals valuables from a retail store in order to obtain cash may engage in
very little contemplation of the potential risks because of the need to alleviate
physically painful withdrawal symptoms. Considerations like these prompt Zimring
and Hawkins (1973) to argue that, rather than deterrence functioning for only one
group of offences and not for others, deterrence is simply more difficult under certain
circumstances than others. In situations where offenders are highly motivated,
making decisions in a context of high emotional arousal and where there are no other
major barriers to carrying out the crime, the deterrent effect of legal punishment is
likely to be significantly reduced regardless of the offence. It is because mala per se
and “expressive” crimes often fit this description, that formal sanctions would
seldom deter these offence types.

The role of informal social control

The distinction between mala per se and mala quia prohibita also highlights the
importance of informal social control in understanding compliance with the law.
Traditionally, deterrence theorists have focused only on the negative consequences
threatened by the criminal justice system, in particular the legal punishment
prescribed by legislatures and/or imposed by the courts. However, other
criminological theories have stressed the role of social influences in controlling
delinquent behaviour and the importance of social and familial norms in explaining
participation in and desistance from crime (e.g. Akers et al. 1979; Durkheim 1964;
Hirschi 1969).

Two such “extralegal” sources of conformity that have been identified in the
deterrence literature are moral commitment to social norms and the threat of social
stigma (Meier & Johnson 1977; Grasmick & Bursik 1990; Baum 1999; Nagin &
Pogarsky 2001). For an individual with a strong belief that illicit conduct is wrong,
the psychological dissonance that flows from violating an internalised norm, and the

36
associated feelings of shame, could operate to discourage offending. Likewise, the
threat of disapproval from peers, family and colleagues may cause an individual to
refrain from committing a crime more than any formal sanction.

While evidence from survey data has demonstrated that these extralegal factors are a
significant deterrent to crime (Grasmick & Bursik 1990; Grasmick, Bursik &
Arneklev 1993; Baum 1999), it is difficult to disentangle the independent effects of
informal social controls on criminal behaviour from those arising from the threat of
formal punishment. For example, Baum (1999) measured the effect of informal
sanctions on drink-driving behaviour by asking survey respondents whether they
agreed with the statement ‘my friends will think I am really stupid if I drove after
drinking’. He found that this factor was more important in deterring drink-driving
than the perceived risk of apprehension created by a police enforcement program
(Random Breath Testing). However, the threat of arrest created by Random Breath
Testing and the threat of peer disapproval for drink-driving may not be factors
operating independently of one another. Peers could react negatively to the fact that a
person drove after they had been drinking because they know that he or she could be
arrested for doing so.

For this reason, Williams and Hawkins (1986) distinguish between social disapproval
which stems from the stigma of arrest and social disapproval which stems from
stigma of the act. Both these types of social stigma are informal social controls that
can operate to prevent individuals from committing an offence. However, the former
has legal sanctions as the source of control and therefore should ultimately be
included as “part of the general deterrence process” (Williams & Hawkins 1986, p.
561). These authors argue that stigma arising from the act is likely to be least
relevant to crimes that lack strong moral condemnation by the general community.
For these offences, disapproval from significant others is more likely to arise from
apprehension and conviction for the offence than from performing the act itself.

To illustrate this concept mathematically, Nagin and Pogarsky (2001) include the
extralegal costs arising from apprehension for an offence as an independent
component of the disutility of the crime:

37
 EU(G)c > P(L)c.U(legal costs + extra legal costs)c.Dt (4)

From equation 4 it can be seen that an increase in P(L)c can trigger both legal and
extralegal consequences. In addition to social disapproval associated with the
potential for arrest, these extralegal sanctions could include such things as reduced
future job opportunities resulting from a criminal conviction or the negative impact
of a criminal conviction on an individual’s professional standing (e.g. registration as
a health professional). Thus, even if the magnitude of legal sanctions for an offence
is nominal, the potential disapproval from others for being arrested for an offence
could nevertheless act to discourage offending.

Feelings of shame from violating an internalised norm can result from a

transgression that goes undetected, making it necessary to account for this additional
nonlegal source of conformance in any proposed model of offender decision-making.
Nagin and Paternoster (1994) do this by adding the term U(Moral Regret)c to the
costs side of the deterrence equation. The ensuing model predicts that, even where
there is no possibility of apprehension or formal punishment, U(Moral Regret)c can
still exceed EU(Gains)c, and subsequently deter an individual from committing an
offence. This accounts for those individuals who, irrespective of instrumental
concerns, will never consider breaking the law.

A further non-instrumental determinant of legal compliance, noted by Tyler (1990),

is an individual’s normative commitment to legal authorities. If a person perceives
legal authorities and the laws they dictate or enforce as fair and legitimate, they will
be more likely to adhere to their specifications. Alternatively, if a person believes
that a particular law is unjust or unfairly enforced, then they are unlikely to modify
their behaviour in accordance with the law. Tyler (1990) differentiates this type of
normative commitment from the normative commitment to personal values or beliefs
described above. He states that “normative commitment through personal morality
means obeying the law because one feels the law is just; normative commitment
through legitimacy means obeying a law because one feels that the authority
enforcing the law has the right to dictate behaviour” (Tyler 1990, p. 4). In this way,
even if an individual is not personally opposed to engaging in a particular behaviour

38
they may still comply with the law because they believe the authority’s power to
prohibit this behaviour is legitimate. Normative commitment would also be
subsumed under the U(Moral Regret)c term added to the costs side of the ledger.

Sherman and his colleagues (1992) describe the relationship between informal and
formal sanctions depicted in equation 4 as the additive hypothesis in general
deterrence. That is, both legal and extralegal consequences deter potential offenders
and together these two factors will have a cumulative effect. However, these authors
also note two other ways, described in the sociological literature, in which legal and
informal threats of punishment can combine to deter offending. The first is the
conditional hypothesis, which proposes that legal threats interact with informal social
controls to create a more powerful deterrent effect than would be expected from
either of these factors in isolation (i.e. a multiplicative relationship between informal
and formal social controls). A corollary of this is that legal punishment can only be
effective if reinforced by informal social controls. The second means by which
formal and informal controls could combine is proposed by the replacement
hypothesis, which expects legal control to be only effective if informal social
controls are absent.

To empirically test these three hypotheses, Sherman et al. (1992) undertook a field
experiment, examining the differential effect of police arrest on the recidivism of
domestic violence perpetrators who had a high “stake in conformity” (i.e. married
and employed offenders), with those who did not. The subsequent analysis found that
police arrest was associated with reduced recidivism for the former offenders, that is
those for whom informal social controls were present, but was associated with
increased recidivism for the latter offenders, thereby offering some support for the
conditional rather than the additive or replacement hypotheses in deterrence. In
describing the implications of these findings, the authors suggest that legal and
informal control should not be thought of as mutually exclusive concepts that can be
readily substituted. For individuals who are strongly bonded to society, informal
sanctions may augment the deterrent effect of legal sanctions, but for those
individuals where these informal controls are absent, formal punishment may prove
ineffective in preventing further offending. In fact, the field experiment described by

39
Sherman et al. (1992) indicates that formal sanctioning may actually serve to
increase the criminal behaviour of some offenders by labelling the individual as
‘deviant’ or ‘criminal’. This labelling process would perhaps be more problematic
for an individual who lacks the available social resources to overcome this type of
social stigmatisation (i.e. those with a low stake in conformity).

Delineating the relative contributions of informal social controls and legal threats to
offender decision-making is a complex undertaking and one which is yet to be
adequately resolved. Sherman’s research does, however, serve to highlight the
interdependent relationship that exists between these two factors. Zimring and
Hawkins (1973) observe “official actions can set off societal reactions that may
provide potential offenders with more reason to avoid conviction than the officially
imposed unpleasantness of punishment” (Zimring & Hawkins 1973, p. 174). These
societal reactions include feelings of embarrassment from the stigma assigned to
arrest by significant others, as well as other informal sanctions, such as “spill-over
effects” from an offender’s friends and family who are adversely affected by the
formal sanction and reduced opportunities for an offender to participate in group
activities (Heckathorn 1990). An emphasis on legal sanctions can also serve to
reinforce social norms that already exist in relation to proscribed behaviours. For
example, the recent focus on the criminal nature of child sex offences aims to
strengthen societal condemnation of these acts and, in so doing, increase anticipated
offender shame associated with committing the crime (e.g. name and shame
campaigns). In this situation, legal punishment operates to amplify if not initiate
extralegal costs and thus can be considered a contributor to the deterrent effect. This
suggests that, at least to some extent, the efficacy of informal social controls in
regulating offender behaviour depends upon the existence of the formal sanctioning
process (Heckathorn 1990).

Further interaction effects in deterrence

Sherman’s research, in short, raises the possibility that threat responsiveness depends
upon a potential offender’s “stake in conformity” or “ties to conventional society”.
Individuals with social responsibilities or strong social networks are more likely to
fear public exposure (in the form of arrest for an offence) than are those for whom

40
these sources of informal control do not exist. Survey research examining sanction
perceptions and offending behaviour, or intentions to offend, have found similarly
heightened deterrent effects of legal sanctions for individuals who have ties to
conventional society (see Nagin 1998).

The presence of informal social controls is just one of several hypothesised

conditions under which punishment may exert a stronger deterrent effect. Other
factors that are thought to condition threat responsiveness include enduring
individual traits, previous experience with legal punishment, and perceptions
regarding the likelihood of being punished for committing an offence. In the
deterrence literature, the relationship between these factors and legal punishment has
been referred to as interaction effects (Homel 1986). Two further interactions
relevant to the current research are: (1) the interaction between individual
characteristics and sanction perceptions and (2) the interaction between perceived
risk of apprehension and perceived severity of punishment.

Individual characteristics and sanction perceptions

Traditional deterrence theory says nothing about individual differences in levels of
responsiveness to legal threats. It maintains that, because all persons are similarly
motivated by pleasure, the costs of punishment should act to deter all people equally.
Accordingly, variations in offending rates will be due to differences in punishment
likelihoods and levels of severity rather than to characteristics of the potential
offender population. This view of criminal behaviour conflicts with other
criminological theories that emphasise the role of individual characteristics in
explaining criminal propensity.

Self-control and impulsivity

One enduring personality trait thought to be pertinent to the effectiveness of legal
sanctions is self-control. Self-control appears most prominently in Gottfredson and
Hirschi’s (1990) General Theory of Crime. According to this theory, people who
commit crime will tend to be “impulsive, insensitive, physical (as opposed to
mental), risk-taking, short-sighted and nonverbal” (Gottfredson & Hirschi 1990, p.
90-91), characteristics which are indicative of a person with low “self-control”.

41
These characteristics are developed at an early age, can be affected by parenting style
and, once established, will remain relatively stable throughout life. Persons with a
low level of self-control tend to engage in behaviours that bring about the immediate
gratification of their wants and desires and, because the negative consequences of
crime are often delayed, this inclines them to involvement in crime. Wilson and
Hernstein (1985) also view criminality, or law abidingness, as the result of enduring
individual differences in criminality, but cite a person’s ability to plan for the future
or to defer gratification (impulsiveness), as the key individual-level factor. This
personality trait manifests itself as a tendency to think in terms of short-term rather
than long-term consequences and, in combination with other stable individual
characteristics, predisposes a person to engage in reckless behaviours, including
those of an illegal nature.

These criminological theories of enduring individual differences, like rational choice

theories, accept that most crimes are motivated by rewards and the gratification of
desires. They also recognise the sensitivity of potential offenders to incentives and
disincentives associated with different criminal opportunities. However, they
anticipate that people who are impulsive, or who have low self-control, will focus on
the short-term benefits that could be derived from the offence rather than any long-
term negative consequences, such as being caught and subsequently punished. In
economic terms, these impulsive individuals would have a high discount rate and
would therefore be less responsive to changes in future legal threats (Kenkel 1993;
Nagin & Pogarsky 2001). In a similar vein, it could be argued that, because persons
who have low self-control are self-centred and insensitive to others, they will have
difficulties in establishing long-term relationships or persisting with education/career
goals and as a consequence will have less to lose by committing the offence than
would others who invest in conventionality (Nagin & Paternoster 1993; Wright et al.
2004). Both these accounts would suggest that those who are more criminally prone,
that is those who are more present-oriented and impulsive, would be less influenced
by the threat of formal punishment.

Wright et al. (2004) offer an alternative hypothesis, suggesting that the criminally
prone persons identified in the above theories may in fact be more likely to be

42
deterred by punishment than others. Consistent with some of the informal social
control literature, these authors suggest that sanction threats will have minimal
influence on individuals who have strong moral beliefs about the appropriateness of
criminal behaviours. Their decisions in relation to the law are guided by moral rather
than instrumental considerations. On the other hand, those who are less inhibited
from crime by moral attitudes or commitments to social norms will be more affected
by calculations of costs and benefits. This implies that impulsive and present-
oriented people, should be the most responsive to changes in punishment variables.

The empirical evidence for each of these theoretical perspectives is not clear-cut.
There is some support for the notion that impulsive individuals, or, in Gottfredson
and Hirschi’s parlance, individuals with low self-control, are more likely to engage
in criminal behaviours (Grasmick et al. 1993; Keane, Maxim & Teevan 1993; Nagin
& Paternoster 1993) and for the notion that these individuals, being more impulsive,
discount future costs associated with offending to a greater extent than other persons
(Nagin & Pogarsky 2001; Pogarsky & Piquero 2004). However, other research
shows that active offenders do consider the risks involved in committing specific
crimes and in continuing their involvement in crime (e.g. Corbett & Simon 1992;
Decker, Wright & Logie 1993; Piquero & Rengert 1999), and that they may be more
sensitive to punishment contingencies than are other persons (Decker, Wright &
Logie 1993; Paternoster & Simpson 1996; Wright et al. 2004). Wright and his
colleagues (2004), for example, found that individuals who have low self-control and
high ‘self perceived criminality’ are more deterred by perceived costs of crime than
those with relatively low scores on these measures. These authors analysed data from
the renowned Dunedin (New Zealand) study, a longitudinal study of 1,000
individuals from birth to age 26. This analysis showed that the perceived risk of
getting caught was significantly and negatively related to self-reported offending
and, further, that this effect was greatest for those with a very low level of self-
control. Meanwhile, perceived risk was not predictive of self-reported crimes for
those found to have a very high level of self-control.

In discussing their results, Wright et al. (2004) emphasise the social-psychological

nature of crime. Any theory of criminal behaviour, these authors propose, should

43
address both individual traits that predispose a person to crime as well as the social
processes or structures impinging upon that individual. Thus, just as psychological
and biological theories of crime need to recognise the sensitivity of criminally prone
individuals to their social environment, so too does deterrence theory need to
consider individual differences in present-orientedness or impulsivity when
determining the role that threatened punishment plays in decisions to offend. All
individuals discount future consequences to some extent. Impulsive individuals may
discount future costs more than do others but this does not mean that they are
incapable of foresight (Wright et al. 2004). In fact, given that criminally prone
individuals are often people who lack internal moral restraint, instrumental
considerations may feature more in their offending decisions than they do in the
decisions of individuals with a high level of self-control.

Previous offending
Past punishment is another factor thought to influence an individual’s sensitivity to
legal punishment. A rational model of offending would predict that people who have
already been formally sanctioned would believe the certainty of punishment to be
higher, compared with those who have had no prior experience of punishment. They
should therefore be more sensitive to future sanction threats. The heightened
deterrent effect created by prior punishment is referred to in the criminological
literature as specific or special deterrence, to distinguish it from general deterrence,
in which people modify their behaviour as a result of observing or otherwise learning
about formal punishment for particular offences.

Stafford and Warr (1993) extend this definition of general and specific deterrence to
incorporate further information about the nature of prior experience with legal
sanctioning. According to these authors, individuals base their judgements about
sanction risks not only on their experience with the sanctioning process but also their
experience in avoiding formal punishment. Escaping detection for an offence can
serve to reduce the perceived certainty that a particular transgression will result in
formal punishment and subsequently increase the likelihood of offending. The more
often punishment has been avoided, the greater the erosion of certainty perceptions
and thus, the greater the chance that an individual will engage in the behaviour.

44
Stafford and Warr also note that one’s own experience with punishment and
punishment avoidance, and the experience of others (vicarious experiences), will
both play a role in the formation of these perceptions. The former, they refer to as
specific deterrence, while the latter they call general deterrence. Thus, rather than
referring to distinct populations (i.e. those who have been previously punished vs.
those who have not), specific deterrence is conceptualised as the deterrent effect of
“direct experience with punishment and punishment avoidance”. General deterrence,
on the other hand, is the deterrent effect of “indirect experience with punishment and
punishment avoidance” (Stafford & Warr 1993, p. 127).

Piquero and Pogarsky (2002) have recently undertaken an empirical test of the
Stafford and Warr model using a scenario-based methodology and found some
confirmatory evidence for the model’s predictions. In particular, they found strong
evidence that both personal and vicarious punishment avoidance is positively related
to offending and that punishment and avoidance experiences influence behaviour via
sanction risk perceptions. They also found strong evidence for the moderating impact
of prior offending, that is, the finding that individuals who report having committed
prior offences (in this study, previously driven whilst drunk) place more weight on
personal experiences in their decisions to offend than on the experiences of others.
Contrary to Stafford and Warr’s hypotheses, however, this study found some
evidence that punishment experience actually encouraged, rather than discouraged,
future offending, with both personal and vicarious punishment experience found to
be positively related to offending likelihood. Though somewhat counterintuitive, this
finding is consistent with previous tests of the Stafford and Warr model (Paternoster
& Piquero 1995; Piquero & Paternoster 1998), as well as more recent survey work on
intentions to offend (Pogarsky & Piquero 2004).

One obvious reason for a counter-productive positive punishment effect is that

previous punishment experience simply identifies those individuals who offend more
frequently because they perceive the risk of apprehension and conviction to be low
(selection effect). However, Piquero and Pogarsky (2002; Pogarsky & Piquero 2004)
suggest two decision-making biases which may also offer a potential explanation for
these results: (1) Self-serving bias and (2) Gamblers fallacy. The self-serving bias

45
refers to the fact that people tend to see themselves in a favourable light when asked
to make judgements. Piquero and Pogarsky (2002) note evidence from the decision-
making literature showing that people generally view themselves as above average
when it comes to several different tasks, such as driving a motor vehicle, managing
people and being productive at work. In relation to criminal decision-making this
could mean that offenders believe they are superior at avoiding detection than others
and the knowledge that someone else has been punished for the same or similar
offences may simply serve to bolster this misplaced belief in their offending prowess.
The gamblers’ fallacy, on the other hand, refers to the faulty assumption that the
probability of an event increases with each period it fails to occur. For a card player,
this may result in a person gambling more money after losing three successive hands
of poker on the assumption that the chance of losing four hands in a row is somehow
reduced after several consecutive losses. Similarly, for an offender, being
apprehended for a crime could be viewed as a rare occurrence and one which is
unlikely to occur again soon after it has happened. Thus, punishment would result in
a reduction in estimated sanction-risk for the next offending occasion and,
consequently, increase the likelihood of further offending.

Tests of these hypotheses provide support for the gamblers’ fallacy theory among
individuals who are less experienced offenders, but not among high-risk individuals
(Pogarsky & Piquero 2003). Indeed research, focusing on experienced offenders has
suggested that offenders do update their certainty perceptions based on new
information in the way anticipated by a rational model of offender decision-making.
One study involving over 1,000 offenders convicted of serious offences found that
felons with higher arrest ratios (i.e. self-reported arrests to self-reported crimes) also
reported a higher risk of incurring further punishment (Horney & Marshall 1992).
This latter study is particularly noteworthy, not only because it examined a sample of
convicted felons, but also because, unlike previous research in this area, punishment
experience was considered in conjunction with offending frequency, thereby
producing a more valid test of the rational-offender model.

Research on the formation of sanction risks is scarce and narrow in scope (Nagin
1998) and it is therefore difficult to draw definitive conclusions about the influence

46
of punishment experience on sanction-risk estimates. However, even if punishment
experience does not translate into higher perceptions of arrest certainty after
sanctioning, this does not mean that offenders with a prior conviction would not be
responsive to changes in legal threats. In fact Homel (1986) found that offenders
with a prior conviction were more likely to modify their offending behaviour after a
significant increase in formal punishment than were other people. This finding is
compatible with the view, expressed earlier, that individuals at high-risk of offending
are less likely to be influenced by informal social controls and therefore perhaps
more responsive to legal threats.

Perceived risk of apprehension and perceptions of punishment

severity
An important contextual factor thought to condition the deterrent efficacy of legal
sanctions is the level of perceived risk of apprehension and conviction within a
certain jurisdiction. As mentioned previously, a central prediction of the subjective
expected utility equation, as applied to deterrence theory, is that the certainty and
severity of legal punishment combine multiplicatively rather than additively. Past
research examining this aspect of deterrence theory consists mostly of experimental
studies in which deterrence variables are manipulated and either prior or future
intentions to offend are measured or, in the case of scenario-based methodologies,
hypothetical likelihoods of offending are reported. Some of this research has found
evidence in favour of the certainty/severity interaction while other studies have
suggested that these two deterrence variables combine in an additive rather than
multiplicative fashion.

Howe and Loftus (1996) conducted three separate experiments in which respondents
were asked to rate the extent to which particular punishment conditions might cause
someone to refrain from committing a fairly serious crime (i.e. the deterrence value
of the punishment threat). The first experiment used a repeated measures design in
which respondents were presented with varying levels of punishment certainty,
severity and celerity for a particular crime. The other two experiments used an
independent groups design with different crime scenarios and sanction risk levels
randomly assigned to each group. Across all scenarios and all designs, robust main

47
effects for punishment certainty and severity were found, as well as significant
interaction effects for these two variables.

Further research with active burglars supports the conclusions of Howe and Loftus.
Decker, Wright and Logie (1993) presented active burglars with a hypothetical
offending scenario in which the punishment risk, the reward to be gained from
committing the crime and the anticipated penalty if caught were varied across
subjects. The offenders were then asked if they would commit a burglary under the
circumstances described. Compared to a matched, non-offender control group, these
offenders were far more willing to commit the hypothetical offence but, nevertheless,
were found to be sensitive to situational contingencies. In particular, the study found
that penalty severity did not, on its own, significantly affect the likelihood of these
offenders stating that they would commit the offence. It was only when the perceived
risks were high and the anticipated penalties severe that stated offending likelihood
was significantly reduced.

Grasmick and Bryjak (1980) provide further support for an interactive deterrence
model. These authors criticised previous perceptual research that had been conducted
in this area because it had assumed that a particular punishment (e.g. one year
imprisonment) would be equally severe to all people, thereby neglecting the
subjective nature of punishment perceptions. To address this limitation, Grasmick
and Bryjak asked their respondents to rate how big a problem the punishment, which
they anticipate the court would impose for a particular offence, would create for
them personally. Using this refined measure of severity, they found an inverse
relationship between perceived severity of punishment and involvement in illegal
behaviour, which was concentrated amongst those respondents who perceived the
chances of being punished to be relatively high.

One major problem with these cross-sectional surveys is that the inverse relationship
between previous offending and sanction risk perceptions could reflect an
experiential rather than a deterrent effect (see Nagin 1998). Research relying on
Grasmick and Bryjak’s (1980) measure of perceived punishment severity but
utilising panel data instead of data on prior criminal involvement (thus predicting

48
subsequent rather than past criminal behaviour) was unable to reproduce the reported
interaction effect (Paternoster & Iovanni 1986). More recent research using scenario-
based methods also failed to find a multiplicative relationship between punishment
certainty and severity (Nagin & Pogarsky 2001).

Outside of the experimental literature, empirical research exploring the interaction

between perceived certainty and severity is limited. This is not surprising given that
the actual probability of being apprehended and convicted for most crime types is
relatively low and that it is extremely difficult for law enforcement agencies to
influence risk perceptions to any great extent. One study which did manage to
investigate the certainty/severity interaction in a field setting was that conducted by
Homel (1986), who investigated the impact of the introduction of Random Breath
Testing (RBT) for drink-driving offences in New South Wales, Australia. RBT had a
considerable impact on the perceived risk of being apprehended for a drink-driving
offence because it ensured that any motorist could be stopped and tested for alcohol
consumption, regardless of whether the attending police officer suspected the
motorist had been drinking. The introduction of this new method of enforcement was
widely publicised, with surveys showing that the vast majority of drivers knew about
the introduction of these new initiatives (Cashmore 1985). Subsequently there was a
significant increase in perceptions of apprehension risk, which resulted in drivers
modifying their behaviour to avoid formal punishment (Homel 1988).

A new offence for drink-drivers apprehended with a Blood Alcohol Concentration

(BAC) above 0.15 was also introduced in New South Wales at the same time as
RBT. This new offence was accompanied by severe penalties. Ten weeks following
the introduction of RBT and the new penalty regime, Homel (1986) asked a sample
of motorists whether they knew about the drink-driving penalty changes. He then
questioned these same respondents six weeks later about their drink-driving
behaviour since the first interview. A logistic regression analysis showed that
respondents who believed drink-driving penalties had increased, at the time RBT was
introduced, were 3.5 times less likely to have driven whilst drunk in the period
spanning the two interviews than those who were not aware of the penalty changes.
Although supportive of deterrence hypotheses, these results should be treated with

49
some caution, given that only 12 of the 175 respondents reinterviewed had reported
driving whilst impaired during the six-week follow-up period. Other evidence,
however, supported the regression analyses (e.g. a significant negative correlation
between knowledge of the penalty increase and modifications to travel behaviours),
leading Homel to conclude, “when the perceived chances of arrest are high,
perceived severity can have a deterrent impact additional to that of arrest certainty”
(Homel 1986, p. 127).

Drink-driving as a case study

This thesis is focused on the deterrent effect of punishment severity in relation to
drink-driving offences. In light of the review we have just conducted, drink-driving
is an ideal offence to use in a case study of deterrence for five main reasons.4

Firstly, it is generally considered that drink-driving offences do not elicit the same
moral condemnation as more traditional crimes, such as murder or break and enter.
Drink-driving is prohibited in most countries, not because it is an act that is
inherently wrong or against the moral code of society, but because alcohol
consumption significantly increases a driver’s risk of incurring an accident and thus
constitutes a behaviour that is a threat to public safety. Accordingly, drink-driving
tends to be classified, in criminal law terms, as a mala prohibita offence, or, in
Zimring and Hawkins’ (1973) terminology, as an instrumental offence. These crime
types have been identified in the deterrence literature as offences less likely to be

4
Given that alcohol can disrupt cognitive processing abilities, the applicability of a rational model of
offending to alcohol-related crimes has been questioned by some academics (e.g. Exum 2002).
However, even if alcohol does have a deleterious effect on an intoxicated driver’s ability to process
information about sanction threats, the fact that drink-driving involves some element of planning
(unlike many opportunistic or impulsive crimes) means that formal punishment can still be influential
in drink-driving decisions. Consistent with Cook’s (1980) “standing decisions” that guide behaviour,
formal sanctions may influence potential offenders through the general stance that they adopt in
relation to drink-driving rather than acting to tip the cost-benefit ratio when a decision to drink and
drive is being made. Homel (1986) draws upon a key concept in prospect theory known as pre-
commitment (Tversky & Kahneman 1981) to further elaborate on this idea. He argues that punishment
can encourage “an individual to take actions in the present to render inoperative an anticipated future
preference” (Homel 1986, p. 33). A driver who has consumed large amounts of alcohol will have
difficulty in rationally processing the utility of various alternatives. However, to avoid any possibility
of experiencing future legal costs the sober driver may decide to employ preventive strategies, such as
leaving the car at home or drinking low-alcoholic drinks. In this way, even if an intoxicated driver
cannot be considered fully rational, the decision to drive and then drink can be modelled as a rational
choice.

50
influenced by informal sources of social control and thus more responsive to threats
of legal punishment.

Secondly, drink-driving is far more common than most other types of criminal
behaviour. In Australia, just over 15 per cent of drinkers aged 14 or over report
having driven a motor vehicle whilst under the influence of alcohol in the last 12
months. In contrast, less than two per cent of drinkers admit to physically assaulting
another person (Australian Institute of Health and Welfare 2002a). Thus, many
people who might never contemplate engaging in other illegal behaviours still decide
to drive home after a ‘night out’ drinking. Surveys suggest that one reason for this is
the belief amongst offenders that drink-driving is widespread (Corbett & Simon
1992; Baum 2000) and is viewed by many drivers as a violation of social
conventions (Elliott 2003). Rather than being viewed as deviant individuals who seek
out offending opportunities, many drink-drivers might be better viewed simply as
risk-takers drawn from the general driving population. As such, we would expect
situational contingencies, including increased threats of formal punishment, to be
more influential for this offence than would be true for other more serious crimes.

Thirdly, relatively precise and inexpensive handheld breath testers have been
developed to assess a person’s blood alcohol level.5 The convenience of these
breathalysers means that they can be widely used by police officers out in the field to
enhance the detection of drink-driving offences. The accuracy of these tests means
that officers can present reliable, objective evidence of a driver’s Blood Alcohol
Concentration (BAC) to the courts, thus boosting the chances of a drink-driving
offence being proven. Many western jurisdictions have further strengthened the
certainty of punishment for these offences by introducing “per se” laws. This type of
legislation removes the necessity to demonstrate intoxication, or intent to drive
whilst drunk, when attempting to prove the offence of drink-driving. In these
instances, all that the courts require is evidence that, at the time of driving a motor
vehicle, the defendant’s BAC was over the prescribed legal limit. Legislation
allowing for the breath testing of motorists, in conjunction with these per se laws,

5
These tests are based on expired air from the lungs. About 1-5% of alcohol absorbed by the body is
expired via the lungs (Jones & Lacey 2001).

51
have ensured that the risk of apprehension and conviction for drink-driving offences
is relatively high in comparison to most other offences. In this context, deterrence
theory would anticipate that the efficacy of punishment severity should be enhanced.

Fourthly, valid and reliable measures of offending (i.e. motor-vehicle crash rates) are
available to evaluate the impact of drink-driving interventions modelled on
deterrence principles (Ross 1984). A common problem associated with traditional
measures of offending, such as recorded crime data, is that they necessarily include
only crimes which come to the attention of officials. Criminologists have long known
that only a small fraction of most criminal acts are ever reported to police.
Furthermore, officially recorded offences can vary as a consequence of interventions
other than those under investigation, which could lead to erroneous conclusions
about the underlying reason for observed changes. Almost all road fatalities and
crashes resulting in serious injury, on the other hand, are brought to the attention of
authorities because of the need to treat victims. The use of road crash data, therefore,
allows for more accurate and reliable conclusions to be made about the effect of a
specific policy under examination (Ross 1984).

Finally, this thesis examines the deterrent efficacy of a substantial increase in the
severity of formal punishment within a real world situation and as such, has
significant implications for public policy. Drink-driving is an important social
problem that results in significant loss of life and which places a substantial burden
on health and criminal justice resources. Over 400 people die, and almost 8,000
persons are hospitalised, for a road injury attributable to alcohol each year in
Australia (Chikritzhs et al. 2000). This translates to just over 17,000 person-years of
life lost from road fatalities and almost 45,000 bed-days required for road-injury
hospitalisations. The social and economic cost of these alcohol-related injuries to the
Australian community is estimated to be over AUS$3.4 billion each year (Collins &
Lapsley 2001). Any strategies that can be demonstrated to reduce the incidence of
this behaviour will therefore be highly relevant for decision-makers working in this
area of public policy.

52
CHAPTER 3. DETERRENCE RESEARCH ON DRINK-
DRIVING

As mentioned in the introduction to this thesis, hypotheses generated by deterrence

theory have been investigated both at the aggregate- and the individual-level of
analysis. Aggregate-level deterrence research consists of two main types: ecological
studies6 and quasi-experimental studies (Nagin 1998). Ecological studies look at
natural variations in sanction levels and crime rates across time or space, while
quasi-experimental studies examine the effect of targeted and specific interventions
on measures of offending behaviour. In the context of drink-driving, ecological and
quasi-experimental studies have mostly relied on motor vehicle crash rates as a
surrogate measure of offending because of the well-established positive relationship
between alcohol consumption and crash risk (Zador 1991; Zador, Krawchuk & Voas
2000). In particular, fatal, serious injury and single-vehicle night-time motor vehicle
crashes, which have been demonstrated to include a greater proportion of alcohol-
related incidents (Evans 1991), are frequently used as a proxy for drink-driving in
aggregate-level studies.

Another substantial deterrence literature that has emerged over the last two decades
emphasises the perceptual nature of the deterrence process and, consequently, the
importance of investigating deterrent effects at the individual-level. This type of
research uses survey data in order to examine perceptions of sanction risk and
sanction severity, and the relationship between these perceptions and self-reported
offending. Most reviewers have considered perceptual studies, in conjunction with
the aggregate-level research described above, to be the mainstay of the deterrence
literature (Nagin 1998; von Hirsch et al. 1999; Weatherburn et al. 2000). However, in
reviewing the deterrent efficacy of punishment severity, a further type of research,
which also features the individual as the unit of analysis, is important: reoffending

6
This type of research has been given various names in the literature including association (e.g. von
Hirsch et al. 1999), correlational (e.g. Ross & McCleary 1983), criminal opportunity (e.g. Cook 1980)
and passive observational (e.g. Cook & Campbell 1979) studies. Generally speaking, they are all
studies that examine how crime rates covary with arrest, conviction or sentencing rates over time or
across jurisdictions, or both of these.

53
studies. Reoffending studies are primarily concerned with the marginal deterrent
effect of particular types of judicial penalties and/or the magnitude of judicial
penalties, and typically rely on official records to measure deterrent outcomes. Their
importance means that they, too, comprise a major component of this literature
review.

Aggregate-level research
Nagin (1998) notes in his recent review of the deterrence literature that the most
important issue for policy makers “is not whether the criminal justice system in its
totality prevents crimes but whether a specific policy, grafted onto the existing
structure, will materially add to the preventative effect” (Nagin 1998, p. 3).
Accordingly, this section will review aggregate-level deterrence research that has
investigated the relative importance of each of the three elements of the deterrence
equation, namely the certainty, severity and celerity of drink-driving punishment. A
summary of these studies, including a description of the methodology they employ
and any significant outcomes found in the research, can be found in Appendix I.

Punishment certainty
Strategies that increase the probability of detection, apprehension and conviction aim
to influence the perceived risk of punishment for drink-driving and in doing so
operate to deter an individual from committing the offence. Increasing the perceived
probability of detection for drink-driving offences can be achieved by boosting the
number of police allocated to the task of drink-driving enforcement, or by adopting
strategies such as roadblocks, sobriety checkpoints or Random Breath Testing
(RBT). The probability of apprehension and/or conviction can also be increased
through legislation that allows police to conduct preliminary breath tests of drivers at
the roadside and/or allows the Blood Alcohol Concentration (BAC) of motorists to
be used as sufficient proof of a drink-driving offences (per se legislation).

Ecological studies
Ecological research has compared the impact of several of these deterrence-based
policies on State-level motor vehicle fatality rates. A feature of the analytical models
utilised in these studies is the inclusion of a broad range of control variables that are

54
thought to have an effect on road fatality rates and/or alcohol consumption, over and
above that created by the deterrence-based policies being examined. Potentially
confounding factors that have been controlled for by econometricians include the
number of kilometres travelled, the number of vehicles on the road, weather
information, traffic density, alcohol sales, the price of alcohol, unemployment rates
and disposable income.

Several of these earlier studies found that proxy variables for the certainty of
punishment were not correlated with vehicle-fatality rates. Wilkinson (1987), for
example, measured differences in law enforcement across USA States in terms of the
probability of arrest (arrests per vehicle mile) and conviction (per cent of arrestees
found guilty or pleading guilty) for drink-driving but found no marginal effect for
these variables on the level of motor vehicle fatalities from 1976 through 1980.
Similarly, Evans, Neville and Graham (1991) report no conclusive evidence that any
specific form of punitive drink-driving legislation implemented in the USA between
1975 and 1986 had a measurable effect on motor vehicle fatalities. However, other
ecological research suggests that State laws facilitating arrests and/or convictions can
deter drink-driving. Saffer and Chaloupka (1989) found that the introduction of
preliminary breath test laws in the USA was associated with a significant reduction
in vehicle-fatality rates from 1980-1985 and Sloan, Reilly and Schenzler (1994)
report that police availability (as measured by police/1,000 population and total
police expenditure/1,000 population) reduced USA State vehicle-fatality rates during
the period 1982 through 1988.

Benson, Rasmussen and Mast (1999) tried to make sense of these disparate findings
by grouping together several proxies for deterrence-based policies in their analysis of
USA vehicle-fatality rates. This enabled the authors to investigate whether a
combination of policies aimed at deterring the drinking driver has an impact on
alcohol-related road fatalities, even when an individual law appears to have no
significant marginal effect. In this analysis, the “driver involvement rate” was used
as a proxy for drink-driving offences, calculated as the number of fatally injured
drivers with a positive BAC (information that was available for only 15 USA States)
divided by the number of drivers in each State. This analysis found that a subset of

55
factors influencing the probability of being stopped (open container laws,
anticonsumption laws7 and police per capita) and/or arrested (illegal per se laws,
preliminary breath test laws and implied consent laws)8 had a significant impact on
driver involvement rates, controlling for other confounding factors (including vehicle
miles travelled per driver, males aged 16-44 per capita, per capita disposable income,
unemployment rate, metropolitan population). Among these variables, those grouped
as influencing the probability of being stopped seemed to be the most important
factor in reducing road fatalities.

It is possible that some ecological studies have failed to find a deterrent effect for
drink-driving policies because the outcome measure used in this research, that is
vehicle-fatality data, is an insensitive measure of offending. These data include
numerous instances of non alcohol-related fatal crashes and exclude many non-fatal
alcohol-related crashes and drink-driving episodes that do not result in a crash. Cook
(1980) observes that “most policy innovations have a small effect at best, and this
effect can easily be submerged and lost in the normal fluctuations of an insensitive
measure” (Cook 1980, p. 252). In the research described above, Benson and his
colleagues (1999) attempted to address this issue by examining only fatal accidents
in which the driver had a positive BAC. However, this measure still excludes both
non-fatal alcohol-related crashes and drink-driving episodes where no crash results.
For this reason, other ecological research has relied on self-reported drink-driving
information contained in survey data as a measure of drink-driving offence rates.

Kenkel (1993) examined data from a USA national health survey to investigate the
deterrent effect of State laws allowing preliminary breath tests and sobriety
checkpoints, as well as laws prohibiting plea-bargaining in drink-driving cases. The
dependent measures examined in this study included both self-reported alcohol
consumption and self-reported drink-driving. The subsequent analysis showed that
each of the policies considered, with the exception of plea-bargaining prohibitions,
was associated with reduced heavy drinking and with reduced drink-driving

7
Open container laws prohibit the possession of any open alcoholic beverage in the passenger area of
a motor vehicle and anti-consumption laws prohibit the consumption of any alcoholic beverage in the
passenger area of a motor vehicle.
8
Implied consent laws dictate that any person who operates a motor vehicle is presumed to have given
his or her consent to a test to determine their Blood Alcohol Concentration.

56
episodes. Moreover, when the effect of the law on heavy drinking was held constant,
there was no evidence for an independent effect of these policies on self-reported
drink-driving. The latter finding suggests that the State laws in question affected
drink-driving behaviour through their impact on alcohol consumption, lending some
credence to the notion that the joint decision to drink and drive can be modelled as a
rational choice (Chaloupka, Saffer & Grossman 1993).

Similarly, Chaloupka and Weschler (1996) considered the effect of drink-driving

policies on self-reported binge drinking amongst young people. They used an index
developed by the USA advocacy group “Mothers Against Drunk Driving” (MADD)
to measure the restrictiveness of each State’s drink-driving laws. This index reflected
factors that increased the probability of arrest, eased standards for arrest and
conviction and raised penalties for a conviction. Utilising a large data set from a
survey of USA college students, the authors concluded that strong State-level
policies related to drink-driving can significantly reduce binge drinking in young
people. In this instance, however, it was the overall ‘package’ of deterrence-based
policies (including both those affecting the certainty and the severity of punishment),
which was found to deter drink-driving (through reduced binge drinking), rather than
any specific individual law.

Reviews of the deterrence literature have highlighted several methodological

shortcomings of studies that rely on natural variations in punishment levels and
crime rates to test casual relationships between these two variables (e.g. Cook 1980;
Nagin 1998; von Hirsch et al. 1999). Nagin (1998) suggests that the most salient of
these is the simultaneity problem: the reciprocal relationship that potentially exists
between crime and sanction levels. Consider a model in which drink-driving arrests
in each State are used to index punishment certainty and State-level alcohol-related
road fatality rates are employed the dependent measure. A significant negative
association between these two variables could be interpreted as evidence that arrests
reduce road fatalities (i.e. Arrests ⇒ Fatalities). However, if a particular jurisdiction
has a relatively high road fatality rate, public demand for the problem to be addressed
may cause authorities to increase the level of police activity targeting these offences
and, consequently, increase the number of arrests. In this instance, road fatalities

57
would have an effect on the number of drink-driving arrests (i.e. Fatalities ⇒
Arrests). This reciprocal relationship between crime rates and sanction levels is often
ignored in ecological studies investigating deterrent effects, resulting in mis-
specified models of deterrence.

A further shortcoming is that ecological studies require the use of an index of the
‘threat level’ achieved by a State’s drink-driving policies yet often the index used in
these analyses is of limited value (Cook 1980). For example, binary variables
representing the existence of sobriety checkpoints or preliminary breath test laws
have been employed in econometric models as indices of the probability of
punishment for drink-driving offences. However, the extent to which these laws are
enforced could vary considerably across jurisdictions and a binary variable would be
unable to capture these differences in degree. Other studies have employed police
expenditure or police allocation per capita as measures of probability of arrest but
these indices are similarly flawed. If police expenditure is equal in two States this
does not necessarily imply that they allocate the same level of resources to the task of
detecting drinking drivers. Indeed, Benson, Mast and Rasmussen (2000) show that,
as police shift resources into the control of violent and property crimes, the
likelihood of a drink-driving arrest decreases significantly. The use of inadequate
estimates of punishment probability is one possible reason why ecological research
has produced such inconsistent findings regarding the deterrent effect of drink-
driving enforcement policies.

Quasi-experimental studies
Quasi-experimental analyses of deterrence-based interventions are thought to be a
more promising methodology because they incorporate many features of a true
experimental design – a well-defined treatment regime, measurement before and
after an intervention and (in many cases) a control group (Ross & McCleary 1983;
Ross 1984; Nagin 1998).9 This means that quasi-experimental studies overcome
several of the threats to inferential validity associated with correlational designs. In
particular, the interrupted times-series technique, which measures the target

9
This experimental design is ‘quasi’ in the sense that there is no control over the behavioural event
which constitutes the intervention variable.

58
behaviour at several points before and after an intervention, has frequently been used
in quasi-experimental research to examine the effect of specific drink-driving
enforcement policies on surrogate measures of offending. This technique addresses
rival explanations for observed effects, such as regression to the mean (extreme
observations that naturally gravitate back to the average over time), maturation
(natural growth processes unrelated to but temporally coincident with the
intervention which could cause a pre-post difference), and instability (fluctuations in
sampling persons or components) (Cook & Campbell 1979). Although the
interrupted time-series analysis cannot itself control for history (some specific event,
unrelated to but temporally coincident with the intervention which could cause a pre-
post difference) it can be supplemented with additional data, such as rates of non
alcohol-related crashes, to address this additional threat to inferential validity (Ross
1984).

A further added advantage of quasi-experimental designs is that by focusing on the

implementation of a specific policy it is possible not only to measure changes in the
target behaviour (i.e. drink-driving offending) before and after the intervention, but
also ascertain the degree to which the new policy has influenced activity levels and
decision-making (Cook 1980). Consequently, it can be determined whether negative
findings are due to the policy or law failing to have a deterrent effect or,
alternatively, failing to affect punishment certainty in the jurisdiction where it has
been implemented. This allows for more informed conclusions regarding the efficacy
of specific deterrence-based interventions for drink-driving offences.

Much of the early quasi-experimental drink-driving research in the deterrence

literature focused on the effectiveness of the Scandinavian approach to controlling
the drink driver because of evidently low drink-driving rates in these countries.
Scandinavian-type laws give police the power to use chemical testing devices to
ascertain a driver’s BAC and allow breath test results to be used as sufficient
evidence of alcohol intoxication. Jurisdictions without this type of legislation
typically rely on behavioural evidence of impairment (or sobriety tests) to discern
whether a driver is under the influence of alcohol. The Scandinavian strategy is
considered to be superior to behaviour-based approaches for apprehending and

59
convicting alcohol impaired drivers because it does not rely on the subjective
judgement of police officers to prove that a drink-driving offence has been
committed.

Ross (1984) conducted a major review of quasi-experimental studies examining the

effectiveness of these types of drink-driving countermeasures in several different
countries. From this review he concluded that the implementation of Scandinavian-
type laws has had a significant impact on alcohol-related traffic injuries and
fatalities. The best example of the deterrent effect of this type of legislation comes
from Ross’ (1984) evaluation of the British Road Safety Act 1967. This piece of
legislation set a blood alcohol limit of 0.08g/100ml and allowed police to demand a
preliminary breath test from motorists who committed a traffic infringement or were
involved in a crash. An interrupted time-series analysis, examining the effects of this
intervention, found a 66 per cent decline in fatal and serious injury crashes occurring
immediately after the enactment of the legislation. This effect was shown to be
particularly strong in drinking hours but not apparent in non-drinking hours. The
effect, moreover, was still evident after controlling for other confounding variables,
including total miles travelled by drivers and variations in alcohol sales over time.
Although the evidence for an initial deterrent effect created by the enactment of this
legislation was strong, Ross observed that the effect dissipated within a few years.
His review of similar drink-driving laws implemented in other western countries
(including New Zealand, France and Canada) generally found that such interventions
were similarly successful in generating an initial deterrent effect but that this effect
decayed over time.

Simply increasing the numbers of police dedicated to drink-driving enforcement has

also been shown, by quasi-experimental research, to have an impact on alcohol-
related road crashes. An enforcement program in Stockton, California, introduced
special patrols on weekend nights dedicated to detecting and arresting drink-drivers
at high-risk times. Although roadblocks, per se legislation and preliminary breath
tests were not available to the Stockton police, there was still a significant impact of
the patrols on drink-driving in this community. The beneficial impact of the police
patrols was evidenced by a 43 per cent reduction in the number of drivers who

60
recorded an illegal BAC in roadside breath test surveys and a 10-15 per cent
reduction in night-time road crashes. Comparisons with night-time crashes in four
control cities showed that crash trends in Stockton were significantly different from
any of these control sites. However, once the extra police were withdrawn from the
task, the observed impact on crashes disappeared (Voas & Hause 1987).

A specific enforcement program aiming to increase both the probability of

apprehension (through preliminary breath testing) and the probability of detection
(through greater use of sobriety checkpoints) provides further evidence of the
deterrent effect of the certainty of punishment. This program was introduced in three
Californian communities and was accompanied by a media campaign which aimed to
increase the perceived risk of punishment by publicising the drink-driving
enforcement efforts of local police. Voas, Holder and Greunewald’s (1997)
evaluation of this program found that there was a significant reduction in single-
vehicle night-time crashes in the three treatment communities, without any similar
change in crash rates in three pre-selected control sites. Additional evidence showed
a significant reduction in the number of drivers with positive BACs in the areas
where enforcement had been boosted. Furthermore, drivers surveyed in the test sites
reported a greater perceived risk of apprehension for drink-driving and a consequent
decrease in the number of times they had driven while drunk.

Research on Random Breath Testing (RBT) in New South Wales, Australia, has
provided perhaps the strongest evidence that enforcement can reduce drink-driving
and subsequent alcohol-related road crashes through the process of deterrence. RBT
was introduced in New South Wales in December 1982 with the specific aim of
increasing the perceived probability of punishment for drink-driving offences by
maintaining high levels of systematic and sustained police efforts against drink-
driving. This enforcement technique resembled those previously discussed, in that it
made use of preliminary breath test and per se laws. A defining feature of RBT
legislation, however, was that it gave police the authority to stop any motorist and
administer a preliminary breath test to them. This method differs from other (less
stringent) testing procedures, such as roadblocks and sobriety checkpoints: where
only drivers who are judged to have been drinking or who are obviously impaired are

61
required to undertake a breath test. The NSW Police aimed to carry out one random
breath test for every three licensed drivers, each year, in order to establish a
relatively high level of probability of detection. The RBT program in New South
Wales was also accompanied by a considerable amount of media coverage and
widespread publicity campaigns designed to augment police efforts by increasing the
perceived risk of detection and apprehension amongst the general public (see
Cashmore 1985).

Subsequent analyses found that the introduction of RBT was accompanied by an

overall 19 per cent reduction in all serious accidents, a 48 per cent reduction in
fatalities and a 26 per cent decline in single-vehicle night-time crashes. Comparable
effects on road fatalities and injuries were also observed in Queensland, Western
Australia and Tasmania after RBT was introduced in these States (Henstridge,
Homel & Mackay 1997). Furthermore, the effect of RBT on drink-driving appeared
to be the direct result of an increase in the perceived risk of apprehension for that
offence. A survey of drivers immediately after the introduction of RBT in New South
Wales found that the perceived chances of arrest increased significantly among NSW
motorists and that these perceptions predicted the number of ways in which
respondents were modifying both their drinking and their driving behaviour (Homel
1986). Similar findings for the effectiveness of police enforcement in reducing
alcohol-related fatalities and injuries are also evident from evaluations of less
intensive RBT operations in the Netherlands (Zaal 1994).

Importantly, unlike other enforcement initiatives, RBT operations in New South

Wales appeared to have a sustained effect on drink-driving rates. Henstridge, Homel
and Mackay (1997) estimated that, at least for single-vehicle night-time accidents,
the impact of RBT took over 10 years to decline to five per cent of its initial
magnitude, meaning that, by 1992, the effect of introducing RBT in New South
Wales was still discernible. The most striking result, however, was that these long-
term effects appeared to depend greatly on enforcement levels (as measured by the
number of breath tests administered by police). Initially the effects of introducing
RBT did wear off, but over the longer term the reductions were maintained because
police increased the number of breath tests they administered to motorists after 1987.

62
In fact the ability of the NSW Police to control drink-driving was shown to be a
direct function of how much effort they put into it. An increase in 1,000 breath tests
per day was found to correspond with a decline of 19 per cent in single-vehicle night-
time crashes. While significant, this relationship between daily testing rates and crash
reductions was not linear, since there was evidence for “diminishing returns” as
enforcement levels increased (Henstridge, Homel & Mackay 1997).

The research conducted by Henstridge and his colleagues shows that the long-term
success of deterrence-related enforcement relies heavily on high levels of visible
police activity to maintain perceptions of risk apprehension at a relatively high level.
Ross and LaFree (1986) suggest that an individual’s perception of risk does not
remain constant but is continually updated on the basis of new experiences. Once
people become aware of an increase in enforcement activity, they initially
overestimate the risk of apprehension but, with repeat successful drink-driving
episodes or reduced publicity of police enforcement efforts, they lower these risk
estimates. Homel (1986) maintains that it is necessary to take account of these
changes in perceptions when attempting to understand the deterrent effect of
enforcement activities. He likens the deterrent effect of RBT to a leaky bucket. That
is, it is a delicate balance between the forces maintaining (i.e. visible police
enforcement) and those tending to erode (e.g. peer pressure, lack of exposure,
successful drink-driving episodes) the perceived risk of apprehension. He concludes
that the “long-term effect of RBT will depend on the relative sizes of the input and
the output effects – in other words, how full the bucket can be kept through police
enforcement” (Homel 1986, p. 136-137).

Punishment severity
So far, we have only discussed the deterrent effect of punishment certainty.
However, deterrence theory also suggests that more severe punishments should
reduce the incidence of offending. Aggregate-level research has investigated this
through two methods. The first involves quasi-experiments capitalising on sudden
changes in statutory penalties in order to investigate whether raising the overall
threat of punishment is an effective deterrent for drink-driving. A limitation of this
research is that sentencing policies often fail to influence the actual penalties

63
convicted offenders receive. The second involves studies of actual sentencing
practice to determine the impact of increased penalties on offending rates. This latter
category includes studies that compare variations in judicial sentencing practices
across jurisdictions, as well as studies that investigate the impact of introducing
mandatory sentencing policies within a jurisdiction.

Increases in the severity of statutory sanctions

One method commonly employed by criminal justice policy makers as a means to
increase the perceived severity of punishment is to raise the overall threat of
sanctions. This can be achieved through laws that increase the magnitude of penalties
for all offenders, for particular offender groups (e.g. repeat offenders) or for certain
offences (e.g. drink-driving offences where the offender records a high BAC).

An evaluation of California’s 1982 driving-under-the-influence (DUI) legislative

reforms provides support for the deterrent effect of increased statutory penalties for
drink-driving offences. These laws introduced more punitive sanctions for drink-
driving offences, including stricter licence restrictions and longer gaol sentences for
repeat offenders. Rogers and Shoenig (1994) examined the effect of these legal
changes on three road crash categories thought most likely to involve alcohol,
namely night-time crashes, single-vehicle night-time crashes and crashes designated
by the police as involving a driver that had been drinking. An interrupted time-series
analysis found significant decreases in these three road crash categories immediately
after the implementation of the new countermeasures. The researchers attempted to
isolate the effects of the reforms from changes in the social climate surrounding
drink-driving produced by the lobby group “Mother’s Against Drunk Driving”
(MADD). This was achieved by including the formation date of MADD as a separate
intervention variable in the time-series model. This additional analysis demonstrated
that, although the effects of the 1982 legislation were embedded in a larger
downward trend, they still remained statistically significant even after accounting for
this trend.

Peck (1991) also examined the impact of the 1982 Californian legislative reforms on
drink-driving recidivism rates in order to investigate the specific deterrent effect of

64
these new laws. He examined the failure rates, for both drink-driving offences and
major traffic offences (such as reckless driving), of two large offender cohorts who
received a drink-driving conviction in either 1980 or 1984. All drink-driving and
major traffic offences committed up until 1989 were included in the analysis and
survival and hazard functions were generated. The analysis showed that the 1984
cohort was less likely to reoffend than the 1980 cohort, both in terms of drink-driving
offences and other major traffic offences. It should be noted, however, that, in
addition to increased penalties, the new laws mandated alcohol-related treatment
programs for some offenders. This additional component of the legislative reforms
could have contributed to the observed reduction in reoffending rates.

An evaluation of a specific legislative amendment to Swedish drink-driving laws in

1994 provides further evidence for the general deterrent effect of more severe
statutory penalties. These legal changes provided for lower BAC levels and increased
penalties for aggravated drink-driving offences (from one year to two year
imprisonment), with the added requirement that imprisonment should be the usual
punishment for this offence. Borschos (2000) examined road traffic crashes between
1986 and 1997 to identify any changes occurring after the introduction of this
legislation, while controlling for other confounding factors, such as per capita
alcohol consumption and traffic density. This analysis revealed a significant
reduction in road fatalities during the study period, over half of which could be
explained by the introduction of the harsher penalties, independently of other factors
included in the model.

A limitation of the quasi-experimental analyses presented above is that the increases

in statutory penalties evaluated by these studies formed part of a larger ‘package’ of
legal interventions aimed at deterring the drink-driver. California’s severe statutory
penalties were introduced at the same time as a 0.1 per se law, a law that increases
the certainty of conviction by removing the need to prove behavioural impairment.
Likewise in Sweden the introduction of more severe penalties coincided with a lower
BAC limit, as well as modifications to punishments for other serious driving
violations and improved police control measures. The fact that several legal
measures were introduced simultaneously makes it difficult to determine which of

65
the interventions produced the desired effect. Since there is strong evidence that
increasing the certainty of punishment can have a substantial deterrent effect on
drink-driving, the findings reported in these evaluations may be due to changes
affecting the likelihood of arrest and conviction, rather than the increased threat of
severe sanctions.

Other research examining increases in the severity of statutory punishment for drink-
driving offences, with no concurrent increase in the certainty of punishment, has
failed to confirm the general deterrent effects of increased statutory penalties.
Beirness, Simpson and Mayhew (1993) assessed the impact of legislation introduced
in Canada in 1985, which established two new offences of ‘impaired driving causing
bodily harm’ and ‘impaired driving causing death’, both which carried maximum
penalties of 10 to 14 years in prison. The legislation also increased fines and
mandated minimum three-month licence suspensions for existing impaired driving
offences.10 An evaluation of these changes showed that, during the 1980s in Canada,
there were significant reductions in several measures of drink-driving behaviour and
its consequences, including alcohol-related fatalities and crashes, drink-driving
arrests, self-reported offending and positive breath tests in roadside surveys.
Unfortunately the downward trend was well established prior to the amendments and
there is therefore no clear evidence that could specifically link these reductions to the
legislative changes introduced in 1985.

Beirness and his colleagues suggest that the absence of any observable impact of the
increase in drink-driving penalties could have been due to low public awareness of
the legislative changes. If drivers are not aware of the increased threat in sanctions
for drink-driving, they are unlikely to alter their perceptions of the risk associated
with this criminal behaviour and therefore would not be expected to be deterred.
Furthermore, the researchers observed that relatively few drivers were charged with
the new offences and, among those who were, many were convicted of a less serious
offence.

10
The offence of impaired driving in Canada is equivalent to drink-driving offences in Australia

66
Ross (1984) also found little evidence for a general deterrent effect of a Finnish law
doubling the maximum statutory penalty for drink-driving from two to four years
imprisonment. An interrupted time-series analysis of road fatality and injury data
showed a statistically significant decline in crashes resulting in an injury but no real
change in single-vehicle road fatalities or total road fatalities. Since the latter two
crash categories more often involve alcohol, Ross concluded that the increase in the
statutory severity of sanctions in Finland did not have observable deterrent
consequences. He did, however, note that, although the legislation doubled the
maximum penalties for drink-driving, the actual penalties meted out by judges did
not change.

We noted in Chapter 2 that the certainty and severity of legal punishment are
predicted by deterrence theory to combine multiplicatively rather than additively. It
follows that increases in perceived severity should exert a bigger effect when
punishment is highly probable. The legislative changes introduced in Finland
occurred well before per se laws or random breath testing had been adopted in that
country. In Canada, a system of random stopping of vehicles is permitted but
suspicion of alcohol intoxication is required before a breath test can be requested.
Furthermore, most drink-driving enforcement in Canada results from an incident,
such as a collision, rather than random stops of drivers (National Highway Traffic
Safety Administration 2000). Thus, any potential deterrent effect of the increased
severity of punishment created by the legislative reforms in Finland and Canada
could also have been moderated by the low certainty of punishment in these
jurisdictions.

The importance of the interaction between perceptions of arrest certainty and

perceptions of penalty severity is illustrated by Hingson et al.’s (1987) evaluation of
changes to statutory penalties in Massachusetts and Maine. In the early 80s both
these USA States introduced new and more severe statutory penalties to increase the
overall threat of punishment for drink-driving offences. In the case of Maine, the
legislation was deemed the “toughest drunk driving law in the nation” (Hingson et al.
1987, p. 593). However, no sustained benefits of the legislative changes, in terms of
reductions in road fatalities or self-reported drink-driving episodes, were observed

67
for either State. Survey data suggested that the null findings might have been
influenced by the low perceived probability of arrest in these States. Pre- and post-
law surveys indicated that, while there were substantial increases in the public
perception that drunk drivers would be arrested, convicted and given severe
penalties, few people believed it was very likely that police would actually stop
drunk drivers. In instances such as these, where there exists a relatively low chance
of being caught, drivers are acting rationally in ignoring the threat of severe
punishment (Ross 1984).

The studies reviewed above reveal mixed evidence as to whether severe statutory
penalties can reduce the incidence of drink-driving through the process of deterrence.
The success of such a strategy however is likely to depend on whether the new
sanctions are implemented as intended, the public is aware that more severe penalties
have been introduced and whether the perceived probability of punishment is high or
low.

Policies impacting sentencing practice

The fact that policies increasing the severity of statutory penalties may not result in
actual changes in sentencing practice, presents a problem for studies examining the
deterrent effect of changes to statutory penalties. As Nagin (1998) observes “actual
policy bears little resemblance to intended policy because the exercise of discretion
by key actors of the criminal justice system drives a wedge between the reality of
policy and its intention as expressed by its formulators” (Nagin 1998, p. 34).
According to deterrence theory, a sudden increase in statutory penalties should
reduce offending by augmenting perceptions of punishment severity if arrested and
convicted for an offence. However, if offenders and the general public come to learn
that the legislative changes have had little impact on actual sentences imposed by the
courts, individuals may readjust their perceptions of punishment severity and any
deterrent effect created by the policy may be diminished. Negative findings of
sanction-based policies could therefore be due to the failure of the legislative reforms
to influence actual sentences rather than the ineffectiveness of more severe
punishment in deterring potential offenders.

68
Other research has overcome this problem by taking advantage of naturally occurring
variations in the sentencing practice of different judges in order to examine the
deterrent efficacy of severe punishment. When individual judges differ significantly
in the penalties they impose and court cases are assigned to judges in an arbitrary
way, a comparison can be made between groups of offenders, who would be
expected to be relatively similar in all respects except for the penalty they receive
from the sentencing judge. Martin, Annan and Forst (1993) conducted one such
‘natural experiment’ to examine the specific deterrent effect of mandatory gaol
sanctions. Capitalising on the fact that some judges did not comply with a judicial
policy in Minnesota mandating a two-day gaol sentence for first time drink-driving
offenders, these researchers compared the recidivism rates of persons sentenced by a
‘gaol’ judge with that of a ‘no gaol’ judge. They found no significant difference
between the two groups of offenders with regard to the number of drink-driving
reconvictions recorded during a 23-month follow-up period.

Research conducted by Ross and Voas (1990) also failed to find an effect for a more
severe judicial sentencing policy that mandated imprisonment. This study compared
drink-driving outcomes of two communities: one where the judiciary had adopted a
harsh sentencing policy of mandatory 15-day gaol sentences (accompanied by heavy
fines and licence restrictions) for first offenders, and another which did not. Since it
was a high profile judge who adopted the harsher sentencing policy, a great deal of
media attention surrounded the new penalties and consequently the policy was
expected to deter drink-driving in the wider community. To assess the general
deterrent effect of the policy the authors relied on several outcome measures of
drink-driving, including roadside breath tests and driver surveys. This analysis found
that the judicial sentencing policy had a strong effect on the punishment expectations
of drivers in New Philadelphia, but there was no difference between the two
communities in the proportion of people self-reporting driving under the influence of
alcohol or those who tested positive to a roadside breath test for alcohol. This finding
conflicts with the prediction of deterrence theory that increases in perceptions of
punishment severity should deter individuals from committing an offence. Again,
however, these researchers warn that the absence of a deterrent effect may have been
due to the low perceived probability of apprehension in this jurisdiction.

69
Another test of the deterrent effect of harsher sentencing practices comes from
evaluations of policies mandating minimum penalties for drink-driving offences.
Mandatory minimum penalties remove some judicial discretion from sentencing
decisions, thereby increasing the likelihood that harsher sanctions established by the
legislature will be implemented as intended. Ecological research examining
variations in road crash fatality rates across USA States provides equivocal evidence
on the effectiveness of State laws mandating specific penalties for drink-driving
offences. Sloan, Reilly and Schenzler (1994), for instance, demonstrated that
mandatory gaol terms, mandatory minimum fines and mandatory minimum periods
of licence disqualification did not succeed in deterring offenders. On the other hand,
Chaloupka, Saffer and Grossman (1993) concluded that mandatory administrative
licence suspensions, high minimum fines and minimum licence suspensions all
significantly reduced road fatalities, but mandatory gaol sentences did not.

As mentioned previously, a shortcoming of these ecological studies is their reliance

on a binary variable to index the threat level achieved by each State’s drink-driving
policies. With respect to mandatory minimum penalties, a dichotomous variable fails
to take account of interstate differences in the actual severity level of penalties
imposed on drink-drivers. Research conducted by Votey and Shapiro (1983)
overcomes this problem by using information on fine amounts, imprisonment lengths
and licence withdrawal periods to estimate the overall expected cost associated with
different types of drink-driving penalties. Including these measures of punishment
severity in an ARIMA time-series model, the researchers examined the impact of
court imposed drink-driving sanctions on fatal and serious injury road crashes in
Sweden from 1976 through 1979, while controlling for a host of confounding
variables. The findings from this multivariate analysis suggested that, of all
sanctions, licence withdrawal appeared to have had the greatest effect in alleviating
road injury in Sweden over the study period, since it appeared to influence both fatal
and serious injury crashes. Evidence for the deterrent effect of gaol sentence and
fines was somewhat less clear. Gaol sentences appeared to be effective in reducing
serious injury crashes but not in reducing fatal crashes, whereas fines had some
impact on fatalities but not on serious injury crashes. It should be noted, however,

70
that when licence withdrawal was included in the models for both classes of crashes,
the impact of gaol sentences and fines was substantially reduced.

The deterrent efficacy of licence sanctions is further evident from the results of a
quasi-experimental study of a mandatory licence suspension law. Blomberg, Preusser
and Ulmer (1987) examined the deterrent effect of a 1982 Wisconsin law that
mandated three- to six-month licence suspensions for all persons convicted of a
drink-driving offence in that State. These researchers compared the 12-month driving
records of a cohort of first time offenders convicted under the old law with a cohort
of first-time offenders convicted under the new law. A significant reduction in both
the number of alcohol-related crashes and drink-driving offences was apparent for
drink-drivers who were convicted under the new law and subsequently had their
licences suspended for 90 days or more.11 This law was also found to have a general
deterrent effect, as evidenced by a significant reduction in ‘alcohol’ road crashes
(single-vehicle injury and fatal crashes involving male drivers which occurred on a
weekend night) immediately after the mandatory penalties were introduced. Again,
however, the evaluation was complicated by the fact that the Wisconsin drink-
driving law contained numerous other provisions regarding the offence and it’s
processing (e.g. the adoption of illegal per se laws), implemented at the same time as
the mandatory suspensions.

Punishment celerity
The final major proposition of classical deterrence theory is that swift punishment –
celerity – can reduce the incidence of a crime. Very few studies have considered
whether temporal delays impact adversely on drink-driving offending and those that
have provide mixed evidence on the effectiveness of swift adjudication. Despite the
limited evidence available for the celerity of punishment, it remains an important
area for deterrence research. If drink-driving recidivism is partly a function of time to
adjudication, reducing court delay or introducing administrative penalties is an
appealing strategy for reducing drink-driving offending.

11
Blomberg, Preusser & Ulmer (1987) report that 45% of convicted Wisconsin drinking drivers lost
their licence in 1981 but in 1982, after the mandatory penalties were introduced, 100% of convicted
drink-drivers lost their licence for at least 90 days.

71
Evaluations of Administrative Licence Revocation (ALR) laws, which authorise
immediate seizure and suspension of a driver’s licence at the point of arrest, provide
perhaps the best evidence for a deterrent effect created by swift adjudication. One
such evaluation was of an administrative driver’s suspension law introduced in
Ontario in 1996, which allowed for a 90-day licence suspension at the time a drink-
driving charge was laid. Evidence for a general deterrent effect of the legal
intervention was confirmed by time-series data showing a significant reduction in the
proportion of fatally injured drivers with positive blood alcohol levels and a logistic
regression analysis which showed a significant decrease in self-reported drink-
driving immediately after the new laws were put into effect (Mann et al. 2000;
Stoduto et al. 2000).

Likewise, Legge and Park (1994) found administrative licence suspension laws to
have a positive impact on alcohol road crash rates. They report the findings of a
pooled cross-sectional time-series analysis of fatal road crashes in the 48 contiguous
USA States, which showed that, of the several deterrence-based variables considered,
administrative laws (alongside per se laws) had the greatest impact on single-vehicle
night-time road fatalities. Voas and Tippets (1999) also report that the introduction of
administrative licence revocation laws contributed significantly to the downward
trend in alcohol-related fatal crashes in the USA from 1982 through 1997. However,
these authors note that the long-term downward trend evident in this analysis is likely
to reflect the growing impact of several new laws over time, as well as other
variables not tested by their model, such as enforcement activity.

The specific deterrent effect of administrative licence suspension laws is less clear.
In their review of this literature, McArthur and Kraus (1999) identified just three
studies that had adequate comparison groups for assessing the specific deterrent
effect of administrative revocation laws. One of these studies found a positive effect
of the laws on recidivism, another found only a short-term (less than two years)
reduction in reoffending, while the third found that the law worked in some States
but not in others. The largest of these studies was the work undertaken by Stewart,
Gruenewald and Parker (1992), examining the impact of administrative licence
suspension laws introduced in Louisiana, Mississippi and North Dakota. Recidivism

72
rates for pre-law and post-law cohorts in each of the three States were compared
using survival analysis and comparisons were made with California, which did not
have administrative licence suspension laws in place. In North Dakota, there was a
significant reduction in drink-driving recidivism rates following the implementation
of the new law. The reduction in recidivism observed in this State was about one-
third, a reduction that was maintained during the three-year follow-up period.
Evidence for a specific deterrent effect of the North Dakota legislative reforms was
further supported by the fact that similar reductions in recidivism were not apparent
in the comparison State, California, over the same period. However, there were no
differences in drink-driving recidivism rates across the pre- and post-law cohorts in
either Louisiana or Mississippi. The researchers suggested that differential police
enforcement amongst the various States may have been a reason for this disparity in
results.

The findings from research evaluating administrative licence suspension laws appear
to support the deterrent effect of punishment celerity but are limited by the fact that
these types of legal reforms influence both the celerity and the certainty of
punishment.12 This characteristic of administrative laws makes it difficult to
determine which component of the deterrence equation is contributing to the
observed effect. In fact, individual-level research, using time to conviction as a
measure of celerity, has failed to confirm the deterrent effect of punishment swiftness
found in evaluations of administrative laws (Yu 1994; Howe & Loftus 1996).
Furthermore, experimental research that has considered not only the temporal delay
in formal punishment but also a potential offender’s preference for delay or
impulsivity, as per economic discounting models, has similarly found little evidence
for punishment celerity affecting offender decision-making (Nagin & Pogarsky
2001). Given what we know about the certainty of punishment, the effects of
administrative licence revocation laws may simply be due to their effect on the
perceived probability of drink-driving punishment, rather than their ability to reduce
time to adjudication.

12
Nichols and Ross (1990) argue that the administrative process is more certain in applying sanctions
due to the differences in standards of proof.

73
Individual-level research
Although aggregate-level effects of deterrence-based interventions are arguably of
greater importance to policy-makers (Gibbs 1975), much of the existing research has
evaluated policies that are either poorly implemented or which vary in intensity
across jurisdictions. Several of the quasi-experimental studies on punishment
certainty, cited above, have managed to overcome these methodological limitations
by focusing on well-defined interventions that have influenced enforcement practice
and risk perceptions as intended. However, several questions regarding the deterrent
efficacy of punishment severity remain unanswered. Specifically, it is unclear from
aggregate-level studies what form drink-driving punishment should take and whether
the magnitude of different penalties has a significant marginal deterrent effect.
Furthermore, aggregate-level studies on punishment severity have generally failed to
examine perceptions of punishment severity and whether these perceptions affect
offending likelihood.

Reoffending studies
Numerous studies of deterrence have explored the effect of judicial penalties on
reoffending rates of drink-drivers. These studies generally compare different degrees
or types of sanctions rather than examining alternatives to legal sanctions or the
imposition of sanctions where none existed previously. Thus, they present evidence
on marginal rather than absolute deterrent effects of punishment.13 One problem
confronting research on the marginal deterrent effect of penalties is that many of the
factors influencing the type and severity of the penalty allocated to a particular
offender can also affect reoffending. For example, offenders sentenced to
imprisonment could have numerous prior convictions of a serious nature and
therefore be at a greater risk of reoffending, irrespective of what penalty they receive.
To conclude that imprisonment is an ineffective deterrent, because recidivism rates
are high amongst this group of offenders, is problematic.

In order to conclusively establish a causal relationship between type and severity of

penalty and subsequent offender behaviour it is necessary to randomly allocate
sanctions to offenders. This ensures that offenders receiving different penalties do
13
A summary of these studies also appears in Appendix I.

74
not vary systematically. One such experimental study examining the deterrent effect
of drink-driving penalties was conducted by Holden (1983) in the USA. For this
study, the court randomly assigned first-time drink-driving offenders to four
experimental conditions: (i) no sanction (control group), (ii) education and therapy,
(iii) supervised probation, and (iv) supervised probation plus education/therapy. No
significant differences in drink-driving rearrest rates between the four groups of
offenders were evident from a two-year follow-up of conviction records, suggesting
that none of the four sanction types examined were more likely to deter drink-drivers
from reoffending. There was, however, some evidence that severe sanctions (in the
form of supervised probation) were more effective in reducing rearrest rates but only
for offenders identified as problematic drinkers.

Although experiments are the most methodologically sound means by which to

investigate the differential impact of penalties on drink-driving recidivism, ethical
concerns and practical difficulties in randomising punishment prevents the vast
majority of researchers using this paradigm in recidivism analyses (Ross &
Blumenthal 1975; Farrington 1983). Most studies have therefore tended to employ
statistical controls to separate out the effects of penalties from those exerted by
offender characteristics. The extent to which valid comparisons between penalties
can be made from such analyses will necessarily depend on appropriate offender
characteristics being identified and adequately measured (Homel 1980). Ideally,
these analyses would include data on a broad range of psychological and sociological
characteristics of offenders. However, reoffending studies have, for the most part,
been limited to information contained in official records, such as age, gender, prior
convictions, BAC at time of arrest and concurrent convictions for other offences.

Reoffending studies relying on these types of statistical controls to account for

potentially confounding factors have found discouraging results for the specific
deterrent effect of gaol sentences. DeYoung (1997) examined the reconviction rates
of all Californian drivers convicted of a drink-driving offence during 1990 and 1991.
Drink-driving reconvictions within 18 months was employed as the dependent
variable in the analysis and a number of demographic and prior personal driving
history measures were included as covariates. DeYoung’s analysis found that first

75
offenders receiving a two-day gaol sentence for drink-driving offences recorded
twice as many drink-driving reconvictions within the 18-month follow-up period, as
offenders allocated other types of punishment.

In his three-year follow-up study of 1,000 drink-drivers convicted in New South

Wales, Homel (1980) found that, controlling for other confounding variables,
imprisonment was no more effective than any other penalty for drink-driving. In fact,
there was some evidence that long periods of imprisonment (especially longer than
six months) actually increased, rather than decreased, the probability of reconviction
for a drink-driving offence. In the ‘natural experiment’ cited earlier, Martin, Annan
and Forst (1993) directly compared the recidivism rates of persons who received a
two-day gaol sentence, under the severe judicial sentencing policy in Minnesota,
with those who did not. This additional analysis generally supported the findings
from their aggregate-level research. Specifically, the results showed that a two-day
stay in gaol was no more effective in reducing the reoffending of first-time drink-
drivers than alternative sanctions, even after controlling for BAC at arrest, age,
gender, prior convictions, seriousness of offence and plea.

Despite the limited deterrent efficacy of imprisonment, there is some evidence that
other penalty types, particularly licence disqualification, can reduce the reoffending
rates of convicted drink-drivers. Evidence from an Australian study, for example,
shows that licence actions are successful, not only in reducing reoffending but also in
limiting the number of road crashes involving convicted drink-drivers. Siskind
(1996) examined the driving records of over 25,000 drivers convicted of a drink-
driving offence in Queensland, who had received at least one licence restriction in
1988. He found evidence for a two-thirds reduction in both crash and drink-driving
offence rates for offenders under disqualification from driving compared with
offenders who had not been disqualified. Moreover, offenders who were
apprehended for a new offence whilst their licence was disqualified were more likely
to be apprehended in the early stages of their disqualification period. This latter
finding would suggest that improvements in driver behaviour resulting from a licence
sanction would be more pronounced for persons disqualified for longer periods of
time.

76
Yu (1994) examined the effect of various penalty types on the recidivism rates of
almost 14,000 drivers with at least one conviction for drink-driving in the State of
New York, USA. He found that drivers who received higher monetary fines had
significantly reduced drink-driving recidivism rates during the three-year follow-up
period. However, there was no evidence that longer periods of licence withdrawal
had a stronger impact on drink-driving recidivism than shorter periods.

A limitation of both these studies is that they failed to adequately control for prior
convictions. The New York study included no information on previous drink-driving
convictions in its recidivism analysis. The Queensland study included a variable
representing any charge for a drink-driving offence in the preceding five years but
did not specify the number of convictions recorded for each offender. Previous
convictions is one of the best predictors of drink-driving reoffending (Ryan et al.
1996) and is also likely to affect both the length and type of penalty imposed by the
court. Research examining the differential impact of penalties on recidivism must
take this fact into account.

A study that did adequately control for previous driving record demonstrates that
more severe licence actions can reduce reoffending, but that high fines and gaol
sentences do not succeed in deterring offenders. Mann et al. (1991) conducted a
three-year follow-up study of approximately 4,000 Ontario drivers who were
convicted of at least one drink-driving offence in 1979. Importantly, this study used
data on the number of convictions recorded for each offender in the five years prior
to the index offence to examine the deterrent effect of different penalty types among
individuals with differing numbers of previous offences. The results of this analysis
indicated that longer periods of licence suspension were associated with significantly
fewer total and alcohol-related crashes for first and multiple offenders, as well as
reduced drink-driving charges for first offenders. On the other hand, higher fines (for
first time offenders) and more days in gaol (for all offenders) were associated with an
increase in the number of crashes and convictions.

77
Reviews of the drink-driving literature on specific deterrence and penalty severity
generally concur with the conclusions of Mann and his colleagues that licence
disqualification is the most effective deterrent for drink-driving and that gaol has
little effect on offender behaviour (Nichols & Ross 1990; Zaal 1994). This is also
consistent with aggregate-level research presented earlier in this chapter on
mandatory licence suspension laws. With gaol being more severe than licence
disqualifications, the superior effect of licence sanctions in reducing offending would
seem at odds with deterrence predictions. However, surveys of offenders have
suggested that a large proportion of drivers fear losing their driver’s licence and the
freedoms that come with it more than they do serving a relatively short imprisonment
term. In particular, persons who rely heavily on their motor vehicle for employment
purposes perceive the threat of licence restrictions as extremely costly (Homel 1986).
It is also worth noting that, rather than eliminating drink-driving completely, licence
restrictions result in an offender driving less often and more carefully during the
period of disqualification than they would otherwise (Ross & Gonzales 1988). Since
many drink-drivers have concurrent convictions for other traffic offences (Homel
1980), these modifications to driver behaviour through fear of further punishment are
likely to have an ancillary effect on other traffic offending, which could also serve to
heighten the overall impact of licence sanctions.

Perceptual studies
Deterrence is essentially a perceptual theory, in which criminal behaviour is thwarted
not by the actual risk of punishment being imposed or the actual severity of
punishment, but rather by what people believe these to be. Recognising this, a large
body of research has emerged investigating the relationship between individuals’
perceptions of sanction risks and their self-reported offending behaviour.

As discussed in Chapter 1, perceptual studies can be broadly classified into three

categories: cross-sectional, longitudinal or panel, and scenario-based (Nagin 1998;
von Hirsch et al. 1999). Cross-sectional studies collect information, at one point in
time, on current perceptions of sanction risks and prior offending behaviour, or
future intentions to offend. Longitudinal or panel studies, on the other hand, measure
perceptions and behaviour in the same sample on two or more occasions. These two

78
types of perceptual research have provided conflicting evidence on the deterrent
effect of formal punishment. Cross-sectional studies have consistently shown
negative associations between self-reported offending and perceived certainty of
arrest or punishment, though the evidence for an effect of perceived punishment
severity is weaker (Grasmick & Bryjak 1980; Grasmick & Bursik 1990; Paternoster
& Simpson 1996). In contrast, panel and longitudinal studies have found little or no
deterrent effect of varying levels of perceived certainty or severity on offender
behaviour (Paternoster et al. 1983; Paternoster & Iovanni 1986; Lanza-Kaduce
1988).

The validity of both these research designs in investigating hypotheses generated by

deterrence theory has, however, been questioned (e.g. Williams & Hawkins 1986;
Nagin 1998; von Hirsch et al. 1999). Cross-sectional designs, in which current
perceptions are used to predict past behaviour, suffer from the problem of causal
ordering. A significant negative relationship between perceptions of apprehension
risk and prior drink-driving episodes could reflect a deterrent effect: people who
perceive the risks as lower will offend more often. Alternatively, it could reflect an
experiential effect: people who offend more often, having escaped detection on
numerous occasions, will perceive the risks as lower. Rational models of offending
would, on the other hand, predict that perceptions are continually updated on the
basis of new information. Thus longitudinal research, which assesses current
perceptions and then observes behaviour at some future point, may be subject to
measurement error, particularly if there is a long delay between the two testing
occasions.

Ideally, deterrence research would measure perceptions of formal punishment at the

point the decision to offend is made (Williams & Hawkins 1986). The difficulties of
achieving this in practice are self-evident. Researchers have therefore attempted to
deal with the temporal lag between measures of perceptions and behaviour through
the use of scenario-based studies. This research design presents participants with
vignettes, describing realistic offending scenarios, and then requires them to indicate
the likelihood that they would offend in the given circumstances. Perceptions of
sanction risks can be measured at the same time as willingness to offend. Although

79
this methodology does not question participants on actual offending behaviour14 it
does succeed in modelling the decision-making process so that perceptions can be
measured concurrently with stated offending likelihoods. Furthermore, as noted by
Nagin (1998), research on situational crime prevention has demonstrated the
importance of the context in which crimes are committed to perceptions about
sanction risks. In drink-driving research, for example, these situational factors could
include such things as distance from destination, amount of alcohol consumed and
inconveniences avoided. Scenario-based research provides detailed information
about the circumstances of the crime and, in doing so, these situational variables are
held constant across the varied conditions or groups.

Previous research using scenario-based designs has provided consistent evidence for
deterrent-like effects of formal sanctions on offending behaviour (e.g. Klepper &
Nagin 1989b; Bachman, Paternoster & Ward 1992; Nagin & Paternoster 1993;
Nagin & Paternoster 1994; Tibbetts 1997). Furthermore, these studies have shown
that the threat of formal punishment has a deterrent effect which is independent of
time-stable individual differences, such as poor self-control and shame-proneness.
Nagin and Paternoster (1993), for example, presented college students with three
scenarios describing the commission of different offences: drink-driving, theft and
sexual assault. Respondents were asked to estimate their chances of arrest or public
exposure (without arrest), the probability of being formally sanctioned by the
criminal justice system (either by way of jail or licence disqualification) or
informally sanctioned by others (e.g. peers, family, the university) if caught, and the
likelihood they would offend in the given scenario. Feelings of shame arising from
the act itself and levels of self-control were also measured in the survey. Subsequent
analyses showed that, even after individual differences in self-control and shame
were taken into account, a significant inverse relationship between intentions to
offend and perceived formal costs was still apparent.

14
Most scenario-based studies do however measure prior offending and have shown a strong, positive
relationship between previous offending episodes and stated likelihood of offending in the described
scenario (e.g. Nagin & Pogarsky 2001). This attests to the veracity of respondents’ answers on the
dependent variable.

80
Using a similar paradigm, Tibbetts (1997) demonstrated that a measure of the overall
threat of formal sanctions (which combined information on both the risk of
apprehension and severity of sanctions) had a significant negative effect on an
individual’s intention to drink-drive, controlling for both low self-control and shame
proneness. Moreover, these external sanctions appeared to be more influential in
drink-driving decisions than in decisions relating to other types of offending
behaviour. Shoplifting, for example, was found to be inhibited primarily by internal
sanctions related to anticipated shame, rather than the risk of formal punishment.

For the most part, however, perceptual research has restricted its focus to the impact
of informal social sanctions on offending. Where survey-based research has
considered formal sanctions it mostly concentrates on the relationship between
offending and the perceived certainty of punishment or the overall threat of
punishment, rather than the independent effect of severity (e.g. Paternoster 1989;
Grasmick & Bursik 1990; Baum 1999). An exception to this general observation is
the recent work conducted by Nagin and his colleagues in the USA (see Nagin &
Pogarsky 2001; Pogarsky 2002; Nagin & Pogarsky 2003). Using scenario-based
designs, these authors experimentally manipulated the severity of formal sanctions
for drink-driving offences by randomly assigning respondents three, six and nine
month licence suspensions. Respondents were then asked to estimate the likelihood
of apprehension if they drove home in these circumstances and the likelihood they
would offend given these conditions. The researchers also collected information on a
host of other variables thought to condition deterrence effectiveness, including
impulsivity and anticipated extralegal consequences. Although subsequent analyses
found some evidence that variations in sanction severity predict stated offending
likelihood, in general these data confirmed the conclusion from ecological and quasi-
experimental studies that the certainty of punishment is a far more robust deterrent
than the severity of punishment.

A shortcoming of Nagin and Pogarsky’s approach to evaluating punishment severity

effects is it assumes that people know the magnitude of the penalty (e.g. the length of
licence disqualification) they would receive if convicted of a drink-driving offence.
Some evidence, however, suggests that public knowledge of criminal penalties is in

81
fact quite poor (Williams, Gibbs & Erickson 1980; Hough & Roberts 1998; Kenkel
& Koch 2001). This failure to consider the link between perceptions of sanction
severity and actual sentencing policy is another notable limitation of the perceptual
deterrence research conducted thus far. As Nagin (1998) points out:

While great effort (in the deterrence literature) has been committed to
analysing the links between sanction risk perceptions and behaviour,
comparatively little attention has been given to examining the origins of
risk perceptions and their connection to actual sanction policy (Nagin
1998, p. 5).

Studies like those above provide information on the likelihood of offending, were a
particular sanction risk perceived by respondents. However, they tell us nothing
about the extent to which changes in criminal justice policy will alter perceptions of
punishment severity. This knowledge gap needs to be addressed before a complete
test of the deterrence model can be realised (Gibbs 1975; Cook 1980).

Miller and Iovanni (1994) attempted to address this problem by examining the
determinants of sanction risk perceptions in relation to intimate partner violence.
They conducted a two-way panel survey of university students, some of whom self-
reported previous experiences with violent conflict in an intimate relationship, while
others reported no prior experience with partner violence. To disentangle experiential
from deterrent effects, the perceived certainty of arrest for partner violence and the
perceived severity of the consequences arising from an arrest were measured at two
time periods conducted three-months apart. Lagged measures of perceived risk and
perceived severity at Time 1 were then included as independent variables in the
regression models predicting risk perceptions at Time 2. Time 1 measures of prior
use of partner violence, as well as measures on constructs derived from social
learning and social control theories, were also independent variables in the study.
These latter variables included measures of the attitudes of significant others toward
violence and prior exposure to models of violence, as well as parental attachment,
peer attachment and commitment to conventional goals.

82
The analysis of these data found variations across gender in terms of the factors
influencing perceived risk. For certainty of arrest, prior use of violence in an intimate
relationship was associated with reduced perceived risk of apprehension, but this
effect was evident only for males. Furthermore, males who believed that partner
violence was illegal (a measure of legal knowledge), perceived arrest for these events
to be more likely. On the other hand, perceptions regarding the severity of
consequences arising from an arrest for partner violence, or the consequences arising
from police being called to an incident, appeared to be more influenced by informal
social controls, such as attachment to parents and peers, but again this effect was
significant only for males.

The study also found some ambiguous and contradictory findings. For example, peer
violence was found to be significantly related to arrest risk for females but in the
opposite direction to that hypothesised. That is, females who reported having more
violent peers perceived the risk of arrest to be higher. Also, contrary to predictions,
beliefs regarding the illegality of partner violence were negatively related to the
perceived severity of consequences arising from an arrest, for the male group. That
is, males who believed partner violence was against the law were less likely to view
arrest as severe. The authors suggest that these unexpected results could be due to the
fact that non offenders may have had difficulty in imagining themselves in a violent
relationship. The nature of partner violence and the low risk of formal intervention
for these type of offences would also mean that informal social controls, such as
moral beliefs toward violence, may be a stronger predictor of perceptions than would
be the case for other offence types. Further investigations of this type with a more
common offence, like drink-driving, would improve our knowledge of relevant
factors influencing risk perception formation.

Significance of the current study

This thesis explores one of the major hypotheses of deterrence theory – that more
severe punishment can have a marginal deterrent effect on offending rates. Despite a
wealth of studies examining this issue, the literature reviewed above reveals mixed
evidence for deterrent-like effects from more severe punishment. Evaluations of
sentencing policies that increase the severity of statutory penalties are an important

83
source of evidence on the deterrent efficacy of punishment severity. However, the
policy impact studies that have been conducted in the drink-driving domain so far
contain several methodological shortcomings that impede our ability to make
confident assessments regarding the effectiveness of this strategy. By identifying and
addressing these weaknesses, this thesis builds on prior deterrence research and, in so
doing, enhances our understanding of the relative importance of punishment severity
in controlling crime. It also explores two areas related to sanctioning policy where
few deterrence researchers have yet ventured: (1) the link between actual and
intended policy and (2) the formation of risk perceptions.

Methodological issues in policy impact studies

Drink-driving policies that stipulate more severe statutory penalties are often
introduced in response to public concern about alcohol-related road crashes or as a
result of extensive lobbying by victim and community groups. To address these
societal concerns, legislatures sometimes implement policies that affect more than
one component of the deterrence equation. The modifications to Californian drink-
driving laws in 1982 exemplify this process. These legal reforms arose partly out of
the publicity generated by the advocacy group MADD, and were aimed at
strengthening the perception amongst drink-drivers that punishment would be
unavoidable and severe (Rogers & Schoenig 1994). To achieve this, penalties for
drink-driving offences were raised substantially. However, 0.1 per se laws and
treatment services for repeat offenders were also introduced at the same time. In
these circumstances it is difficult to disentangle the effect of increased penalty
severity from other changes made to drink-driving laws. While it is reassuring to
know that increasing the overall threat of formal punishment exerts a substantial
deterrent effect, this research is unable to provide specific information for policy
makers and legislators as to which aspect of the legislative ‘package’ is most
effective.

Many evaluations of drink-driving interventions have also been carried out in

jurisdictions where the certainty of punishment is relatively low. The certainty and
severity of legal punishment are predicted by deterrence theory to combine
multiplicatively rather than additively, thus severity is most likely to matter when

84
punishment is highly probable. In North America, where much drink-driving
research has been conducted, drink-driving enforcement takes the form of sobriety
checkpoints or roadblocks. This is a limited version of RBT that requires reasonable
suspicion of drinking before an officer can obtain a breath sample from a driver.
While checkpoints have been shown to reduce drink-driving and associated alcohol-
related road crashes under certain conditions (Shults et al. 2001), the reduced risk of
being tested for alcohol impairment (Homel 1988) and the infrequent use of
checkpoints (DeJong & Hingson 1998; Shults et al. 2001) suggest that the perceived
risk of apprehension in these jurisdictions is unlikely to be sustained at a high level.
However, in Australian States, such as New South Wales (where the case study for
this thesis is situated), enforcement of drink-driving through RBT has continued at a
relatively high level since its introduction in the early 1980s. As a consequence,
significant reductions in alcohol-related road crashes (Homel 1994; Span 1995; Span
& Stanislaw 1995; Henstridge, Homel & Mackay 1997) and self-reported drink-
driving episodes (Homel 1986; Homel, Carseldine & Kearns 1988), as well as
increases in perceptions of apprehension risk (Homel 1986), have been observed in
these jurisdictions. An investigation of increased penalties in this context, where
there exists a relatively high level of visible enforcement of drink-driving offences,
would permit more informed conclusions regarding the effectiveness of severe
punishment.

Poor implementation of sentencing policies is a further reason why several drink-

driving studies have failed to find deterrent effects for severe sanctions (Ross 1984;
Mayhew, Beirness & Simpson 1995; Nagin 1998). Ross and Voas (1990), for
example, found that appeals quadrupled when a New Philadelphian judge adopted a
harsh sentencing policy providing for jail terms and the elimination of plea-
bargaining in all drink-driving cases within his jurisdiction. Similarly, Mayhew,
Beirness and Simpson (1995) found that between 20 and 40 per cent of offenders
charged with a newly introduced drink-driving offence (i.e. impaired driving causing
harm or death), which incurred severe penalties upon conviction were eventually
convicted of a less serious offence that had lower prescribed penalties. Voas (2001)
also reports that many drink-drivers escape licence suspension in the USA because
the courts want to encourage offenders to plead guilty and/or to avoid calls for jury

85
trials. Increasing penalty severity may not have a deterrent effect if actual sentencing
policy does not change. If offenders get off more lightly than they had originally
anticipated they may adjust their estimates of sanction risk downward, thus reducing
any initial specific deterrent effect created by the sanction policy. This may also have
implications for the general deterrent effect of more severe penalties by decreasing
the perceived credibility of drink-driving sanctions as other people become aware of
this leniency in sentencing practice. Only once a policy prescribing increased
statutory penalties has been shown to affect sentencing practice, can its impact on
subsequent criminal behaviour be considered.

Finally, since increasing the statutory penalties for an offence is posited to prevent
future legal violations by influencing perceptions of sanction severity, there must be
some minimum level of general awareness before a sanctioning policy can be
expected to have a deterrent-like effect. Several studies, particularly those evaluating
police interventions (Ross 1984; Cashmore 1985; Voas, Holder & Gruenewald
1997), have demonstrated that broadcasting the implementation of new drink-driving
initiatives can greatly enhance deterrence outcomes. However, amendments to
statutory penalties are generally not supported by similar enthusiastic publicity
campaigns and therefore often are not widely known in the general community.
Under these circumstances, perceptions of sanction severity would vary little as a
result of changes in punishment levels and, as a consequence, offending rates would
remain relatively unaffected.

Knowledge gaps in the deterrence literature

Nagin (1998) identifies four important research areas that have thus far received little
attention in the deterrence literature but which need to be addressed before confident
conclusions regarding the deterrent efficacy of various policy options can be drawn.
These priority areas are: (1) long term effects (2) the implementation of policies
across different population units (3) the gap between intended and actual policy
which means laws are generally not administered as intended (4) the relationship
between risk perceptions and actual sanctioning policy. While addressing all four of
these areas is beyond the scope of the current research, this thesis will bring evidence
to bear on the latter two areas highlighted as priorities for deterrence research.

86
CHAPTER 4. INCREASED DRINK-DRIVING
PENALTIES IN NEW SOUTH WALES: A Case Study

The purpose of the Traffic Amendment (Penalties and Disqualification)

Bill is to make our roads safer by implementing new and increased
penalties for certain serious traffic offences. Adoption of the legislation
will give effect to new penalties that aim to reflect the safety implications
of the offences, decreases in real value of fines since the current
provisions were put in place and increased community concern about
road safety issues, particularly when driver behaviour is regarded as
highly irresponsible. The increased penalties greatly enhance the
deterrent effect of our road penalties and will help to improve road safety
(Mr Scully; Minister for Transport, and Minister for Roads, Second
Reading of the Traffic Amendment (Penalties and Disqualifications) Bill,
21/05/98).

On September 30 1998, the NSW Traffic Amendment (Penalties and

Disqualifications) Act was introduced by the NSW State Government to enhance the
deterrent effect of drink-driving penalties and improve road safety. This legislation
essentially doubled the maximum penalties for almost all drink-driving offences.15
Various members of the NSW Parliament voiced their opposition to these legislative
amendments during the Second Reading of the NSW Traffic Amendment (Penalties
and Disqualifications) Bill. Concerns highlighted by these parliamentarians focused
mainly on the financially excessive nature of the penalty increases and the potential
benefits that the State Government would reap from any subsequent increase in
revenue. However, the opposition also claimed that the imposition of severe
punishment is unlikely to influence offending behaviour to any great extent. This
point was echoed by other non-government organisations, including The Law
Society of New South Wales – who, in a written response to the Government’s
proposal, stated “the available evidence shows that the most effective deterrent is the

15
A small number of drink-driving offences, such as occupying a seat next to a learner driver while
intoxicated were not subject to the penalty changes.

87
fear of being caught rather than the severity of penalties for speeding, drinking and
drug driving” (Mr Souris, Deputy Leader of the National Party; Second reading of
Traffic Amendment (Penalties and Disqualifications) Bill to the Legislative
Assembly, 03/06/98).

Despite this opposition, the imposition of severe penalties for drink-driving is an

extremely popular policy amongst NSW residents. In a survey, measuring public
support for policies designed to address alcohol-related harm, almost nine in 10
persons aged 14 or over supported more severe penalties for drink-driving offences
(Australian Institute of Health and Welfare 2002b). Thus, the strengthening of legal
sanctions to counter this criminal behaviour was at least reflective of public opinion
at the time.

This thesis examines whether these changes in the severity of drink-driving penalties
in New South Wales had the intended deterrent effect on drink-driving rates in this
State. As will be described in this Chapter, this case study represents a good
opportunity to test deterrence theory predictions regarding changes to penalty
severity because several of the preconditions for deterrence success have been met. If
more severe penalties are a deterrent for drink-drivers then we would expect the
effect of harsher penalties to be maximised under these conditions. This chapter also
describes the two quasi-experimental studies that were undertaken to investigate the
impact of the penalty changes: an interrupted time-series analysis of road crash rates
and a recidivism analysis comparing drink-drivers convicted pre-1998 with those
convicted post-1998. Before describing the methodology used in these studies, it is
important to gain an understanding of the context in which this case study is situated
and the sentencing policy that forms the independent variable in this research. The
following section therefore discusses drink-driving legislation in New South Wales
and the amendments that were made to statutory penalties for drink-driving offences
in 1998.

88
Drink-driving legislation in New South Wales and the 1998
penalty changes
In New South Wales, drink-driving encompasses four distinct offences based on the
alleged offender’s BAC at the time of the infringement.16 These four offences
include a special-range Prescribed Concentration of Alcohol (PCA) offence17 with a
BAC limit of 0.02g/100ml,18 a low-range PCA offence with a BAC limit of
0.05g/100ml, a mid-range PCA offence with a BAC limit of 0.08g/100ml and a high-
range PCA offence with a BAC limit of 0.15g/100ml.19 For each of these offences
the culpable act is defined as operating a motor vehicle with a BAC in excess of the
specified limit. Thus there is no need to prove behavioural impairment resulting from
alcohol intoxication. Drink-driving laws such as these, where evidence of an illegal
BAC is sufficient to prove the offence, are known in the drink-driving literature as
per se laws. They are also strict liability offences, since there is no mental element
(mens rea) required for the offence to occur. The term “per se” refers to the fact that
neither mens rea nor behavioural impairment due to intoxication needs to be proven
to successfully prosecute this offence.

In 1998, the penalties prescribed for these offences were defined under section 4E of
the NSW Traffic Act 1909. This legislation prescribed maximum monetary fines for
each of the four PCA offence categories and maximum gaol sentences for mid- and
high-range PCA offences. The magnitude of these penalties varied as a function of
the PCA category within which the current conviction fell and whether the current
conviction was deemed to be a repeat offence20 (though the maximum gaol term did

16
These offences were originally defined under section 4E of the NSW Traffic Act 1909 but can now
be found under section 9 of the NSW Road Transport (Safety and Traffic Management) Act 1999.
17
PCA is the offence category for drink-driving in NSW.
18
Special range offences apply to special category drivers and supervisors. This primarily includes
holders, and supervisors, of a learner licence or a first-year provisional licence, under 25 drivers who
have had a licence for less than 3 years, persons not authorised to drive in NSW, drivers of heavy
vehicles and drivers of public vehicles or vehicles that carry dangerous loads.
19
More recently, an offence for the Novice PCA range was introduced in NSW (see NSW Road
Transport (Safety and Traffic Management) Amendment (Alcohol) Act 2004). This offence applies to
learner or provisional drivers and prohibits these drivers from driving with any alcohol present in their
blood. Since this legislation was not implemented until 2004, after the completion of the current
analysis it won’t be considered in the remainder of the thesis.
20
Penalties for a second or subsequent offence are applicable if a person has been convicted of a
drink-driving or other major offence in the previous 5 years. Major offences primarily consist of
driving offences resulting in death or bodily harm to another person and any of the four PCA offences
(see Part 1 of NSW Road Transport

89
not vary by number of offences for mid-range offences). All PCA offences also
attracted some form of licence disqualification under this Act. Maximum licence
disqualification periods were specified for each PCA offence category, as were
automatic disqualification periods to be applied in the absence of a specific court
order. Minimum licence disqualification periods were specified for all PCA
offences21 with the exception of low- and special-range offences in cases where it
was an offender’s first offence. Furthermore, any person charged with a high-range
PCA offence could have their licence suspended by a police officer within 48 hours
of being charged and the suspension remained in effect until a court determined the
charge.

With the implementation of the NSW Traffic Amendment (Penalties and

Disqualification) Act 1998, these statutory penalties underwent several changes.
Table 4.1 shows the penalties applicable to drink-driving offences before these
legislative amendments came into effect and the penalties applicable after these
changes. As seen here, the legislative amendments significantly increased the
magnitude of almost all penalties for drink-driving offences. These harsher penalties
included doubling the maximum gaol terms applicable to mid-range and high-range
PCA offences and doubling the maximum monetary fines for all PCA offences.
Mandatory three-month minimum licence disqualification periods were also
introduced for first offenders charged with a special-range or low-range PCA offence
(offences which previously had no such mandate) and minimum licence
disqualification periods were doubled for all other PCA offences. Furthermore, the
legislative amendments allowed for the licences of mid-range offenders to be
immediately suspended by police until the offender’s appearance in court, a
regulation that had previously only been applicable to high-range PCA offences, and
introduced a new licence disqualification penalty of up to five years for offenders
who accumulated three or more convictions within a five-year period (these

(General) Act 1999).

21
Although there were other legislative provisions that allowed a court to avoid disqualification,
notably section 556A of the NSW Crimes Act 1900 which is now equivalent to the section 10 of the
NSW Crimes (Sentencing Procedure) Act 1999.

90
offenders would have to be declared an “habitual offender” by the court to be eligible
for this new penalty).22

Table 4.1: Statutory penalties for drink-driving offences in New South Wales before
and after the 1998 statutory amendments (penalty changes are in bold text)
PCA offence Penalty type First offence Second and
subsequent offences
pre-1998 post-1998 pre-1998 post-1998
Special range PCA Max. fine $500 $1,100 $1,000 $2,200
(BAC .02 to < .05)
Licence
Disqualification
- min. none 3 months 3 months 6 months
- max. 3 months 6 months Unlimited Unlimited
- auto* none 6 months 12 months 12 months
Low range PCA Max. fine $500 $1,100 $1,000 $2,200
(BAC .05 to < .08)
Licence
Disqualification
- min. none 3 months 3 months 6 months
- max. 6 months 6 months Unlimited Unlimited
- auto* none 6 months 12 months 12 months
Mid range PCA Max. fine $1,000 $2,200 $1,000 $3,300
(BAC .08 to < .15)
Max. gaol term 6 months 9 months 6 months 12 months
Licence
Disqualification
- min. 3 months 6 months 6 months 12 months
- max. Unlimited Unlimited Unlimited Unlimited
- auto* 12 months 12 months 3 years 3 years
Immediate No Yes No Yes
suspension**
High range PCA Max. fine $1,500 $3,300 $2,000 $5,500
(BAC .15 &
above)
Max. gaol term 9 months 18 months 12 2 years
months
Licence
Disqualification
- min. 6 months 12 months 1 year 2 years
- max. Unlimited Unlimited Unlimited Unlimited
- auto* 3 years 3 years 5 years 5 years
Immediate Yes Yes Yes Yes
suspension**
* “Automatic” is the disqualification period that applies in the absence of a specific court order
** Suspension of licence within 48 hours of being charged for this offence and until the charge is
determined by the court

22
The NSW Traffic Act 1909 was repealed on 1st December 1999. Maximum monetary penalties and
gaol terms applicable to drink-driving offences in NSW can now be found in the NSW Road Transport
(Safety and Traffic Management) Act 1999. Relevant licence sanctions can be found in the NSW Road
Transport (General) Act 1999.

91
The legislative amendments did not, however, affect the power of police to conduct
Random Breath Testing (RBT) to ascertain a driver’s BAC level. In New South
Wales, in accordance with section 4E of the NSW Traffic Act 1909 (now Division 3
of the NSW Road Transport (Safety and Traffic Management) Act 1999), police
officers can request a preliminary breath test from any person driving a vehicle,
attempting to start a motor vehicle or supervising a learner licence while the learner
is driving. When requesting a breath sample from a driver, the police officer does not
have to have any suspicion that the person has been drinking. This section of the Act
also permits police to request or signal a driver to stop his or her vehicle for the
purpose of a preliminary breath test and if a driver is found to be over the legal limit
or refuses to undergo a breath test, the police officer has the authority to immediately
arrest the person and take them into custody. Once in police custody an evidential
breath analysis is conducted by an authorised police officer specially trained in
breath analysis techniques. So long as the breath analysis is conducted within two
hours of the original incident, the reading at the time of the evidential test is deemed
to be the BAC recorded at the time of the incident. It is an offence for a driver to
refuse to undergo a preliminary breath test23 or to refuse to stop if requested to do so
by police. Failure to submit to an evidential breath analysis and refusing to stop are
offences attracting the same penalties as a high-range PCA offence.

The case study

The 1998 changes to drink-driving penalties in New South Wales, which are
described above, represent a critical case in testing propositions of deterrence theory
because (1) the legislation substantially raised the severity of formal punishment
without any concurrent changes to punishment likelihood and (2) the offence and the
jurisdiction to which the penalties relate meet several of the conditions necessary for
optimising deterrent effects.

A key assumption of deterrence theory is that more severe punishment can prevent
future offending. The NSW legislative changes to statutory penalties for drink-
driving represented a substantial increase in the threat of formal punishment for this

23
Unless the person is unable to do so on medical grounds.

92
offence. The sentencing policy essentially doubled the prescribed maxima and
minima for drink-driving offences, and these were crimes that already attracted
relatively harsh sanctions. Also the penalty changes, unlike others previously
evaluated in the deterrence literature, were not accompanied by any legislative
measure designed to increase the perceived risk of apprehension. The policy, thus,
affected only one element of the deterrence equation, meaning that any subsequent
effect on offending can more confidently be attributed to increases in the level of
punishment severity.

There are also thought to be several different circumstances within which deterrence
predictions are thought to hold true. As highlighted in Chapter 2, the presence (or
lack) of informal social controls is one important factor thought to condition threat
responsiveness. For crimes that lack strong moral condemnation by the general
community, social disapproval from apprehension and conviction for an offence is
thought to be more influential in offending decisions than social stigma arising from
performing the act itself. Given that drink-driving offences generally do not elicit the
same level of moral condemnation as more traditional crimes, we would anticipate
deterrence-based interventions aimed at drink-drivers to be more efficacious.

South (1998) identifies a further six conditions necessary for deterrence to succeed in
influencing offender behaviour:
(1) The proscribed behaviour must be adequately defined
(2) Potential offenders must believe they cannot avoid apprehension if caught
(3) Sufficiently severe penalties must be attached to the offence
(4) The targeted behaviour must be the result of a deliberate decision made in a
reasonably rational state
(5) Potential offenders must perceive a high risk of being detected if they
commit the offence
(6) The punishment prescribed for the offence must be known and be
unavoidable
The offence of drink-driving meets the first four of these six criteria. As discussed in
Chapter 2, per se laws have removed any doubt about what constitutes an offence of
this type and technological advancements in breath testing devices have significantly

93
improved the chances of apprehension and conviction for a drink-driving offence. In
most cases, drink-driving also involves an element of planning, given that the drink
driver must choose to both drive to the drinking establishment and to consume
alcohol. This behaviour can therefore be considered the result of a deliberate
decision, rather than an impulsive or an automatic reaction. Furthermore, over the
last two or more decades legislatures have tended to adopt a tough stance on drink-
driving and related behaviours. Consequently, it is an offence that, in many States
and Provinces, attracts quite severe sanctions relative to other traffic violations.

The final two conditions for deterrence success articulated by South are also present
in the particular jurisdiction that has been selected for this case study:
• Potential offenders must perceive a high risk of being detected if they commit
the offence
As discussed in previous chapters, RBT is an essential component of drink-
driving enforcement in Australia and has been successful in creating a public
perception that the likelihood of detection for these offences is high. This
goal has not been achieved to the same extent in many other western
countries. Evidence from an international survey shows that eight out of 10
Australians report having ever been checked by the police for the presence of
alcohol but only 63 per cent of Canadians, 29 per cent of Americans and 16
per cent of respondents from the United Kingdom report personal experience
with police checks for drink-driving (Williams, Fergusson & Camissa 2000).
The introduction of RBT in the Australian State of New South Wales was
particularly successful in raising perceptions of punishment certainty and was
demonstrated to have a long-term impact on alcohol-related road crashes
(Henstridge, Homel & Mackay 1997). Evidence suggests that this high level
of drink-driving enforcement is still present today. A 2001 survey of NSW
drivers found that 71 per cent had seen an RBT unit in operation during the
previous six months and almost one in five recalled being personally breath
tested during this six-month period (Mitchell-Taverner 2002a). If deterrence
theory is right, increasing punishment severity should have a suppression
effect on drink-driving in this context, ceretis paribus.

94
 • The punishment prescribed for the offence must be known and be
unavoidable
The severity-based policy being examined in this research, introduced
mandatory penalties for drink-driving offences by stipulating minimum
licence disqualification periods for all drink-driving offences. To ensure that
the message was conveyed to drivers that severe punishment for this offence
is unavoidable, the statutory amendments made no exceptions for any
particular offences or types of offenders.24 Whether this threat was
successfully translated into actual sentencing practice and adequately
conveyed to the target audience are questions that will be explored in greater
detail throughout this thesis. The intention of the policy, however, was to
create a perception that severe punishment will be imposed on all persons
detected for a drink-driving offence.

The 1998 sentencing policy introduced in New South Wales, Australia, thus provides
a good test of one of the main propositions of deterrence theory: that an increase in
the severity of punishment can reduce the likelihood of offending.25

Measuring policy effectiveness - The quasi-experiment

A major step in designing and conducting a case study in order to test deterrence
hypotheses is defining the units of analysis. In this case study, the unit of analysis
could be the entire State, or each individual within the State. The former is
appropriate when we seek to answer questions regarding the policy’s effect on
offending rates at the aggregate level. The latter is appropriate when we seek
information about a policy’s influence on individual risk perceptions and behaviour.
An aggregate-level analysis is essential in determining the social policy implications
of alterations to punishment severity (Gibbs 1975) and therefore comprises a major

24
Note that in some Australian jurisdictions (e.g. Qld) drivers can be given a licence restriction as
opposed to a licence disqualification which allows certain offenders to use their vehicle at particular
times (e.g. to go to work). This type of licence sanction would reduce perceived punishment severity
and raise doubts about the certainty of punishment being imposed. Licence restrictions are not
currently available as a sentencing option in NSW.
25
Being a single-case study, one question that may arise in relation to this work is how generalisable
these findings are to other areas of offending or other types of offenders. This is not, however, the
intended aim in case study research. As Yin (1994) notes “case studies, like experiments, are
generalisable to theoretical propositions and not to populations or universes” (Yin 1994, p. 10).

95
component of this research. This takes the form of a quasi-experiment examining the
impact of the 1998 legislative amendments on alcohol-related road crash rates (Study
1) and drink-driving recidivism rates (Study 2) in New South Wales. If the policy did
achieve its deterrence objectives, then we would expect to see evidence from these
two studies for:
1. A significant reduction in alcohol-related road crashes immediately after the
introduction of more severe statutory penalties in 1998
2. A significant reduction in the likelihood of reoffending and the time taken to
reoffend for drink-drivers convicted after 1998.

The next sections describe the purpose of these two studies and the methodologies
employed in this research.

Alcohol-related road crashes - Study 1

The first study relies on statistics of alcohol-related road fatalities and serious crashes
to measure changes in drink-driving offending rates after the penalty changes were
introduced in New South Wales in 1998. Road crash data are frequently used in
drink-driving studies to index offending levels because of the significant positive
relationship that exists between crash risk and alcohol consumption. Case-control
studies have consistently shown that the risk of incurring a road crash increases as a
function of a driver’s Blood Alcohol Concentration (BAC). Birrell (1974), for
example, estimates that, at a BAC of 0.05g/100ml, the risk of crashing is about
double what would be expected at a zero BAC. At a BAC of 0.10g/100ml it is about
seven times greater and at 0.15g/100ml BAC it is about 25 times greater than the risk
of crashing at zero BAC. The degree of crash severity is also significantly related to
levels of alcohol intoxication, with more severe crashes recording a greater
proportion of intoxicated drivers (Evans 1991; Zador 1991; Zador, Krawchuk &
Voas 2000).

Crash data is also considered a more reliable and well-documented source of

information about drink-driving offending than other readily available measures. A
common problem associated with traditional measures of offending, such as arrest
statistics, is that they necessarily include only crimes which come to the attention of

96
officials. However, criminologists have long known that only a small fraction of
most criminal acts are ever reported to police. Moreover, officially recorded offences
can vary as a consequence of interventions other than those under investigation,
which could lead to erroneous conclusions about the underlying reason for observed
changes. For example, variations in recorded drink-driving arrests or positive breath
tests may simply reflect changes in enforcement practice rather than actual variations
in drink-driving behaviour. Almost all road fatalities and crashes resulting in serious
injury, on the other hand, are reported to authorities because of the need to treat
victims (Ross 1984). The use of road crash data, therefore, allows for more accurate
and reliable conclusions to be made about the impact of a specific policy on
offending rates in that jurisdiction.

Although a strong association exists between alcohol consumption and injury

crashes, a large proportion of injury crashes are not alcohol-related and therefore
would not be expected to vary as a function of drink-driving legal innovations. For
this reason, more refined measures of alcohol-related crashes need to be employed to
investigate trends in drink-driving. Ideally, trends in road crashes where the driver
records a BAC over the legal limit would be used for this purpose. However, the
BAC of drivers is not always reliably recorded for non-fatal road crashes. This is
particularly true for the current case study. In New South Wales, despite legislative
provisions requiring that all drivers involved in a crash resulting in injury are tested
for the presence of alcohol, 28 per cent of all injury crashes occurring between 1994
and 2001 had no information available on the BACs of the drivers involved. This
figure increases in the latter years of this study period, peaking at 39 per cent in
2001.26 These missing data would undoubtedly affect the usefulness of recorded
BACs as a measure of drink-driving trends.

Two surrogate measures of alcohol-related road crashes, all fatal crashes and single-
vehicle night-time crashes (i.e. crashes occurring between 8pm and 4am which
involve only one vehicle; Voas, Holder & Gruenewald 1997), are therefore employed
in this study. Both of these categories include a greater proportion of alcohol-related

26
These figures are derived from the data set obtained from the NSW RTA for the interrupted time-
series analysis.

97
incidents, with 22 per cent of fatal crashes and 40 per cent of single-vehicle night-
time crashes occurring between 1994 and 2001, where alcohol involvement was
known, having at least one controller with a BAC over the legal limit (cf. nine per
cent for all injury crashes and 29 per cent for all night-time crashes). These two
surrogate measures of alcohol-related crashes also have frequently been used in
drink-driving studies to investigate the effect of deterrence-based initiatives (Ross
1984; Hingson et al. 1987; Henstridge, Homel & Mackay 1997; Voas, Holder &
Gruenewald 1997).

The interrupted time-series analysis

An interrupted time-series analysis is used in Study 1 to examine changes in crash
rates after the 1998 legislative reforms took effect. Interrupted time-series designs
infer the effects of a treatment by comparing measures of the outcome variable many
times before the treatment with measures taken many times after the treatment has
been implemented. A model can be constructed from the pre-observational data as to
what the trends in the outcome measure are likely to be in the absence of any
intervention. The actual trends in the outcome measure can then be compared with
those in the predicted model to see whether they differ significantly. The extent to
which they do differ may then be attributable to the intervention (Cook & Campbell
1979).

To undertake this analysis, unit record crash data for a period of eight years, from
January 1 1994 through December 31 2001, were extracted from the Traffic Accident
Database System (TADS). This database, which is maintained by the NSW Roads
and Traffic Authority (RTA; the motor vehicle licensing authority in New South
Wales), includes information for all crashes in New South Wales reported to police
in which a person was killed or injured, or at least one motor vehicle was towed
away.27 No identifying information, such as names or vehicle registration numbers,
was obtained. Consistent with previous Australian research (Henstridge, Homel &
Mackay 1997; Chikritzhs et al. 2000), only crashes resulting in the death or injury of
at least one person involved in the collision are considered. These crashes, defined

27
Two additional criteria for the inclusion of crashes into the RTA data collection are (a) that the
crash occurred on a road open to the public and (b) that the crash involved at least one moving road
vehicle.

98
here as injury crashes, are thought to be both more reliably reported and more likely
to be alcohol-related.28

Since injury crash numbers are likely to vary as a function of the number of vehicles
on the road, monthly crash rates per 100,000 motor vehicles registered in New South
Wales were examined in order to control for road usage.29 Vehicle registration data is
not, however, reported on a monthly basis in New South Wales. Instead, the RTA
reports two six-monthly snapshots of the number of vehicles registered in New South
Wales as at the June 30 and December 31. Monthly vehicle registrations were
extrapolated for the current analysis from these six-monthly data.30 Using interrupted
time-series analysis, the monthly rate of injury crashes31 per 100,000 vehicles was
thus examined to determine whether the legal intervention was associated with a
reduction in crash rates. The primary aim of using an interrupted time-series analysis
in this instance is to isolate the effects of the legal changes from other trends or
variability in the crash data.

Additional interrupted time-series analyses are carried out for the two ‘alcohol-
related’ road crash categories, that is single-vehicle night-time and fatal crashes.
Again, crash rates per 100,000 vehicles registered are considered in order to control
for changes in road usage. In an attempt to control for potentially confounding
variables, such as weather conditions, interrupted time-series analyses are also
carried out for a low alcohol-involvement crash category (or control series), namely
vehicle-to-vehicle daytime crashes (i.e. multiple-vehicle crashes occurring between

28
Unlike previous road crash research conducted in NSW (Henstridge, Homel & Mackay 1997;
Chikritzhs et al. 2000), serious injury crashes could not be used for the current study. This is due to
the fact that, since 1997, NSW police no longer make the distinction between serious and other injury
when recording details about a road crash. This information, therefore, was not available in TADS for
all crashes recorded during the study period.
29
Registrations were considered to be a more accurate measure of road usage than licences on issue
since many people have driver’s licences for identification purposes, no longer drive or do not have
access to a vehicle.
30
Excluded from the vehicle registration data were registrations of trailers and plant equipment. It
should also be noted that the RTA changed their registration data collection from December 1994 and
onwards by excluding ‘outstanding registrations’. These are registrations that have been transferred to
another person but are yet to be confirmed with the RTA by the new registrant. This accounts for
anywhere between 30,000 and 50,000 registrations each year. Since this only represents 1% of all
registrations in the State it was decided that this discrepancy was not sufficient to bias the crash rates
and therefore, 1994 registration statistics could be used for the analysis.
31
Fatal crashes are included in the injury crash category.

99
9am and 3pm; Henstridge, Homel & Mackay 1997). For this crash category, only
one per cent of crashes occurring between 1994 and 2001, where alcohol
involvement was known, had at least one controller who recorded an illegal BAC.
Trends in high alcohol-involvement crash rates are then compared with trends in
crash rates for the low alcohol-involvement category. Since the legal intervention
would be expected to affect only drink-driving offending, any apparent impact of the
policy should be less apparent for crashes less likely to involve alcohol.

Ninety-six monthly observations were used as data points in the series (57 pre-
intervention observations and 39 post-intervention observations), which is considered
an adequate number of time-points for an interrupted time-series analysis (Cook &
Campbell 1979; Box, Jenkins & Reinsel 1994). The intervention point for the change
in penalties was October 1998, the first full month following the enactment of the
Traffic Amendment (Penalties and Disqualifications) Bill on 30 September 1998.32 A
Dickey-Fuller test for the presence of a unit root was initially conducted for each
crash series to ascertain whether a regression procedure could be utilised in the
interrupted time-series analysis (Koop 2000). Estimates of the Augmented Dickey
Fuller test statistic (ADF) were obtained using E-Views4. For each series the test
statistic was significant at the 0.05 level (injury rate => ADF = -7.026, fatal rate =>
ADF = -7.257, single-vehicle night-time rate => ADF = -4.615, multiple-vehicle
day-time rate => ADF = -5.023), indicating that the null hypotheses, in regard to the
presence of a unit root, could be rejected.33 Therefore it was concluded that all four
data series were trend stationary and that an Ordinary Least Squares regression
approach was an appropriate procedure to adopt for the interrupted time-series
analyses.

32
It could be argued that July 1 1998 should be used as the intervention point since the legislative
changes were announced by the government in late June and assented to on the July 2 1998. However
further interrupted time-series analyses using July 1998 as the intervention point revealed no change
in the significance of the intervention variables and only slight changes in the size of the estimated
coefficients in the model.
33
For the injury and multiple-vehicle day-time series the Augmented Dickey-Fuller (ADF) test was
based on the last 56 observations only. The reason for not using the first part of the series concerned
the change in the level of the series after the intervention. A visual inspection of the time series plot
suggested that the change in the level of the series in the latter half of the study period might be large
enough to overwhelm the identification statistics, which could lead to incorrect specification of the
model (Enders 1995).

100
Examination of the autocorrelation function plots of each data series indicated that
all four series had seasonal autocorrelation. To control for this autocorrelation, 11
dummy variables corresponding to months of the year were included as independent
variables in the models. September was used as the referent category in the model
since it was thought to be the month that would be least likely to be affected by either
school/public holidays or changes in rainfall. The regression model also contained
variables for:
1. The underlying trend
2. The level of the series from October 1998
3. The underlying trend from October 1998

The two intervention terms included in the model to quantify the impact of the new
policy on crash rates were the variables for the level of the series and the underlying
trend of the series from October 1998. The first of these variables modelled an
intervention effect where the level of crash rates changed immediately after the more
severe penalties were introduced. This term was a binary variable corresponding to
the implementation of the new penalties in October 1998. The second of these
variables modelled an intervention effect where there was a change in the underlying
trend of the data series after the legal intervention. This variable was coded 0 for the
first 58 months prior to the new penalties being introduced and 1-39 thereafter. It is
the significance and magnitude of the estimated parameters for these two terms,
which are of most interest in this analysis. If the more severe penalties introduced in
New South Wales did deter drinking drivers we would expect to see either one or
both of these terms statistically significant and have negative coefficients in the
regression model.

Once the parameters of the model were estimated, diagnostic checks of the residuals
were conducted to test for the presence of autocorrelation. This was done by
examining the autocorrelation function (ACF) and partial autocorrelation function
(PACF) plots for the estimated models, as well as Ljung-Box Q-statistics and their p-
values.34

34
The Q-statistic at lag k is a test statistic for the null hypothesis that there is no autocorrelation up to
order k.

101
Drink-driving recidivism rates - Study 2
Study 2 compares the reoffending patterns of two cohorts: one group consisting of
offenders convicted of a drink-driving offence in New South Wales in 1997 (pre-law
cohort), the other consisting of offenders convicted of a drink-driving offence in New
South Wales in 1999 (post-law cohort). The overall aim of this study is to examine
whether drink-driving offenders sentenced after the new more severe penalties were
introduced, differ significantly in their rates of recidivism from those sentenced
under the old penalty system.

This study builds on Study 1 in two important ways. Firstly, it provides an additional
source of data to measure the effectiveness of the legislative reforms on drink-
driving offending. One concern with relying on motor-vehicle crash rates as a
measure of drink-driving offending is that small changes in offending rates may not
be evident in these aggregate-level data. This is because many drink-driving episodes
do not result in road crashes and subsequently would not be captured in monthly
crash data. Null findings from the interrupted time-series analysis may, therefore, be
due to the policy having no discernable impact on drink-driving offending or,
alternatively, due to the measure of offending not being sensitive enough to detect
any effect. Study 2 attempts to rule out the latter interpretation.

Secondly, if the results of the interrupted time-series analysis do show that the
introduction of the new law had a significant effect on crash rates, Study 2 can
investigate the mechanism by which the law has impacted offending. For example,
did the legislative changes reduce the probability of offending, decrease the time to
first offence or reduce offending amongst certain groups of offenders (e.g. high-
range PCA offenders)? This study also examines the specific deterrent effect of the
policy – that is, whether offenders, with a previous conviction for a drink-driving
offence, are responsive to changes in punishment variables.

A threshold question when examining the effect of changes to sentencing law on

recidivism is whether the legal changes resulted in any increase in the severity of
penalties actually imposed by the courts. If they did not affect sentencing practice,

102
any change in recidivism cannot be attributed to the specific deterrent effect of the
new laws. In addition to sentencing practice it is also important to consider whether
the legislation had any negative impact on the prosecution of drink-driving offences.
As discussed in Chapter 3, circumvention of the more severe penalties could reduce
the perceived certainty of punishment and subsequently diminish any impact the
sentencing policy may hope to have in deterring offenders. Given this, before
assessing changes in reoffending rates, an analysis of drink-driving prosecutions and
average drink-driving penalties imposed by magistrates in the local courts, before
and after the legislative amendments, is undertaken.

Measures of recidivism

In Study 2, the number of convictions for a new offence is used to index recidivism.
Reconviction rates represent only one of several measures that could be used to
assess reoffending. A number of other drink-driving studies have relied on crash
statistics as an indicator of reoffending because of its close relationship with drink-
driving and its public health significance. However, as mentioned previously,
missing data on driver BAC, particularly in the latter years of the study period,
means that these data would not be a reliable indicator of recidivism. Although
more refined surrogate measures like single-vehicle night-time crashes could be
used, the total number of these crashes occurring in New South Wales each year is
relatively small (approximately 500 annually). The number of convicted drink-
drivers incurring this type of crash would be even smaller, making it difficult to
detect any significant change in reoffending rates after the penalty increase.

An alternative measure of offending is participant self-report. This measure of

recidivism is advantageous in that it allows for a relatively accurate estimate of the
‘true’ rate of reoffending. Court or crime data, on the other hand, only reflect those
offending episodes that come to the attention of law enforcement agencies. The
retrospective nature of the current study did not permit the use of self-report but an
estimate of the true number of new offences committed by each offender is not
necessary for this study. The question of interest in this analysis is whether the pre-
law cohort differs significantly from the post-law cohort in their pattern of

103
offending and all that is needed to answer this is an unbiased indicator of
reoffending (Luke & Lind 2002).

The issue then becomes whether offenders from one group are more likely to be
arrested and convicted for a new offence than offenders in the other. It would be
reasonable to assume that there would be little difference between the two cohorts in
terms of the probability of conviction for a new offence since the legislative reforms
were intended to affect only the severity of punishment. However, the probability of
arrest remains a potential source of bias if drink-driving enforcement levels changed
substantially over the study period.

It should be noted here that the actual probability of arrest for drink-driving is
extremely low: estimated to be somewhere between one in 82 trips, based on data
from self-report surveys (Liu et al. 1997), and one in 300 trips based on roadside
breath-test surveys (Voas & Hause 1987). Furthermore, RBT, which is the primary
means of enforcement of drink-driving offences in New South Wales, aims not to
catch more offenders but to increase the perceived likelihood of apprehension for
drink-driving amongst the general motorist population through highly visible and
unpredictable road-side testing (Homel 1988). Thus, even if the number of breath
tests were substantially different during the two follow-up periods this would not be
expected to affect the actual probability of arrest substantially enough to render
conviction rates unreliable. Increased breath testing by NSW Police during the study
period could, however, have an additional deterrent effect on drink-driving
offending (whether measured by road crash or conviction statistics) and therefore
will need to be taken into account when interpreting the findings from Study 1 and
Study 2.

The recidivism analysis

Study 2 employs a non-equivalent control group design to examine changes in

reconviction rates. In these types of studies, the responses of a treatment group and
comparison group are measured before and after an intervention. The treatment
group in Study 2 consists of drink-drivers convicted after the increase in statutory

104
penalties (post-law cohort) and the comparison group consists of drink-drivers
convicted before the penalty changes (pre-law cohort). The validity of inferences
made from this design relies on the treatment and comparison groups being as near
identical as possible, but since there is no random allocation to groups there still
exists the possibility that one group differs systematically from the other. If the
initial differences between the two groups are sufficiently large, any observable
effect of the intervention may be due to selection bias. Therefore, attempts are made
to select groups that would be relatively similar in all respects other than the
variable of interest. Where this cannot be achieved, statistical controls can be
incorporated into the analysis in order to account for any pre-existing differences
between the groups.

Data for the two offender groups being compared were drawn from NSW Bureau of
Crime Statistics and Research Reoffending Database for this analysis. This database
has assigned a unique identifier, based on several decision criteria, for every
individual convicted of an offence in the NSW Local Courts since 1995. This
identifier allows court appearances to be aggregated across years for each offender,
giving an overall total of the number and type of offences for which any individual
has been convicted (see Weatherburn, Lind & Hua 2003 for more discussion on
matching criteria employed in the construction of this database).

The pre-law cohort consisted of individuals convicted of a Prescribed Concentration

of Alcohol (PCA) offence during the period January 1 1997 through December 31
1997, while the post-law cohort was individuals convicted of a PCA offence during
the period January 1 1999 through December 31 1999. Offenders in the pre-law
cohort, who reappeared for a drink-driving offence in 1999, were excluded from the
1999 cohort to avoid duplication. The follow-up period in which reoffending is
measured commences from the date of the first PCA conviction and continues for at
least 36 months post-conviction. For the pre-law cohort, the cut-off date was
December 31 2000 and for the post-law cohort, December 31 2002. For offenders
convicted earlier in the year of interest the actual follow-up period exceeds 36
months, which will be taken into account in the survival analysis. A three-year

105
follow-up period has previously been estimated to capture approximately 60 per
cent of offenders eventually reconvicted for a drink-driving offence (Homel 1980).

Two measures of reoffending are used in this study to examine the effectiveness of
the new penalties: (1) the proportion of offenders who were convicted of a new
drink-driving offence during the follow-up period and (2) the number of days to first
reconviction for a drink-driving offence. These measures of offending have been
used successfully in previous offender studies to establish differences in recidivism
(Peck 1991; Stewart, Gruenewald & Parker 1992; Voas, Holder & Gruenewald
1997; Lind et al. 2002; Luke & Lind 2002). Frequency of offending is also a useful
measure of recidivism. However, in order to minimise the effect of other contextual
variables on offending rates the two groups selected for the analyses were close in
temporal proximity. Some of the follow-up period for the pre-law cohort would
therefore have overlapped with the implementation of the new penalties, rendering
the frequency comparison between groups problematic.

A reconviction for drink-driving is defined in this analysis as any finalised court

appearance35 for a PCA offence during the follow-up period. Date of finalisation is
used to identify the time at which the new offence occurred. It was not possible to
use arrest date or offence date in this analysis since many of these dates are missing
from the NSW Bureau of Crime Statistics and Research’s Local Courts database.
Although there would be a long delay between the offence date and the finalisation
date, this would bias the results only if drink-driving offences were finalised sooner
for one of the two groups. The average time from offence date (where these dates
are available) to court finalisation date for drink-driving offences is therefore
examined for each cohort.

Using these court recorded offence data, two comparisons are made between the
offender cohorts. Firstly, the proportion of offenders in each group who have been
reconvicted for a drink-driving offence during a three-year follow-up period is

35
A finalised appearance is a group of one or more offences, against a single individual or a company,
which are finalised by the Local Court on a single day.

106
compared.36 This analysis provides a descriptive measure of the presence or absence
of a reduced percentage of recidivists after the introduction of the more severe
statutory penalties. A shortcoming of this analysis is that it assumes the two offender
cohorts do not differ in any systematic way. Systematic group differences can be
addressed, however, by conducting a logistic regression analysis. In this analysis the
offender is the unit of analysis and the outcome is a binary variable, coded for
whether or not the offender reappeared before the court during the follow-up period.
A dummy variable is included in the model to designate whether an offender was
sentenced prior to or after the legislative reforms took effect. Controlling for other
confounding influences, such as age, gender and area of residence, differences
between the offender groups can then be examined.

Secondly, a survival analysis is conducted to compare the two groups on their time to
first offence. Some individuals in the pre- and post-law cohorts will not be convicted
of a second offence during the follow-up period, giving rise to censored data, and the
follow-up periods for offenders in each group will also differ in length. These
characteristics of the data point to survival analysis as being the most appropriate
technique for the analysis. This approach provides an estimate of the likelihood of
reoffending at any given time during the follow-up period. For this study, if the
increased statutory penalties have the effect of reducing recidivism, then it should
take longer on average for the post-law cohort to be convicted of a new PCA offence
than the pre-law cohort.

In the survival analysis, Kaplan-Meier survival functions are generated for the two
offender groups. This procedure produces a survival curve showing the proportion of
each group ‘surviving’ beyond a given duration of time, that is, the proportion of
each group that has not reappeared for a new offence. The two survival curves can
then be directly compared, using nonparametric tests (e.g. log-rank tests), in order to
determine whether the two groups differ significantly in terms of the time taken, on
average, to reappear before the court. Alternatively, the two samples can be
compared using Cox Proportional Hazard models. This method examines the

36
Note that in the logistic regression analysis the outcome measures is any reconviction within three
years of the index offence. However in the survival analysis the outcome is any reconviction before
the cut-off date specified for the relevant cohort.

107
proportion of each group reoffending at various points or intervals of time following
the originating offence. More importantly, this analysis permits the use of covariates
to adjust for a priori differences between groups. Therefore, it will be possible to test
for significant differences in recidivism rates between the pre- and post-law groups
while controlling for other factors which may affect reoffending (e.g. age, gender and
area of residence).

Two issues that need to be considered for the survival analysis are: the effect of
imprisonment on time available to reoffend during the follow-up period and the
assumptions underlying the proportional hazards model. While offender ‘free’ time
during the follow-up period may be somewhat reduced by a custodial sentence, very
few drink-drivers in New South Wales receive an imprisonment term. This is
particularly true for offenders convicted of low-range and special-range PCA
offences, since there is no prescribed gaol term for these offences. However,
imprisonment is a possibility for offenders convicted of mid- and high-range PCA
offences. Therefore, survival analyses, excluding offenders who were sentenced to
imprisonment, are also conducted to confirm the results obtained from the larger
dataset. In any analysis it is also important to consider the assumptions
underpinning the estimated models. A major assumption of a Cox Proportional
Hazards Model analysis is that the “hazard (i.e. risk of failure) for any one
individual is proportional to the hazard for any other individual, where the
proportionality constant is independent of time” (Kleinbaum 1996, p. 133). This
assumption will be tested in this study by graphing the log-log survival plots for the
estimated model to ensure that the plots do not cross for one or more categories of
predictors (Kleinbaum 1996).

Perceptual research - Study 3

An essential factor to consider in any investigation of deterrent effects is people’s
perceptions of formal punishment. Ideally, this research would therefore have
measured knowledge about the increases in statutory penalties for drink-driving
offences contemporaneously with the implementation of the legislative changes, to
investigate whether the legislative reforms had the intended effect on severity
perceptions. The retrospective nature of this project prevented this analysis from

108
being undertaken. What could be measured, however, was the extent to which
knowledge of the sanctions currently applicable to drink-driving offences in New
South Wales predict the likelihood of committing this offence. For this reason a
scenario-based study (Study 3) was designed to examine whether people who have
more accurate knowledge of the current sanctions for drink-driving offences in New
South Wales are less likely to state that they would drink and drive under certain
conditions.

Study 3 also contributes to the broader deterrence literature by examining the

determinants of sanction perceptions, specifically, the influence of legal knowledge
on perceptions of sanction severity. As noted previously in Chapters 1 and 3,
perceptual studies examining deterrent effects have focused, almost exclusively, on
the relationship between risk perceptions and intentions to offend or risk perceptions
and actual offender behaviour. By restricting their focus in this way, these
investigations neglect to consider an assumption of the deterrence doctrine which is
essential for policy, that being the link between information known to an individual
and their judgment of sanction risks (Piquero & Pogarsky 2002). This gap in the
literature was highlighted by Nagin (1998) in his recent review of the deterrence
research, but has been noted by scholars over the last two decades or more as a
neglected area of research (e.g. Cook 1980). Miller and Iovanni (1994) also echo this
concern when they note “one must move beyond the focus on the effects of perceived
risk and must investigate the determinants of perceived risk themselves” (Miller &
Iovanni, p. 282). By examining the factors influencing sanction risk perceptions,
particularly the influence of legal knowledge on severity perceptions, this scenario-
based research aims to bridge the gap between policy and behaviour.

Hypotheses
As discussed in Chapter 3, the validity of cross-sectional and longitudinal surveys in
investigating hypotheses generated by deterrence theory has been questioned (Nagin
1998). For this reason a scenario-based design was employed to examine the
relationship between legal knowledge and offending likelihood. In this survey,
participants are presented with a vignette, describing a realistic drink-driving

109
scenario, and are then asked to indicate the likelihood they would offend under the
circumstances described.

The scenario-based survey undertaken for this thesis was designed to test four
specific hypotheses generated by deterrence theory:

1. Increased perceived risk of arrest reduces offending likelihood

One of the most robust findings from the deterrence literature is that the perceived
certainty of apprehension can affect the likelihood of offending (Hingson et al. 1987;
Homel 1988; Henstridge, Homel & Mackay 1997; Voas, Holder & Gruenewald
1997; Nagin & Pogarsky 2001). Several scenario-based studies have tested
punishment certainty by having respondents estimate, on a continuum, the likelihood
of apprehension and conviction for an offence. However, from a policy perspective it
is important to establish whether police practice can successfully manipulate
perceptions and subsequent behaviour. That is, can previous experience of RBT
patrols in an area influence a person’s current decision to drink and drive via
increased perceptions of apprehension risk. For this reason, the level of RBT activity
was varied across scenarios and its subsequent effect on stated offending likelihood
examined. Deterrence theory would anticipate that those allocated the scenario with a
high-level of police activity would be less likely to state that they would offend.

2. Enhanced knowledge of drink-driving penalties reduces offending likelihood

Given that we were unable to investigate whether people knew about the 1998 drink-
driving policy, one of the key objectives of the scenario-based study was to examine
knowledge of the current drink-driving penalties in New South Wales. One point to
note here is that penalties applicable to drink-driving can take two forms: statutory
penalties prescribed by the legislature and actual penalties imposed by the courts.
This distinction is important in a deterrence context because perceptions of
punishment severity are thought to depend not only on the magnitude of the
prescribed penalty but also the likelihood of being sanctioned if convicted (von
Hirsch et al. 1999). In this way, a person may know that severe statutory maximums
are in place for an offence yet still offend because they believe that they will receive
a much lighter penalty if caught. Thus, the survey tested knowledge of both statutory

110
and actual penalties for drink-driving offences. Utilising natural variation in levels of
knowledge we then explored whether those individuals who were better informed
about drink-driving penalties were less likely to report they would offend.

3. Increased perceptions of sanction severity reduces offending likelihood

Within a deterrence framework we would anticipate that knowledge of legal
penalties influences behaviour by raising perceptions of sanction severity and
subsequently, the anticipated costs associated with the act. Unless knowledge of
penalties alters perceptions of penalty severity then any relationship found between
knowledge and behaviour cannot be attributed to deterrence. This study therefore not
only tested respondents’ knowledge of the current penalties applicable to drink-
driving offences but also had them rate the severity of these penalties. Using a
measure employed in other perceptual research (Homel 1986; Nagin & Paternoster
1993), respondents were asked ‘how much of a problem’ the penalties would create
for them personally. This question was designed to capture the subjective nature of
severity perceptions, as they relate specifically to formal punishment prescribed by
the law (statutory penalties) and imposed by the criminal justice system (actual
penalties).

4. Perceptions of sanction severity matter more when punishment is perceived to be

certain
According to the deterrence model, a multiplicative relationship exists between the
certainty and the severity of punishment. Thus punishment severity matters more in
situations where offenders perceive it as likely they will be apprehended for the
offence. While making intuitive sense, empirical research exploring this aspect of
deterrence theory provides mixed evidence on the existence of such a relationship
(Homel 1986; Howe & Brandau 1988; Howe & Loftus 1996; Grosvenor, Toomey &
Wagenaar 1999). The survey formally tests the interaction between perceptions of
punishment certainty and severity by including an interaction term in the regression
models predicting stated offending likelihood.

111
Proposed model of the deterrence process
Figure 4.1 presents a proposed model of the deterrence process that provides a
framework for analysing the survey data. This model is loosely based on Homel’s
(1986) deterrence model describing the effect of RBT on future drink-driving
behaviour (see Homel 1986, p. 41). Additional pathways have been included in this
revised version to describe the relationship between knowledge of penalties for
drink-driving offences (statutory and actual penalties) and decisions to offend.
Underpinning this model is the basic assumption from subjective expected utility
theory that formal punishment acts to reduce an offending behaviour by increasing
the disutility (or costs) of the act. Accordingly, an individual will be deterred by the
extent to which the anticipated costs associated with the behaviour outweigh the
benefits. Expanding upon these basic concepts, the model also incorporates the
influence of anticipated informal sanctions, that arise from a drink-driving arrest, on
the decision to offend. Thus, in addition to the costs potentially imposed by the
criminal justice system, decisions to drink and drive will be affected by informal
sanctions from peers, family and the wider community when the offence is publicly
exposed.

Not all pathways described in the model are being investigated in Study 3 (the boxes
in bold denote the variables being measured in the survey) and some interaction
effects are implied. Despite these limitations, Figure 4.1 is a useful graphical
depiction of the deterrence hypotheses described above that are being investigated in
this study. Firstly, it proposes that individuals must have knowledge of the applicable
penalties for drink-driving offences before they can be deterred by the formal
punishment prescribed for this offence. While seemingly commonsensical, this
relationship between knowledge and behaviour is one that is generally assumed
rather than empirically tested in deterrence investigations. Secondly, for knowledge
of drink-driving penalties to influence behaviour it must increase an individual’s
perceptions of sanction severity. Unless a significant relationship is found between
knowledge and perceptions any relationship evident between knowledge and
behaviour may be spurious. Thirdly, exposure to police activity will influence an
individual’s perceptions of apprehension certainty, thereby lowering the likelihood of
drink-driving.

112
There are several other factors that could impact on both knowledge of drink-driving
penalties and perceptions of arrest certainty, which will also be considered in this
scenario-based research. Most importantly, consistent with previous research in the
deterrence domain, the model proposes that prior offending, alcohol consumption
and prior involvement in alcohol-related crashes could be determinants of the level
of legal knowledge. Marginal offenders, that is people who have previously driven
while drunk, who are at high-risk of acute alcohol-related harm or who have
previously experienced an alcohol-related road crash, will be more motivated to seek
out and retain relevant information about the formal costs associated with drink-
driving behaviours and therefore be more knowledgeable of drink-driving laws
(Kenkel & Koch 2001). People who refrain from offending on moral grounds, on the
other hand, are less likely to change their beliefs on the basis of this information and
will potentially be less knowledgeable in this area. For these individuals, the model
assumes that extralegal constraints, such as moral adherence to the law or feelings of
guilt, reduce the benefits gained from committing the act and subsequently reduce
the likelihood of offending.

113
Figure 4.1: Proposed model of the deterrence process

Prior offending/ Prior convictions/

alcohol Knowledge of Media publicity/
consumption/ prior statutory friends, family,
accident penalties acquaintances
involvement

Informal costs Perceptions of Perceived

arising from penalty severity applicability of
apprehension
statutory penalties

Total anticipated formal Total benefits

Decision to
costs associated with associated with
drink & drive
DD (disutility) DD (utility)

Perceived risk of Media publicity/ Moral inhibitions

apprehension friends, family, associated with
acquaintances drink-driving

Enforcement Prior successful drink-

activity driving episodes

114
Participants and methodology
Students attending the University of New South Wales participated in the survey,
which had received approval from the University of New South Wales Human
Research Ethics Committee. These students were recruited from several criminal
law,37 engineering and general studies38 classes to ensure that a broad cross-section
of students studying various disciplines was included in the study. Students were
considered a relevant target group since young people are at greater risk of injury
from alcohol-related road crashes (Chikritzhs et al. 2000) and of drinking alcohol at
acute-risk levels for alcohol-related harm (Chikritzhs et al. 2003). Furthermore, these
students are also likely to be more knowledgeable of driving laws applicable in New
South Wales, given that many would have obtained their full licence in recent years.
The survey was self-completed by the students after the researcher gave a brief
introduction and explained the questionnaire format (the participant information
sheet and drink-driving questionnaire can be found in Appendix II).

In the survey, knowledge of the drink-driving penalties currently in force in New

South Wales was tested through a series of multiple-choice questions asking about
the applicability of penalties and the statutory maximums and minimums that are
prescribed for drink-driving offences. Given that the vast proportion of drink-drivers
appearing before the Local Courts in New South Wales are persons convicted of a
low- or mid-range drink-driving offence (NSW Bureau of Crime Statistics and
Research 2002b), the questions asked in the survey pertained only to these two
offences. As an example, the question for a low-range drink-driving offence was as
follows:

37
Students in the criminal law classes were in their first year of study and therefore would not be
expected to be any more knowledgeable on drink-driving penalties than other types of students. Later
analyses confirmed that law students were, on average, no more knowledgeable about the penalties
than other participants.
38
All University of New South Wales students are required to complete four general studies courses
to fulfil the requirements of their degree. Most faculties offer these courses and students must attend a
course offered by a school other than their own. For this research, students from a general studies
course offered by the law and medical faculties were approached to participate in the survey.

115
 Here are some questions about what the law can and cannot do when a
person is convicted of drink-driving in NSW. Please answer all questions
in the order in which they appear by ticking the appropriate box.

If a person was caught driving with a blood alcohol level just over the
legal limit (i.e. just over 0.05g/100ml) and it was their first drink-driving
conviction….

Participants were then asked to indicate whether a person could be fined, have their
driver’s licence disqualified or be goaled and, if these penalties were applicable, they
were asked to indicate the magnitude of each sanction.

Participants who correctly indicated the exact penalty applicable to the drink-driving
offence described (i.e. either low- or mid-range drink-driving offences) received no
points. For those correctly nominating that a particular penalty could be applied to
the offence but did not know the magnitude of that penalty, a number of points based
on their underestimation/overestimation were allocated. Those who incorrectly
thought that a penalty was not applicable to the offence received the maximum
points. Points for each response were totalled and participants were given an overall
score reflecting the number of mistakes they made in answering these questions.39

In a similar way, participants were asked a series of multiple-choice questions on the

actual penalties they would receive were they caught and convicted for a low- or
mid-range drink-driving offence in New South Wales. By way of example, the
question pertaining to actual penalties for a low-range drink-driving offence was as
follows:

The previous questions asked about the minimum and maximum penalties,
allowed by NSW law, for drink-driving offences.

39
This scoring process replicated a methodology previously used by Kenkel and Koch (2001) in their
research on deterrence and knowledge of the law.

116
 Suppose you were caught with a blood alcohol limit just over the legal
limit (i.e. just over 0.05g/100ml). Give your best estimate as to what the
actual penalty (or penalties) would be if you were convicted of your first
drink-driving offence.

Again, participants indicated whether they would be fined, have their licence
disqualified or be gaoled and the magnitude of each of these penalties. In order to
assess the accuracy of these responses, the multiple-choice answers were compared
to the average penalties offenders received upon conviction for a low- or mid-range
drink-driving offence in NSW Local Courts, between 1999 and 2001.40 A composite
score was then assigned to reflect participant knowledge of the actual penalties
imposed for drink-driving offences in New South Wales.

Participants were then randomly allocated one of three scenarios, each varying in the
level of RBT activity described, but following the general format shown below:
Suppose you drove by yourself one evening to meet some friends at a pub.
You didn’t intend to drink much alcohol but by the end of the evening,
you’ve had enough drinks so you’re pretty sure your blood alcohol level is
above the legal limit. You haven’t seen police conducting Random Breath
Tests (RBT) in this area for some time. You live about 10kms away and
you have to be at work early the next morning. You can either drive home
or find some other way home, but if you leave the car at the pub, you will
have to return early the next morning to pick it up.41
The sentence italicised in the above scenario is the information that was varied across
the three certainty conditions. ‘You haven’t seen police conducting Random Breath
Tests (RBT) in this area for some time’ was the description given to participants
allocated the low-risk of apprehension scenario. Participants allocated the medium-
risk scenario were told ‘You have often seen police conducting Random Breath Tests
(RBT) in this area’ and participants allocated the high-risk scenario were told ‘You

40
These data were obtained from the Local Courts Database maintained by the NSW Bureau of Crime
Statistics and Research.
41
Nagin and Pogarsky (2001) previously used a similar scenario to examine the relationship between
punishment costs and intentions to offend amongst a sample of university students in the USA.

117
have often seen police conducting Random Breath Tests (RBT) in this area and since
it is a Friday night you know that there will be more police on the roads’.

After reading the scenario, participants were asked to estimate the likelihood, on a
scale of 0 (not at all) to 100 (definitely), that they would drive home under the
circumstances provided. To confirm that the three different scenarios did affect
perceptions of apprehension certainty, participants were also asked to estimate the
likelihood, on a scale of 0 (not at all) to 100 (definitely), of being caught and
convicted if they did drive home. Finally, to obtain subjective ratings of penalty
severity, participants were asked to indicate how big a problem, on a scale of 0 (no
problem) to 100 (a very big problem), the drink-driving penalties would create for
them. This latter measure is one that has been used in previous research to gauge
subjective estimates of punishment severity (Grasmick & Bryjak 1980; Homel 1986;
Nagin & Paternoster 1993).

Prior deterrence research has suggested an important role for extralegal factors, such
as “moral inhibitions”, “moral attachments”, “conscience” and “shame”, in offending
decisions. To account for these extralegal costs when modelling deterrence, some
researchers have included a moral component to the costs side of the expected utility
equation (e.g. Nagin & Paternoster 1994), such that the decision rule becomes –
offend if:
p.U(Benefits) > p.U(Sanctions) + U(Moral Regret)
However, Nagin and Pogarsky (2001) account for these same moral costs through the
utility function: that is, as moral inhibitions associated with offending increase,
U(Benefits) decrease. Thus, for persons who are so opposed to breaking the law that
they will not offend under any circumstances it can be assumed that the crime is of
no benefit to that individual (i.e. U(Benefits) = 0). Nagin and Pogarsky (2001) have
identified these individuals who are morally opposed to offending by asking survey
respondents in scenario-based research to report the likelihood of offending if there
was no possibility of being sanctioned in the given circumstances. Participants who
answer ‘0’ to this question as well as ‘0’ to the question on offending likelihood can
be assumed to be insensitive to changes in formal sanctions and can therefore be
excluded from the analysis. This methodology was also used in the current survey to

118
remove from the sample, those individuals for whom drink-driving would derive no
benefit.

Demographic information thought to influence offending decisions and/or

knowledge of the law was also collected in the survey. This information included
age, gender, university degree currently enrolled in, number of times respondent
drove a motor vehicle after drinking too much alcohol, number of convictions for a
drink-driving offence, individual and relatives involvement in alcohol-related crashes
and frequency of high-risk drinking in previous 12 months. These data were
collected to verify that groups receiving the three different scenarios, which contain
varying levels of apprehension risk, did not differ in any systematic way. It was also
necessary to include these measures as control variables in the regression analysis,
given that legal knowledge and severity perceptions were not experimentally
manipulated in the scenarios.

Before commencing data collection, the scenario-based survey was piloted with a
sample of third-year criminology students from the University of New South Wales
(n=26). This pilot was undertaken to ensure that the drink-driving scenario presented
in the survey was realistic and that it elicited a number of non-zero offending
probabilities. Perceptions of apprehension risk were also compared across the three
scenario conditions to confirm that the different levels of police activity described,
affected perceptions of punishment certainty as intended. As a result of this pilot
several minor changes were made to the design of the scenario. Most importantly, the
wording contained in the medium- and high-risk scenarios was changed from ‘you
remember seeing police conducting RBTs of drivers in this area recently’ and ‘you
remember seeing police conducting RBTs of drivers in this area recently and since it
is a Friday night you know that there will be more police on the roads’, respectively,
to ‘you have often seen police conducting Random Breath Tests (RBT) in this area’
and ‘you have often seen police conducting Random Breath Tests (RBT) in this area
and since it is a Friday night you know that there will be more police on the roads’.
These alterations were made to the scenario design after observing little difference in
risk perceptions between the low- and medium-risk scenarios.

119
CHAPTER 5. IMPACT OF THE 1998 PENALTY
CHANGES ON OFFENDING RATES IN NEW SOUTH
WALES

As discussed in Chapter 4, two aggregate-level studies were undertaken to examine

the impact of the penalty changes made to drink-driving offences in New South
Wales, in 1998.

Study 1 consists of interrupted time-series analyses of four road crash series: all
injury road crashes, fatal crashes, single-vehicle night-time crashes and multiple-
vehicle day-time crashes occurring in New South Wales between 1994 and 2001.
Fatal and single-vehicle night-time crashes are the two surrogate measures of
alcohol-related road crashes used in this study and multiple-vehicle day-time crashes
is the control series against which the alcohol-related series are compared. The
second study involves an analysis of the recidivism rates of two drink-driver cohorts:
one convicted before the penalty changes in 1998 and the other convicted after the
penalty changes. Reoffending is measured as any new court appearance for a drink-
driving offence recorded during the follow-up period and both the probability of
offending and the time to first new offence are compared across the two cohorts.

If more severe penalties have an effect on offending we would expect to see a

significant reduction in alcohol-related road crashes and drink-driving recidivism
rates after the implementation of the new penalty regime in 1998. This Chapter
presents the results of Study 1 and Study 2 and discusses their implications in light of
current knowledge about deterrence and the context in which the penalties were
implemented.

Study 1 - Road crash analysis

Overall, 164,573 motor vehicle crashes, in which a person was either injured or
killed, occurred on NSW roads between 1994 and 2001, with the vast majority (97.4

120
per cent) of these being non-fatal crashes. These 164,573 crashes resulted in a total
of 4,684 fatalities and 210,084 casualties.

All road crashes resulting in injury

The first crash series examined in this study consists of all road crashes recorded in
New South Wales from 1994 to 2001, that resulted in the death or injury of at least
one person. These crashes are presented in Figure 5.1 by the year in which they
occurred. As shown in this figure, injury crashes remained fairly stable in New South
Wales from 1994 to 1998, with the exception of a notable decrease in 1997.
However, from 1999 through 2001 there was an increase in the total number of
injury crashes recorded each year.

Figure 5.1: Total injury crashes in NSW, 1994-2001

25,000
22,704
21,852
19,488 19,366 19,388 19,886
19,306 18,378
20,000
Total number

15,000

10,000

5,000

0
1994 1995 1996 1997 1998 1999 2000 2001

To examine trends in injury crashes, controlling for road usage, the rate of monthly
injury crashes per 100,000 vehicles registered in New South Wales was calculated as
the dependent variable for the interrupted time-series analysis.42 The final regression

42
Another technique to control for the number of vehicle on the roads would be to use the amount of
petrol sold in NSW each month as the denominator for monthly crash rates. Automotive petrol sales
data were therefore sought from the Commonwealth Department of Industry, Sciences and Resources
for the period 1994 to 2001. Interrupted time-series analyses were then carried out for each of the four
crash series previously examined using monthly crashes per kilolitres of petrol sold as the dependent

121
model estimated for these data is shown in Table 5.1. Included in this model is a term
for the underlying trend in injury crashes, a term for the change in the underlying
trend from October 1998 and dummy variables to control for seasonal
autocorrelation in the data.43 The term modelling a change in the level of the injury
crash series was not significant (t=0.011, p=0.991) and was therefore excluded from
the final model.

The estimated model accounted for 71 per cent of the total variance in injury crash
data but an examination of the autocorrelation and partial autocorrelation function
plots revealed significant autocorrelation in the residuals at lag 3. Re-estimating the
regression model with the inclusion of a third order autocorrelation term (AR3),
resulted in trend terms that had near identical magnitudes and statistical significance
to those shown in Table 5.1. The Ordinary Least Squares estimates shown in this
table were therefore retained.

Table 5.1: Linear Regression Model for Injury Crashes per 100,000 vehicles
registered, New South Wales, 1994-2001
Variable Coefficient Standard t p value
error
Intercept 49.484 0.867 57.075 < 0.001
Underlying trend -0.090 0.014 -6.570 < 0.001
Change in underlying 0.261 0.031 8.355 < 0.001
trend from Oct. ‘98
Seasonal terms Jan. -2.424 1.013 -2.392 0.019
Feb. -0.691 1.013 -0.682 0.497
Mar. 5.632 1.013 5.562 < 0.001
Apr. 0.231 1.012 0.228 0.820
May 5.287 1.012 5.224 < 0.001
Jun. 1.512 1.012 1.494 0.139
Jul. 1.968 1.012 1.945 0.055
Aug. 3.331 1.012 3.293 0.002
Oct. 1.907 1.012 1.885 0.063
Nov. 3.532 1.012 3.491 0.001
Dec. 3.413 1.012 3.373 0.001
Notes: R2 = 0.707 & Durbin-Watson = 1.633,44 significant autocorrelation at lag 3 p = 0.005

variable. The results of these supplementary analyses were similar to those found using vehicle
registrations as the denominator.
43
Note that September is the referent month for the seasonal variables.
44
Durbin-Watson is a test statistic for the presence of first order serial correlation. If the value of the
test statistic is close to 2 then the model is considered a good fit to the data.

122
As seen in Table 5.1, the coefficient for the underlying trend is negative and
statistically significant (t=-6.57, p<0.001). This indicates that there was a general
decreasing trend in injury crash rates over the study period of 0.09 crashes, per
100,000 vehicles registered, per month. The term for the change in trend, post
intervention, was also significant, indicating that there was a significant change in
this downward trend in injury crashes from October 1998 onwards (t=8.36, p<0.001).
The magnitude of this change can be calculated by adding together the coefficients
for the initial decreasing trend and the increasing trend post intervention. Thus, after
the legal policy had been implemented, it was estimated that injury crashes per
100,000 vehicles registered increased in New South Wales at a rate of 0.15 per
month.

The significant increase in injury crashes after the intervention apparent from the
results of the interrupted time-series analysis, is displayed graphically in Figure 5.2.
This figure presents the injury crash rate per month from 1994 through 2001, with a
regression line showing how the underlying trend in crash rates changed after the
new drink-driving penalties were introduced in September 1998. To construct this
graph, the seasonal variables were removed from the regression model and the
underlying trend and change in the underlying trend were estimated. By plotting the
observed crash rates against the estimated values from this model, a clearer picture of
the trend in injury crashes over the study period emerges.

123
 Figure 5.2: Monthly injury crashes per 100,000 vehicles
registered incorporating a fitted trend line, NSW, 1994-
2001

60
Penalties increase
Rate per 100,000 vehicles

55

50

45
Injury rate
40 Trend line
35

30
Ju 4

Ju 5

Ju 6

Ju 7

Ju 8

Ju 9

Ju 0

Ju 1
Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja
l-9

l-9

l-9

l-9

l-9

l-9

l-0

l-0
n-

n-

n-

n-

n-

n-

n-

n-
9

0
4

1
Alcohol-related road crash rates
As noted previously, two surrogate measures of alcohol-related crashes, all fatal
crashes and single-vehicle night-time crashes, were employed in Study 1 to better
estimate the impact of the penalty increase on drink-driving offending. Figure 5.3
presents the raw number of each of these crash types recorded in New South Wales
each year, from 1994 through to 2001. As seen here, the yearly number of fatal
crashes remained relatively stable over this eight-year period, ranging from a high of
563 fatal crashes in 1995 to a low of 486 fatal crashes in 2001. At the same time, the
number of single-vehicle night-time crashes appeared to be steadily declining from
1994 through 1997. However, after a low of 1,427 crashes in 1997, single-vehicle
night-time crashes began to increase in the latter half of the study period, with
crashes peaking at 1,643 in 2001.

124
 Figure 5.3: Monthly fatal and monthly single-vehicle
night-time crashes, NSW, 1994-2001

2,000
1,800 1637 1643
1603 1575 1594
1505 1535
1,600 1427
1,400 SVNT
Total number

1,200 Fatal
1,000
800
553 563 538 525 543
600 491 506 486
400
200
0
1994 1995 1996 1997 1998 1999 2000 2001

Table 5.2 and Table 5.3 display the estimated regression models for fatal crashes per
100,000 vehicles registered and single-vehicle night-time crashes per 100,000
vehicles registered, respectively. Again, the only intervention variable included in the
final models is a term for the change in the underlying trend from October 1998
onwards, given that there was no significant change in the level of these series post
intervention (for fatal crashes; t=-0.12, p=0.91 and for single-vehicle night-time
crashes; t=-0.15, p=0.88). Both the regression model for fatal crash rates and the
regression model for single-vehicle night-time crash rates contained no
autocorrelation (as assessed by the Durbin-Watson statistic and the autocorrelation
and partial autocorrelation plots) and these models accounted for 37 per cent and 49
per cent of the total variance in crash rates, respectively. Taken together, this
suggests that the estimated regression models for both alcohol-related series could be
considered a good fit to the data.

125
Table 5.2: Linear Regression Model for Fatal Crashes per 100,000 vehicles
registered, New South Wales, 1994-2001
Variable Coefficient Standard t p value
error
Intercept 1.285 0.091 14.155 < 0.001
Underlying trend -0.004 0.001 -3.086 0.003
Change in underlying 0.002 0.003 0.629 0.531
trend from Oct. ‘98
Seasonal terms Jan. 0.044 0.106 0.418 0.677
Feb. 0.063 0.106 0.594 0.554
Mar. 0.310 0.106 2.927 0.004
Apr. 0.155 0.106 1.466 0.147
May 0.362 0.106 3.411 0.001
Jun. 0.115 0.106 1.087 0.280
Jul. 0.163 0.106 1.535 0.129
Aug. 0.080 0.106 0.756 0.452
Oct. 0.265 0.106 2.498 0.015
Nov. 0.246 0.106 2.317 0.023
Dec. 0.301 0.106 2.844 0.006
Notes: R2 = 0.373 & Durbin-Watson = 2.319; no autocorrelation or partial autocorrelation in
residuals

Table 5.3: Linear Regression Model for Single-Vehicle Night-Time Crashes per
100,000 vehicles registered, New South Wales, 1994-2001
Variable Coefficient Standard t p value
error
Intercept 4.017 0.137 29.424 < 0.001
Underlying trend -0.009 0.002 -3.950 < 0.001
Change in underlying 0.012 0.005 2.407 0.018
trend from Oct. ‘98
Seasonal terms Jan. -0.183 0.160 -1.146 0.255
Feb. -0.167 0.159 -1.050 0.297
Mar. 0.145 0.159 0.912 0.365
Apr. -0.064 0.159 -0.403 0.688
May -0.083 0.159 -0.519 0.605
Jun. -0.305 0.159 -1.913 0.059
Jul. 0.013 0.159 0.082 0.935
Aug. -0.043 0.159 -0.269 0.789
Oct. 0.261 0.159 1.641 0.105
Nov. 0.318 0.159 1.995 0.049
Dec. 0.698 0.159 4.380 < 0.001
Notes: R2 = 0.489 & Durbin-Watson = 2.034; no autocorrelation or partial autocorrelation in
residuals

The model for fatal crash rates shown in Table 5.2 reveals that there was a downward
trend in this crash series in New South Wales over the period 1994 through 2001.
The magnitude of this decreasing trend was in the order of 0.004 fatal crashes, per

126
100,000 vehicles registered, each month (t=-3.09, p=0.003). This significant
downward trend was also observed for all injury crashes over the same time period.
However, for fatal crash rates, the term modelling the change in this underlying
trend, post intervention, was not statistically significant (t=0.63, p=0.53). This
suggests that the implementation of the new laws in September 1998 had no impact
on the general downward trend in fatal crash rates in New South Wales between
1994 and 2001.

The data series representing single-vehicle night-time crash rates also showed a
significant downward trend, prior to the intervention, of 0.01 crashes, per 100,000
vehicles registered, (t=-3.95, p<0.001) per month. However, unlike fatal crash rates,
there was evidence from the regression model that there was a significant change in
this underlying trend from October 1998 onwards (t=2.41, p<0.05). The positive
coefficient for the term modelling the change in the underlying trend indicates that,
after the penalties for drink-driving offences were increased in September 1998,
single-vehicle night-time crashes began to rise. The magnitude of this change was in
the order of 0.003 additional crashes, per 100,000 vehicles registered, per month.
Given that single-vehicle night-time crashes contain a greater proportion of alcohol-
related incidents, this significant increase in crash rates raises some doubts about the
hypothesised deterrent effect of the 1998 sanctioning policy.

Figure 5.4 and Figure 5.5 display the monthly rates for the fatal and single-vehicle
night-time crash series, as well as regression lines showing the underlying trend and
the change in trend from October 1998. Again, these graphs were constructed by
plotting the actual monthly crash rates against the predicted monthly crash rates,
where the predicted values are estimated from the regression model with seasonal
variables removed. These figures illustrate the decline in both alcohol-related crash
series over the initial part of the study period and the significant increase in single-
vehicle night-time crash rates after the implementation of the new penalties, an
increase which is not apparent in the fatal crash series.

127
 Figure 5.4: Monthly fatal crashes per 100,000 vehicles
registered incorporating a fitted trend line, NSW, 1994-
2001

2.5
Rate per 100,000 vehicles

Fatal rate
2
Penalties increase Trend line

1.5

0.5

0
Ju 4

Ju 5

Ju 6

Ju 7

Ju 8

Ju 9

Ju 0

Ju 1
Ja

Ja 4

Ja 5

Ja 6

Ja 7

Ja 8

Ja 9

Ja 0
l-9

l-9

l-9

l-9

l-9

l-9

l-0

l-0
n-

n-

n-

n-

n-

n-

n-

n-
9

1
Figure 5.5: Monthly single-vehicle night-time accidents
per 100,000 vehicles incorporating a fitted trend line,
NSW, 1994-2001

5.5
SVNT rate
Rate per 100,000 vehicles

5
Trend line
4.5

3.5

3
Penalties increase
2.5

2
Ju 4

Ju 5

Ju 6

Ju 7

Ju 8

Ju 9

Ju 0

Ju 1
Ja

Ja 4

Ja 5

Ja 6

Ja 7

Ja 8

Ja 9

Ja 0
l-9

l-9

l-9

l-9

l-9

l-9

l-0

l-0
n-

n-

n-

n-

n-

n-

n-

n-
9

128
Control series
The control series in Study 1 was the number of multiple-vehicle crashes occurring
between 9am and 3pm (of which approximately one per cent are alcohol-related).
Figure 5.6 presents the raw number of these types of crashes occurring in New South
Wales each year between 1994 and 2001. Following a similar pattern to all injury
crashes, the yearly number of multiple-vehicle day-time crashes appeared to be
relatively stable in New South Wales between 1994 and 1998. However, in the latter
half of the study period these crashes began to increase in number, particularly in the
final years of the eight-year period (i.e. 2000 and 2001).

Figure 5.6 Multiple-vehicle day-time crashes in NSW,

1994-2001

7,000
6,077
6,000 5,731

5,004 5,030 5,098 5,051 5,186

4,813
5,000
Total number

4,000

3,000

2,000

1,000

0
1994 1995 1996 1997 1998 1999 2000 2001

The estimated regression model for the rate of multiple-vehicle day-time crashes per
100,000 vehicles registered is shown in Table 5.4. Included in this model are terms
for the underlying trend in this crash series, the change in this trend after the
penalties were raised in September 1998 and 11 dummy variables to control for
seasonal autocorrelation in the data. Again, the regression model revealed that the
intervention had no statistically significantly impact on the level of the series (t=-
0.03, p=0.98) and, therefore, this variable was excluded from the final model. The
model shown in Table 5.4 accounts for 58 per cent of the variance in the control
series and from an examination of the autocorrelation and partial autocorrelation

129
plots, as well as the Durbin-Watson test statistic, the model contained no significant
autocorrelation.

The significant, negative coefficient for the term modelling the underlying trend in
these data indicates that, over the eight-year period, there was a decreasing trend in
multiple-vehicle day-time crashes of 0.02, per 100,000 vehicles registered, per month
(t=-4.14, p<0.001). From October 1998 onwards, however, there was a statistically
significant change in this downward trend, with multiple-vehicle day-time crashes
beginning to increase at a rate of 0.05 crashes, per 100,000 vehicles registered, per
month (or a 4.5 per cent change from October 1998).

Table 5.4: Linear Regression Model for Multiple-Vehicle Day-Time Crashes per
100,000 vehicles registered, New South Wales, 1994-2001
Variable Coefficient Standard t p value
error
Intercept 12.515 0.312 40.122 < 0.001
Underlying trend -0.020 0.005 -4.141 < 0.001
Change in underlying 0.073 0.011 6.469 < 0.001
trend from Oct. ‘98
Seasonal terms Jan. -0.112 0.365 -0.306 0.760
Feb. -0.946 0.364 -2.596 0.011
Mar. 0.972 0.364 2.668 0.009
Apr. 0.212 0.364 0.581 0.563
May 0.964 0.364 2.646 0.010
Jun. 0.237 0.364 0.650 0.517
Jul. 1.067 0.364 2.932 0.004
Aug. 1.084 0.364 2.979 0.004
Oct. 0.529 0.364 1.454 0.150
Nov. 0.344 0.364 0.946 0.347
Dec. 0.854 0.364 2.344 0.022
Notes: R2 = 0.576 & Durbin-Watson = 1.775; no autocorrelation or partial autocorrelation in
residuals

The significant increase in the control series post intervention is displayed

graphically in Figure 5.7. This figure presents the monthly rates of multiple-vehicle
day-time crashes and a fitted trend line derived from an estimated regression model
with the seasonal terms removed. As with the trends observed for the two other non-
fatal crash series, rates of multiple-vehicle day-time crashes were decreasing up until
the implementation of the intervention, at which point these crash rates reversed
direction and began to rise.

130
 Figure 5.7: Monthly multiple-vehicle day-time crash
rates per 100,000 vehicles registered incorporating a
fitted trend line, NSW, 1994-2001
MVDT rate
Trend line
16
Rate per 100,000 vehicles

Penalties increase
15
14
13
12
11
10
9
8
Ju 4

Ju 5

Ju 6

Ju 7

Ju 8

Ju 9

Ju 0

Ju 1
Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja
l-9

l-9

l-9

l-9

l-9

l-9

l-0

l-0
n-

n-

n-

n-

n-

n-

n-

n-
9

0
4

1
Summary
To summarise so far, the interrupted time-series analyses found statistically
significant increases in injury crash rates, single-vehicle-night-time crash rates and
multiple-vehicle day-time crash rates after the more severe penalties for drink-
driving offences were introduced in New South Wales in 1998.45 The only crash
series in which there was no significant increase during the latter half of the study
period was monthly fatal crashes.

What is of most interest for our test of deterrence is whether there was any change in
alcohol-related crashes after the legislative amendments were enacted in New South
Wales. To determine this, two surrogate measures of alcohol-related crashes, single-

45
The magnitude of these effects should warrant some caution when interpreting the public health
significance of these findings. For single-vehicle night-time crashes, the observed increase in crashes
after the policy was implemented was 0.012 crashes each month for every 100,000 vehicles registered
in New South Wales. With at least three and a half million motor vehicles registered in New South
Wales during this time, this increase equates to five additional single-vehicle night-time crashes
occurring each year. In a State of six and a half million persons this rise in crash rates, although
statistically significant, may seem relatively small. However, it is arguable that any increase in the risk
of injury or death can be considered an important finding.

131
vehicle night-time crashes and fatal crashes, were examined. The differing trends in
these two series during the latter half of the study period were therefore unexpected.

Here it should be noted however that fatal crashes in New South Wales have been
decreasing at a fairly constant rate over the last two decades (Australian Transport
Safety Bureau 2001). One of the reasons for this reduction is certainly the
introduction of RBT (Henstridge, Homel & Mackay 1997) but this phenomena of
falling road mortality rates is not exclusive to New South Wales, nor indeed to
Australia, but has been observed in many Western countries during the second half
of this century, despite a concurrent increase in the number of vehicles on the road
(Australian Transport Safety Bureau 2001). It is possible that this general downward
trend in road fatalities is not due to fewer people being involved in serious, and
potentially fatal, road crashes but due to other factors, such as enhancements in
automotive design, road engineering and trauma care (Nathens et al. 2000),
improving the chances of a crash victim surviving. If the lethality of road crashes has
reduced in more recent times, then single-vehicle night-time road crashes resulting in
an injury might be the best indicator of changes in alcohol-related road crashes,
particularly when measuring changes over the short term. It is also the crash category
that contains a greater proportion of drivers with an illegal BAC and therefore should
be more sensitive to changes in drink-driving levels.

The increased drink-driving penalties introduced in 1998 were therefore expected to

have the biggest impact (in the form of a significant reduction) on single-vehicle
night-time crashes. The control series, multiple-vehicle day-time crashes, was
expected to be the least affected because fewer of these crashes involve alcohol. The
findings from the interrupted time-series analyses described above, which suggest
that there was a significant increase in both single-vehicle night-time and multiple-
vehicle day-time crash rates after the 1998 penalty changes, are therefore, prima
facie, inconsistent with deterrence theory.

When interpreting these results however it needs to be recognised that the

significance test in an interrupted time-series analysis is not a test of ‘cause’ (Cook &
Campbell 1979). While the above analyses did find a statistically significant change

132
in these two non-fatal crash series, at the specified time point, the analysis does not
offer any explanations for the change. This means that any other number of factors,
apart from the sentencing policy, may have led to the observed increase in crashes
post intervention. In fact sensitivity analyses, in which the intervention point was
placed six months prior to the penalty increase being implemented, resulted in the
same conclusions. Thus, rather than the sentencing policy actually increasing road
crashes, it is more likely that the intervention occurred at, or around, the time that
crash rates were already on the rise in New South Wales.

Alcohol-related crashes versus control series

Further support for the conclusion that the drink-driving penalty changes did not
contribute to the rise in road crashes is apparent when we compare the post-
intervention increase in single-vehicle night-time crashes with the increase in
multiple-vehicle day-time crashes. If the penalty changes had a negative effect on
alcohol-related crashes then we would expect to see a greater increase in our
surrogate measure of alcohol-related road crashes after the changes were enacted
than in the control series. However, Figures 6.1 and 6.2 suggest that multiple-vehicle
day-time road crashes (the control series) may have been increasing at a greater rate
after the intervention point than single-vehicle night-time road crashes. To more
formally test this possibility, the following regression model was estimated:

Y = b0 + b1GROUP + b2TREND + b3TRENDCH + b4TREND2 + b5TRENDCH2

where Y is the crash rates for both multiple-vehicle day-time crashes and single-
vehicle night-time crashes; GROUP is a binary variable representing the two crash
categories (coded 0 for multiple-vehicle day-time rates and 1 for single-vehicle
night-time crash rates); TREND is a linear variable to model the underlying trend in
the series, TRENDCH is a binary variable to model the change in the underlying
trend after October 1998 and TREND2 and TRENDCH2 are interaction terms for
GROUP*TREND and GROUP*TRENDCH. Dummy variables corresponding to the
months of the year were also included in the final model to deseasonalise the data.

133
The variable of interest in this analysis is the last term in the regression equation, that
is the interaction between group membership and the change in the underlying trend
post intervention (GROUP*TRENDCH). This term essentially compares the rate of
increase after the intervention across the two crash categories (single-vehicle night-
time crash rates and multiple-vehicle day-time crash rates), taking into account other
trends or variability in the data series. The regression approach adopted here was
considered appropriate because there was no serial autocorrelation in either the
single-vehicle night-time or the multiple-vehicle day-time crash series and it would
be reasonable to assume that these types of crashes can be considered two separate
observations at a single point in time.46

The final regression model, which is shown in Table 5.5, reveals that the interaction
term for group by trend change (GROUP*TRENDCH) is statistically significant (t=-
4.76, p<0.001). This suggests that the change in the underlying trend from October
1998 (i.e. the post-slope of the series) was significantly different for the two crash
categories. The estimated post-slope for multiple-vehicle day-time crashes from the
regression equation is 0.054 and the post-slope for single-vehicle night-time crash
rates is 0.004. This significant finding provides some evidence that alcohol-related
crashes were increasing at a slower rate after the 1998 sentencing policy was
introduced than were non alcohol-related crashes.

46
While collinearity diagnostics suggested that there was some multicollinearity in the model, both
the condition index and variance proportion estimates were within an acceptable range (Kleinbaum,
Kupper & Muller 1988). Comparisons of the predicted rates for each crash series with the observed
rates also confirmed the validity of the estimated model.

134
Table 5.5: Linear Regression Model for Multiple-Vehicle Day-Time and Single-
Vehicle Night-Time Crash rates, NSW, 1994-2001
Variable Coefficient Standard t p value
error
Intercept 12.854 0.217 59.176 < 0.001
GROUP -8.976 0.222 -1.002 < 0.001
Underlying trend (TREND) -0.024 0.004 -5.681 < 0.001
Change in underlying trend 0.078 0.010 8.160 < 0.001
from Oct. ’98 (TRENDCH)
Group*Underlying trend 0.014 0.006 2.365 0.019
(TREND2)
Group*Change in -0.064 0.013 -4.759 < 0.001
underlying trend from Oct.
’98 (TRENDCH2)
Seasonal terms Jan. -0.144 0.218 -0.660 0.510
Feb. -0.557 0.218 -2.554 0.012
Mar. 0.565 0.218 2.590 0.010
Apr. 0.078 0.218 0.352 0.725
May 0.448 0.218 2.057 0.041
Jun. 0.026 0.218 -0.119 0.906
Jul. 0.546 0.218 2.505 0.013
Aug. 0.523 0.218 2.401 0.017
Oct. 0.394 0.218 1.809 0.072
Nov. 0.326 0.218 1.497 0.136
Dec. 0.763 0.218 3.503 0.001

Preliminary conclusions
To assess changes in drink-driving levels after the sentencing policy was
implemented in New South Wales, single-vehicle crashes occurring at night were
selected as one of the surrogate measures for alcohol-related road crashes. This series
was chosen because a greater proportion of these crashes involve drivers with illegal
BACs. However, it needs to be remembered that over half of the crashes included in
this category are unrelated to alcohol. Thus, if some other unmeasured factor was
driving an increase in non alcohol-related crashes then it would be reasonable to
expect some residual effect to also be evident in this surrogate measure. The
comparison between the control and alcohol-related crash series described above
provides some support for this assertion.

If the observed rise in single-vehicle night-time crashes was driven mostly by the
increase in non alcohol-related crashes, then two possible conclusions regarding the
policy’s effect on alcohol-related crashes can be drawn from the findings presented

135
above: (1) that the increase in statutory penalties had a beneficial impact on alcohol-
related road crashes but the effect was swamped by the secular rise in non alcohol-
related crashes occurring at the same time or (2) that the increase in statutory
penalties for drink-driving had no beneficial impact on alcohol-related road crashes.

Given that single-vehicle night-time crashes were found to be increasing at a slower

rate post-intervention than multiple-vehicle day-time crashes, it is possible that there
was a significant reduction in alcohol-related road crashes after the intervention point
but that this decrease was overwhelmed by the rise in non alcohol-related crashes
occurring at the same time. This would suggest that the sentencing policy introduced
in New South Wales in 1998 had a deterrent effect on drink-driving but that our
surrogate measure of alcohol-related crashes was not sensitive enough to detect the
change. However, there may be other factors that could potentially account for the
differential rise in these two crash series. One possibility is that there was an increase
in traffic density in the latter half of the study period, which resulted in more road
crashes. Because traffic is heaviest during the day, this change would have a greater
effect on multiple-vehicle day-time crashes than on single-vehicle night-time
crashes. Thus, it also remains possible that the sentencing policy failed to have any
impact on alcohol-related road crashes and that the differential rise in alcohol and
non-alcohol crash rates found in this study was simply due to other factors.

Other potentially confounding factors such as traffic density that could account for
the differential rate of increase in the two crash series are difficult to measure at the
aggregate level. Nevertheless, the second quasi-experimental study undertaken for
this thesis will be able to shed further light on the findings from these time-series
analyses. Recalling from Chapter 4, Study 2 compares the recidivism rates of drink-
drivers convicted before the penalty changes with the recidivism rates of drink-
drivers convicted after the rise in penalties. If a significant difference in reoffending
is found in this second study, and it is in the direction that deterrence theory would
predict, then it remains possible that alcohol-related road crashes decreased after the
penalty changes but that this reduction was swamped by the rise in non alcohol-
related crashes. If, on the other hand, there are no significant differences found
between these two offender groups, or if the differences are not in the expected

136
direction, then we can rule this out as a potential explanation for the results from
Study 1. An examination of the impact of the penalty increase on drink-driving
reoffending rates is therefore necessary before we can make confident conclusions
regarding the deterrent effect of the 1998 sentencing policy.

Study 2 - Analysis of recidivism rates

As discussed in Chapter 4, before changes in recidivism can be considered it is
important to establish that the 1998 statutory amendments were successfully
translated into sentencing practice. If these legal changes failed to affect the severity
of penalties actually imposed by the courts, any reduction in recidivism cannot be
attributed to the specific deterrent effect of the new laws. In addition, it is also
important to consider whether the legislation had any negative impact on the
prosecution of drink-driving offences. Circumvention of the more severe penalties by
players within the judicial system could reduce the perceived certainty of punishment
and subsequently diminish any impact the sentencing policy may hope to have in
deterring offenders. The following section examines both these issues by comparing
all drink-driving offences prosecuted in the NSW Local Courts in 1997 with those
prosecuted in 1999. These data were extracted from the Local Courts Database
maintained by the NSW Bureau of Crime Statistics and Research (BOCSAR).

Implementation of statutory amendments

There are essentially five ways in which the intended deterrent effects of the new
sentencing laws for drink-drivers could have been undermined: (1) a reduction in
drink-driving charges brought before the courts, (2) an increase in court delay, (3) a
reduction in guilty pleas, (4) a reduction in proven offences or (5) no subsequent
change to the severity of drink-driving penalties imposed by the courts (MacCoun
1993; Nagin 1998).

As shown in Table 5.6, the number of drink-driving prosecutions and the proportion
of these that were proven in court increased after the penalties were raised. On the
other hand, there was also no significant change in the average time from the offence
being committed to the matter being finalised in the Local Court. There was a
significant decrease from 1997 to 1999 in the proportion of offenders who pleaded

137
guilty for a drink-driving offence, however this decrease was mostly due to more
offenders not turning up before the court to submit a plea rather than an increase in
the proportion of offenders pleading not guilty.47 Taken together, these data provide
no evidence to suggest that the new sentencing laws were undermined by factors
associated with the prosecution of drink-driving offences in NSW courts.

Table 5.6: Drink-driving offences prosecuted in NSW Local Courts, 1997 & 1999
1997 1999 % change
Total prosecutions (n) 19,237 23,285 +21.0
Average time to finalisation (days) 72.9 71.5 -1.9
Offences with a guilty plea (%) 91.2 88.1 -3.4
Offences proven (%) 97.9 98.2 +0.3

Importantly, the new drink-driving laws did increase the severity of penalties that
drink-driving offenders received upon conviction. As seen from Table 5.7, average
gaol terms, fines and licence disqualification periods imposed by the courts for drink-
driving offences increased after the implementation of the new legislation in 1998.
Statistical tests showed these increases in the magnitude of penalties to be
significant.48

Table 5.7: Penalties imposed for drink-driving offences in NSW Local Courts, 1997
& 1999
Penalty type 1997 1999 % change
Gaol % gaoled 1.7 1.8 +5.9
Average (mths) 4.3 5.1 +18.6
Fine % fined 73.9 71.6 -3.1
Average ($) 513.2 752.5 +46.6
Licence disqualification % disqualified 82.2 80.0 -2.7
Average (mths) 12.3 14.3 +16.3

However, the magnitude of this rise in penalties was not uniform across the State
(see Table 5.8). In Sydney metropolitan courts the average gaol term rose by 28 per
cent, the average fine by 51 per cent and the average licence disqualification period
by nine per cent. For non-Sydney courts, the average gaol term did not change
significantly from 1997 to 1999 but the average fine increased by 40 per cent and the

47
Offences proven (x2=5.57, df=2, p=0.02); time to finalisation (t=0.72, df=42,519, p=0.47); offences
with a guilty plea (x2=123.74, df=2, p=0.001).
48
Gaol (t=-3.43, df=737, p<0.001); Fine (t=-57.04, df=30,279, p<0.001); Licence Disqualification
(t=-13.36, df=33,736, p<0.001).

138
average licence disqualification period by 24 per cent.49 This differential rise in the
severity of drink-driving penalties across the State is an important outcome to
consider in the recidivism analysis. The harsher penalties imposed on non-Sydney
offenders could affect their perceptions of the costs associated with drink-driving and
therefore have a greater impact on their offending behaviour. Furthermore, with less
access to alternative forms of transport, people living outside of major cities are
heavily reliant on motor vehicles and thus may be more likely to perceive licence
disqualification as a relatively severe sanction. Given this, a greater reduction in
reoffending rates for offenders residing in country and regional areas would be
anticipated.

Table 5.8: Penalties imposed for drink-driving offences in NSW Local Courts broken
down by court location, 1997 and 1999
Sydney courts non-Sydney courts
Penalty type 1997 1999 % 1997 1999 %
change change
Gaol % gaoled 1.2 1.3 +8.3 2.2 2.4 +9.1
Average 4.6 5.9 +28.3 4.2 4.6 +9.5
(mths)
Fine % fined 76.5 75.5 -1.3 71.6 67.1 -6.3
Average ($) 520.5 788.0 +51.4 506. 707. +39.7
1 0
Licence % 85.0 83.0 -2.4 79.6 76.5 -3.9
disqualification disqualified
Average 12.8 14.0 +9.4 11.8 14.6 +23.7
(mths)

The only trend toward greater leniency in sentencing was a statistically significant
but small decrease in the proportion of offenders who received a licence
disqualification and in the proportion of offenders who were fined in 1999 (see Table
5.7).50 This decrease in the proportion of drink-drivers disqualified was due to more
drink-drivers having their conviction dismissed under Section 10 of the NSW Crimes
(Sentencing Procedures) Act51 in 1999 than in 1997. This increase in Section 10

49
Variations across localities in the average licence disqualification were still apparent when
differences in the type of drink-driving offences prosecuted in these courts (i.e. high-, mid-, low- or
special-range PCA) were taken into account.
50 2
x =8.00, df=1, p=0.005
51
Section 10 of the NSW Crimes (Sentencing Procedures) Act 1999 permits a court, in certain
circumstances, to direct that a relevant charge be dismissed for a guilty offender or that the guilty
offender be conditionally discharged. In the case of drink-driving offences this means that the licence

139
dismissals was most apparent for offences that had minimum licence disqualification
periods introduced in 1998, that is low- and special-range drink-driving offences (see
Figure 5.8). In contrast there was little change between 1997 and 1999 in the
imposition of Section 10s for those convicted of a mid- or high-range offence,
offences that already had prescribed minimum licence disqualifications prior to the
current legislative amendments.52 These data lend some support to the claim that the
removal of judicial discretion through the introduction of statutory minima resulted
in magistrates circumventing the severe penalties through the imposition of
alternative sanctions.

Figure 5.8: Percentage of proven drink-driving

offences dismissed without a conviction (s10), by
offence type and year

35.0 1997
30.0 28.1 1999
24.2
25.0
percentage

20.0 18.4 17.4

15.1
15.0
9.5
10.0 6.5 7.0
5.0

0.0
Special Low Mid High
Drink-driving offence type

In summary, the 1998 legislation resulted in a significant increase in the average

penalties imposed for drink-driving offences without having any negative impact on
the prosecution of drink-driving offences. The only setback in terms of the deterrence
aims of the legislation was a statistically significant, though small, decrease in the
proportion of offenders being disqualified from driving. However, four out of every
five drink-driving offenders convicted in 1999 did receive a licence disqualification,

disqualification provisions under the NSW Road and Transport (General) Act also are not applied.
This section of the NSW Crimes (Sentencing Procedures) Act 1999 is equivalent to Section 556A of
the NSW Crimes Act 1900 which was repealed in 1999.
52
Special-range PCA (x2=8.36, df=1, p=0.004); Low-range PCA (x2=13.65, df=1, p<0.001); Mid-
range PCA (x2=3.45, df=1, p=0.063); High-range PCA (x2=1.04, df=1, p=0.307).

140
which, on average, was longer in duration than licence disqualifications given to
their 1997 counterparts. Since evidence suggests that longer periods of licence
disqualification can have a specific deterrent effect on drink-driving offending
(Blomberg, Preusser & Ulmer 1987; Mann et al. 1991; Siskind 1996), the 1998
legislative amendments still had the potential to reduce recidivism rates in New
South Wales.

Characteristics of offender cohorts

To compare recidivism rates before and after the penalty changes, data for two
offender cohorts were extracted from BOCSAR’s Reoffending Database: one
consisting of individuals convicted of drink-driving in 1997 (pre-law cohort) and the
other consisting individuals convicted of drink-driving in 1999 (post-law cohort). A
total of 18,017 persons were included in the pre-law cohort and 20,759 persons were
included in the post-law cohort. All these offenders had been convicted of at least
one drink-driving offence in 1997 or 1999, respectively. Another 697 offenders were
excluded from the post-law cohort since they had previously been convicted of a
drink-driving offence in 1997.

Table 5.9 presents descriptive data on the characteristics of offenders included in

each cohort. These data show that the 1997 offender group, that is those convicted
under the old penalty system, were more likely to be male, Aboriginal,53 reside
outside of the Sydney metropolitan area, be convicted of a high-range PCA offence,
have a previous conviction for a drink-driving offence, as well as previous
convictions for other offences, and have concurrent offences finalised at the time of
conviction, in comparison to their 1999 counterparts. These differences in offender
characteristics across the two cohorts underline the need to take into account pre-
existing differences between the groups when examining reoffending rates.

53
Care should be taken when comparing Aboriginality across cohorts since a greater proportion of
offenders convicted in 1997 had missing information on this variable.

141
Table 5.9: Comparison of offender characteristics for drink-driving offenders
included in the 1997 and the 1999 cohorts
Offender characteristics 1997 1999 p-value54
n % n %
Gender
Male 15,521 86.1 17,661 85.1 0.003*
Female 2,496 13.9 3,098 14.9
Aboriginality
Non-Aboriginal 13,632 75.7 18,391 88.6 <0.001*
Aboriginal 1,264 7.0 1,294 6.2
Unknown 3,121 17.3 1,074 5.2
Age
<25 5,322 29.5 6,035 29.1 0.535
25-29 3,232 17.9 3,730 18.0
30-34 2,531 14.0 2,846 13.7
35-39 2,155 12.0 2,516 12.1
40-49 2,860 15.9 3,404 16.4
50+ 1,875 10.4 2,221 10.7
Unknown 42 0.2 7 0.1
Area of residence
Sydney metropolitan area 8,614 47.8 11,054 53.2 <0.001*
Rest of NSW 8,480 47.1 8,831 42.5
Other55 923 5.1 874 4.2
Offence type
High PCA 5,455 30.3 5,360 25.8 <0.001*
Other 12,562 69.7 15,399 74.2
PCA convictions in previous 2 yrs
0 17,067 94.7 20,118 96.9 <0.001*
1 922 5.1 629 3.0
2 28 0.2 12 0.1
Convictions for any other offence in
previous 2 yrs
0 15,365 85.3 18,041 86.9 <0.001*
1 1,876 10.4 1,933 9.3
2 523 2.9 501 2.4
3+ 253 1.4 284 1.4
Concurrent offences
0 13,348 74.1 16,054 77.3 <0.001*
1 2,750 15.3 2,736 13.2
2+ 1,919 10.7 1,969 9.5
TOTAL 18,017 100.0 20,759 100.0
*Significant at the 0.05 level

54
Gender (x2=8.95, df=1); Aboriginality (x2=45.25, df=1); Age (x2=4.10, df=5); Area of residence
(x =117.85, df=2); Offence type (x2=95.26, df=1); PCA convictions in previous 2yrs (x2=118.78,
2

df=2); Convictions for any other offence in previous 2yrs (x2=23.70, df=3); Concurrent offences
(x2=56.107, df=2).
55
This category includes offenders who, at the time of conviction, were living interstate (92.5%), in a
State institution (2.7%), overseas (1.3%) or who had no fixed abode (1.3%). It also included a small
number of offenders for whom there was no information available on the area of residence (2.1%).

142
Probability of reappearing for a drink-driving offence
The first measure of offending considered in this analysis is the probability of
reappearing in the Local Court for a drink-driving offence within three years of the
initial drink-driving conviction. The percentage of drink-drivers appearing before the
courts for a new drink-driving offence by area of residence56 and year of conviction
is shown in Table 5.10. The most notable feature of this table is that the vast majority
of drink drivers, regardless of where they lived or when they were convicted, did not
reappear for a drink-driving offence during the follow-up period. In fact, about 90
per cent of persons in all four groups had no new appearances for drink-driving
within three years of the reference offence.

Table 5.10: Number and percentage of drink-driving offenders who reappeared for a
drink-driving offence during the 3-year follow-up by offender group and area of
residence
Sydney Non-Sydney
1997 1999 1997 1999
New PCA n % n % n % n %
appearance
No 7,831 90.9 10,062 91.0 7,576 89.3 7,929 89.8
Yes 783 9.1 992 9.0 904 10.7 902 10.2
Total 8,614 100.0 11,054 100.0 8,480 100.0 8,831 100.0

From Table 5.10 it can also be seen that there is little difference between Sydney
offenders convicted in 1997 and Sydney offenders convicted in 1999, in terms of the
percentage who reappeared for a drink-driving offence within three years of
conviction. The difference between 1997 and 1999 in the proportion reappearing was
slightly greater for offenders residing elsewhere in New South Wales but statistical
tests showed this difference to be non-significant.57 However, as described
previously, the differences in offender characteristics apparent between the two
offender cohorts mean that a direct comparison of these percentages could be biased.
Thus logistic regression techniques were applied to compare adjusted differences in
reoffending rates across groups.

56
Persons residing interstate, in State institutions, overseas or who had no fixed abode were excluded
from the analyses.
57
Sydney offenders (x2=0.079, df=1, p=0.783); non-Sydney offenders (x2=0.922, df=1, p=0.345).

143
The main effects model from the logistic regression analysis is shown in Table 5.11.
As seen in this table, offender group, area of residence, gender, age, Aboriginality,
prior convictions for non drink-driving offences58 and concurrent offences were
predictive of reappearing for a drink-driving offence within the three-year follow-up
period. The type of drink-driving offence (high-range v. other) and prior convictions
for a drink-driving offence were not independently predictive of reappearing and
therefore these variables were excluded from the final model.

Table 5.11: Odds ratios for risk of appearing for a new drink-driving offence – Main
effects model
Variable Odds 95% Confidence p-value
ratio Interval’s
lower upper
Offender group
1999 v. 1997 0.88 0.82 0.94 < 0.001
Area of residence
Non-Sydney residence v. Sydney 1.14 1.06 1.22 < 0.001
Gender
Female v. male 0.75 0.67 0.83 < 0.001
Aboriginality
Aboriginal v. non-Aboriginal 1.31 1.16 1.48 < 0.001
Age
25-29 v. <25 0.76 0.69 0.84 < 0.001
30-34 v. <25 0.81 0.73 0.91 < 0.001
35-40 v. <25 0.77 0.69 0.87 < 0.001
40-49 v. <25 0.75 0.67 0.84 < 0.001
50+ v. <25 0.50 0.43 0.58 < 0.001
Convictions for any other offence in previous 2 yrs
1 (other) prior v. none 1.46 1.32 1.62 < 0.001
2 (other) priors v. none 1.61 1.35 1.91 < 0.001
3+ (other) priors v. none 1.53 1.21 1.93 < 0.001
Concurrent offences
1 v. none 1.13 1.03 1.25 0.012
2+ v. none 1.22 1.09 1.36 < 0.001

As described previously, the marginal increase in the length of licence

disqualification periods after the implementation of the sentencing policy was found
to be greater for offenders residing outside of the Sydney metropolitan area than was
the case for their city counterparts. In order to examine whether the deterrent effect

58
This variable excluded prior convictions for a PCA offence to avoid multicollinearity problems.

144
of the penalty changes was greater for drink-drivers residing from non-Sydney areas
an interaction term for year of conviction by area of residence was included in the
logistic regression model.59 This analysis found evidence for a significant interaction
between these two variables (p=0.037) after controlling for other confounding
factors. Table 5.12 shows the odds ratios for this interaction effect adjusted for the
covariates presented in Table 5.11. For offenders residing in the Sydney metropolitan
area there was no significant difference in the odds of reappearing for a new drink-
driving offence when comparing drink-drivers convicted in 1997 with those
convicted in 1999. However, the odds of reappearing did decrease for the non-
Sydney drink-drivers who received the more severe penalties. The odds for non-
Sydney drink-drivers convicted in 1997 were 23 per cent higher than Sydney drink-
drivers convicted in the same year, but by 1999, Non-Sydney drink-drivers had the
same odds of reappearing as the Sydney drink-drivers convicted in 1997.

Table 5.12: Adjusted odds ratios for risk of appearing for a new drink-driving
offence – Interaction model
Variable Odds 95% Confidence
ratio Interval’s
lower upper
Sydney 1999 v. Sydney 1997 0.95 0.86 1.04
Non-Sydney 1997 v. Sydney 1997 1.23 1.11 1.37
Non-Sydney 1999 v. Sydney 1997 1.00 0.91 1.11

These differences across offender cohorts can be more clearly understood by

comparing the probability of reappearing for a new drink-driving offence. In order to
calculate probabilities60 from the logistic regression output a base model for
comparison needs to be specified. In this instance a male, non-Aboriginal offender,
who is aged less than 25 years and has no prior convictions or any concurrent
offences, was chosen. Figure 5.9 displays the probability of reappearing before the
court for a new drink-driving offence, given this set of offender characteristics, by

59
The Hosmer and Lemeshow Goodness-of-fit test statistic for this interaction model was not
significant at the 0.05 level (chi-square =11.747; p=0.163). This means we retain the null hypothesis
that there is no difference between the observed and predicted values of the dependent variable and,
thus, can conclude that the model’s estimates fit the data at an acceptable level.
60
Probabilities were estimated from the regression model using the following equation; ln⎛⎜ p ⎞⎟ = α
⎝ 1 − p⎠
+ β 1 X 1 + β 2 X 2 + β 3 X 3 + . . . + β k X k , where p is the probability of reappearing for a new drink-
driving offence and Xk are the predictor variables included in the regression model.

145
area of residence and year of conviction. As seen from this figure the probability of a
young, male, non-Aboriginal, first offender from outside of the Sydney metropolitan
area reappearing for a new drink-driving offence was three percentage points less
after the penalties were raised than before (14 per cent in 1997 v. 11 per cent in
1999). This equates to a reduction of 21 per cent in the probability of reappearing for
a drink-driving offence after the statutory penalties were increased. In contrast, there
was no change in the probability of non-Sydney drink-driver reoffending given this
set of offender characteristics.

Figure 5.9: Probability of reappearing for a new drink-

driving offence by offender residence and year of
conviction; male, non-Aboriginal, aged <25, with no
prior or concurrent offences

0.16
0.14
0.14

0.12 0.11 0.11

0.11
0.1
Probability

0.08

0.06 Sydney
Non-Sydney
0.04

0.02

0
1997 1999

Time to first drink-driving offence

Time to first reappearance for a drink-driving offence was the second measure of
reoffending used in the recidivism analysis. Survival time for non-recidivists was
calculated as the number of days from the finalisation date of the reference drink-
driving offence until the cut-off date specified for the relevant offender groups (i.e.
December 31 2000 for the 1997 cohort and December 31 2002 for the 1999 cohort).
In the case of recidivists, survival time was the number of days from the finalisation
date of the reference drink-driving offence until court finalisation of the new drink-

146
driving offence. Calculating survival time in this way gives rise to the following
mean times61 to first offence for Sydney and non-Sydney drink-drivers:
Sydney drink-drivers convicted in 1997: 1,370 days
Sydney drink-drivers convicted in 1999: 1,370 days
Non-Sydney drink-drivers convicted in 1997: 1,357 days
Non-Sydney drink-drivers convicted in 1999: 1,361 days
Log-rank tests of equality for the survival functions of the two offender cohorts (i.e.
1997 and 1999) showed no significant differences in the time to first drink-driving
reappearance for both drink-drivers residing in the Sydney metropolitan area and
offenders residing elsewhere in New South Wales.62

At this point, other explanatory variables related to offending have not been
controlled for. In order to take account of pre-existing differences between offender
groups, a Cox proportional hazards model was used to fit data on time to first drink-
driving reappearance. The main effects model, as shown in Table 5.13, found
significant effects for offender group, area of residence, gender, Aboriginality, age,
prior convictions for a non drink-driving offence,63 prior convictions for a drink-
driving offence and concurrent offences. Current conviction for a high-range PCA
offence was not significant and was therefore excluded from the final model.

61
Median survival times could not be calculated because more then 50% of offenders in each group
had not reappeared for a drink-driving offence by the end of the follow-up period.
62
Sydney drink-drivers (log-rank: p=0.744); non-Sydney drink-drivers (log-rank: p=0.477).
63
Graphing the hazard function for the ‘previous convictions’ category indicated that the hazard
function for the 2 and 3+ categories crossed, suggesting that the proportional hazards assumption is
not met for this predictor variable. The 2 and 3+ categories were therefore combined for the Cox
regression analyses.

147
Table 5.13: Hazard ratios for time to new appearance for a drink-driving offence –
Main effects model
Variable Hazard 95% Confidence p value
ratio Interval’s
lower upper
Offender group
1999 v. 1997 0.90 0.85 0.96 0.001
Area of residence
Non-Sydney residence v. Sydney 1.11 1.05 1.19 < 0.001
Gender
Female v. male 0.73 0.66 0.81 < 0.001
Aboriginality
Aboriginal v. non-Aboriginal 1.25 1.12 1.38 < 0.001
Age
25-29 v. <25 0.78 0.71 0.85 < 0.001
30-34 v. <25 0.81 0.73 0.89 < 0.001
35-40 v. <25 0.77 0.69 0.85 < 0.001
40-49 v. <25 0.77 0.70 0.84 < 0.001
50+ v. <25 0.51 0.44 0.58 < 0.001
PCA convictions in previous 2 yrs
Prior PCA v. none 1.15 1.01 1.32 0.032
Convictions for any other offence in previous 2 yrs
1 (other) prior v. none 1.38 1.27 1.51 < 0.001
2 (other) priors v. none 1.49 1.32 1.69 < 0.001
Concurrent offences
1 v. none 1.10 1.01 1.20 0.026
2+ v. none 1.20 1.09 1.32 < 0.001

Once again an interaction term was included in the model to examine the differential
effect of the sentencing policy across different areas of New South Wales. This
analysis found a significant interaction effect between offender group and area of
residence (p=0.011). Table 5.14 shows the hazard ratios for this interaction effect
after adjusting for the covariates presented in Table 5.13. These hazard ratios
indicate that, controlling for other confounding variables, non-Sydney drink-drivers
had a decreased risk of reappearing for a new drink-driving offence after the
penalties were raised but the reappearance rate for Sydney drink-drivers remained the
same. Before the penalties were raised, non-Sydney offenders had a reappearance
rate that was 21 per cent higher than Sydney drink-drivers but after the penalty
changes there was no significant difference between non-Sydney and Sydney
offenders.

148
Table 5.14: Adjusted Hazard ratios for time to new appearance for a drink-driving
offence – Interaction model
Variable Hazard 95% Confidence
ratio Interval’s
lower upper
Sydney 1999 v. Sydney 1997 0.98 0.90 1.07
Non-Sydney 1997 v. Sydney 1997 1.21 1.11 1.33
Non-Sydney 1999 v. Sydney 1997 1.01 0.93 1.11

Figures 5.10 and 5.11 display the survival curves, adjusted for the explanatory
variables included in the Cox regression model, for Sydney and non-Sydney
offenders, respectively.64 These adjusted survival curves show the proportion of each
group who had not yet reappeared before court for a drink-driving offence plotted
against the number of days in the follow-up period. Again base model needs to be
specified to calculate the survival functions for each group. Our base model is a
male, non-Aboriginal drink-driver aged less than 25 years, who had no prior
convictions or concurrent offences. As seen from these figures, offenders who reside
outside of the Sydney metropolitan area and who were convicted of drink-driving in
1999 generally remained offence-free for longer than non-Sydney drink-drivers
convicted of an offence in 1997. In contrast, there was almost no difference in the
time to first new drink-driving offence for Sydney drink-drivers convicted after the
penalties were raised compared to those convicted prior to the legislative
amendments.

64
The adjusted survival curves were estimated separately for Sydney and non-Sydney offenders.
However, the estimates derived from these new models, in comparison to the full interaction model
reported in this Chapter, were almost equal.

149
 Figure 5.10: Time to first offence, Sydney drink-drivers

1.00
0.98
0.96 1997
0.94 1999
0.92
Proportion

0.90
0.88
0.86
0.84
0.82
0.80
0 200 400 600 800 1000 1200 1400
Days to first reappearance for drink-driving

Figure 5.11: Time to first offence, non-Sydney drink-

drivers

1.00
0.98
0.96 1997
0.94 1999
Proportion

0.92
0.90
0.88
0.86
0.84
0.82
0.80
0 200 400 600 800 1000 1200 1400
Days to first reappearance for drink-driving

A major assumption of the Cox proportional hazards model employed in this analysis
is that the hazard ratio for one category is proportional to the hazard ratio for any
other category of an explanatory variable. One method for checking whether this

150
assumption is satisfied is to plot the log-log survival curves65 for the different
categories of explanatory variables. Parallel survival curves for each category
indicates that the proportional hazards assumption is satisfied and the estimated
model can be considered valid (Kleinbaum 1996). This procedure was undertaken for
each of the predictor variables appearing in the Cox regression models shown above.
These curves were parallel for all predictor variables indicating the proportional
hazards assumption is met for the predictors included in the final model.

Summary and discussion of recidivism analysis

The results from Study 2 provide some evidence for a beneficial effect of the 1998
sentencing policy on drink-driving recidivism. At least for offenders residing outside
of the Sydney metropolitan area, drink-drivers sentenced before the statutory
penalties were increased, had higher odds of reappearing for a new offence and
reoffended sooner than drink-drivers sentenced after the penalties were raised. This
effect was not apparent for Sydney drink-drivers. Furthermore, examination of the
implementation of the sentencing policy, indicates that increases in court-imposed
penalties for drink-driving offences coincided with the legislative changes and that
negative effects on prosecution rates, conviction rates, guilty pleas and court delay
were not apparent.

Prima facie these results would suggest that harsher penalties can decrease rates of
drink-driving reoffending, at least for one specific group of offenders. However,
three other explanations for this effect need to be discounted before such a
conclusion can be reached.

The first of these is changes to drink-driving enforcement practice during the follow-
up period. If the perceived risk of apprehension was greater after the penalties were
increased then this could account for the observed reduction in recidivism. To
examine this possibility, data on the number of breath tests administered to motorists
over the period 1994 through 2001 were obtained from the NSW Police. These data
showed that the number of breath tests conducted by police did increase in the later

65
The log-log survival curve is a transformation of an estimated survival curve that results from
taking the natural log of an estimated survival probability twice.

151
years of the study period (i.e. 2001 and 2002). However, the overall number of
breath tests in the first year of follow-up for the 1999 cohort was still at a slightly
lower level in comparison to the first year of follow-up for the 1997 cohort
(1,961,448 v. 2,077,034 tests). Restricting the outcome used in the logistic regression
analysis to any reappearance within the first 12 months after conviction (instead of
three years) resulted in the same conclusion that non-Sydney drink-drivers from the
1999 cohort had reduced odds of reoffending (note that this issue of enforcement of
drink-driving offences is discussed in greater detail in the next Chapter).

A second plausible explanation for the findings from Study 2 is that the reduction in
reoffending observed for drink-driving offences simply reflects a secular downward
trend in all offending rather than a specific deterrent effect created by the penalty
increase. Additional logistic regression and survival analyses, comparing recidivism
rates before and after 1998, were therefore conducted for two groups of non drink-
drivers: offenders convicted of a property crime and offenders convicted of a violent
crime.66 Property and violence offences were chosen as comparisons because these
crime types were unaffected by the penalty changes implemented in 1998 and are
less likely to be influenced by changes in police practice than other offences (e.g.
drug offences). These analyses showed a significant reduction for violent offences in
the odds of reappearing for a new offence within three years of conviction and also a
significant increase, between 1997 and 1999, in the time it took this group to
reoffend. However, unlike drink-driving offences, this reduction in reoffending did
not vary by an offender’s area of residence. No differences in reoffending were
apparent for persons convicted of a property offence in 1997 compared to those
convicted of a property offence in 1999.

It could be argued that the observed increase in time to reoffend for the 1999
offender cohort stemmed not from a deterrent effect but rather an incapacitation
effect. That is, it could be argued that drink-drivers convicted in 1999 took longer to
reappear before the court for a new drink-driving offence than their 1997

66
Comparisons of reoffending rates for driving offenders could not be undertaken since statutory
penalties for driving offences were increased at the same time as drink-driving penalties. For practical
reasons, offenders who appeared before the Local Courts on every third day, and were convicted of a
property or violence offence, were considered here.

152
counterparts because they were sentenced to longer custodial terms. This argument
can be rejected, however, because the proportion of drink-drivers that were
imprisoned for the reference offence was only one per cent in both cohorts and the
average custodial sentence received was only around four and a half months. It is
true that the actual ‘free’ time that this small group of offenders had to appear before
the courts for a new offence during the follow-up period was reduced after 1998.
However, using proportional hazard models to fit data on time to first offence,
excluding offenders who were imprisoned, confirmed the significant interaction
effect between year of conviction and area of residence, controlling for other
explanatory variables.

The data from Study 2 therefore indicate that the changes to drink-driving penalties
in New South Wales in 1998 were associated with a significant reduction in the rate
of reoffending for offenders from country and regional areas. When considering the
importance of these findings, however, it needs to be noted that the overall effect of
the increased penalties on recidivism rates was relatively small. The probability of a
drink-driver reoffending was reduced by just three percentage points in non-Sydney
locations. Given such a small effect size from what was essentially a doubling of the
statutory penalties for all drink-driving offences, and keeping in mind the associated
costs with administering the new penalty regime, the value of the tougher penalties
strategy in controlling drink-driving remains questionable.

It is possible, however, that Study 2 underestimated the overall impact of the 1998
sentencing policy on reoffending. In order to minimise the potential effect of other
contextual variables on reoffending rates (e.g. changes in police activity), the two
cohorts selected for comparison were relatively close in temporal proximity. This
meant that some of the follow-up period for offenders included in the 1997 cohort
extended beyond the introduction of the new policy. A general deterrent effect
created by the penalty increase therefore had potential to influence the rate at which
this group reoffended. Any bias relating to this aspect of the research design would
have been towards the null hypothesis. Thus, the overall conclusions from the
analysis remain valid, but it is still possible that the ‘true’ magnitude of the group
difference in reoffending was greater than the current research suggests. It is difficult

153
to ascertain the extent to which the announcement of the drink-driving penalty
changes may have affected the reoffending rate of drink-drivers included in the 1997
cohort. However, as will be seen in the next chapter, the 1998 increase in drink-
driving penalties received very little publicity, therefore any underestimate of the
policy’s impact on recidivism in Study 2 is likely to be minimal.

It is also possible that the impact of the 1998 legislation would have been greater if
licence disqualifications were more systematically applied for drink-driving offences.
While the legislative amendments had the effect of increasing the average licence
disqualification for drink-driving offences across the State, 20 per cent of guilty
offenders still escaped licence disqualification on being found guilty of a drink-
driving offence (via a Section 10 dismissal) despite the existence of these mandatory
minima (for more details on trends in Section 10 dismissals in New South Wales see
Moffatt, Fitzgerald & Weatherburn 2004). Ensuring that almost all offenders are
recipients of a licence disqualification once found guilty for a drink-driving offence
could have amplified the effects found here.

Conclusions from the quasi-experimental studies

To summarise so far, Study 1 found a significant increase in one of the surrogate
measures used to index alcohol-related road crashes (single-vehicle night-time
crashes) after the harsher penalties were introduced in New South Wales in
September 1998. Further exploratory analyses suggested that this increase was
mostly driven by a rise in non alcohol-related road crashes that occurred at or around
the same time as the penalty changes. From these findings two possible conclusions
about the sentencing policy could be drawn: (1) that there was a reduction in alcohol-
related crash rates after the penalty changes but that this reduction was overwhelmed
by the rise in non alcohol-related crashes and therefore could not be detected in our
surrogate measure or (2) there was no change in alcohol-related road crashes after the
sentencing policy was implemented.

Study 2 showed that average gaol terms, fines and licence disqualification periods for
drink-driving offences significantly increased after the statutory penalties were raised
in September 1998, with some variations in the magnitude of this rise across the

154
State. There was also little evidence for a negative effect of the sentencing changes
on the prosecution of drink-driving offences. There was a small reduction in the
proportion of drink-drivers who received a licence penalty after the legislative
amendments were enacted. However the vast proportion of drink-drivers appearing
before the courts in 1999 did have their licence disqualified and for longer than their
1997 counterparts. The results of the logistic regression and survival analyses found
a significant reduction in the probability of drink-drivers reoffending and the time to
first new drink-driving offence after these tougher penalties were enacted. This
reduction in reoffending was small and specific to offenders residing in country and
regional areas.

In light of the findings from Study 2 it remains possible that there was a significant
reduction in alcohol-related crashes after the penalty changes but, because of the
overwhelming rise in non alcohol-related crashes around the same time, this decrease
could not be detected in our analysis of single-vehicle night-time crash rates (Study
1). However, the results of Study 2 also suggest that if there were a reduction in
alcohol-related crashes immediately after the policy was implemented, the magnitude
of this effect would have been relatively small.

These findings indicate that the doubling of statutory penalties for drink-driving
offences in New South Wales in 1998 and the subsequent increase in average court
penalties for these offences resulted in only a small reduction in offending. The
question then becomes why didn’t the policy have a greater impact on drink-driving
rates in New South Wales? Does punishment severity matter little in offender
decision-making or are people who have previously been sanctioned for an offence
more sensitive to changes in punishment variables? Before the latter questions can be
addressed, we need to investigate other factors that may have affected the deterrent
efficacy of the penalty changes within this real-life setting. Therefore, in the next
Chapter we will examine the context in which the 1998 sentencing policy was
implemented.

155
CHAPTER 6. CONTEXT OF POLICY
IMPLEMENTATION

Chapter 3 identified several reasons why a sentencing policy that raises punishment
severity may have minimal influence on offending rates. These include: (1) the
policy failing to alter actual sentencing practice (2) the penalty increase having
countervailing effects on the prosecution and conviction of the targeted offences (3)
the perceived risk of apprehension being relatively low in the jurisdiction where the
new law is implemented and (4) the target audience being unaware that formal
punishment for the offence has changed.

The data provided in Chapter 5 demonstrated that the magnitude of drink-driving

penalties increased significantly after the statutory amendments were enacted in
1998. On the whole, there was also no indication that the penalty changes had
negative, countervailing effects on the prosecution and conviction of drink-driving
offences. There was some evidence that a proportion of drivers escaped the
mandatory minimum licence disqualifications prescribed by the new laws, which
could have contributed somewhat to the diminished effect of the policy on
reoffending. However, the vast majority of convicted drink-drivers were recipients of
more severe penalties after the legislation was implemented and yet there was still
only a small observed effect of these changes on offending levels.

Two other reasons for the failure of the 1998 sentencing policy to exert a greater
deterrent effect on drink-driving remain. Firstly, the perceived risk of apprehension
may have altered over the course of the study period. Henstridge and his colleagues
(1997) have shown a direct relationship between the number of breath tests that
police undertake and subsequent reductions in alcohol-related road crashes. If the
NSW Police altered their enforcement activity at or around the time the new
penalties came into operation then this could have undermined the intended deterrent
effect of the penalty changes. Secondly, it is possible that many drink-drivers failed
to change their offending behaviour because they were simply unaware that drink-
driving penalties had increased. These two accounts, for the relatively small effect

156
observed in the quasi-experiment, are explored in greater detail in the following
sections.

Enforcement of drink-driving offences

As described in Chapter 2, deterrence theory anticipates that the severity of
punishment matters more when punishment is highly probable. Given the relatively
high level of drink-driving enforcement in New South Wales, through RBT, it was
thought that raising the severity of drink-driving penalties should have a greater
effect in this State than would be the case for many other jurisdictions. It is therefore
essential to consider any variations in the level of RBT at or around the time the
penalties were increased in New South Wales in 1998 before any definitive
conclusions can be reached about the effectiveness of punishment severity.

To investigate this issue further the number of breath tests administered by NSW
Police and the number of police-recorded drink-driving incidents (or charges), were
examined over the eight-year period covered by the quasi-experimental study.67 Data
for 2002 are also included in this analysis, given that the follow-up period in which
reoffending was measured for the post-law cohort fell within this year. Drink-driving
charges are included here as an indicator of enforcement activity, rather than as a
measure of offending because, as discussed in Chapter 4, the actual probability of
arrest for drink-driving is extremely low.

Figure 6.1 displays both the number of breath tests administered by NSW Police
each month from April 1994 through December 2002 and the number of drink-
driving charges recorded from January 1995 through December 2002. As seen in this
figure, these two indicators appear to follow a similar pattern during the earlier years
of the study period but from 1998 the series begin to diverge, with the number of
breath tests dropping as drink-driving charges begin to rise. Interrupted time-series
analyses (using a regression approach) were conducted for each of these data series,
67
Monthly breath tests administered by NSW Police between April 1994 and December 2002 were
examined. However, the COPS (Computerised Operational Police System) database, used by NSW
police to record drink-driving incidents, was only introduced in April 1994 and there is some concern
about data reliability for the first few months following the statewide rollout. Therefore, only drink-
driving incidents recorded during the period 1995 through 2002 are considered here. The NSW Police
supplied breath test data and police charge data were obtained from the NSW Bureau of Crime
Statistics and Research (see NSW Recorded Crime Statistics 1998 to 2002).

157
with the intervention point corresponding to the implementation of the new
penalties.68 For breath tests, this analysis showed a significant decreasing trend prior
to the penalty change, followed by a drop in breath tests that coincided with the
intervention point and then a significant change in the underlying trend, as the
number of breath tests began to increase from October 1998. In contrast, there was a
small, but significant, upward trend in the number of drink-driving charges before
the penalty change, then a significant rise in charges at the time the penalties were
raised but no change in the underlying upward trend thereafter. In summary, at the
time drink-driving penalties were increased in New South Wales, there were fewer
overall breath tests being conducted by police but there were more people being
caught for drink-driving offences.69

Figure 6.1: Monthly police breath tests and recorded

drink-driving incidents, NSW, Apr1994-Dec2002

3000 450000
Drink-driving incidents
400000
2500 Breath tests
350000

Number of breath tests

Number of incidents

2000 300000

250000
1500
200000

1000 150000

100000
500
50000

0 0
O -01
Ap -01
O -02
Ap
O -94
Ap -94
O -95
Ap -95
O -96
Ap -96
O -97
Ap -97
O -98
Ap -98
O -99
Ap -99
O -00
Ap -00
ct

ct

ct

ct
ct

ct

ct

ct

ct
r

r
r

-0
2

68
A regression approach was adopted for the interrupted time-series analyses, after a Dickey-Fuller
test confirmed that both series were trend stationary. The intervention point was October 1998, the
first full month following the enactment of the Traffic Amendment (Penalties and Disqualifications)
Bill. Eleven dummy variables corresponding to the months were also included in the models to
control for seasonal autocorrelation in the data. Diagnostic checks of the residuals suggested that the
models contained no significant autocorrelation. The final models are included in Appendix III.
69
Note that multivariate modelling of the relationship between breath testing rates and alcohol-related
crashes was not possible due to the variability in the breath testing data and the lack of data points in
both series to enable the use of quarterly data.

158
At first glance these results would seem somewhat incongruous, given both breath
tests and drink-driving charges represent levels of police enforcement. However,
what appears to have happened over this time period (and this was verbally
confirmed by Commander Ron Sorrenson, head of Traffic Services at the NSW
Police when data on breath tests were sought for this study) is that the police
switched their enforcement strategy from one of general deterrence, with stationary
RBTs in highly visible locations, to the targeting of high-risk times and locations
(e.g. weekend nights and major roads) for drink-driving. Thus, while they were not
conducting as many breath tests post September 1998, the police were in the right
places, at the right times, and were therefore catching more offenders.

The probable impetus for this switch in tactics (and the subsequent increase in drink-
driving charges) was the launch of a NSW Government initiative, known as the
Enhanced Enforcement Campaign. This was a statewide program, managed by the
RTA, which funded police Local Area Commands (LACs) that wished to enhance
the enforcement operations in their local area (including those targeting drink-
drivers). The program commenced operation in NSW in March 1998, and by the
middle of 1998 many applications for funding had been received and approved by
the RTA (Barry Wirrick, Enhanced Enforcement Program Manager, RTA; personal
communication 2003). LACs that were successfully granted funds used these monies
to set up random breath testing sites late at night/early in the morning and during
holiday periods when a greater proportion of drink-driving occurs.

Because raising the perceived certainty of apprehension amongst drink-drivers is

associated with reductions in offending (Voas & Hause 1987; Henstridge, Homel &
Mackay 1997; Voas, Holder & Gruenewald 1997), the increase in drink-driving
charges and the decrease in breath tests had the potential to affect offending
behaviour, albeit in opposing directions. It is possible that the reduced level of police
breath testing from 1997 onwards had a deleterious effect on the perceptions of arrest
risk held by the general driving population and that this weakened the deterrent effect
of the 1998 sanctioning policy. It is also possible, however, that the targeting of high-
risk times and locations increased the perceived risk of apprehension, particularly

159
amongst frequent offenders, and thus contributed somewhat to the decrease in
reoffending apparent in Study 2.

Police breath tests and public perceptions of enforcement activity

There is good reason to believe that the drop in total breath tests administered by
NSW Police compromised the deterrent effect of the 1998 penalty changes. Previous
research conducted in New South Wales has demonstrated a direct relationship
between the overall number of breath tests conducted by police and subsequent
reductions in alcohol-related road crashes (Henstridge, Homel & Mackay 1997).
Thus, even modest reductions in breath tests during the study period could have
reduced perceptions of apprehension risk. Due to the retrospective nature of the
quasi-experimental design, perceptions of apprehension risk before and after the
drink-driving penalties were increased could not be assessed. However, data bearing
upon this issue was available from an annual, national survey conducted by the
Australian Transport Safety Bureau (ATSB).

The ATSB Community Attitudes Survey is a telephone survey, of Australian

residents aged 15 years and over, which measures respondents’ attitudes toward a
range of road safety issues (see Mitchell-Taverner 1994; 1995; 1996; 1997; 1998;
1999; 2000; 2002a; 2002b). Importantly, for this thesis, the survey asks people in all
Australian States about their exposure to random breath tests and their perceptions of
RBT activity. Figures 6.2 and 6.3 present NSW data, from this survey, for each year
between 1994 and 2002. Shown in Figure 6.2 is the proportion of NSW respondents
that thought RBT activity had increased, and the proportion that thought RBT
activity had decreased, in the previous two years. In Figure 6.3, is the proportion of
the NSW sample that reported seeing RBT in operation in the last six months and the
proportion reporting having been personally breath tested during this same period. In
both figures the number of breath tests administered by NSW Police each year during
the study period is also plotted for comparative purposes.

160
 Figure 6.2: Yearly breath tests and perceptions of
enforcement activity, NSW, 1994-2002
Increased
45 Decreased 3000000

40 Total yearly breath tests

2500000
35

30 2000000

No. breath tests

Percentage

25
1500000
20

15 1000000

10
500000
5

0 0
1994 1995 1996 1997 1998 1999 2000 2001 2002

Figure 6.3: Yearly breath tests by reported experience

with RBT, NSW, 1994-2002
Seen in operation
90 Tested 3000000

80 Total yearly breath tests

2500000
70

60 2000000 no. breath tests

Percentage

50
1500000
40

30 1000000

20
500000
10

0 0
1994 1995 1996 1997 1998 1999 2000 2001 2002

The NSW sample included in this survey is relatively small (n equals approximately
250 for the years shown here) and, therefore, these data are quite variable across the
study period. Despite this variability, there does appear to be some congruency

161
between perceptions of drink-driving enforcement reported in the survey each year
and trends in the number of breath tests recorded by police. When comparing these
data, it should be remembered that the survey inquired about the level of RBT
activity occurring during the previous two years or exposure to RBT in the previous
six months. Thus, we would expect to see a lag between breath tests administered by
police and perceptions recorded in the survey. Accordingly, from Figure 6.2 it can be
seen that the peak in breath tests evident in 1996, corresponds to fewer people
reporting a decrease, and more people reporting an increase, in police activity in
1997. As the number of breath tests begins to decline after 1996, the proportion
perceiving a decrease in RBT activity commences to rise each year, reaching a peak
of 21 per cent in 2001.

The relationship between trends in experiences with RBT activity and trends in
police breath tests is a little less clear. Figure 6.3 shows a peak in 1997 (25 per cent)
in the proportion reporting having been breath tested, as well as a peak in the same
year (79 per cent) in the proportion reporting that they had seen RBT in operation,
which closely corresponds to the peak in breath tests administered by NSW Police in
1996. Then in the 1998 survey, we see a drop in RBT experience, and RBT exposure,
which is consistent with the decreasing trend in the number of breath tests after 1996.
However, reported experience with RBT testing jumps up in 1999 but drops off in
the remaining years of the study period. Meanwhile the proportion exposed to RBT
steadily increases from 1998 until 2002.

Overall, these data clearly show that after the statutory penalties were raised in 1998,
more NSW residents perceived that RBT activity had decreased and fewer people
had seen RBT in operation. In 1997, 12 per cent of the respondents surveyed thought
that RBT activity in New South Wales had decreased, but by 1999 this percentage
had risen to 20 per cent, or one in five. Similarly, in 1997, 79 per cent of respondents
reported having seen RBT in operation, but by 1999 only 67 per cent reported
exposure to RBT operations in New South Wales. Given strong evidence for a
relationship between drink-driving enforcement levels and subsequent reductions in
alcohol-related crashes, this decline in exposure to police activity raises some
concern about the level of perceived certainty of apprehension at the time the

162
sentencing legislation was implemented. Reductions in the intensity of drink-driving
enforcement may also have contributed somewhat to the increase observed in non
alcohol-related crashes after the intervention in 1998. Although RBT is specifically
designed to increase the perceived risk of apprehension for drink-driving offences,
the fact that it involves many police in highly visible locations means that it may also
have an ancillary effect in deterring (or not) other traffic offending, particularly drug
driving and speeding.

Police targeting of high-risk times and locations

While the reduction in the number of breath tests may have diminished the general
deterrent effect of the statutory changes, it might be thought that the hotspotting
techniques adopted by police contributed to the reduction in drink-driving recidivism
observed after the penalty changes. Upon closer examination of drink-driving
charges across different areas of New South Wales, this explanation for the findings
from Study 2 appears unlikely. Despite being a statewide campaign, the enhanced
enforcement program was taken up more readily by some LACs than by others
(Barry Wirrick, Enhanced Enforcement Program Manager, RTA; personal
communication 2003). This disparity across areas is clearly shown in Table 6.1,
which presents a breakdown of the number of drink-driving charges by Statistical
Division, from 1997 through 2002. As shown here, all Statistical Divisions in New
South Wales experienced an increase in drink-driving charges over this six-year
period. The areas that experienced the greatest increase in charges during this time
were located in the Sydney Metropolitan region. Yet the reduction in reoffending,
observed in Study 2, was found only for drink-drivers residing in country and
regional areas.

163
Table 6.1: Number of police recorded drink-driving incidents and percentage change
from 1997 to 2002 by Statistical Division where incident occurred, New South
Wales.
Statistical Division 1997 1998 1999 2000 2001 2002 %
change
97-02
Sydney 7,559 9,279 12,099 12,126 12,009 12,795 69%
Inner Sydney 805 961 1,202 1,374 1,459 1,745 117%
Eastern Suburbs 391 426 494 622 695 702 80%
St George- Sutherland 682 763 909 1,058 1,140 1,313 93%
Canterbury – 315 340 499 452 501 514 63%
Bankstown
Fairfield – Liverpool 492 666 1,062 969 946 1,249 154%
Outer South West 515 891 972 750 677 693 35%
Sydney
Inner Western Sydney 164 171 289 332 336 384 134%
Central Western 412 458 574 626 596 536 30%
Sydney
Outer Western Sydney 746 999 1,216 1,416 1,054 885 19%
Blacktown – Baulk. 562 891 1,081 826 808 894 59%
Hills
Lower Northern 680 797 1,359 981 1,139 909 34%
Sydney
Hornsby – Kuring-gai 273 296 574 560 710 670 145%
Northern Beaches 745 846 976 993 826 1,089 46%
Gosford – Wyong 777 774 892 1,167 1,121 1,212 56%
Hunter 1,797 1,674 1,929 2,303 2,742 3,000 67%
Illawarra 891 1,043 1,143 1,150 1,166 1,402 57%
Richmond –Tweed 967 1,024 1,148 1,270 1,383 1,415 46%
Mid-North Coast 1,078 1,171 1,286 1,257 1,309 1,163 8%
Northern 794 835 882 955 997 977 23%
North Western 596 576 751 754 658 708 19%
Central West 657 728 905 905 832 823 25%
South Eastern 875 957 1,204 1,127 1,225 1,201 37%
Murrumbidgee 640 713 787 941 860 861 35%
Murray 563 522 662 602 618 612 9%
Far West 90 111 234 167 133 130 44%
NSW TOTAL 16,507 18,633 23,028 23,557 23,933 25,087 52%
Source: NSW Bureau of Crime Statistics and Research, NSW Recorded Crime Statistics 1998 to
2002.

It is true that many country and regional areas also experienced noticeable increases
in drink-driving charges after 1998 and is conceivable that police targeting of high-
risk times and locations in these areas had a greater marginal deterrent effect on
offending than in the Sydney Metropolitan areas. Further data would be needed on
perceptions of enforcement activity across different areas of New South Wales and
the localities where breath tests were administered to investigate this possibility. This

164
information, however, is not available. Nor is there any research on the relative
effectiveness of “hotspotting” in deterring different types of drink-drivers (i.e. city
versus country). It is therefore difficult to assess the extent to which the switch in
enforcement tactics, by the NSW Police, may have contributed to the reduction in
drink-driving recidivism observed in Study 2. All that can be said is that changes in
perceived certainty of drink-driving punishment cannot entirely be ruled out as a
potential explanation for our findings.

Media coverage of statutory penalty increase

A central question in this thesis is whether the 1998 increase in statutory penalties for
drink-driving offences, which occurred in New South Wales, deterred offending by
increasing perceptions of sanction severity. We should therefore not only examine
changes in drink-driving behaviour but also assess people’s beliefs about the severity
of drink-driving punishment before and after the implementation of the new laws.
Unfortunately, the current investigation, being a retrospective study of the legislative
changes, does not permit this analysis. As an alternative we examine media coverage
of drink-driving and related issues, at and around the time of the introduction of the
new policy, in order to gain insight into the extent to which the increased threat was
adequately communicated to the target audience.

Newspaper coverage of penalty changes

To examine the extent of print media covering the 1998 rise in drink-driving
penalties, an interactive media database, Factiva, was used to gather any relevant
newspaper articles published in National or NSW Metropolitan newspapers, between
June 1 1997 and June 30 1999. Factiva combines media sources from both the Dow
Jones Interactive (Fairfax holdings) and Reuters Business Briefing databases and
provides full text access to news from more than 8,000 international sources. The
publications of most relevance to the current analysis (and which are included in this
database) are The Sydney Morning Herald, The Sun Herald, The Australian, the
Australian Financial Review, The Daily Telegraph, the Sunday Telegraph, the
Newcastle Herald and the Illawarra Mercury.

165
NSW country newspapers could not be accessed because the information gathered
from this historical search was limited to the sources contained in the Factiva
database. Although this necessarily restricts the analysis, the National newspapers
and daily Sydney Metropolitan newspapers that are included here, (in particular The
Australian, Sydney Morning Herald and Daily Telegraph) are also widely distributed
throughout New South Wales and therefore would be accessed by many living in
country and rural areas. Furthermore, while the quantity of drink-driving articles may
possibly be greater in country newspapers, there is no reason to suspect that the
content would differ to any considerable degree.

This database search identified a total of 448 newspaper articles that included the two
keywords of ‘drink driving’ and ‘NSW’. Only 139 of these specifically referred to
issues associated with the offence of drink-driving in New South Wales. Figure 6.4
displays the monthly distribution of these drink-driving articles across the two-year
period examined. Also shown in this figure is the month in which drink-driving
penalties were raised in New South Wales. It can be seen from this figure that in
1997 and 1998, drink-driving articles appeared most frequently in June and
December, which corresponds to the Christmas/New Year holiday period and the
Queen’s birthday holiday in June. There was also a peak in newspaper articles in the
month immediately following the implementation of the statutory amendments to
drink-driving penalties (October 1998), which was not evident for the same month in
the previous year.

166
 Figure 6.4: Number of drink-driving articles appearing
in NSW metropolitan papers each month, Jun1997-
Jun1999

18 17
16 Penalties increase
16
14 13
Number of articles

12
12 11

10 9

8 7
6 6
6 5 5
4 4 4 4
4 3 3 3
2 2
2 1 1 1
0 0
0
8

99
97

98

99
97

98

8
8

9
7

8
-9
-9

-9
r-9

r-9
-9

-9

b-
n-

n-

n-
g-

g-
b
ec

ec
ct

ct
Ap

Ap
Fe
Ju

Fe

Ju

Ju
Au

Au
O

O
D

A coding system was developed to extract further relevant information from these D

139 articles. This coding system was similar to that used by Cashmore (1985), in her
analysis of the impact of RBT in New South Wales, and essentially consisted of three
categories of information. The first category concerned the physical characteristics of
the article and included data such as the name of the newspaper, page number, size of
the article (in number of words) and the headline. The second category referred to
sources cited in the article, such as police, government officials/departments and
non-government organisations. The third category concerned the type of issues
discussed in the article: (1) Description of drink-driving penalties (2) Reference to
drink-driving penalty increase in 1998 (3) Description of Random Breath Testing
operations (4) Other issues related to drink-driving and road safety. The coding also
indicated whether these topics were the main or secondary issue of the article and
whether the article contained an overall evaluative element (i.e. whether negative or
positive).

As can be seen in Table 6.2, the majority of articles discussing drink-driving offences
over the period examined appeared in National and Sydney Metropolitan papers,
with almost 40 per cent published in either The Daily Telegraph or Sunday

167
Telegraph but a substantial minority came from Wollongong and Newcastle
newspapers, which service much smaller populations. Table 6.2 also shows that the
NSW Police was most often the source for information on drink-driving and related
issues discussed in these articles, with the courts and the NSW Government being the
next most frequent information source. The size of the drink-driving articles
published ranged from 64 to 1,985 words in length, with a median of 346 words.
Approximately one-third of the 139 drink-driving articles gathered in the search
appeared on the first to the third page of the newspaper (only seven were front-page
articles).

Table 6.2: Newspaper outlets and sources used in drink-driving

articles, June 1997 – June 1999
No. %
Newspaper
The Australian 13 9.4
Sydney Morning Herald 23 16.5
Daily Telegraph 34 24.5
Sun Herald 5 3.6
Sunday Telegraph 18 12.9
Sunday Age 1 0.7
Newcastle Herald 16 11.5
Illawarra Mercury 29 20.9
Total 139 100.0
Source
Police 66 47.5
Courts 27 19.4
Government officials/departments 22 15.8
Political parties 3 2.2
Staysafe parliamentary committee 2 1.4
Non-government associations 6 4.3
Other 2 1.4
Total 139 100.0

The number of references to each of the four content areas into which the articles
were collapsed is shown in Table 6.3. Also indicated in this table is whether each
content area was discussed as a main or secondary topic of the article. As shown
here, only 15 articles describing the penalty increase for drink-driving offences in
1998 were published in NSW Metropolitan newspapers over the two years examined.
The penalty changes were the main topic discussed in just seven of these. Most of the
articles referring to the new laws were published either at the time the statutory

168
amendments came into effect or in the two weeks following its enactment (i.e.
September 30 1998 – October 14 1998) but two coincided with the announcement of
the penalty changes by the State Government in June 1998. A further 51 articles
discussed drink-driving penalties in New South Wales in more general terms, usually
with reference to a specific drink-driving case prosecuted in the courts. However, the
majority of drink-driving articles concerned RBT operations or road crashes arising
from episodes of drink-driving.

Table 6.3: Type of reference by content area, June 1997 – June 1999
Main Secondary Total
Content Area n % n % n %
Drink-driving penalties 37 73 14 28 51 100
Drink-driving penalty 7 47 8 53 15 100
increase in 1998
RBT operations 26 46 31 54 57 100
Other issues 39 75 12 23 52 100

Table 6.4: Nature of evaluation by content area, June 1997 – June 1999
Negative Positive Total
Content Area n % n % n %
Drink-driving penalties 26 79 7 21 33 100
Drink-driving penalty 4 80 1 20 5 100
increase in 1998
RBT operations 3 15 17 85 20 100
Other issues 12 48 13 52 25 100

If the articles identified in the database search contained an evaluative component

they were also coded as negative or positive depending on the general view
expressed in relation to the content area discussed (see Table 6.4). As seen from
Table 6.4, many of the articles that discussed drink-driving penalties during this
period expressed a negative view of the legal implications for these offences. Often
these newspaper articles implied that drink-drivers once apprehended frequently
escape the severe penalties that accompany this offence, conveying the impression
that the courts are ‘soft’ on these types of crimes (e.g. “Soft pca sentences slammed
by Collins”, Newcastle Herald, 27th June 1998).

This was particularly true around the middle of 1998, only a few weeks after the
State Government announced their intentions to increase drink-driving penalties.

169
During this period, a former NSW Minister for Corrective Services was charged with
a high-range drink-driving offence. Instead of receiving the harsh penalties that were
prescribed for this offence, the former politician escaped a criminal conviction (via
the Section 10 equivalent available to magistrates at that time) and was given a $500,
12-month good behaviour bond (The Sydney Morning Herald, 18th June 1998; The
Sydney Morning Herald, 19th June 1998). Around the same time, the media also
highlighted a similar case, where a police officer charged with high-range drink-
driving escaped conviction (The Daily Telegraph, 24th June 1998; Newcastle Herald,
25th June 1998). Yet another occurred in the latter part of 1998, in which a Supreme
Court judge escaped conviction after pleading guilty to driving with a BAC that was
twice the limit (The Daily Telegraph, 26th November 1998; Illawarra Mercury, 27th
November 1998; The Daily Telegraph 27th November 1998).

Several articles suggested that these cases illustrated the preferential treatment in
sentencing accorded to high-profile and wealthy people found guilty of a drink-
driving offence: “Three times the limit, but Yapsley keeps licence” (The Sydney
Morning Herald, 18th June 1998), “When justice is no longer” (The Daily Telegraph,
23rd June 1998), “Celebrity drunks ‘get off lightly’/ One law for the rich and one for
the rest of us” (Sunday Telegraph, 28th June 1998) and ‘Penalties should be
mandatory’ (Sunday Telegraph, 28th June 1998). This negative view conveyed in
relation to drink-driving penalties could have reduced the deterrent effect of the 1998
sentencing reforms. As noted in Chapter 2, an individual’s normative commitment to
legal authorities is a non-instrumental determinant of legal compliance (Tyler 1990).
The impression conveyed by the media during 1998 that drink-driving laws are not
equally enforced amongst all members of the driver population could have reduced
the perceived legitimacy of authority and, consequently, resulted in fewer drivers
being willing to modify their behaviour in accordance with the law. These newspaper
articles also highlight a court’s ability to dismiss a drink-driving offence without a
conviction if the person is considered to be of ‘good character’. Knowledge of this
possible exemption from criminal penalties could also serve to reduce the perceived
certainty of conviction for drink-driving in New South Wales. If this were the case,
the negative publicity would have substantially undermined the deterrent effect of the
severe penalties for drink-driving.

170
There was also a general tendency by the media to view the 1998 penalty changes in
a negative light. The implementation of the statutory amendments coincided with the
Labour Day holiday weekend in New South Wales and, over this holiday weekend,
NSW Police reported an increase in charges for drink-driving and charges for other
major traffic offences. Several newspapers took this as evidence that the more severe
penalties introduced by the State Government in September had had little impact on
driver behaviour: “Demerits damned” (The Australian, 6th October 1998), “Tougher
laws no deterrent” (Illawarra Mercury, 6th October 1998) and “Shocking behaviour
on roads” (Newcastle Herald, 6th October 1998).

A further point to note here is that the few articles that did describe the penalty
changes for drink-driving tended to focus on penalty changes for other driving
offences, such as speeding. For example, headlines used to describe the 1998 penalty
changes included “Big jump in fines for traffic offences” (The Sydney Morning
Herald, 5th June 1998), “Serious speeders to lose licences” (The Daily Telegraph, 5th
June 1998) and “Speeding points doubled” (The Daily Telegraph, 29th September,
1998) and similarly, the headline used to discuss the impact of these more severe
penalties: “Double demerits damned” (The Australian, 6th October, 1998). Many
drink-driving strategies that have attempted to modify behaviour have stressed the
criminal nature of the act of drink-driving and the criminal conviction that could
result if apprehended. Yet the media consistently packaged drink-driving along with
other non-criminal driving offences. This may lead many people to view drink-
driving as just another traffic offence which results in nothing more than
administrative licence sanctions or monetary penalties.

RTA advertising campaigns and other publicity

The small amount of print media devoted to the drink-driving penalty changes was
reflected in the attention the legislative amendments received elsewhere in the media,
in particular on radio. Data pertaining to radio coverage of the sentencing policy
were obtained from an on-line transcript library maintained by the Australian media
company Rehame. The information that could be gleaned from this database was
extremely limited. The records that needed to be accessed were over five years old

171
and for this reason transcripts were not available from the media agency. However,
summaries of announcements were provided, as were details of the time, date and
station on which the announcement was made. Given the restricted information that
was available, the search was limited to records between January 1998 and March
1999 and ‘drink-driving’ was used as the keyword. This search returned a total of 49
references, only four of which appeared from the summary to be specifically
discussing the 1998 penalty changes in New South Wales. Relevant details for these
four radio announcements are shown in Table 6.5. Although little information was
available on the content of the announcement, the few records retrieved in the search
are indicative of how little publicity the 1998 sentencing policy received in the NSW
media.

Table 6.5: Summaries of radio references to 1998 penalty increase, January 1998 –
March 1999
Date Time Station Announcer Summary
05/05/1998 4:00pm Sydney Radio news The opposition has accused the
2GB government of revenue raising by
lifting the penalties for speeding and
drink-driving. Interview with Shadow
Minister for Roads, George Souris.
05/05/1998 5:13pm Sydney Mike Carlton Discussion on tough new penalties for
2UE driving offences. Interview with NSW
Minister for Roads and Transport, Carl
Scully.
06/05/1998 5:44am Sydney Alan Jones Announcer comments on new penalties
2UE for driving offences
06/05/1998 7:08am Sydney Alan Jones Announcer comments on new penalties
2UE for driving offences

Information was also sought from the NSW Roads and Traffic Authority on any
media campaigns launched to communicate the changes to drink-driving penalties to
the general driver population. The road traffic licensing authorities were aware of
only one publicity campaign (‘Unintended Consequences’) that was run in
conjunction with the penalty increase. Unfortunately, these authorities could only
provide limited information on the campaign’s details (John Bruton, Corporate
Policy & Communications, RTA; personal communication 2002).

This campaign was a two-part television advertisement that attempted to convey the
serious implications arising from driving with a blood alcohol over the legal limit.

172
The first advertisement depicted a male driver who, after drinking with his friends at
a local hotel, drove into a young girl crossing the road. The advertisement was
extremely emotive, showing the child incurring horrific injuries from the crash and
the mother’s subsequent reaction to the injuries sustained by her child. The
advertisement also showed the informal sanctions (e.g. shame and embarrassment)
that can arise from this offence, with the driver having to return home to his wife and
young daughter after the crash to explain what had just occurred. The second
advertisement utilised the same characters and extended the drink-driving scenario to
a courtroom setting. The scene depicted the driver being sentenced by a magistrate
for the drink-driving accident in which the young girl was killed. A gaol sentence
was imposed and, after the magistrate read out the penalty, court officers
immediately escorted the driver out of the courtroom. The scene also showed the
distress of the driver’s wife and young daughter upon hearing that the driver would
be gaoled for the offence.

The second advertisement in this two-part series was shown on NSW television in
late 1998 and early 1999, after the introduction of the new, more severe penalties.
While this publicity would have been relatively successful in conveying the serious
consequences that can arise from an episode of drink-driving, it did not specifically
describe the 1998 penalty changes or convey to the ‘average’ drink-driver the
penalties attached to these offences. Instead, it highlighted the penalties that would
be imposed in what many drivers would see as an extreme scenario. The
consequences depicted in the advertisement may therefore not have had enough
personal relevance to exert a deterrent effect.

What went wrong?

Criminal justice policies that raise statutory maximums often fail to reduce offending
because they are not successfully translated into sentencing practice. This can happen
because statutory maximums are raised but there is no subsequent change in the
actual penalties imposed by the court or because players within the criminal justice
system circumvent the more severe penalties by imposing alternative sanctions. The
1998 statutory amendments to drink-driving penalties did result in larger fines and
longer licence disqualification periods for the majority of drink-drivers convicted

173
after the penalty changes came into effect, yet only a small change in offending
levels was observed.

Over the period in which drink-driving behaviour was measured in the quasi-
experimental study, NSW Police appeared to alter their enforcement tactics from one
where they predominantly deployed stationary RBTs to highly visible locations, to
the more frequent use of “hotspotting” techniques that target high-risk drink-driving
times and locations. The extent to which these changes in enforcement practice
affected potential offenders perceptions of apprehension certainty is unclear.
However, survey evidence suggests that, after 1998, fewer people had seen RBTs in
operation and more people believed that enforcement activity targeting drink-drivers
had decreased. While it is possible that “hotspotting” may have contributed
somewhat to the decrease in recidivism found in Study 2, it is also possible that the
drop-off in police breath tests reduced the perceived risk of apprehension and thus
the deterrent efficacy of the new laws.

The extent to which the penalty changes affected people’s perceptions of sanction
severity is also unclear from the quasi-experiment. As noted earlier, the retrospective
nature of this study made it impossible to measure severity perceptions pre- and post-
intervention. However, the media analysis presented above revealed that very little
publicity accompanied the enactment and the announcement of the drink-driving
penalty increases. Furthermore, the media attention the statutory amendments did
receive was ineffectively communicated. It was packaged with other non-criminal
driving offences and was surrounded by negative publicity regarding the certainty of
severe penalties being imposed by the courts. This, too, may have undermined the
deterrent efficacy of the new penalties.

The current chapter illustrates the difficulties that arise when attempting to test the
deterrence model in a field setting, particularly where the research is conducted
retrospectively and a quasi-experimental methodology must be employed. Attempts
were made to introduce statistical controls but, in a quasi-experimental design, it is
impossible to control for all extraneous variables. Further work is therefore necessary
at the individual level, to strengthen our conclusions regarding the deterrent efficacy

174
of punishment severity. An individual-level analysis will allow for more effective
measurement of the policy-perception-behaviour link, in an experimental setting
where it is possible to control for other contextual influences.

175
CHAPTER 7. KNOWLEDGE OF PENALTIES,
PERCEPTIONS OF SANCTION SEVERITY AND
DRINK-DRIVING

The aim of policies that raise the severity of statutory penalties is often simply to
communicate the fact that severe punishment accompanies violations of the law
(Tonry 1996). Since they endeavour to modify behaviour by influencing perceptions
of sanction severity rather than by imposing harsher punishments these policies are
consistent with deterrence principles. In order to achieve positive outcomes in
relation to offending however, the general public has to be adequately informed of
the legislative changes. As demonstrated in the previous chapter, statutory increases
are often implemented without sufficient media attention or publicity and thus, are
unlikely to affect offender behaviour.

To date, little research has been conducted on factors that influence the formation of
risk perceptions (Nagin 1998). The research that has been conducted focuses mostly
on the role of individual attitudes, previous offending or social influences, rather than
on the effect of criminal justice policy. Study 3 aims to fill this knowledge gap by
investigating whether knowledge of criminal penalties influences perceptions of
sanction risk and, consequently, decisions about whether to offend. Specifically, this
study uses a scenario-based methodology to determine (1) whether people who have
more accurate knowledge of the legal sanctions currently applicable to drink-driving
offences in New South Wales perceive the penalties to be more severe and (2)
whether they say they are less likely to offend, compared with those who have little
or no knowledge of the law.

Study 3 also aims to further elaborate on the findings from the quasi-experimental
study reported in Chapter 5. Study 1 and 2 it will be recalled, found only small
reductions in measures of offending after the drink-driving penalties were increased
in 1998. The media analysis discussed in Chapter 6 suggested that one possible
reason for this weak effect is that the public may have had insufficient knowledge of

176
the penalty changes. If knowledge does influence perceptions of punishment severity
and decisions to offend then inadequate threat communication is a plausible
explanation for the small deterrent effect observed. Alternatively, if there is no such
relationship, the efficacy of raising punishment severity as a method for deterring
offenders remains in doubt.

Formation of risk perceptions

Pogarsky, Piquero and Paternoster (2004) describe deterrence as a two-stage
decision-making process. In the first stage external factors affect a potential
offender’s perceptions regarding sanction risks. In the second, these risk perceptions
affect offending behaviour (see Linkage 1 and Linkage 2 in Figure 7.1). The
perceptual stage is a dynamic process, whereby perceptions of threats are continually
updated on the basis of new information acquired from external sources or from
one’s own offending experiences. The second stage of the deterrence process (the
behavioural linkage) occurs contemporaneously with the offending opportunity. If,
when an offending opportunity arises, present perceptions are that the costs
associated with the act are too great, the expected utility is reduced and deterrence
ensues. Importantly, this two-stage process does not imply that the
perceptual/behavioural linkages are sequential. Instead, the behavioural linkage is
thought to be “nested within” a continual process of perception modification and
evolution (Pogarsky, Piquero & Paternoster 2004). This means that perceptions have
the potential to influence offending behaviour but criminal conduct and its
consequences also have the potential to affect threat perceptions.

Figure 7.1: Two linkages in the deterrence process (from Pogarsky, Piquero &
Paternoster 2004)

Objective properties of punishment

Committing crimes Perceptions of Offending
Consequences of crime punishment behaviour
Other events and experiences
Linkage 1 Linkage 2
Perceptual linkage Behavioural linkage

Pogarsky, Piquero and Paternoster (2004), along with other scholars, note that
deterrence research thus far has focused almost exclusively on the behavioural

177
linkage (Cook 1980; Miller & Iovanni 1994; Nagin 1998) and has neglected to
investigate how perceptions of sanction risks are formed. To address this limitation
of prior deterrence research, Pogarsky and his colleagues (2004) tested the model
shown in Figure 7.1, using data from a panel-based survey of 1,530 high school
students in the USA. This survey consisted of two waves of data collection
conducted approximately 12 months apart. At both survey administrations (Time 1
and Time 2) participants were asked to estimate the likelihood of being caught by
police for three different offences (shoplifting, vandalism, marijuana use). The
number of times participants were arrested between Time 1 and Time 2 was also
measured in the survey, as was offending experiences: including both the number of
times between Time 1 and Time 2 that a participant committed the offence being
contemplated (specific offending) and the number of times a participant committed
other similar offences (general offending). In addition, participants were also asked
about the offending experience of their peers between Time 1 and Time 2. Pogarsky
and his colleagues included this latter measure in the analyses to reflect vicarious
punishment avoidance.

Analysis of these data found some support for the proposed two-stage process. Both
the number of arrests and the extent of peer offending between the first and
subsequent interview waves were found to be significantly associated with changes
in perceived certainty of arrest over that period. The direction of these relationships
was consistent with deterrence predictions. That is, increased arrests between Time 1
and Time 2 were associated with increased perceptions of arrest certainty at Time 2.
Having more peers who had offended, on the other hand, reduced the perceived
likelihood of arrest at Time 2. The extent to which these experiences influenced risk
perceptions also appeared to depend upon the perceived certainty of arrest prior to
the new information being received. Arrest had little impact on individuals who
already had relatively high estimates of punishment certainty at Time 1. It appeared
that punishment avoidance for these offenders was the most influential factor in
altering perceptions. On the other hand, the risk perceptions of individuals who had
low estimates of arrest certainty at Time 1 were more affected by punishment
experiences. Pogarsky, Piquero and Paternoster suggest that this may reflect the
disparity between participant’s expectations of punishment at Time 1 and

178
punishment experience or avoidance in the intervening period. If punishment
experiences are contrary to what they expected then participants will update their
perceptions of sanction risk accordingly.

These findings are generally consistent with other research examining the effect of
offending experiences on perceptions of punishment certainty, several of which were
described in detail in Chapter 2 (see Paternoster & Iovanni 1986; Piliavin et al. 1986;
Horney & Marshall 1992). However, this study only investigated punishment
certainty. It remains to be seen whether the formation of severity perceptions follows
a similar pattern to that observed for perceived certainty of punishment.

Legal knowledge as a factor in the deterrence process

This Chapter focuses on the question of whether knowledge of the statutory penalties
prescribed by the legislature or knowledge of the actual penalties imposed by the
courts influences severity perceptions and subsequent decisions to drink and drive.
The model being used to inform this research was introduced in Chapter 4 and is re-
presented here as Figure 7.2. As indicated previously, not all pathways shown in this
model are being investigated in the scenario-based study (those appearing in bold
text identify the variables that are being measured) and some interactions are
implied. Nevertheless, Figure 7.2 is a useful graphical depiction of the deterrence
hypotheses being investigated in Study 3 and provides a framework for the analysis
and interpretation of data collected from participants.

The two linkages of the deterrence process described by Pogarsky and his colleagues
(2004), namely the perceptual and behavioural linkages, are an additional feature of
Figure 7.2. Because a cross-sectional design is being employed in Study 3, this
research will have to examine the link between knowledge and perceptions (Linkage
1) concurrently with the link between perceptions and behaviour (Linkage 2). This
means that the dynamic process by which risk perceptions are updated on the basis of
new information cannot be measured in this study. Nevertheless, the cross-sectional
approach was considered appropriate given that this is one of the first investigations
of the relationship between legal knowledge, severity perceptions and behaviour
using a scenario-based research design. In the absence of evidence that current legal

179
knowledge affects perceptions of sanction severity, then measuring changes in risk
perceptions over time becomes unnecessary.

Furthermore, Pogarsky and his colleagues (2004) note that perceptions of

apprehension risk may be far more dynamic than perceptions regarding the severity
of formal punishment because certainty perceptions are, by their very definition,
probabilistic. This view is consistent with the work undertaken by Miller and Iovanni
(1994) described in Chapter 3. These authors found evidence from their panel-based
survey that prior use of violence in an intimate relationship had a negative effect on
perceptions of arrest certainty for courtship violence. However, there was no
significant relationship between prior violence use and severity ratings of the
problems arising from an arrest for this type of offence. This would suggest that a
cross-sectional study would be appropriate when investigating factors affecting
perceptions of punishment severity.

Consistent with Pogarsky, Piquero and Paternoster (2004), the proposed deterrence
model depicted in Figure 7.2 is premised on the idea that the objective properties of
formal punishment influence behaviour by modifying perceptions of punishment
severity. The model however includes other factors that may influence an
individual’s knowledge of the law, including previous offending experiences, the
level of offending-risk (as indicated by alcohol consumption levels and involvement
in alcohol-related accidents) and social influences from peers, family and the media.
These potential determinants of legal knowledge, as well as previous research
examining the relationship between legal knowledge and offending, are discussed in
greater detail in the following section.

180
Figure 7.2: Proposed model of the deterrence process

Prior offending/ Prior convictions/

alcohol Knowledge of Media publicity/
consumption/ prior statutory friends, family,
accident penalties acquaintances
involvement

Linkage 1

Informal costs Perceptions of Perceived

arising from penalty severity applicability of
apprehension Linkage 1 statutory penalties

Linkage 2 Linkage 2

Total anticipated formal Decision to Total benefits

costs associated with drink & drive associated with
DD (disutility) Linkage 2 DD (utility)

Linkage 2

Perceived risk of Media publicity/ Moral inhibitions

Linkage 1
apprehension friends, family, associated with
acquaintances drink-driving

Linkage 1 Linkage 1

Enforcement Prior successful drink-

activity driving episodes

181
Previous research
Within a rational choice paradigm of decision-making, knowledge of the law
logically precedes perceptions of sanction severity. As von Hirsch et al. (1999)
observe, “a potential offender can not fear consequences of which he is unaware”
(von Hirsch et al. 1999, p. 46). Previous research examining the level of legal
knowledge held by members of the general public suggests that people are often
poorly informed when it comes to the laws and criminal penalties applicable in their
State, District or County (e.g. Williams, Gibbs & Erikson 1980; Snortum & Berger
1989; Hough & Roberts 1998; Kenkel & Koch 2001). This research seems difficult
to reconcile with drink-driving studies that have found an association between the
introduction of various drink-driving laws, such as preliminary breath test or
mandatory licence disqualification laws, and subsequent reductions in aggregate-
level measures of offending (e.g. Saffer & Chaloupka 1989; Evans, Neville &
Graham 1991; Kenkel 1993; Benson, Rasmussen & Mast 1999). If people are
generally poorly informed about criminal laws and penalties, why does the
implementation of these drink-driving laws reduce offending?

Part of the explanation, Kenkel and Koch (2001) propose, lies in the fact that
particular groups of people are more knowledgeable about the law than others. From
a general population survey conducted in the USA, Kenkel and Koch (2001) found
that those at greater risk of drink-driving, that is heavy drinkers and prior offenders,
knew more about the applicability of drink-driving laws in their State than did other
respondents in the sample. Peer group behaviour was also an important predictor of
knowledge. Individuals who have peers that tend to limit their drinks when driving
know more about drink-driving laws, than those individuals whose peers abstain
from drinking. Thus even if the general public has a poor knowledge of drink-driving
laws, those who are being targeted by such legislation demonstrate a better
understanding of the legal consequences of drink-driving.

Although Kenkel and Koch (2001) identified potential factors influencing legal
knowledge, they failed to consider whether those who are more knowledgeable about
the laws are less likely to offend. Previous research examining the relationship

182
between legal knowledge and offending behaviour provides mixed results. Williams,
Gibbs and Erickson (1980) found some evidence for an inverse relationship between
perceived statutory maximums and aggregate-level crime rates. These authors
surveyed 2,400 adult residents in Tuscon, Arizona, on their perceptions of the
maximum and the applicability of five different penalty types (e.g. whether or not a
person can be imprisoned or whether or not a person can be fined) for nineteen
different types of crime. These perceptions were found to be moderately correlated
with actual maximums stipulated by the law. Furthermore, the analysis showed
strong negative correlations between perceived statutory maximums and official
crime rates in that area, suggesting that enhanced public knowledge of criminal
penalties is associated with reduced criminal activity.

However, further analyses of these data revealed that, when a measure of public
disapproval of these crimes was included in the analytical model, the significant
relationship between perceived maximum penalties and crime rates disappeared
(Williams & Gibbs 1981). This, the authors propose, demonstrates that the inverse
relationship between perceived maximums and actual crime rates is not due to
deterrence, but rather the shared normative values of the public and legislators in
relation to these types of crime. It should be noted here, however, that the data used
in this analysis came from a survey examining a range of criminal offences, most of
which were serious (e.g. first-degree murder, first-degree rape and aggravated
assault). Serious offences have traditionally been thought of as less susceptible to
deterrence-based initiatives because they are already supported by the moral code
society (Andenaes 1966; Zimring & Hawkins 1973). Yet no distinction is made in
the analysis between these crimes and others that are more instrumentally motivated
(e.g. petty theft) and thus more likely to be affected by formal sanctions.
Furthermore, as discussed in Chapter 3, ecological designs such as this suffer from
several methodological weaknesses, including reverse causality, which preclude
confident conclusions regarding deterrent effects.

Homel’s longitudinal study of drink-drivers in NSW (previously referred to in

Chapter 2; Homel 1986) overcomes the limitations of William et al.’s ecological
research and presents some support for the notion that improved legal knowledge can

183
result in modifications to behaviour, particularly for those who are more likely to
offend. When RBT was introduced in NSW a high-range drink-driving offence that
carried harsh penalties was also introduced. To examine the impact of this offence,
Homel (1986) asked drivers whether they knew about the penalty for this offence and
also asked about the number of modifications that they had made to their driving
behaviour since its introduction. He then reinterviewed these drivers six weeks later
to measure any changes in drink-driving behaviour during the follow-up period. His
results indicated that those who knew about the penalty increase were not only more
likely to modify their drinking and driving behaviour at the first interview but were
also less likely to drink and drive during the six-week follow-up period (this was
independent of the effects of RBT on drink-driving behaviour). Furthermore, these
changes in drink-driving behaviour were more apparent for respondents who had a
previous conviction for drink-driving than for those without a conviction.

Despite this initial evidence for a deterrent effect, Homel (1986) also found that two
other measures of perceived punishment severity used in the survey did not
significantly predict changes in offending behaviour. He asked participants to
imagine the punishment they would receive if caught for drink-driving and then rate
(on a five-point scale) how big a problem this punishment would be for them. He
also asked participants to rate their chances (on a five-point scale) of receiving no
penalty from the court if arrested for drink-driving. Logistic regression analyses
found neither of these indicators to be significant predictors of having driven in the
intervening period between the two interviews.

This latter finding seems somewhat surprising, given the significant relationship
between knowledge and behaviour, however it should be noted that just 12 (6.9 per
cent) out of the 175 participants reinterviewed in the second survey reported
offending during the follow-up period. This low base rate would reduce the power to
detect significant effects in the regression analysis and could explain why the study
failed to find a relationship between the measures of severity perceptions and
behaviour. A further possibility is that the categorical variable used in the analysis to
measure perceived severity may not have sufficiently captured individual variation in
severity perceptions. A continuous measure of severity perceptions as is sometimes

184
used in deterrence research to measure perceived likelihood of punishments, would
improve the power of the analysis to detect significant effects if they exist.

The scenario-based study - Study 3

The primary aim of Study 3 is to examine the policy-perception-behaviour link and
in doing so, address an important knowledge gap identified in the deterrence
literature. Specifically this research investigates the relationship between knowledge
of the drink-driving penalties currently applicable in New South Wales, perceptions
of punishment severity and willingness to offend in a given set of circumstances.
This research will also allow for more confident conclusions regarding the deterrent
efficacy of punishment severity to be drawn from the results of the case study.
Chapter 6 suggested that a possible reason for the small effect of the 1998 penalty
changes was that the sentencing policy was inadequately publicised. Examining the
influence of current knowledge of drink-driving penalties on willingness to offend
will allow us to determine the extent to which this account for the small observed
deterrent effect is valid. This research also builds on the quasi-experimental research
in that it allows the level of apprehension risk to be experimentally manipulated
across groups. Chapter 6 demonstrated that the level of enforcement activity
associated with drink-driving offences in NSW altered over the period in which
drink-driving was measured for the quasi-experimental study. This modification to
RBT activity could have potentially affected offending behaviour at the same time
that the penalties were raised. In this scenario-based research, however, the effect of
punishment severity on offending decisions can be examined at various levels of
perceived certainty of apprehension.

Hypotheses
As outlined in Chapter 4, the scenario-based study was designed to test four specific
hypotheses generated by deterrence theory:

1. Increased perceived risk of arrest reduces offending likelihood

To investigate whether police practice can influence behaviour through altering
perceptions of arrest certainty, the level of RBT activity was varied across scenarios.
The three conditions of punishment certainty varied in the scenario were: ‘You

185
haven’t seen police conducting Random Breath Tests (RBT) in this area for some
time’ (low-level of enforcement activity); ‘You have often seen police conducting
Random Breath Tests (RBT) in this area’ (medium-level); and ‘You have often seen
police conducting Random Breath Tests (RBT) in this area and since it is a Friday
night you know that there will be more police on the roads’ (high-level). Deterrence
theory would anticipate that participants allocated the scenario with a high-level of
police activity would be less likely to state that they would offend under the
circumstances described.

2. Enhanced knowledge of drink-driving penalties reduces offending likelihood by

increasing perceptions of sanction severity
Natural variation in levels of legal knowledge was exploited in Study 3 by way of a
series of multiple choice questions asking about the penalties for low- and mid-range
drink-driving offences. These questions pertained to both the maximum penalties
prescribed by the law and the actual penalties the individual would receive from the
court if found guilty of drink-driving. We would anticipate that individuals who are
more knowledgeable about either actual or statutory drink-driving penalties, would
be less likely to state they would offend in the given scenario. Within a deterrence
framework, legal knowledge can only affect behaviour by altering perceptions of
punishment severity. For this reason, we would expect a significant relationship, not
only between knowledge of penalties and drink-driving behaviour, but also between
legal knowledge and severity perceptions.

3. Increased perceptions of sanction severity reduces offending likelihood

Perceptions of sanction severity were measured by asking respondents to indicate
‘how much of a problem’ the penalties would create for them personally. This
question has been employed in previous perceptual research (Grasmick & Bryjak
1980; Homel 1986; Nagin & Paternoster 1993) in order to capture the subjective
nature of severity perceptions. However, unlike previous studies, participants in
Study 3 are asked to rate their perceptions regarding penalty severity on a scale of 0
(no problem) to 100 (a very big problem). This continuous variable was employed in
an attempt to measure more accurately the range of severity perceptions held by
survey participants. Again, consistent with deterrence predictions, we would expect

186
that participants who perceive the penalties as more severe to state that they are less
likely to offend given the conditions described in the scenario.

4. Perceptions of sanction severity matter more when punishment is perceived to be

certain
Study 3 also formally tests the interaction between certainty and severity by
including an interaction term in the regression models predicting stated offending
likelihood. One condition thought necessary to maximise the deterrent efficacy of
severe sanctions is a relatively high-level of perceived risk of detection and
apprehension. If this is true, then we would anticipate that perceptions of sanction
severity matter most in the offending decisions of participants allocated the scenario
with a high-level of police activity targeting drink-driving.

The model depicted in Figure 7.2 also describes several other factors that may
influence the formation of risk perceptions and decisions to offend. In particular, it
includes extralegal consequences arising from apprehension for a drink-driving
offence as an independent factor affecting the disutility of the crime. This is
consistent with Williams and Hawkins (1986) proposal that informal social controls
arising from arrest and conviction should be included as “part of the general
deterrence process” (Williams & Hawkins 1986, p. 561). This additional influence
on offending decisions is not, however, investigated in Study 3. Much of the
perceptual research that has previously been undertaken in the deterrence domain has
examined the influence of informal social controls on offending decisions. This
research has made an important contribution to the deterrence literature by
demonstrating that informal social sanctions, under some conditions, can operate like
legal threats in preventing offending (e.g. Grasmick & Bursik 1990; Grasmick,
Bursik & Arneklev 1993; Baum 1999).70 However, perceptual research has said little
about the policy-behaviour link upon which deterrence-based criminal justice
interventions are modelled. Unless a relationship exists between actual sentencing
policy and offending behaviour then the efficacy of raising the threat of formal
punishment to prevent future offending remains in doubt. This area of research is one

70
Though it should be noted that moral inhibitions associated with offending are accounted for in the
analysis by excluding individuals who would not offend even where there is no possibility of
punishment if caught.

187
of the gaps in the deterrence literature identified by Nagin (1998) and one that needs
further exploration given the results obtained from the quasi-experimental study
described in Chapter 5.

Results of Study 3
The following sections present the results of Study 3. Sample demographics and
measures on the independent variables relevant to our hypotheses are initially
described and then bi-variate relationships between these independent variables and
the dependent variable (i.e. the stated likelihood from 0 to 100 of driving home under
the prescribed conditions) are examined. Following this, the results of a linear
regression analysis (estimated using Ordinary Least Squares) in which the
independent variables were regressed against the dependent variable are presented
and any findings pertinent to the deterrence hypotheses being tested are discussed.
Finally, the assumptions underpinning the estimated regression models are tested and
further necessary exploratory analyses are undertaken.

Sample demographics
The sample for Study 3 consisted of 445 participants recruited from various legal,
engineering and general studies classes at the University of New South Wales. Just
over half (56 per cent) of the participants recruited for this study were female. The
average age of participants was 21 years, and 91 per cent were aged between 18 and
25 years. Twenty-six per cent of the sample reported that they had, on at least one
occasion, driven after drinking too much alcohol but only three participants reported
a previous conviction for a drink-driving offence. One in ten participants had
themselves been involved in an alcohol-related crash or knew a relative who had.
Over one-quarter of participants reported that they had consumed alcohol at an acute-
risk level71 at least once weekly in the previous 12 months and 10 per cent reported
drinking at these levels on more than one occasion every week.

71
Drinking alcohol at or above the Australian low-risk guidelines for acute harm, that is, more than
four standard drinks for females and more than six standard drinks for males (National Health and
Medical Research Council 2001)

188
Independent variables
Perceptions of apprehension certainty
As described previously, respondents were randomly assigned one of three scenarios
varying in the level of RBT activity described; low (n=146), medium (n=148) and
high (n=150). Figure 7.3 shows the mean estimates of apprehension certainty across
these three groups. There was a significant difference in perceived certainty of
apprehension between the low- and high-risk groups (t=-2.140, df=235, p=0.033), as
well as between the low and medium groups (t=-3.147, df=241, p=0.002), but no
significant difference between the medium and high groups was found (t=0.822,
df=240, p=0.412). For this reason the medium and high groups were combined in
later analyses. The participants assigned to these two groups showed no significant
differences in terms of age, gender, alcohol consumption, prior offending or prior
accident involvement.72

Figure 7.3 Mean estimates of certainty of apprehension and

conviction by scenario group

50
45
44.5
Av. perceived certainty of

43.7
40
35
apprehension

34.8
30
25
20
15
10
5
0
low medium high

Scenario

Knowledge of statutory penalties

Generally, participants had a good understanding of the types of penalties that are
prescribed by legislation for drink-driving offences according to the recorded BAC

72
Age, t=-0.986, df=362, p=0.325; gender, x2=0.547, df=1, p=0.459; alcohol consumption, t=0.693,
df=363, p=0.489; prior offending, t=-0.641, df=364, p=0.522; accident involvement, t=-0.508, df=364,
p=0.612.

189
level, but did not have a good understanding of statutory maxima and minima
relevant to particular offences.

Knowledge of the applicability of fines was high, with over 95 per cent of the sample
correctly reporting that a person could be fined for a low- or mid-range drink-driving
offence in NSW. However at least half of the respondents underestimated the
maximum fine for both these offences. Most participants knew that gaol terms were
not applicable to low-range drink-driving offences but 43 per cent incorrectly
believed that gaol terms were not applicable to mid-range drink-driving offences.
The results from the survey, moreover, suggest that knowledge of mandatory
minimum licence disqualification periods for drink-driving offences was quite low.
Fifty-five per cent of participants did not know about minimum disqualification
periods for low-range drink-driving offences (minima that were introduced in NSW
as part of the 1998 legislative changes to driving penalties) and 35 per cent did not
know about the minima for mid-range drink-driving offences.

Participants were assigned a score based on the number of errors they made in
relation to questions on statutory penalties for drink-driving offences (mistake1). A
low score represented few errors and thus good knowledge of the applicable statutory
penalties, while a high score represented many errors, or poor knowledge. The
frequency distribution for these scores is shown in Figure 7.4.

190
 Figure 7.4: Frequency distribution of scores on
knowledge of statutory penalties

50
45
40
35
No. of persons

30
25
20
15
10
5
0
0

8
10

12

14

16

18

20

22

24

26

28
mistakes

Knowledge of actual penalties

As with statutory penalties, a large proportion of participants (57 per cent and 34 per
cent for low-range and mid-range respectively) underestimated the actual average
fine imposed on people convicted of drink-driving offences. Most, however, knew
that they would be fined (98 per cent for both low- and mid-range) and knew that
they would not be gaoled (90 per cent low-range; 73 per cent mid-range). On the
other hand, a substantial proportion of participants (40 per cent) mistakenly thought
that they would not have their licence disqualified for a low-range drink-driving
offence. For a mid-range drink-driving offence, nearly all of the participants (98 per
cent) correctly thought that their licence would be disqualified, but only one-third
reported correctly that the average disqualification period would be nine months or
more. Again, based on the answers to these questions, a knowledge score (mistake2)
was calculated, the frequency distribution of which is shown in Figure 7.5.

191
 Figure 7.5: Frequency distribution of knowledge of actual
penalties

60

50

40
No. of persons

30

20

10

0
10

12

14

16

18

20

22

24
0

mistakes

Perceptions of penalty severity

Figure 7.6 displays the frequency distribution of the ratings of penalty severity for all
participants. The distribution of severity ratings shown here suggests that the vast
majority of participants perceived the penalties to be very severe, with almost one-
third rating the severity of drink-driving penalties at 100. Only three respondents
reported that the penalties would be no problem at all for them.

192
 Figure 7.6: Frequency distribution of severity scores

140

120

100
No. of respondents

80

60

40

20

0
0 10 20 30 40 50 60 70 80 90 100
Severity of penalties

Predicting stated likelihood of offending

Estimates of stated offending likelihood
When presented with the drink-driving vignette, almost one-third of participants
stated that they would not drive home under the circumstances described but almost
one-quarter stated that there was a 50 per cent chance or greater that they would do
so. If there were no possibility of being caught and punished for the offence, 18 per
cent of participants said that they would still not drive home, but 18 per cent said
they would definitely drive home. Respondents who would not drive home even if
there were no chance of being caught or convicted can be considered a group who is
morally opposed to drink-driving (Nagin & Pogarsky 2001). This group was
therefore excluded from further analyses (79 respondents in total). The frequency
distribution of the dependent variable (excluding these “nondeterrable” respondents)
is shown in Figure 7.7. The average reported offending likelihood in this reduced
sample was 37 per cent, with a median of 30 per cent (SD=32.1).73

73
It is worth noting that these estimates of reported offending likelihood are slightly higher than those
obtained by Nagin and Pogarsky (2001) in their scenario-based drink-driving study using a student
sample from a US university (av. 31%; median 20%).

193
 Figure 7.7: Frequency distribution of offending likelihood
estimates

70
60
50
No. of persons

40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Likelihood of offending

Zero order correlations

Table 7.1 shows descriptive measures and zero order correlations for the variables of
interest in the study. As seen from this table, there are a number of significant
bivariate relationships between the dependent and the independent variables and
among the independent variables themselves. Firstly, there is a moderate, positive
correlation between mistakes made with regard to statutory sanctions for drink-
driving offences and mistakes made with regard to the actual penalties imposed by
the courts. However, only the latter variable has a significant positive correlation
with stated offending likelihood. This indicates that those who made more mistakes
when asked about the actual penalties drink-drivers receive upon were more likely to
offend in the scenario. There are also significant relationships between stated
offending likelihood and the two deterrence variables measured in the survey. The
direction of these zero-order correlations is negative, indicating that persons who
were assigned to the low certainty of apprehension group and those who perceive the
penalties to be less of a problem are more likely to report that they would drink and
drive under the circumstances described in the scenario.

194
Earlier we hypothesised that knowledge of penalties would influence offending
likelihood through its impact on perceptions of penalty severity. Consistent with this
prediction, the zero-order correlations between ratings of sanction severity and both
knowledge scores are negative. Thus, the more mistakes a respondent made on the
questions tapping knowledge of drink-driving penalties in NSW, the lower the rating
of punishment severity. Note, however, that these correlations are weak and only the
relationship between severity perceptions and knowledge of actual penalties is
statistically significant at the 0.05 level.

There is a positive correlation between prior drink-driving episodes and three other
independent variables; age, number of alcohol-related crashes and frequency of high-
risk drinking. Participants who report having frequently driven after drinking too
much alcohol were older, had been involved in more alcohol-related crashes and
drank alcohol at high-risk levels more frequently. We would expect those who drink
and drive more often to be involved in more alcohol-related crashes and to drink
alcohol more frequently at high-risk levels. The positive correlation between age and
prior offending may seem contrary to the results of many criminological studies, but
it should be remembered that the prior offending variable in this study measures the
number of occasions ever driven whilst drunk. Older people in this sample would
have had more opportunity to offend than younger persons and would therefore be
more likely to record numerous drink-driving episodes.

Prior offending also has a significant positive correlation with stated likelihood of
offending in the hypothetical scenario. This indicates that those who had previously
offended were also more likely to state that they would offend in the circumstances
described in the scenario. While not surprising, this latter finding does attest to the
truthfulness of participant responses on the dependent variable.

The number of previous drink-driving episodes is negatively correlated with both of

the variables measuring legal knowledge. This suggests that persons who offend
more often know more about the statutory penalties applicable to drink-driving
offences (mistake1) and about the court penalties imposed on drink-drivers
(mistake2). A similar relationship is evident for levels of alcohol consumption and

195
legal knowledge, with those who frequently drink at high-risk levels making fewer
mistakes about drink-driving penalties. These findings are consistent with the results
from Kenkel and Koch’s (2001) research, which showed that people who had
previously driven whilst drunk knew more about illegal per se laws and also that
heavy drinkers make fewer mistakes about laws related to licence suspension.

Table 7.1: Zero order correlations between independent and dependent variables
(n=366)
x1 x2 x3 x4 x5 x6 x7 x8 x9 y
Age (x1) 1.0
Gender (x2) -.0 1.0
Prior .21* -.07 1.0
offending
(x3)
Accident .07 -.03 .17* 1.0
involvement
(x4)
Alcohol .14 -.04 .46* .11* 1.0
Consumption
(x5)
Mistake1 (x6) -.02 .02 -.16* .03 -.24* 1.0
Mistake2 (x7) -.03 .01 -.16* .03 -.28* .38* 1.0
Certainty gp .05 -.04 -.01 -.01 -.04 -.01 .03 1.0
(x8)
Severity (x9) .05 .01 .04 -.11* .11* -.09 -.18* -.01 1.0
Offending .03 .03 .23* .07 .03 .10 .16* -.12* -.11* 1.0
likelihood
(y)
Mean 21.11 .52 1.58 .16 2.13 13.44 7.19 .67 76.24 37.0
Standard 3.34 .50 5.18 .58 1.76 5.21 3.14 .47 24.09 31.7
Dev.
* Significant at the 0.05 level
** Where ordinal variables are correlated spearman’s correlation coefficient is presented

Linear regression models including knowledge and certainty variables

To predict stated likelihood of drink-driving in the scenario, controlling for the
influence of the factors appearing in Table 7.1, multivariate analyses were
undertaken using Ordinary Least Squares regression. Three separate models were
estimated and the details of these models are shown in Table 7.2. Model 1 includes
the composite variable measuring knowledge of statutory penalties, Model 2 includes
the composite variable measuring knowledge of actual penalties and Model 3
includes a categorical variable measuring estimates of perceived penalty severity. All

196
three models also contain a dichotomous variable measuring perceived certainty of
apprehension (low vs. high) and measures on other relevant control variables.

Table 7.2: Linear regression models for stated likelihood of offending

Model 1 Model 2 Model 3
b p b p b p
CERTAINTY
High v. low -10.17 < 0.01 -10.31 < 0.01 -10.32 < 0.01
(3.39) (3.37) (3.42)
KNOWLEDGE OF PENALTIES 0.88 < 0.01
(Statutory penalties) (0.32)
KNOWLEDGE OF PENALTIES 2.04 < 0.01
(Actual penalties) (0.53)
SEVERITY
Medium v. Low -6.28 0.11
(3.86)
High v. Low -7.78 0.06
(4.06)
DRINK-DRIVING FREQUENCY 2.37 < 0.01 2.39 < 0.01 2.23 < 0.01
(no. of times) (0.47) (0.46) (0.47)
RISKY ALCOHOL
CONSUMPTION
Monthly v. Rarely/never 5.21 0.18 5.72 0.14 4.26 0.27
(3.89) (3.81) (3.88)
Weekly v. Rarely/never -3.46 0.42 -2.00 0.64 -5.16 0.22
(4.24) (4.23) (4.23)
ALCOHOL-RELATED 0.61 0.83 0.41 0.88 0.72 0.80
ACCIDENT FREQUENCY (no. (2.80) (2.77) (2.82)
of times)
AGE -0.18 0.71 -0.16 0.74 -0.11 0.83
(0.49) (0.48) (0.49)
GENDER -0.99 0.76 -0.82 0.79 -0.75 0.82
(3.20) (3.17) (3.22)
Notes: Standard error of estimates appear in parentheses. For all models n=364 (note that two cases
had missing data for age).

Perceptions of apprehension certainty

As seen in Table 7.2, the coefficient for perceived certainty of apprehension is
negative and statistically significant (at α=0.01) in all three models. The magnitude
of this certainty effect is such that respondents who received the high-risk of
apprehension scenario were, on average, 10 per cent less likely to report that they
would drive home than were respondents who received the low-risk scenario.

197
Knowledge of penalties
Table 7.2 also shows significant positive coefficients for both knowledge variables.
This indicates that the more mistakes respondents made with regard to the drink-
driving penalties, the more likely they were to say they would offend in the
hypothetical scenario, controlling for other variables in the model. In the case of
statutory penalties, an increment of five on the knowledge scale increases the stated
probability of drink-driving by a little over four per cent. An increment of five on the
knowledge scale for actual penalties increases the stated probability of drink-driving
by 10 per cent. The difference in regression coefficients for the knowledge variables,
as well as the difference in variance accounted for by Models 1 (R2=0.106) and 2
(R2=0.124), indicates that knowledge of actual penalties is a better predictor of stated
offending likelihood than knowledge of statutory penalties.74

If the knowledge effect found in the above analysis was due to deterrence then we
would also expect to see a significant relationship between legal knowledge and
perceptions of punishment severity. Table 7.1 shows that there were in fact negative
correlations between ratings of penalty severity and both knowledge scores, however
only the relationship between severity perceptions and knowledge of actual penalties
imposed on drink-drivers was statistically significant. This accords with the
regression models shown in Table 7.2, which suggest that knowledge of actual
penalties imposed for drink-driving is a better predictor of reported offending
likelihood than is knowledge of statutory sanctions.

Perceptions of penalty severity

Main effects for severity perceptions on stated offending likelihood were also
considered. Due to the highly skewed distribution of the severity estimates, it was not
possible to include severity perceptions in the linear regression analysis as a
continuous variable. Thus, a new categorical variable was created for this analysis
which took one of three values: low (0-70), medium (71-90), and high (91-100).
Approximately one-third of respondents in the sample fell into each of these three
categories. The average reported likelihood of offending for each of these severity

74
Achen (1982) also suggests looking at the b coefficient multiplied by the mean for the
corresponding independent variable to determine the level of importance of variables in the regression
analysis. This calculation also found knowledge of actual penalties to be a better predictor.

198
groups across the two certainty conditions is presented in Figure 7.8. This figure
shows that respondents assigned to the low-risk of apprehension scenario were more
likely to state that they would offend than respondents assigned to the high-risk
condition. Conversely respondents who rated the penalties as a ‘very big’ problem
were less likely to state that they would offend compared to those who considered
the penalties to be less of a problem. Inclusion of these subjective ratings of penalty
severity in the linear regression models described previously produced a small effect
for severity group membership (see Model 3 in Table 7.2). Respondents in both the
medium and high severity groups reported that they were less likely to offend in the
scenario than those in the low severity group (medium group b=-5.49, p=0.15; high
group b=-7.51, p=0.07) but only the difference between the high severity and the low
severity groups approached significance at the 0.05 level.

Figure 7.8: Likelihood of offending by perceptions of

penalty severity and apprehension risk

50.0
45.0
Av. likelihood of offending

40.0
35.0
30.0
25.0
20.0
15.0
low risk
10.0
high risk
5.0
0.0
low medium high
perceptions of severity

Interaction between certainty and severity

The interaction between punishment certainty and severity is also of importance to
our hypotheses. Perceptions of punishment severity, in other words, should matter
more in offending decisions when punishment is highly probable. Looking back at
Figure 7.8 we see that, on average, participants allocated to the low-level of RBT
activity scenario stated, that they were more likely to offend in the circumstances
described than those allocated to the high-level scenario. Participants in both these
groups who perceived the penalties to be less of a problem stated that they were more

199
likely to offend. However, if severity were more likely to matter at high levels of
punishment certainty than we would expect to see divergence of the two lines shown
in this figure with the severity effect being greater for the high-risk of apprehension
group. As seen from this graph the two lines are parallel. Moreover when product
terms for the severity and certainty groups are included in the regression models
there is no evidence for a significant interaction effect.75 These findings are
inconsistent with the assumption that the relationship between perceived risk of
apprehension and perceived punishment severity is multiplicative.

Testing the assumptions of linear regression

The regression models described above show a significant relationship between
knowledge of drink-driving penalties (both actual and statutory penalties) and stated
offending likelihood, as well as between the perceived risk of detection and stated
offending likelihood. While these two findings are supportive of our deterrence
hypotheses, they need to be treated with some caution because the residual diagnostic
plots for all three models indicate lack of homogeneity of variance. These plots
suggest that the models tend to overestimate low offending probabilities and
underestimate high probabilities.76 This violates the assumption of linear regression
that the residuals should be dispersed randomly throughout the range of the estimated
dependent variable.77 This assumption violation could mean that (1) there is an
interaction effect not included in the model or (2) independent variables predictive of
the outcome measure have been omitted from the model.

75
The interaction between certainty and severity was modelled using the following regression
equation; Y = α + β 1 X 1 + β 2 X 2 + β 3 X 3 + β 4 X 1 X 2 + β 5 X 1 X 3 . . . + β k X k , where Y is the stated
likelihood of offending in the scenario, X1 is certainty group (high v. low), X2 is a binary variable
representing inclusion in the medium severity group (SEVERITY1), X3 is a binary variable
representing inclusion in the high severity group (SEVERITY2) and X 1 X 2 and X 1 X 3 are interaction
terms for GROUP*SEVERITY1 and GROUP*SEVERITY2. Other predictor variables were
also included in the regression model. The final model found that neither of the two interaction terms
in this model were significant (p=0.816; p=0.696, respectively). The incremental change in R2 was
also small, increasing from 0.097 to 0.098.
76
Note that similar results were evident even after excluding respondents who reported a zero
likelihood of offending in the scenario.
77
Transformation of the dependent variable did not improve the model fit. However, similar results to
those from the linear regression analysis were found from a logistic regression analysis where the
dependent variable was 50% or greater chance of offending in the given scenario. This type of
analysis is not subject to the same assumptions as a linear regression analysis.

200
To investigate the first of these possibilities further exploratory analyses were
conducted, partitioning out the effect of the punishment variables for different sub-
samples. Given previous evidence in the literature that people who are at greater risk
of offending are more sensitive to changes in punishment variables (e.g. Homel
1986; Decker, Wright & Logie 1993; Wright et al. 2004), high-risk drinkers and
those reporting prior offending episodes were examined separately. The mean plot
graphs shown below provide some evidence for an interaction effect between risky
alcohol consumption and perceived risk of apprehension (Figure 7.9), as well as
between prior offending and perceived risk of apprehension (Figure 7.10). The
relationship is such that the effect of the perceived certainty of punishment on stated
offending likelihood appears to be more pronounced for the high-risk groups (i.e.
frequent risky drinking and greater than five previous offending occasions).

Figure 7.9: Likelihood of offending by certainty group

and drinking status

60 never/not often (n=201)

monthly/1 a wk (n=119)
50 frequently (n=45)
Av. likelihood of offending

40

30

20

10

0
low medium

certainty group

201
 Figure 7.10: Likelihood of offending by certainty and
prior offending episodes

100 no priors (n=257)

90 1-5 priors (n=88)
80
>5 priors (n=21)
Av. likelihood offending

70
60
50
40
30
20
10
0
low high
Certainty group

To test the interaction between drinking status and the deterrence variables measured
in the survey, the sample was divided into three groups, based on self-reported
frequency of drinking. Separate regressions of the dependent variable (i.e. stated
offending likelihood) on the remaining independent variables were then carried out.78
The results of these analyses are shown in Tables 7.3 and 7.4 (note that non-
significant control variables have been removed). The most notable feature of both
these tables is that Model 3, which includes only participants who report drinking at
acute-risk levels on a weekly basis, accounts for substantially more of the variance in
the dependent variable than do Models 1 or 2 (i.e. R2=0.32 v. R2=0.06 & R2=0.12,
respectively). Furthermore, examination of the residual plots for Model 3 reveals no
serious departures from linear regression assumptions, indicating that the model is a
relatively good fit for stated likelihood of offending.

78
Including dummy variables in the regression equations to represent this interaction increased R2
from 0.12 to 0.14, but tests of this incremental change were non-significant (F=1.64 < Fc=2.37).
However, given that there were only small numbers of persons in the high-risk group (n=45) it is
possible that there was insufficient power to detect a significant interaction effect using this approach.

202
Table 7.3: Linear regression models for stated likelihood of offending specified for
different drinking groups
Model 1 Model 2 Model 3
(rarely/never n=201) (regular n=119) (frequent n=45)
b p b p b p
CERTAINTY
High v. low -7.12 0.13 -10.10 0.08 -25.34 0.01
(4.69) (5.77) (9.73)
MISTAKE1 (STAT. 1.23 < 0.01 0.07 0.91 1.50 0.16
PENALTIES) (0.41) (0.57) (1.05)
DRINK-DRIVING 1.75 0.15 2.38 < 0.01 2.29 < 0.01
FREQUENCY (no. of times) (1.20) (0.66) (0.68)
R-square 0.06 0.12 0.32
Notes: Standard Error of estimates appear in parentheses

Table 7.4: Linear regression models for stated likelihood of offending specified for
different drinker groups
Model 1 Model 2 Model 3
(rarely/never n=201) (regular n=119) (frequent n=45)
b p b p b p
CERTAINTY
High v. low -7.43 0.11 -9.49 0.09 -22.76 0.02
(4.66) (5.63) (9.34)
MISTAKE2 (ACTUAL. 2.50 < 0.01 1.62 0.06 2.02 0.17
PENALTIES) (0.74) (0.87) (1.44)
DRINK-DRIVING 1.88 0.12 2.53 < 0.01 2.21 < 0.01
FREQUENCY (no. of times) (1.19) (0.65) (0.68)
R-square 0.07 0.14 0.32
Notes: Standard Error of estimates appear in parentheses

In terms of the independent variables, we see that the effect for punishment certainty
is much larger for frequent risky drinkers than it was for the full sample. Model 3
shows that frequent drinkers who were assigned to the high-risk of apprehension
scenario were, on average, 25 per cent less likely to state that they would offend than
frequent drinkers assigned to the low-risk scenario. However, knowledge of penalties
is no longer a statistically significant predictor of stated offending likelihood for this
subgroup. This could be because there is an interaction between drinking status and
knowledge, such that knowledge is more predictive of offending at low-frequency
drinking. Alternatively it could be due to the small sample size (n=45) (i.e. there is
insufficient power to detect an effect). Since the beta coefficients for the knowledge
variables in Model 3 are similar to those reported for the full sample, the latter
explanation would seem to be the most accurate interpretation of this result.

203
In Models 1 and 2, R2 is relatively low and the variable representing perceived risk
of apprehension is not significant in the case of the low-risk drinkers. Furthermore,
the residual plots for both of these models display a similar pattern to that seen for
the full sample, that is overestimation of the low offending probabilities and
underestimation of high probabilities. This suggests that it may not be fear of
punishment that is influencing the offending decisions for these low-risk groups but
some other variable that has been omitted from the model.

Similar results as those described above were found when the sample was separated
on the prior offending variable. That is, the effect of perceived risk of apprehension
on stated offending likelihood was much greater for those participants who had
driven home on numerous occasions after drinking too much alcohol than for those
who rarely offended. The variance accounted for by the regression models was also
greater for the frequent offender group and the residual plots showed no serious
violations of regression assumptions. To minimise repetition, however, the linear
regression models for these supplementary analyses have been placed in Appendix
III.

Summary and discussion of survey findings

Study 3 confirms previous research findings in showing that perceived certainty of
apprehension can affect the likelihood of offending. Simply knowing that RBTs have
been operating within the immediate area on a frequent basis is enough to reduce the
chances that someone will drink and drive. While there was an overall effect of
apprehension risk it is also worth noting that there was no difference in perceived
certainty of apprehension between the medium- and high-risk scenarios. The high-
risk scenario attempted to mimic a police blitz or hotspotting operation that tasked
more police to the job of breath testing drivers on weekends. Contrary to expected
utility predictions, this increase in police numbers did not significantly raise
perceptions of apprehension risk any more than that created by frequent sightings of
police in the area.

This unexpected finding can perhaps be explained by Sherman’s (1990) idea of

“ambiguity aversion” – in situations of uncertainty people will tend to play it safe

204
and overestimate the risks associated with getting caught. Since the day of the week
was not identified in the medium-risk scenario, respondents may have assumed the
worst and estimated the risk as if it were a weekend and more police would be
patrolling. This suggests that keeping potential offenders guessing about the certainty
of being breath-tested on any one day may be just as useful in drink-driving
enforcement as increasing the actual risk of being tested at certain high-risk times
and locations.

A further finding from this scenario-based research is that punishment certainty

appears to be more of a deterrent for people who are at greater risk of offending. This
result is consistent with previous research conducted in the deterrence domain,
including studies that have focused on drink-driving. For example, Grosvenor,
Toomey and Wagenaar (1999) conducted a cross-sectional survey of adolescents
(12th grade students) asking them about their drink-driving behaviour in the previous
30 days. These students were also asked the penalty they thought they would receive
if caught drink-driving, the chances of being caught by the police for drink-driving
and their drinking behaviour in the previous two weeks. Controlling for other
individual demographic characteristics, perceived certainty of apprehension was only
an effective deterrent for ‘binge drinkers’ (i.e. those reporting drinking five or more
drinks on one occasion in the previous two weeks). Among non-binge drinkers there
was no relationship between drink-driving and perceived certainty of apprehension
for an offence. Similarly, in our sample, participants allocated the high-risk of
apprehension scenario were, on average, 10 per cent less likely to state they would
offend. However, for participants who frequently drink alcohol at high-risk levels,
the reported offending likelihood was reduced by up to 25 per cent in the scenario
where the perceived risk of apprehension was high.

Data from this study also provide some support for the notion that increased
knowledge of applicable penalties is associated with reduced offending likelihood
under scenario-based conditions. The regression analyses described above showed
that participants who made more mistakes about the penalties applicable to drink-
driving offences in New South Wales were more likely to state that they would
offend under the circumstances described in the scenario than were participant who

205
made fewer mistakes about the sanctions. Interestingly, however, these analyses
suggest that statutory penalties are not the most influential factor in perceptions of
punishment costs. Instead, severity perceptions appear to be more closely related to
the penalties people think they would receive if convicted for drink-driving rather
than to the maximum penalties prescribed by the legislature. This was demonstrated
by the significant bivariate relationship between severity perceptions and knowledge
of actual penalties for drink-driving, as well as evidence indicating that knowledge of
actual drink-driving penalties is a better predictor in the linear regression models
predicting stated offending likelihood. This result is not a particularly surprising
finding but it does have some important implications for criminal justice policy. It
suggests that simply raising the maximum or minimum penalties will not be enough
to prevent future offending. For deterrence to occur, the public needs to know about
the penalty changes and believe that this will result in more severe penalties if
convicted.

This need to increase awareness of drink-driving penalties is underlined by the fact

that, overall, knowledge of applicable penalties in NSW was relatively poor amongst
survey participants. A substantial proportion of participants was not aware that there
are minimum licence disqualification periods prescribed for drink-driving offences in
this State and, though many knew that they could be fined for drink-driving, most
participants significantly underestimated the monetary penalties that would ensue.
While it is acknowledged that the sample recruited for this survey is not
representative of the general driving population in NSW, these results mirror those
found from a large telephone survey of licensed drivers conducted by the NSW
Roads and Traffic Authority in 2004. This survey involved Computer Administered
Telephone Interviews with 1,143 drivers aged 17 to 69 years who reported having
consumed alcohol at least once a month in the previous 12 months. Analysis of the
data from these interviews revealed that 35 per cent of the drivers were not aware
that they could lose their licence if charged with drink-driving, while 73 per cent
were not aware that they could receive a fine of over $1,000. Furthermore, 40 per
cent of the drivers surveyed were not aware that a drink-driving charge could result
in a criminal record (Bryant, Hawkins & ACNielsen 2004).

206
So where does this leave the NSW Government’s drink-driving policy which raised
the statutory penalties for drink-driving offences in 1998? If the public was better
informed would the implementation of the policy have led to a reduction in drink-
driving or alcohol-related road crashes? The significant effect of knowledge found in
this study, as well as the generally low level of knowledge about drink-driving
penalties, would suggest that greater publicity of the policy may have decreased the
likelihood of some people drink-driving. As discussed in Chapter 6, the publicity that
surrounded the penalty increase in NSW was sparse and sometimes negative. If a
well-coordinated media campaign had been launched by the Government to inform
the public of the 1998 legislative amendments, we may have seen a greater reduction
in drink-driving offending levels after the penalty changes.

However, although the knowledge variables were significant in the regression

analyses shown above, only tentative conclusions about the relationship between
legal knowledge, perceptions of punishment severity and offending decisions can be
reached on the basis of these data. Only one of the legal knowledge variables was
significantly correlated with perceptions of penalty severity and, although
statistically significant, the relationship in question was weak. Furthermore, when
subjective ratings of penalty severity were included in the regression models
predicting stated offending likelihood, only the difference between the low and high
severity groups was close to significance and no interaction effect between perceived
certainty and severity of punishment was apparent. We also cannot ignore the fact
that diagnostic checks of the linear regression models suggested the omission of an
important variable from the analysis (at least for the full sample).

Very little research has been conducted thus far on the role that knowledge of
criminal penalties plays in risk perceptions and deterrence. In fact, the current study
appears to be the first attempt using the scenario-based methodology to investigate
the extent to which legal knowledge can predict offending behaviour. The validity of
the measures used in the current research to assess respondents legal knowledge need
to be tested further before any definitive conclusions can be made. Given that
knowledge of statutory penalties was predictive of stated likelihood of offending but
not correlated with perceptions of sanction severity, it is possible that some

207
construct, other than a participant’s understanding of the criminal penalties
prescribed for drink-driving offences was being measured by the multiple-choice
questions asked in the survey. The accuracy of responses to these questions could be
indicative of levels of general knowledge or intelligence rather than specific
knowledge relating to maximum and minimum penalties. The former may also be
associated with participants’ own beliefs about what the penalties should be for
drink-driving offences rather than what they actually know the sanctions to be.

It is also possible, however, that the weak association between legal knowledge and
severity perceptions is due to the fact that the measure used to estimate perceived
sanction severity was capturing more than just the disutility of State-imposed
punishment. The question assessing perceived penalty severity was intentionally
phrased in such a way as to invite participants to include in their estimates only the
formal costs associated with a conviction for drink-driving. Many participants,
despite these instructions, may have incorporated other informal costs into their
estimates of penalty severity, such as the cost of a conviction on future job prospects.
If this were the case, then the significant relationship between knowledge of statutory
penalties and stated offending likelihood may still be evidence for a deterrent effect,
even though no association was found between statutory knowledge and perceptions
of penalty severity.

Nevertheless, what is clear from the results shown above is that the effect of
perceived certainty of apprehension on stated offending likelihood was much more
robust than that found for perceptions of penalty severity. This finding - that
certainty matters more than severity - is a recurring theme throughout the perceptual
research and indeed much of the fieldwork conducted in the deterrence domain.

There are two potential reasons why survey research has failed to find a significant
effect of the perceived severity of criminal punishment. Firstly, as mentioned
previously in this thesis, being apprehended for an offence imposes more costs on an
individual than the sanction imposed by the criminal justice system. Survey-based
research suggests that there is potential for informal sanctions from peers and family,
who might learn of the arrest, to deter offenders through evoking feelings of shame

208
and embarrassment (e.g. Miller & Iovanni 1994; Nagin & Pogarsky 2001). Unlike
the formal sanctions imposed by the criminal justice system, these informal costs
associated with arrest are likely to be incurred almost simultaneously with the
offending episode. Given that people tend to discount future costs at a greater rate
than they do current costs, it is possible that the informal costs arising from an arrest
may be more immediate for many potential offenders than those associated with
formal sanctioning (von Hirsch et al. 1999).

The second reason is that there may be insufficient variation in perceptions of the
severity of criminal punishment to detect a significant effect. As seen from Figure
7.6, most participants in our survey perceived criminal penalties as being very
severe. Over two-thirds of the respondents rated the penalties for drink-driving, on a
scale of 100, at 70 or above. One-third rated them at 100. In contrast, even for the
high-risk scenario, the average perceived risk of apprehension was estimated at less
than 50 per cent. Again, it should be recognised that this is not a representative
sample of NSW drivers. However, if a similar distribution of perceptions of penalty
severity exists in the broader population then this suggests that enhanced RBT
activity may have greater scope to influence offending decisions than do policies
aiming to increase perceptions of sanction severity.

Limitations of Study 3
Two further limitations of Study 3 should be kept in mind when assessing the
implications of these findings for deterrence theory and criminal justice policy.

Firstly, the survey sample used in this research is not representative of the wider
driver population. A student-based sample was recruited for Study 3 on the premise
that young people more frequently engage in risky drinking and driving behaviour.
However, it is possible that for many of these university students, factors additional
to those threatened by the criminal justice system would also be predictive of their
likelihood of engaging in illegal behaviours. Indeed, other perceptual deterrence
research has shown that people who have strong ties to conventional society are more
likely to factor informal social sanctions (i.e. feelings of shame and embarrassment)
arising from public exposure of the act itself into their decisions to offend, rather

209
than the formal costs associated with legal sanctioning once caught (Nagin &
Paternoster 1994; Nagin 1998). Further exploratory analyses in Study 3, which
partitioned out the effect of the deterrence variables for ‘high-risk’ sub samples (i.e.
risky drinkers and prolific offenders), improved the amount of variance accounted
for by the estimated models. But the power to detect significant effects in these
analyses, at least for perceptions of sanction severity, was substantially reduced by
the small number of participants meeting the ‘high-risk’ classification. The current
research should therefore be extended to include surveys of active or convicted
drink-drivers when trying to establish the importance of legal sanctions in models of
offender decision-making.

Secondly, in the scenario-based design used in Study 3, participants are asked only to
estimate the likelihood that they would offend under the conditions described. Thus
we are examining the relationship between legal knowledge and hypothetical
offending rather than legal knowledge and actual episodes of drink-driving. A
longitudinal study that measures participants’ knowledge of formal sanctions and
then subsequent drink-driving occasions during a certain follow-up period would
further expand on the work described here.

210
CHAPTER 8. SUMMARY AND DISCUSSION

The major aim of this thesis has been to assess the deterrent efficacy of increases in
the severity of criminal punishment. A considerable literature already exists on this
topic. However, as highlighted in the introductory chapter to this thesis and in
Chapter 3, the available evidence is far from conclusive. The current study improved
on prior research by examining changes in the threat of formal punishment in a
context where deterrent effects created by modifications to the severity of legal
sanctions should be optimised.

The case study selected for this research was a sentencing policy that substantially
raised the severity of statutory penalties for drink-driving offences in New South
Wales, Australia, in 1998. Drink-driving is an ideal offence to use in a case study of
deterrence because it is an offence that (1) is instrumentally motivated and not
viewed as inherently wrong or against the moral code of society (2) has a high risk of
apprehension, prosecution and conviction (3) nevertheless results in very few
offenders being imprisoned (thus minimising incapacitation effects) and (4) has valid
and reliable measures of offending available. The specific policy examined in this
case study also was thought to provide a more definitive test of deterrence
hypotheses than previous investigations of the severity of formal punishment because
it was (1) implemented in a jurisdiction where the perceived risk of detection for
drink-driving offences is relatively high (2) intended to affect only the severity of
formal punishment but not the probability of punishment and (3) designed to make
licence sanctions mandatory, and therefore unavoidable, for all persons convicted of
drink-driving in New South Wales.

Furthermore, the quasi-experimental methods used in this research were able to

overcome several limitations of previous ecological investigations of punishment
severity because they involved a specific well-defined intervention, included
comparison groups to control for other possible influences on crime rates and, in the
case of the interrupted time-series analyses, measured crime at many points before
and after the intervention (thereby making it possible to distinguish deterrent effects

211
from other fluctuations in offending levels). Another strength of this thesis is that
data from the quasi-experiment were supplemented by data collected in a scenario-
based survey. This survey explored an area of deterrence which, to date, has received
little attention in the empirical literature, namely the relationship between sentencing
policy, sanction perceptions and criminal behaviour.

This chapter summarises the results from the three studies undertaken to examine the
marginal deterrent effect of the changes made to drink-driving penalties in New
South Wales. The findings from these studies are presented within the broader
context of deterrence theory and the implications for criminal justice policy and
future research are discussed.

The deterrent effect of increased drink-driving penalties in

New South Wales

Changes in drink-driving behaviour after the penalty increase

Two aggregate-level studies were undertaken to determine the impact of the 1998
penalty changes on drink-driving offending rates in New South Wales. Study 1
assessed changes in alcohol-related road crashes that were contemporaneous with the
penalty increase. Study 2 examined variations in reoffending rates before and after
the legislative changes were implemented. It was expected that, if the sentencing
policy had the intended deterrent effect on drink-drivers, then a reduction should be
apparent in both alcohol-related road crashes and drink-driver recidivism rates after
the penalties were raised.

The interrupted time-series analyses of road crash data (Study 1) showed that, prior
to the penalty changes being implemented in New South Wales, there was a
significant downward trend in two surrogate measures of alcohol-related road
crashes: single-vehicle night-time crashes and crashes resulting in a fatality. After the
enactment of the new penalty regime in September 1998, there was a significant
change in one of these surrogate measures, with the monthly rate of single-vehicle
night-time crashes beginning to rise. The magnitude of this increase (though

212
relatively small) appeared to be inconsistent with the hypothesised deterrent effect of
the new sentencing policy.

Further analyses, however, indicated that non alcohol-related road crashes (as
measured by multiple-vehicle day-time crashes) were also increasing after the
intervention point and, more importantly, appeared to be increasing at a greater rate
than single-vehicle night-time crashes. This additional observation offers a possible
explanation for the unexpected results found in Study 1. Single-vehicle night-time
crashes are more likely to involve alcohol than other types of crashes but a
substantial proportion of these road crashes are unrelated to alcohol. As such, if there
was an overall rise in non alcohol-related crashes in New South Wales after the
penalties were increased in 1998, then it would be reasonable to expect some residual
effect to also be evident in our surrogate measure of alcohol-related crashes. Two
possible conclusions regarding the effect of the sentencing policy on alcohol-related
road crashes could be drawn from these findings: (1) there was a reduction in
alcohol-related crash rates after the penalty changes but this reduction was
overwhelmed by the rise in non alcohol-related crashes and therefore could not be
detected in our surrogate measure or (2) there was no change in alcohol-related road
crashes after the sentencing policy was implemented.

Study 2 compared the probability of reoffending and the time to new offence for two
drink-driver cohorts, one consisting of persons convicted of drink-driving in the year
before the penalty changes and the other convicted in the year after. Data from this
study showed that, controlling for other potentially confounding variables, offenders
convicted under the new penalty regime were less likely and took longer to reappear
before the court for a new drink-driving offence than those convicted prior to the
legislative amendments. This effect, however, was only apparent for drink-drivers
residing in country and regional areas and the size of the effect was small, with the
probability of reoffending being reduced by just three percentage points. No
differences in reoffending were apparent for drink-drivers residing in the Sydney
metropolitan region.

213
The significant reduction in reoffending after the drink-driving penalties were
increased in 1998, which was found in Study 2, is consistent with the possibility that
the tougher penalties resulted in fewer alcohol-related road crashes but the surrogate
measure used in Study 1 was unable to detect this effect. However, the fact that
drink-driving recidivism was reduced by only a small amount and only for offenders
residing outside the Sydney Metropolitan area suggests that any decrease in alcohol-
related road crashes after the intervention is likely to have been relatively small.
Thus, the results from these quasi-experimental studies provide only limited support
for a marginal deterrent effect created by the more severe sanctions.

Implementation of the 1998 sentencing policy

Previous studies have identified poor policy implementation as a potential
explanation for the failure of tougher sentencing policies to achieve their deterrence
aims. Changes to statutory punishments have, in other words, failed to affect the
actual penalties convicted offenders receive from the court (e.g. Ross 1984;
MacCoun 1993; Nagin 1998) or have produced unintended effects that undermine
the certainty of apprehension, prosecution and conviction of the targeted offence
(e.g. Barber & Wilson 1968; Ross & Voas 1990; Beirness, Simpson & Mayhew
1993; Mayhew, Beirness & Simpson 1995; Voas 2001). A comparison of drink-
driving prosecutions and court imposed sanctions before and after the penalty
changes in New South Wales suggested that the 1998 sentencing policy examined in
this thesis was, for the most part, implemented as intended by its formulators.

Additional analyses, reported in connection with Study 2, for example, showed that
average drink-driving penalties imposed by NSW courts in 1999 were significantly
higher than those imposed in 1997. Gaol terms were found to have increased by 19
per cent (or almost one month), fines by 47 per cent (or approximately $240) and
licence disqualification periods by 16 per cent (or two months). In fact, for drink-
drivers sentenced in non-Sydney courts, the increase in average licence
disqualification periods was found to be even greater at 24 per cent (or almost three
months). Given previous research suggesting that licence disqualification is the most
effective sanction for deterring drink-drivers (Nichols & Ross 1990; Zaal 1994), the
longer disqualification periods imposed by magistrates from non-Sydney courts may

214
go some way in explaining the reduced recidivism of offenders residing in country
and regional areas.

In terms of the prosecution and conviction of drink-driving offences, there was also
generally no evidence for the new sentencing laws being undermined by players
within the criminal justice system. Comparing prosecutions in 1997 with those in
1999 showed a significant increase in the number of drink-drivers brought before the
courts, a slight increase in the proportion of offenders found guilty and no change in
court delay or ‘not guilty’ pleas being submitted to the court. The only setback in
terms of the deterrence aims of the legislation was a small but statistically significant
increase in the proportion of drink-drivers who escaped a licence disqualification
upon being found guilty. This was particularly evident for offences where mandatory
minima were introduced as part of the 1998 legislative changes (i.e. special- and low-
range drink-driving offences). If all convicted drink-drivers received a licence
disqualification as intended by the sentencing policy then we may have seen a greater
impact of the penalty increase on reoffending rates. Nevertheless, four out of five
drink-drivers convicted under the new penalty regime did receive a licence
disqualification, which was, on average, longer than their 1997 counterparts.

Two further possible reasons why the sentencing policy failed to have a larger impact
on drink-driving rates in New South Wales were also explored in Chapter 6. The first
of these concerns the level of enforcement at and around the time that the legislative
amendments were enacted. Modifications to the perceived risk of detection and
apprehension have been shown to have a significant effect on offending rates of
drink-drivers (e.g. Ross 1984; Voas & Hause 1987; Henstridge, Homel & Mackay
1997; Voas, Holder & Gruenewald 1997). Thus any substantial reduction in drink-
driving enforcement practices over the study period could have compromised the
intended deterrent effect of the penalty changes. From an examination of breath
testing data, it would seem that the intensity of drink-driving enforcement did drop
off in New South Wales around the time that the drink-driving penalties were raised.
From 1994 through 1996 the total number of breath tests conducted by NSW Police
increased from 2.4 to 2.7 million but from 1997 onwards they began to decline,
dropping to a low of 1.9 million tests in 1999 and 2.0 million tests in 2000. This

215
decline in yearly testing rates was also reflected in people’s perceptions of drink-
driving enforcement activity, with fewer residents from New South Wales reporting
in the 1999 ATSB survey that they had seen RBTs in operation and more reporting
that they thought RBT activity had decreased.

For a sanctioning policy to meet its deterrence objectives, perceptions of sanction

severity must be increased and this can only occur if the target population is aware of
the policy change (McLean, Kloeden & McCaul 1991; Voas, Holder & Gruenewald
1997; Nagin 1998). A media analysis of the publicity surrounding the penalty
changes revealed that, from the beginning of June 1997 to the end of June 1999, 139
drink-driving articles appeared in New South Wales Metropolitan newspapers but
only 15 of these described the details of the 1998 penalty increases. Furthermore,
where the sentencing policy was discussed it appeared to be ineffectively
communicated. It was packaged with more minor traffic offences and was
accompanied by other drink-driving articles that questioned the certainty of more
severe penalties being imposed by the courts. Also, over the same time period, there
was very little discussion of the penalty changes on radio. Road traffic licensing
authorities were aware of only one publicity campaign, an emotive TV
advertisement, that was run in conjunction with the penalty increase. Although public
risk perceptions could not be assessed in this study, the scarce and sometimes
negative publicity that the policy received, raises the possibility that public
awareness of the new penalties was not sufficient to produce any real change in the
level of drink-driving amongst the general motoring population.

Knowledge of drink-driving penalties and perceptions of sanction

severity
Aggregate-level analyses of changes in offending rates comprised a major
component of this thesis. These types of investigations are essential in assessing the
extent to which deterrence theory can be applied to criminal justice policy. As Tittle
(1980) notes, “in the final analysis, laboratory manipulations are too artificial and
removed from everyday life to satisfy serious scholars who try to judge the work of
the criminal justice system” (Tittle 1980, p. 386). However, our research

216
demonstrates the inherent difficulties in measuring deterrent effects in a real world
situation where the researcher has little or no control over the behavioural event:

The problem is this, of course, that useful comparisons of a rate of crime

before and after the changes in punishment policy depend, in each case,
on the assumption that any changes noted in the rate of a particular
threatened behaviour can be attributed to a specific shift in a facet of
punishment policy, and to that change exclusively. The world rarely
holds still for the researcher in this way (Morris & Zimring 1969, p.
145).

Reduced levels of drink-driving enforcement activity and poor knowledge of the

legislative amendments remain potential explanations for the minimal impact of the
1998 sentencing policy on drink-driving offending rates. Yet while both are plausible
accounts for the small effect observed in the quasi-experimental research there is no
way of knowing whether the marginal deterrent effect of the policy would have been
any greater had the penalty changes been more widely known and enforcement
intensity did not change during their implementation. The only way to assess this
possibility was to examine the relationship between sentencing ‘policy’, sanction
perceptions and behaviour in a setting where other potentially confounding
contextual variables could be controlled.

The third study in this research attempted to achieve this aim by seeing whether
current knowledge of drink-driving penalties influenced perceptions of the severity
of drink-driving punishment and the stated likelihood of offending in a hypothetical
scenario. It was anticipated that those who knew more about the current penalty
regime in New South Wales would perceive drink-driving punishment as more
severe and thus would be less likely to state they would offend. Again, however,
there was only minimal evidence for the hypothesised deterrent effect of penalty
severity on willingness to drive while intoxicated. Instead, consistent with previous
research, the deterrent effect observed for increases in the perceived risk of
apprehension was much more robust (Paternoster 1989; Grasmick & Bursik 1990;
Nagin & Pogarsky 2001).

217
Study 3, for example, found a significant relationship between legal knowledge and
offending, as evidenced by the fact that respondents who made more mistakes in a
test of knowledge about drink-driving penalties were more likely to state that they
would drive home in the scenario described, even after controlling for other
confounding variables. Moreover, knowledge of the actual penalties that drink-
drivers receive upon conviction was found to be a better predictor of behavioural
intentions than knowledge of the statutory maxima and minima stipulated by drink-
driving legislation. Prima facie, these results are consistent with deterrence
hypotheses and would lead one to conclude that, had the 1998 sentencing policy
introduced in New South Wales been more widely publicised, the penalty increases
may have had a greater impact on levels of offending. This conclusion is supported
by the fact that, consistent with data from other surveys conducted in New South
Wales (Bryant, Hawkins & ACNielsen 2004), the level of knowledge of drink-
driving penalties found in Study 3 was generally quite low.

It should be noted, however, that legal knowledge is thought to deter offending only
through its impact on perceptions of sanction severity. Our analyses of the
relationship between knowledge scores and ratings of penalty severity in Study 3
revealed weak associations between these variables. Only the correlation between
knowledge of actual penalties and perceived severity of drink-driving penalties was
significant at the 0.05 level. Furthermore, when severity perceptions were included in
the regression models predicting stated likelihood of offending, only the difference
between those respondents rating penalty severity as low and those rating it as very
high (between 91 and 100 out of 100) was approaching significance. There was no
significant interaction effect between perceived risk of apprehension and perceptions
of penalty severity.

These somewhat contradictory findings from the scenario-based survey may be due
in part to measurement error in the knowledge variable. The multiple-choice
questions employed in this research were designed to assess current knowledge of
the applicable penalties for drink-driving offences. However it is possible that the
knowledge scores were measuring factors other than legal knowledge that have a

218
negative relationship with offending likelihood but are not related to perceptions of
punishment severity. For example, individuals who have a higher level of
intelligence may be more skilled at answering these types of questions but also be
less likely to commit a criminal offence.

An alternative explanation for the study findings is that respondents incorporated

other informal costs arising from apprehension and conviction for a drink-driving
offence into their ratings of penalty severity. Previous perceptual research has clearly
demonstrated that anticipated social sanctions, like embarrassment and shame, can
operate to deter some individuals from offending (e.g. Grasmick & Bursik 1990;
Grasmick, Bursik & Arneklev 1993; Nagin & Pogarsky 2001). If these additional
costs were included in respondents’ ratings of severity then we would not necessarily
expect to find a very strong relationship between legal knowledge and severity
perceptions. The significant effect of knowledge found in Study 3 would therefore
need to be further tested before confident conclusions regarding the role of legal
knowledge in deterrence can be drawn.

The only true clear result to emerge from the scenario-based survey is that increases
in the perceived risk of apprehension reduce the likelihood of drink-driving, a finding
that is consistent with the extant research on deterrence (e.g. Homel 1986; Nagin &
Pogarsky 2001; Nagin & Pogarsky 2003). Simply knowing that police have been
conducting frequent RBTs in the immediate area reduced stated likelihood of
offending, in the described scenario, by as much as 10 per cent. For individuals who
had previously driven whilst drunk or who frequently drank alcohol at high-risk
levels, willingness to offend was reduced by an even greater amount (up to 25 per
cent) when the perceived risk of apprehension was high. This suggests that
perceptions regarding the risk of being punished may be more important in the
offending decisions of high-risk individuals than they are for other groups of
potential offenders.

Implications for theory and practice

The case study selected for examination in this thesis was considered a good real-life
test of deterrence hypotheses because many of the preconditions for deterrence

219
success were thought to be in place. However, a number of contextual factors were
identified in this research that could have served to undermine the penalty changes
and which therefore preclude definitive conclusions regarding the deterrent efficacy
of the legislative reforms. Perhaps the most important of these is poor policy
implementation: the possibility that the increase in formal punishment had minimal
impact on offending because it was either inadequately communicated to the target
audience or implemented in such a way that the actual threat was not raised to the
intended level. By addressing these barriers to successful implementation early on in
policy development it may be possible to enhance the effectiveness of deterrence-
based interventions. With this in mind, the next section of this Chapter discusses
some strategies that could be employed to increase the credibility of sanction-based
policy.

Translating policy into a credible threat

(T)he success of any deterrence process will be determined by the degree
to which this message is successfully transmitted to the population of
offenders (Geerken & Gove 1975, p. 499; cited in Miller & Iovanni
1994).

Survey data on public awareness of penalties for different types of crimes show that
legal knowledge is generally quite poor. People often do not know whether particular
types of penalties are applicable to certain offences or they substantially
underestimate the magnitude prescribed by legislation (e.g. Williams, Gibbs &
Erickson 1980; Snortum & Berger 1989; Kenkel 1993; Hough & Roberts 1998;
Kenkel & Koch 2001). It is not surprising, then, that changes to penalties frequently
result in little or no change in offender behaviour. Complete and accurate knowledge
of criminal penalties is not necessarily a precondition for deterrence to work.
However, if sentencing policies are to improve compliance with the law, there must
at least be a general perception that the threat of punishment has increased. This can
only be achieved if appropriate publicity and education campaigns accompany any
changes that are made to the severity of statutory punishments. While seemingly
commonsensical, this step is one that is often ignored during the development and
implementation of new sentencing regimes.

220
An example whereby publicity has improved the effectiveness of modifications to
the severity of formal punishment is presented by Beha (1977). This research
examined the impact of a sentencing policy dictating mandatory minimum prison
terms for defendants found guilty of illegally carrying a firearm in Massachusetts.
This case study is particularly noteworthy because the sentencing changes were
accompanied by publicity that not only educated citizens about the legal changes but
also emphasised the certainty of the increased penalties being imposed, with the
overall message conveyed to the public being that, once caught, “nobody can get you
out” (Beha 1977, p. 98). Relevant media attention to the legislative amendments was
generated by the office of the “sponsor” of the legislation (the then-Speaker of the
House), as well as by an advertising agency. Furthermore implementation of the
changes was postponed long enough for this information to be adequately conveyed
to the community. Beha’s subsequent examination of the impact of the firearm
legislation showed a dramatic decrease in the number of arrests and prosecutions for
the target offence after the penalty rise, a decrease which did not appear to result
from changes in enforcement or prosecutorial practices. There was also a substantial
improvement in compliance with ancillary firearms laws, especially those dealing
with the licensing and identification of personally owned firearms.

Intensive media campaigns publicising sentencing changes can also have unintended
beneficial effects on perceptions of the certainty of apprehension and conviction.
Blomberg and his colleagues (1987), for example, considered the effect of a publicity
campaign accompanying a new law mandating licence disqualifications for first-time
drink-drivers on perceptions of the certainty and severity of drink-driving
punishment. These authors found that media attention surrounding the new laws
significantly improved general awareness of the mandatory licence sanctions, with
knowledge of the law amongst survey respondents increasing from 50 per cent to
over 80 per cent. However, they also found that survey participants reported an
increase in the number of arrests and conviction for drink-driving after the penalty

221
changes were publicised, despite the media campaign failing to refer to this aspect of
punishment and there being no change in arrest or conviction rates.79

An alternative to mass media education campaigns is to communicate the increased

threat directly to those most at risk of offending. This approach was used to great
effect in the Boston gun control intervention, referred to earlier in this thesis, known
as “Operation Ceasefire” (Braga et al. 2001; previously cited in Chapter 1). This
intervention aimed to generate a strong message to chronic gang offenders that
violence would no longer be tolerated and that formal punishment would
immediately follow any such incidents. When violence did occur, several legal
strategies were used to formally sanction the gang members involved. Other gangs
were then informed that “this gang did violence, we responded with the following
actions, and here is how to prevent anything similar from happening to you” (Braga
et al. 2001, p. 200). Communication of the increased threat was achieved by way of
formal meetings between members of the working group and gang members, through
police, probation and outreach worker contacts with the gang members and through
meetings with individuals in juvenile detention facilities. This program proved to be
a success in deterring offenders, with evidence for substantial reductions in levels of
gun crime in the Boston community and, in particular, gun crime involving young
people.80

Operation Ceasefire and Beha’s (1977) study of the changes to Massachusetts

firearm laws highlight another important requirement if a sentencing policy is to pose
a credible threat: the belief that harsh punishment will be imposed once apprehended
for an offence. The scenario-based survey described in Chapter 7 indicates that
perceptions of sanction severity are more closely associated with knowledge of the

79
The reasons for these changes in perceived certainty of punishment are unclear. One possibility is
that the publicity of the new penalties simply served as a reminder that criminal justice agencies are
targeting drink-driving and, for those who choose to offend, formal punishment will ultimately ensue.
These “spillover” effects do, however, present an additional challenge for researchers who are
attempting to delineate the ‘cause’ of any resultant change in behaviour given the overwhelming
evidence for the deterrent effect of increases in the perceived certainty of punishment.
80
The extent to which this strategy can be used on a larger scale with more high volume crimes like
drink-driving is unclear. It is possible that the credibility of the increased threat would be
compromised once the size of the would-be offender population becomes too large and scarce
resources are overstretched (Nagin 1998). The Boston intervention does, however, provide some
promising evidence for improved communication of the formal consequences associated with
offending altering perceptions and subsequent behaviour, even amongst persistent offenders.

222
actual penalties offenders receive upon conviction, than with the prescribed
maximum statutory penalties. This suggests that, simply informing the public that
formal penalties have increased, is insufficient for deterrence to occur. Offenders
also need to believe that, once introduced, the court will impose these harsher
penalties on all persons apprehended and prosecuted for the target offence. Strategies
employed to improve public knowledge of penalty changes must therefore be
designed to create this perception that severe punishment will be imposed.

The media analysis presented in Chapter 6 suggests that achieving this change in
perceptions is not necessarily an easy task. Much of the publicity drink-driving
penalties receive in the media comes from cases where the court has failed to impose
a penalty or has imposed a penalty that is viewed as particularly light. Focusing on
these isolated but high-profile cases undermines the deterrent effectiveness of court-
imposed sanctions. As the comments of the Chief Justice of New South Wales below
illustrate, this problem appears to be relevant, not only to drink-driving offences, but
to criminal offences more broadly, particularly in the jurisdiction where the case
study for this thesis was situated:

A key objective of criminal sentencing is, of course, deterrence. For

deterrence to work, potential offenders must have an understanding of
the likely consequences of criminal conduct. If, as I believe is the case,
media reporting gives excessive emphasis to light sentences and gives
the impression that such sentencing is typical, when it is not, then
deterrence will not work (Spegilman 2005, p. 3).

These remarks suggest that one way to improve the deterrent effect of penalties is
not to increase them but to communicate the current threat of formal punishment for
particular offences in a more effective way or more directly to the target audience.
This could be by providing offenders, at the time of conviction, with information
regarding the actual penalty they will receive if they reappear before the court or,
alternatively, by actively publicising the average penalties being imposed in the
courts in order to counterbalance media coverage of atypical cases.

223
The latter strategies are particularly attractive given that increases in the severity of
statutory punishment can sometimes result in little or no change in the actual
penalties that convicted offenders receive. The case study described in this thesis
provides some evidence highlighting this possibility. The 1998 sentencing policy
introduced minimum licence disqualification periods for two drink-driving offences
which previously had not attracted such a mandate and doubled the minimums
already applicable to other offences. These amendments resulted in an increase the
average length of licence disqualification periods but did not appear to affect the
proportion of drink-drivers whose licence was disqualified. In fact there was a slight
decrease in the percentage of drink-drivers who had their licence disqualified after
the penalty changes, an effect that was much more pronounced for offences where
mandatory minimums had been introduced. The drink-driving literature offers strong
evidence for a specific deterrent effect of licence disqualification, with drink-drivers
who are disqualified from driving being involved in fewer alcohol-related crashes
and recording fewer offences (e.g. Votey & Shapiro 1983; Blomberg, Preusser &
Ulmer 1987; Mann et al. 1991; Siskind 1996). Thus, the failure of 1998 sentencing
policy to affect the extent to which licence disqualification is applied to drink-driving
offences is an obvious reason for the small marginal deterrent effect observed in the
quasi-experimental study.

Where there are wide disparities in sentencing (see Moffatt, Fitzgerald &
Weatherburn 2004) defendants may come to believe that they can reduce the severity
of punishment if they can get their case listed before a ‘lenient’ judge. Greater
consistency in the sentencing of drink-drivers may therefore also enhance the
credibility of the threat generated by formal punishment. This could potentially be
achieved through the development and application of sentencing guidelines for these
types of offences (as has since happened in New South Wales),81 or, through better
judicial education on the deterrent efficacy of licence sanctions. As discussed in
Chapter 3, there is good evidence starting to emerge from North America showing
that administrative licence suspension laws can have a beneficial impact on alcohol-

81
Application by the Attorney General under Section 37 of the Crimes (Sentencing Procedure) Act for
a Guideline Judgment Concerning the Offence of High Range Prescribed Concentration of Alcohol
Under Section 9(4) of the Road Transport (Safety and Traffic Management) Act 1999 (No. 3 of 2002)
[2004] NSWCCA 303.

224
related road crashes and, to some extent, drink-driving recidivism rates. The
deterrence mechanism by which this effect is achieved is not yet clear. Nevertheless,
these laws have the potential to increase the perceived certainty of licence sanctions
for drink-driving offences without necessarily changing the level of judicial
discretion available in the sentencing process. An alternative approach therefore
would be to introduce administrative penalties.

The marginal deterrent efficacy of punishment severity

It is true that the “implementation problem” limits our ability to draw definitive
conclusions regarding the deterrent efficacy of the penalty increase examined in this
thesis. With greater consistency in sentencing and greater publicity of the policy, the
tougher penalties could have had a larger impact on drink-driving levels in New
South Wales. It is also possible that the deterrent effect of the harsher punishment
could have been further enhanced if NSW Police maintained a high level of
enforcement around the time of the legislative reforms.

Despite these limitations however, the results of our quasi-experimental research are
consistent with those from many other aggregate-level studies reported in the
deterrence literature (see von Hirsch et al. 1999 for a review), in suggesting that
penalty increases have limited marginal deterrent efficacy. The results from our
scenario-based study (Study 3), which was able to examine punishment severity
effects whilst controlling for other potentially confounding factors, is further support
for this conclusion. Here it was found that the deterrent effect of the perceived
severity of punishment was much less robust than the deterrent effect of the
perceived risk of detection: a finding that is consistent with the extant perceptual
research on deterrence. Thus, even if sentencing policies could be translated into a
credible threat, the marginal deterrent effect of tougher penalties might still be
relatively small.

There are at least three possible reasons identified in the research literature why
policies that raise the level of punishment severity might have limited efficacy in
deterring offenders: (1) the vast majority of potential offenders are deterred by the
anticipated informal social sanctions associated with public exposure of the offence

225
rather than the formal punishment prescribed by legislatures (2) there is a fixed cost
associated with arrest for an offence and the penalties incurred later in the criminal
justice process are mostly discounted by potential offenders (3) a severity threshold
exists beyond which changes in criminal penalties will have no additional impact on
offending likelihood. These three accounts for weak severity effects, and their
relevance to the results from the case study, are discussed below.

Since the introduction of RBT in New South Wales and Australia there has been a
strong emphasis on the serious risk that drink-drivers pose to themselves and to other
drivers on the road. For example, a recent national anti-drink-driving advertisement
used the slogan “if you drink and drive you are a bloody idiot”, which was
accompanied by graphic images of a road crash victim. It is possible that this
portrayal of a drink-driver as a ‘killer drunk’ has resulted in drink-driving behaviours
being more widely condemned in the community and the social stigma attached to
the behaviour therefore being more pronounced. With this increased stigmatisation of
drink-driving behaviours, anticipated feelings of shame from violating an
internalised norm would become more influential in deterring potential drink-drivers
and the sanctions attached to formal apprehension and conviction for the offence less
efficacious. Thus, during the early stages of RBT in Australia drink-driving levels
may have decreased primarily because of fear of formal punishment but over time a
social norm has developed that drink-driving is wrong and voluntary compliance
with the law has improved (Ross 1984).

While social stigma associated with the act is undoubtedly important in the offending
decisions of some would-be drink-drivers, the significant effect of punishment
certainty found in the scenario-based research described in Chapter 7 suggests that
the “fear of arrest” is still a powerful deterrent for many people. In Study 3,
respondents who were told that police had been conducting RBTs in the immediate
area, on a frequent basis, were less likely to state that they would offend in the
hypothetical scenario than respondents who were told that police had not been
conducting RBTs in the area for some time. On the other hand, only weak deterrent
effects were found for the perceived severity of formal punishment for drink-driving
offences. This suggests that there are additional costs associated with public

226
exposure of the offence through conviction or mere apprehension which are serving
to discourage many people from offending and which are independent of the formal
costs prescribed by the criminal justice system.

As discussed in previous chapters, these additional “extralegal” costs associated with

an arrest would incorporate the sense of shame or feelings of embarrassment evoked
by the knowledge that significant others will learn of the offence but might also
include anticipated nonlegal costs such as reduced future job opportunities because
of a criminal conviction or the negative impact of a criminal conviction on an
individual’s professional standing (e.g. registration as a health professional). For our
student-based sample in the scenario-based study, as is perhaps true for the broader
driver population, it is likely that these extralegal costs play an important role in their
decisions to offend. The severity of legal punishment will then matter only to the
extent that it can have an additional effect over and above this fixed cost associated
with an arrest. Given also that informal social sanctions would be incurred almost
simultaneously with an arrest, unlike criminal penalties which are a future contingent
cost that tends to be discounted (von Hirsch et al. 1999), the marginal deterrent
efficacy of sanction severity will be limited in many cases.

Nevertheless, there are those individuals who frequently engage in risky behaviour
(as do their peers) and who are less influenced by informal social sanctions. For these
people (as argued in Chapter 2) punishment contingencies are expected to play a
greater role in deterring further offending than extralegal costs. This is one possible
explanation why the deterrent effect of the 1998 sentencing policy was observed only
for a small group of ‘high-risk’ offenders, that is convicted drink-drivers. Drink-
drivers with a previous conviction, being less constrained by informal social controls,
may have modified their driving behaviour to a greater extent than did other NSW
drivers once the new, more severe drink-driving penalties were introduced.

Even amongst this group of high-risk offenders, however, the marginal deterrent
efficacy of increases in the severity of formal punishment could still be somewhat
limited, as the relatively small change in reoffending observed in Study 2 would
suggest. Previous research has shown that impulsive individuals, being more present-

227
oriented, tend to discount future contingencies to a greater extent than other people
(Nagin & Pogarsky 2001; Pogarsky, Piquero & Paternoster 2004) and that these
same individuals are also more likely to commit crime (Grasmick et al. 1993; Keane,
Maxim & Teevan 1993; Nagin & Paternoster 1993). As such, many offenders would
place greater weight on the immediate benefits that could be gained from committing
a crime than on the discounted formal costs threatened by the criminal justice
system, when assessing the utility of different offending opportunities. Thus, even if
many drivers with a previous conviction or a prior history of drink-driving were to
perceive an increase in the formal costs associated with this offence, the fact that
these increased costs would be substantially delayed, if indeed they are incurred at all
(given that formal punishment is contingent on apprehension), means that a
substantial proportion of this offender group might still choose to engage in the
proscribed behaviour.

A further constraint on the efficacy of punishment severity, relevant to both ‘high-

risk’ offenders and the broader offender population, is the possibility of diminishing
returns at the extreme ends of the punishment severity continuum. According to
subjective expected utility theory any increment in the severity of criminal penalties
should raise perceptions of punishment severity, thereby increasing the disutility of
offending and helping to tip the cost/benefit ratio. However, the relationship between
actual penalty severity and perceived penalty severity is potentially non-linear. For
example, two months in gaol may be perceived as twice as severe as one month but
two years in gaol perceived as only 50 per cent more severe than one year. This
would imply that the marginal deterrent effect of penalty changes on rates of
offending will not simply be a function of the magnitude of the change. Instead, the
response function relating punishment severity and crime rates might resemble more
of a concave shape, like that shown in Figure 8.1. In this case, more severe penalties
would still be associated with reductions in crime levels but the response to an
incremental increase (the marginal effect) will vary depending upon the severity of
the sanctions being altered. Small increments in less severe penalties would result in
large changes in crime rates (P0 → P1), but the same increment in very severe
penalties would result in lesser reductions (P1 → P2).

228
 Figure 8.1: The marginal deterrent effect of changes
in penalty severity

Crime
levels

P0 P1 P2 Penalty
magnitude

This issue of diminishing marginal returns for sentencing policies has particular
relevance to drink-driving offences in New South Wales. The penalties stipulated for
this offence are the most severe of any driving offence in this State and are tougher
than those legislated by many other Australian jurisdictions. Thus, if potential
offenders already regarded the penalties as exceedingly severe before 1998, then
doubling these penalties is unlikely to have much additional, observable effect on
their behaviour.

A further point to note here is the possibility that this relationship between actual and
perceived severity of formal punishment also is not continuous. Von Hirsch and his
colleagues (1999) suggest that there may be a point or limit beyond which
punishment severity no longer matters. Once this “severity threshold” has been
reached or surpassed, further changes in penalty severity will no longer affect
perceptions and people will begin to focus more on the risk of being caught when
deciding to offend than on the formal costs that would be imposed upon conviction.
The results from Study 3 provide some evidence that this “severity threshold” is
being approached for drink-driving penalties, at least for the young people sampled
in the this survey. When asked to indicate how much of a problem the current
penalties for drink-driving offences would be for them, most of the survey
participants judged the penalties to be very severe. Over two-thirds of the
respondents rated the penalties (on a scale of 0 to 100) at 70 or above, one-third rated

229
them at 100. While this is not a representative sample of drink-drivers from New
South Wales, if a similar distribution of perceptions of penalty severity exists in the
broader offender population then policies aiming to increase perceptions of sanction
severity will have limited potential to further influence offending decisions.

Future directions for deterrence research and drink-driving

policy
This thesis has highlighted several areas of deterrence that require further
investigation and has also offered some insight into effective, and less effective,
strategies for addressing drink-driving and related issues. These future directions for
research and policy are summarised in the following sections in conjunction with
other non deterrence-based approaches that could also be employed to reduce the
harm arising from drink-driving behaviours.

Future research
As several deterrence scholars have noted, very little is known about the factors
influencing perceptions of risks, in particular the relationship between these
perceptions and sanction-based policy (Miller & Iovanni 1994; Nagin 1998; von
Hirsch et al. 1999; Pogarsky, Piquero & Paternoster 2004). The scenario-based
research presented in this thesis examined one possible determinant of these
perceptions, namely knowledge of legal penalties. As described previously, this
study presents some evidence for a significant relationship between legal knowledge
and stated offending likelihood but the correlations between perceptions of
punishment severity and knowledge were weak, and only in the case of an
individual’s knowledge of actual penalties was this relationship significant. Further
research validating the measures used in this scenario-based research is therefore
needed in order to verify whether, controlling for other potentially confounding
variables, knowledge of criminal justice policies and practices has the potential to
affect risk perceptions and subsequent behaviour.

This type of research should, however, be extended to include active offenders. With
just a few exceptions (e.g. Horney & Marshall 1992; Wright et al. 2004), one of the
major limitations of the work conducted in perceptual deterrence research thus far is

230
the use of student-based samples or general population surveys to examine
deterrence hypotheses. Much of this work has found that informal social controls
play a major role in decisions to offend. This is not surprising given that these
respondents would typically have many close personal relationships and strong ties
to conventional society and therefore have more to ‘lose’ if their transgression is
publicly exposed. On the other hand, there is a proportion of the population for
whom these broader social controls are absent or ineffective and who, it has been
argued in this thesis, may be more responsive to incentives and disincentives. These
“marginal offenders” are the target group for deterrence-based interventions and,
therefore, should also be the major focus of research investigating the link between
sanctioning policy, risk perceptions and offending likelihood or intentions to offend.

Further work is also needed on the extent to which improved communication of the
current threat posed by the criminal justice system can serve to enhance deterrent
outcomes. Research across various jurisdictions has shown that public knowledge of
the penalties applicable to proscribed behaviours is generally quite low (Williams,
Gibsb & Erickson 1980; Snortum & Berger 1989; Hough & Roberts 1998; Kenkel &
Koch 2001). It is possible, as demonstrated in Study 3 of this thesis and in other
work conducted in the drink-driving domain (see Kenkel & Koch 2001), that people
who are being targeted by such laws have a better understanding of the legal
consequences associated with different criminal behaviours. However, media
reporting of atypical cases where penalties are light or of the inconsistency in
sentencing practice could serve to undermine the perceived certainty that these
penalties will be applied if apprehended for the offence. Thus, research needs to
establish the extent to which the threat of formal punishment can be made more
credible by communicating, directly to those at most risk of engaging in the
behaviour, the actual penalties that will be imposed if one appears, or reappears,
before the court for a particular offence.

Improving offender knowledge of the legal penalties imposed by the courts will,
however, only succeed in achieving a reduction in offending if perceptions of
punishment severity increase and, further, increase enough to persuade some
offenders that the costs associated with the behaviour have become too great. It is not

231
yet clear, however, from the available research on punishment severity and
deterrence, just how much more punishment or what type of punishment is needed
for this to occur. Small increases in the severity of formal punishment may have little
effect on perceptions and offending levels. Large changes may have a greater impact
on subjective utilities but might also depend on the severity level of the penalties
being modified. If the current penalties applicable to an offence were already
perceived as severe, only minimal, marginal deterrent effects might be achieved and
if the penalties have already surpassed a “severity threshold” then it is possible that
no change in crime levels would be observed. Alternatively, particular types of
formal punishment may be perceived as more severe or more inconvenient than
others and as such, certain penalties might be expected to be more efficacious in
deterring offenders. Future deterrence research should therefore aim to more
accurately quantify the relationship between penalty severity/type and crime levels.

Policy recommendations
Keeping in mind these limitations of the existing research on deterrence and formal
punishment, the current work suggests two deterrence-based approaches that could
be used to minimise the harm associated with drink-driving.

Firstly, there is strong evidence for the marginal deterrent efficacy of increasing the
perceived risk of drink-driving punishment. The enforcement of drink-driving
offences has, in the words of Sherman and his colleagues (1998), been “one of the
great success stories of world policing” (Sherman et al. 1998, p. 17). This success
has been achieved not by increasing the actual probability of arrest for this offence
(the size of the potential offender population would significantly reduce the
credibility of such an approach) but by increasing the perceived risk of detection and
apprehension. According to deterrence theorists, these risk perceptions are not stable
but are continually updated on the basis of new experiences, which are both personal
and vicarious in nature (e.g. Ross & LaFree 1986; Pogarsky, Piquero & Paternoster
2004). Encountering enforcement activity, or learning about it from others, will
cause people to initially overestimate the risk of apprehension, but with repeat
successful offending episodes there is a chance that people will begin to lower their
perceptions of the probability of arrest over time. Sherman (1990) suggests that the

232
way to prevent this decline in risk perceptions is to keep people continually guessing
about what the actual risk of detection might be at any one moment. In these
situations of uncertainty, he anticipates that people will tend to play it safe and
overestimate the risk of detection, thus decreasing the likelihood of offending if and
when such an opportunity arises.

The original version of RBT, adopted in Australia, was modelled on these basic
principles of deterrence, with police enforcement of drink-driving consisting of
breath testing operations situated in highly visible locations. Most importantly, the
police continually varied the places where the testing sites were located and the times
that the testing was undertaken, thus making it difficult for drivers to predict the
chance of encountering RBTs on the roads at any particular point in time. The
current research and literature review suggests that this strategy needs to continue as
the dominant force in this area of drink-driving policy, if enforcement is to succeed
in achieving a general deterrent effect. It is, as Homel (1986) warned almost 20 years
ago, essential that policy makers and police authorities do not allow this general
deterrence model of RBT to be superseded by the tendency of law enforcement
officers to want to catch more offenders, as appeared to have happened in New South
Wales toward the end of the 1990s. One way by which this could be achieved is
through routine monitoring of breath testing operations and assessment of their effect
on driver perceptions and/or behaviours. This information could then be provided on
a regular basis to all officers involved in the enforcement of drink-driving to
reinforce the preventative value of undertaking random breath tests of drivers and to,
ultimately, shape future planning of enforcement activities in this area.

Secondly, the formal threat posed by the criminal justice system should be
communicated more effectively to high-risk offenders, who may be less influenced
by the informal social sanctions attached to a drink-driving arrest. This need not be
through increases in the severity of statutory sanctions, particularly in light of
possible countervailing effects that could result from the implementation of such a
strategy, but rather transforming the current threat of formal punishment into one
which is more credible. Improving knowledge of the actual penalties that will be
applied for drink-driving offences and increasing the perceived certainty that these

233
harsh penalties will be imposed if caught, particularly amongst individuals at-risk of
drink-driving (e.g. young, heavy drinkers and previous offenders), could curtail the
frequency of offending. Even so, the consistency with which licence sanctions are
imposed by magistrates for drink-driving offences also would need to be improved,
through sentencing guidelines or even the introduction of administrative penalties,
for this approach to be most convincing.

As discussed previously, strategies focusing on the severity of legal sanctions could

be relatively limited in their ability to reduce offending levels, particularly if
penalties are already severe. Therefore other drink-driving countermeasures that
target ‘hard-core’ or persistent offenders should also be considered. One such
strategy is to combine traditional punishments for repeat offenders with alcohol
treatment programs. Wells-Parker et al. (1995), who undertook a meta-analysis of
215 independent evaluations of drink-driving remediation programs, estimates that
alcohol treatment (e.g. psychotherapy, counselling, education) combined with licence
suspensions could reduce recidivism by between seven and nine per cent compared
to licence sanctions alone. Alcohol treatment programs identified as the most
effective were those that were well-structured, combined several treatment
modalities and provided follow-up services after treatment completion (Wells-Parker
2000). These findings should not, however, be interpreted as evidence that treatment
programs should be mandatory for all convicted drink-drivers. Instead, given the
relatively low recidivism rate overall amongst these offenders, alcohol treatment
should be considered simply as an another option for dealing with the hard core
drink-driver for whom other more traditional methods may prove ineffective.

Another drink-driving intervention that has been shown to be relative successful in

reducing the offending of repeat, high-risk offenders is alcohol ignition interlocks.
These devices are fitted to the ignition of a drink-driver’s vehicle and prevent a
vehicle from being started until a low alcohol or alcohol-free breath sample has been
provided. Once the ignition has been started further tests are required every few
minutes, thereby preventing a sober person from starting the engine on behalf of the
impaired driver. Voas et al. (1999) reviewed eight studies evaluating the
effectiveness of these sanctions and found that offenders who were ordered to use

234
alcohol interlocks were less likely to be reconvicted for a drink-driving offence than
were offenders who received only a licence sanction upon conviction. Despite these
promising findings, however, these reductions in offending were evident only during
the period in which the device was operational. Once the interlock was removed,
there were no significant differences in recidivism between those offenders who used
an interlock and those who received more conventional punishments from the courts.
Though it should also be noted that the relative risk of crash involvement for
recidivist drink-drivers is more than double that for other drivers (Fetherson, Lenton
& Cercarelli 2002). Therefore any modification in their drink-driving behaviour
would be of public health significance.

Finally, as discussed at the outset of this thesis, punishment is not the only
mechanism by which policy affects offending behaviours, and accordingly criminal
justice interventions should not be considered as the only means by which drink-
driving can be controlled. Other more macro-level policies also have the potential to
affect levels of drink-driving through their impact on alcohol consumption and
should therefore be considered alongside police enforcement and court-imposed
punishments.

One such policy shown to be particularly effective in reducing alcohol-related road

crashes is taxation on alcoholic beverages. Ecological studies conducted in the USA
have demonstrated that, net of other factors, increases in alcohol excises are
associated with substantial reductions in alcohol-related road fatality rates
(Chaloupka, Saffer & Grossman 1993), and, particularly, motor vehicle fatalities
involving young people (Saffer & Grossman 1987). Kenkel (1993) estimates from
his research that a 10 per cent increase in the price of alcohol would reduce the
probability of drink-driving in the USA by about seven per cent for males and eight
per cent for females, and that the impact of this change on young drivers would be
even greater. Other empirical research has also shown that increases in the physical
availability of alcohol are associated with increased levels of alcohol-related harm,
including harms associated with drink-driving. Gruenewald and his colleagues
(1996), for example, demonstrated in their cross-sectional research conducted in the
USA that alcohol-related road crashes are more likely to occur where there is a high

235
density of alcohol outlets. Meanwhile, Australian quasi-experimental research has
shown that alcohol-related crashes and drink-driving incidents increase with longer
hours (Chikritzhs, Stockwell & Masters 1997) and more days (Smith 1988) of
alcohol sale. Thus, measures designed to decrease the physical availability of alcohol
could also have an additional beneficial effect on the level of drink-driving in a given
community.

The research presented throughout this thesis suggests that deterrence-based

strategies, particularly those focusing on the severity of formal punishment, may
have limited potential to achieve further reductions in alcohol-related road injuries
and fatalities, at least within the Australian context. As such, the non-deterrence
based strategies, briefly outlined above, should also be given due consideration when
identifying future priorities in this area of public policy.

236
References

Achen, C.H. 1982, Interpreting Regression, Sage Publications, Beverley Hills.

Akers, R.L., Krohn, M.D., Lanza-Kaduce, L. & Radosevich, M. 1979, ‘Social
learning and deviant behavior’, American Sociological Review, vol. 44, pp.
635-655.
Andenaes, J. 1966, ‘The general preventive effects of punishment’, University of
Pennsylvania Law Review, vol. 114, no. 7, pp. 949-983.
Andenaes, J. 1971, ‘Deterrence and specific offenses’, The University of Chicago
Law Review, vol. 38, pp. 537-553.
Andenaes, J. 1974, Punishment and Deterrence, University of Michigan Press, Ann
Arbor.
Australian Institute of Health and Welfare 2002a, 2001 National Drug Strategy
Household Survey: First results, Australian Institute of Health and Welfare,
Canberra.
Australian Institute of Health and Welfare 2002b, 2001 National Drug Strategy
Household Survey: State and Territory supplement, Australian Institute of
Health and Welfare, Canberra, ACT.
Australian Transport Safety Bureau 2001, Alcohol and Road Fatalities, Australian
Transport Safety Bureau, Canberra, ACT.
Bachman, R., Paternoster, R. & Ward, S. 1992, ‘The rationality of sexual offending:
Testing a deterrence/rational choice conception of sexual assault’ Law and
Society Review, vol. 26, pp. 343-372.
Barber, R. N. & Wilson, P. 1968, ‘Deterrent aspect of capital punishment and its
effect on conviction rates: The Queensland experience’, Australian and New
Zealand Journal of Criminology, vol. 1, no. 2, pp. 100-108.
Baum, S. 1999, ‘Self-reported drink driving and deterrence’, The Australian and New
Zealand Journal of Criminology, vol. 32, no. 3, pp. 247-261.
Baum, S. 2000, ‘Drink driving as a social problem: Comparing the attitudes and the
knowledge of drink driving offenders and the general community’, Accident,
Analysis and Prevention, vol. 32, pp. 689-694.

237
Becker, G. S. 1968, ‘Crime and punishment: An economic approach’, Journal of
Political Economy, vol. 76, pp. 169-217.
Beha, J. A. 1977, ‘"And nobody can get you out"; The impact of a mandatory prison
sentence for the illegal carrying of a firearm on the use of firearms and on the
administration of criminal justice in Boston - Part 1’, Boston University Law
Review, vol. 57, pp. 96-146.
Beirness, D. J., Simpson, H.M. & Mayhew, D.R. 1993, Assessment of the Impact of
the 1985 Amendments to the Drinking and Driving Section of the Criminal
Code of Canada, Department of Justice Canada, Ottawa.
Benson, B. L., Mast, B.D. & Rasmussen, D.W. 2000, ‘Can police deter drunk
driving?’, Applied Economics, vol. 32, pp 357-366.
Benson, B. L., Rasmussen, D.W. & Mast, B.D. 1999, ‘Deterring drunk driving
fatalities: An economics of crime perspective’, International Review of Law
and Economics, vol. 19, pp. 205-225.
Beyleveld, D. 1979, ‘Identifying, explaining, and predicting deterrence’, British
Journal of Criminology, vol. 19, no. 3, pp. 205-224.
Birrell, J. L. 1974, Drinking, Driving and You, Sun Books, Melbourne, Victoria.
Blomberg, R. D., Preusser, D.F. & Ulmer, R.G. 1987, Deterrent Effects of
Mandatory Licence Suspension for DWI Conviction, National Technical
Information Service, Springfield, Virginia.
Blumstein, A., Cohen, J. & Nagin, D. 1978, Deterrence and Incapacitation:
Estimating the Effects of Criminal Sanctions on Crime Rates, Report prepared
for the National Research Council's Panel on Research on Deterrent and
Incapacitative Effects 1978, National Academy of Sciences, Washington DC.
Borschos, B. 2000, An evaluation of the Swedish drunken driving legislation
implemented on 1/02/94, paper presented at the 15th International Conference
on Alcohol, Drugs and Traffic Safety, May 22 - 26, Stockholm, Sweden.
Box, G. E. P., Jenkins, M. & Reinsel, G.C. 1994, Time-Series Analysis: Forecasting
and Control, Prentice Hall, New Jersey.
Braga, A. A., Kennedy, D.M., Waring, E.J. & Piehl, A.M. 2001, ‘Problem-oriented
policing, deterrence, and youth violence: An evaluation of Boston's operation
ceasefire’, Journal of Research in Crime and Delinquency, vol. 38, no. 3, pp.
195-225.

238
Briscoe, S. 2004a, ‘Raising the bar: Can increased statutory penalties deter the drink-
driver?’, Accident, Analysis and Prevention, vol. 36, no. 5, pp. 919-929.
Briscoe, S. 2004b, ‘The impact of increased drink-driving penalties on recidivism
rates in NSW’, Alcohol Studies Bulletin, no. 5, NSW Bureau of Crime
Statistics and Research, Sydney, NSW.
Bryant, P., Hawkins, A. & ACNielsen 2004, Drink-driving attitudes and self-
reported behaviours in NSW drivers, paper presented at the Road Safety
Research, Policing and Education Conference, Nov. 14 – 16, Perth, WA.
Carroll, J. S. 1978, ‘A psychological approach to deterrence: The evaluation of crime
opportunities’, Journal of Personality and Social Psychology, vol. 36, no. 12,
pp. 1512-1520.
Cashmore, J. 1985, The Impact of Random Breath Testing in NSW, NSW Bureau of
Crime Statistics and Research, Sydney, NSW.
Chaloupka, F. J., Saffer, H. & Grossman, M. 1993, ‘Alcohol control policies and
motor-vehicle accidents’, Journal of Legal Studies, vol. 22, pp. 161-186.
Chaloupka, F. J. & Wechsler, H. 1996, ‘Binge drinking in college: The impact of
price, availability and alcohol control policies’, Contemporary Economic
Policy, vol. 14, no. 4, pp. 112-124.
Chambliss, W. 1967, ‘Types of deviance and the effectiveness of legal sanctions’,
Wisconsin Law Review, Summer Issue, pp. 703-719.
Chan, J. B. L. & Oxley, D. 2004, ‘The deterrent effect of capital punishment: A
review of the research evidence’, Contemporary Issues in Crime and
Criminal Justice, no. 84, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
Chikritzhs, T., Stockwell, T., Heale, P., Dietze, P. & Webb, M. 2000, National
Indicators Project Technical Report No. 2: Trends in Alcohol-Related Road
Injury in Australia, 1991-1997, National Drug Research Institute and Turning
Point Alcohol and Drug Centre Inc., Curtin University of Technology, Perth,
WA.
Chikritzhs, T., Catalano, P. Stockwell, T., Donath, S., Young, D. & Matthews, S.
2003, Australian Alcohol Indicators, 1990-2001; Patterns of alcohol Use and
Related Harms for Australian States and Territories, National Drug Research
Institute, Curtin University of Technology, Perth, WA.

239
Chikritzhs, T. N., Stockwell, T.R. & Masters, L. 1997, Evaluation of the public
health and safety impact of extended trading permits for Perth hotels and
night-clubs, National Drug Research Institute, Curtin University of
Technology, Perth, WA.
Clarke, R. V. & Cornish, D.B. 1985, ‘Modeling Offenders' Decisions: A Framework
for Research and Policy’, in M. Tonry & N. Morris (eds), Crime and Justice:
A Review of Research, vol. 6, The University of Chicago Press, Chicago, pp.
147-185.
Clarke, R. V. & Felson, M. 1993, ‘Introduction: Criminology, Routine Activity, and
Rational Choice’ in R.V. Clarke & M. Felson (eds), Routine Activity and
Rational Choice: Advances in Criminological Theory, vol. 5, Transaction
Publishers, New Brunswick, New Jersey, pp. 1-14.
Collins, D. J. & Lapsley, H.M. 2001, Counting the cost: Estimates of the Social Cost
of Drug Abuse in Australia in 1998-99, Commonwealth Department of
Health and Ageing, Canberra, ACT.
Cook, P. J. 1980, ‘Research in criminal deterrence: Laying the groundwork for the
second decade’, in N. Moms & M. Tonry (eds), Crime and Justice: A Review
of Research, vol. 2, University of Chicago Press, Chicago pp. 111-268.
Cook, T. D. & Campbell, D.T. 1979, Quasi-Experimentation: Design and Analysis
Issues for Field Settings, Houghton Mifflin Company, Boston.
Corbett, C. & Simon, F. 1992, ‘Decision to break or adhere to the rules of the road,
viewed from the rational choice perspective’, British Journal of Criminology,
vol. 32, no. 4, pp. 537-549.
Decker, S., Wright, R. & Logie, R. 1993, ‘Perceptual deterrence among active
residential burglars: A research note’, Criminology, vol. 31, no. 1, pp. 135-
147.
DeJong, W. & Hingson, R. 1998, ‘Strategies to reduce driving under the influence of
alcohol’, Annual Review of Public Health, vol. 19, pp. 359-378.
DeYoung, D. J. 1997, ‘An evaluation of the effectiveness of alcohol treatment, driver
license actions and jail terms in reducing drunk driving recidivism in
California’, Addiction, vol. 92, no. 8, pp. 989-997.

240
Douglas, M. & Wildavsky, A. 1982, Risk and Culture: An Essay on the Selection of
Technical and Environmental Dangers, University of California Press,
London, England.
Durkheim, E. 1964, The Rules of Sociological Method, Free Press, New York.
Ehrlich, I. 1973, ‘Participation in illegitimate activities: A theoretical and empirical
investigation’, The Journal of Political Economy, vol. 81, no. 3, pp. 521-565.
Ehrlich, I. 1975, ‘The deterrent effect of capital punishment: A question of life and
death’, The American Economic Review, vol. 65, no. 3, pp. 397-417.
Elliott, B. 2003, Deterrence theory revisited, paper presented at the Road Safety
Research, Policing and Education Conference, Sept. 24 - 26, Sydney, N.S.W..
Enders, W. 1995, Applied Econometric Time Series, Wiley & Sons, New York.
Evans, L. 1991, Traffic Safety and the Driver, Van Nostrand Reinhold, New York.
Evans, W. N., Neville, D. & Graham, J.D. 1991, ‘General deterrence of drunk
driving: Evaluation of recent American policies’, Risk Analysis, vol. 11, no.
2, pp. 279-289.
Exum, M. L. 2002, ‘The application and robustness of the rational choice perspective
in the study of intoxicated and angry intentions to aggress’, Criminology, vol.
40, pp. 933-966.
Farrington, D. 1983, ‘Randomized experiments on crime and justice’, in M. Tonry &
N. Morris (eds), Crime and Justice: A Review of Research, vol. 4, University
of Chicago, Chicago, pp. 257-307.
Farrington, D. & Langan, P.A. 1992, ‘Changes in crime and punishment in England
and America in the 1980s’, Justice Quarterly, vol. 9, no. 1, pp. 5-46.
Fetherson, J., Lenton, S. & Cercarelli, L.R. 2002, The Repeat Drink Drivers Study,
National Drug Research Institute, Curtin University of Technology, Perth,
WA.
Gibbs, J. P. 1975, Crime, Punishment and Deterrence, Elsevier Scientific Publishing
Company, New York.
Gottfredson, M. & Hirschi, T. 1990, A General Theory of Crime, Stanford University
Press, Stanford, California.
Grasmick, H. G. & Bryjak, G.J. 1980, ‘The Deterrent Effect of Perceived Severity of
Punishment’, Social Forces, vol. 59, no. 2, pp. 471-491.

241
Grasmick, H. G. & Bursik, R.J. 1990, ‘Conscience, significant others, and rational
choice: Extending the deterrence model’, Law and Society Review, vol. 24,
no. 3, pp. 837-861.
Grasmick, H. G., Bursik, R. J. & Arneklev, B.J. 1993, ‘Reduction in drunk driving as
a response to increased threats of shame, embarrassment, and legal
sanctions’, Criminology, vol. 31. no. 1, pp. 41-67.
Grasmick, H. G., Tittle, C.R., Bursik, R.J. & Arneklev, B.J. 1993, ‘Testing the core
empirical implications of Gottfredson and Hirschi's general theory of crime’,
Journal of Research in Crime and Delinquency, vol. 30, no. 1, pp. 5-29.
Grosvenor, D., Toomey, T.L. & Wagenaar, A.C. 1999, ‘Deterrence and the
Adolescent Drinking Driver’, Journal of Safety Research, vol. 30, no. 3, pp.
187-191.
Gruenewald, P.J., Miller, A.B. & Roeper, P. 1996, ‘Access to alcohol: geography
and prevention for local communities’, Alcohol Health and Research World,
vol. 20, pp. 244-251.
Heckathorn, D.D. 1990, ‘Collective sanctions and compliance norms: A formal
theory of group-mediated social control’, American Sociological Review, vol.
55, pp. 366-384.
Henstridge, J., Homel, R. & Mackay, P. 1997, The long term effects of RBT in 4
Australian states: A time series analysis, Federal Office Of Road Safety,
Canberra, ACT.
Hingson, R., Heeren, T., Kovenock, D., Mangione, T., Meyers, A., Morelock, S.,
Lederman, R. & Scotch, N. A. 1987, ‘Effects of Maine's 1981 and
Massachusetts' 1982 Driving-Under-the-Influence legislation’, American
Journal of Public Health, vol. 77, no. 5, pp. 593-597.
Hirschi, T. 1969, Causes of Delinquency, University of California Press, Berkeley.
Holden, R. T. 1983, ‘Rehabilitative sanctions for drunk driving: An experimental
evaluation’, Journal of Research in Crime and Delinquency, vol. 20, pp. 55-
72.
Homel, R. 1980, Penalties and the Drink Driver: A Study of One Thousand
Offenders, NSW Bureau of Crime Statistics and Research, Sydney, NSW.
Homel, R. 1986, Policing the Drinking Driver: Random Breath Testing and the
Process of Deterrence, Federal Office of Road Safety, Canberra, ACT.

242
Homel, R. 1988, Policing and Punishing the Drinking Driver: A Study of General
and Specific Deterrence, Springer-Verlag, New York.
Homel, R. 1994, ‘Drink-driving law enforcement and the legal blood alcohol limit in
NSW’, Accident, Analysis and Prevention, vol. 26, no. 2, pp. 147-155.
Homel, R. 2004, ‘Drinking and driving’, in T. Babor, R. Caetano, S. Casswell, G.
Edwards, N. Giesbrecht, K. Graham, J. Grube, P. Gruenewald, L. Hill, H.
Holder, R. Homel, E. Osterberg, J. Rehm, R. Room & I. Rossow (eds),
Alcohol: No Ordinary Commodity, Oxford University Press, London, pp.
157-172.
Homel, R., Carseldine, D. & Kearns, I. 1988, ‘Drink-driving countermeasures in
Australia’, Alcohol, Drugs and Driving, vol. 4, no. 2, pp. 113-144.
Horney, J. & Marshall, I.H. 1992, ‘Risk perceptions among serious offenders: The
role of crime and punishment’, Criminology, vol. 30, no. 4, pp. 575-594.
Hough, M. & Roberts, J. 1998, Attitudes to punishment: findings from the British
Crime Survey, Home Office, London, England.
Howe, E. S. & Brandau, C.J. 1988, ‘Additive effects of certainty, severity and
celerity of punishment on judgments of crime deterrence scale value’,
Journal of Applied Social Psychology, vol. 18, no. 9, pp. 796-812.
Howe, E. S. & Loftus, T.C. 1996, ‘Integration of certainty, severity and celerity
information in judged deterrence value: Further evidence and methodological
equivalence’, Journal of Applied Social Psychology, vol. 26, no. 3, pp. 226-
242.
Johnson, E. & Payne, J. 1986, ‘The decision to commit a crime: An information-
processing analysis’, in D. B. Cornish & R. V. Clarke (eds), The Reasoning
Criminal: Rational Choice Perspectives on Offending, Springer-Verlag, New
York.
Jones, R. K. & Lacey, J.H. 2001, Alcohol and Highway Safety 2001: A Review of the
State of Knowledge, National Highway and Traffic Safety Administration,
Washington, DC.
Keane, C., Maxim, P.S. & Teevan, J.J. 1993, ‘Drinking and driving, self-control and
gender: testing a general theory of crime’, Journal of Research in Crime and
Delinquency, vol. 30, no. 1, pp. 30-46.

243
Kenkel, D. S. 1993, ‘Drinking and driving and deterrence: the effectiveness and
social costs of alternative policies’, Journal of Law and Economics, vol. 36,
pp. 877-913.
Kenkel, D. S. & Koch, S.F. 2001, ‘Deterrence and knowledge of the law: The case of
drunk driving’, Applied Economics, vol. 33, pp. 845-854.
Kleinbaum, D. G. 1996, Survival Analysis: A Self-Learning Text, Springer-Verlag,
New York.
Kleinbaum, D. G., Kupper, L.L. & Muller, K.E. 1988, Applied Regression Analysis
and Other Multivariable Methods, Kent Publishing Company, Boston.
Klepper, S. & Nagin, D.S. 1989a, ‘Tax compliance and perceptions of the risks of
detection and criminal prosecution’, Law and Society Review, vol. 23, no. 2,
pp. 209-240.
Klepper, S. & Nagin, D.S. 1989b, ‘The deterrent effect of perceived certainty and
severity of punishment revisited’, Criminology, vol. 27, no. 4, pp. 721-746.
Koop, G. 2000, Analysis of Economic Data, Wiley & Sons, Chichester.
Lanza-Kaduce, L. 1988, ‘Perceptual deterrence and drinking and driving among
college students’, Criminology, vol. 26, no. 2, pp. 321-341.
Legge, J. S. & Park, J. 1994, ‘Policies to reduce alcohol-impaired driving: Evaluating
elements of deterrence’, Social Science Quarterly, vol. 75, no. 3, pp. 594-606.
Levitt, S. D. 1996, ‘The effect of prison population size on crime rates: Evidence
from prison overcrowding litigation’, The Quarterly Journal of Economics,
vol. 111, May Issue, pp. 319-350.
Lind, B., Weatherburn, D., Chen, S., Shanahan, M., Lancsar, E., Haas, M. & De
Abreu Lourenco, R. 2002, New South Wales Drug Court Evaluation: Cost-
Effectiveness, NSW Bureau of Crime Statistics and Research, Sydney, NSW.
Liu, S., Siegal, P., Brewer, R., Mokdad, A., Sleet, D. & Serdula, M. 1997,
‘Prevalence of alcohol-impaired driving: Results from a national self-reported
survey of health behaviors’, Journal of American Medical Association, vol.
277, no. 2, pp. 122-125.
Luke, G. & Lind B. 2002, ‘Reducing juvenile crime: Conferencing versus court’,
Crime and Justice Bulletin, no. 69, NSW Bureau of Crime Statistics and
Research, Sydney, NSW.

244
MacCoun, R. J. 1993, ‘Drugs and the law: A psychological analysis of drug
prohibition’, Psychological Bulletin, vol. 113, no. 3, pp. 497-512.
Mann, R. E., Smart, R. G., Stoduto, G., Adlaf, E., Vingilis, E. R., Beirness, D. J. &
Lamble, R. 2000, ‘Changing drinking-and-driving behaviour: The effects of
Ontario's administrative driver's licence suspension law’, Canadian Medical
Association Journal, vol. 162, no. 8, pp. 1141-1142.
Mann, R. E., Vingilis, E. R., Gavin, D., Adlaf, E. & Anglin, L. 1991, ‘Sentence
severity and the drinking driver; relationships with traffic safety outcomes’,
Accident, Analysis and Prevention, vol. 23, no. 6, pp. 483-491.
Martin, S. E., Annan, S. & Forst, B. 1993, ‘The special deterrent effects of a jail
sanction on first-time drunk drivers: A quasi-experimental study’, Accident,
Analysis and Prevention, vol. 25, no. 5, pp. 561-568.
Mayhew, D. R., Beirness, D.J. & Simpson, H.M. 1995, Impact of the 1985
Amendments to the Drink Driving Section of the Criminal Code of Canada,
Proceedings of the 13th International Conference on Alcohol, Drugs and
Traffic Safety, Aug 13-18, Adelaide, SA.
McArthur, D. L. & Kraus, J.F. 1999, ‘The specific deterrence of administrative per
se laws in reducing drunk driving recidivism’, American Journal of
Preventive Medicine, vol. 16, supp. 1, pp. S68-S75.
McLean, A. J., Kloeden, C.N. & McCaul, K.A. 1991, ‘Drink-driving in the general
night-time driving population, Adelaide 1989’, Australian and New Zealand
Journal of Public Health, vol. 15, no. 3, pp. 190-193.
Meier, R. F. & Johnson, W.T. 1977, ‘Deterrence as social control: The legal and
extralegal production of conformity’, American Sociological Review, vol. 42,
April Issue, pp. 292-304.
Miller, S. L. & Iovanni, L. 1994, ‘Determinants of perceived risk of formal sanction
for courtship violence’, Justice Quarterly, vol. 11, no. 2, pp. 281-312.
Mitchell-Taverner, P. 2002a, Community Attitudes to Road Safety: Community
Attitudes Survey Wave 14, 2001, Australian Transport Safety Bureau,
Canberra, ACT.
Mitchell-Taverner, P. 2002b, Community Attitudes to Road Safety: Community
Attitudes Survey Wave 15, 2002, Australian Transport Safety Bureau,
Canberra, ACT.

245
Mitchell-Taverner, P. 2000, Community Attitudes to Road Safety: Community
Attitudes Survey Wave 13, 2000, Australian Transport Safety Bureau,
Canberra, ACT.
Mitchell-Taverner, P. 1999, Community Attitudes to Road Safety: Community
Attitudes Survey Wave 12, Australian Transport Safety Bureau, Canberra,
ACT.
Mitchell-Taverner, P. 1998, Community Attitudes to Road Safety: Community
Attitudes Survey Wave 11, Federal Office of Road Safety, Canberra, ACT.
Mitchell-Taverner, P. 1997, Community Attitudes to Road Safety: Community
Attitudes Survey Wave 10, Federal Office of Road Safety, Canberra, ACT.
Mitchell-Taverner, P., Adams, K. & Hejtmanek, S. 1996, Community Attitudes to
Road Safety: Community Attitudes Survey Wave 9, Federal Office of Road
Safety, Canberra, ACT.
Mitchell-Taverner, P., Adams, K. & Hejtmanek, S. 1995, Community Attitudes to
Road Safety: Community Attitudes Survey Wave 8, Federal Office of Road
Safety, Canberra, ACT.
Mitchell-Taverner, P., & Adams, K. 1994, Community Attitudes to Road Safety:
Community Attitudes Survey Wave 7, Federal Office of Road Safety,
Canberra, ACT.
Moffatt, S., Fitzgerald, J. & Weatherburn, D. 2004, ‘Sentencing drink-drivers: The
use of dismissals and conditional discharges’, Contemporary Issues in Crime
and Criminal Justice, no. 81, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
Morris, N. & Zimring, F.E. 1969, ‘Deterrence and corrections’, The Annals of the
American Academy of Political and Social Science, vol. 381, pp. 137-146.
Nagin, D. S. 1998, ‘Criminal Deterrence Research at the Outset of the 21st Century’,
in M. Tonry (ed) Crime and Justice: A Review of Research, vol. 23,
University of Chicago Press, Chicago, pp. 1-42.
Nagin, D. S. & Paternoster, R. 1993, ‘Enduring individual differences and rational
choice theories of crime’, Law and Society Review, vol. 27, no. 3, pp. 467-
496.

246
Nagin, D. S. & Paternoster, R. 1994, ‘Personal capital and social control: The
deterrence implications of a theory of individual differences’, Criminology,
vol. 32, no. 4, pp. 581-598.
Nagin, D. S. & Pogarsky, G. 2001, ‘Integrating celerity, impulsivity, and extralegal
sanction threats into a model of general deterrence: Theory and evidence’,
Criminology, vol. 39, no. 4, pp. 865-891.
Nagin, D. S. & Pogarsky, G. 2003, ‘An experimental investigation of deterrence:
Cheating, self-serving bias and impulsivity’, Criminology, vol. 41, no. 1, pp.
167-193.
Nathens, A., Jurkovich, G., Jurkovich, G., Cummings, P., Rivara, F. & Maier, R.
2000, ‘The effect of organized systems of trauma care on motor vehicle crash
mortality’, Journal of American Medical Association, vol. 283, no. 15, pp.
1990-1994.
National Health and Medical Research Council 2001, Australian Alcohol Guidelines:
Health Risks and Benefits, Commonwealth of Australia, Canberra, ACT.
National Highway Traffic Safety Administration 2000, On DWI Laws in Other
Countries, National Highway Traffic Safety Administration, Washington DC.
Nichols, J. L. & Ross, H.L. 1990, ‘The effectiveness of legal sanctions in dealing
with drinking drivers’, Alcohol, Drugs and Driving, vol. 6, no. 2, pp. 33-60.
NSW Bureau of Crime Statistics and Research 2003, New South Wales Recorded
Crime Statistics 2002, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
NSW Bureau of Crime Statistics and Research 2002a, New South Wales Recorded
Crime Statistics 2001, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
NSW Bureau of Crime Statistics and Research 2002b, New South Wales Criminal
Courts Statistics 2001, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
NSW Bureau of Crime Statistics and Research 2001, New South Wales Recorded
Crime Statistics 2000, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.

247
NSW Bureau of Crime Statistics and Research 2000, New South Wales Recorded
Crime Statistics 1999, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
NSW Bureau of Crime Statistics and Research 1999, New South Wales Recorded
Crime Statistics 1998, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
NSW Bureau of Crime Statistics and Research 1998, New South Wales Recorded
Crime Statistics 1997, NSW Bureau of Crime Statistics and Research,
Sydney, NSW.
Paternoster, R. 1989, ‘Decisions to participate in and desist from four types of
common delinquency: Deterrence and the rational choice perspective’, Law
and Society Review, vol. 23, no. 1, pp. 7-40.
Paternoster, R. & Iovanni, L. 1986, ‘The Deterrent Effect of Perceived Severity: A
Reexamination’, Social Forces, vol. 64, no. 3, pp. 751-777.
Paternoster, R. & Piquero, A. 1995, ‘Reconceptualizing deterrence: An empirical test
of personal and vicarious experiences’, Journal of Research in Crime and
Delinquency, vol. 32, no. 3, pp. 251-286.
Paternoster, R., Saltzman, L.E., Chiricos, T. G. & Waldo, G. P. 1982, ‘Perceived risk
and deterrence: Methodological artefacts in perceptual deterrence research’,
The Journal of Criminal Law and Criminology, vol. 73, no. 3, pp. 1238-1258.
Paternoster, R., Saltzman, L.E., Waldo, G. P. & Chiricos, T. G. 1983, ‘Perceived
risks and social control: Do sanctions really deter?’, Law and Society Review,
vol. 17, no. 3, pp. 458-479.
Paternoster, R. & Simpson, S. 1996, ‘Sanction threats and appeals to morality:
Testing a rational choice model of corporate crime’, Law and Society Review,
vol. 30, no. 3, pp. 549-583.
Peck, R. C. 1991, ‘The general and specific deterrent effects of DUI sanctions: A
review of California's experience’, Alcohol, Drugs and Driving, vol. 7, no. 1,
pp. 13-42.
Peterson, R. D. & Bailey, W. C. 1988, ‘Murder and capital punishment in the
evolving context of the Post-Furman Era’, Social Forces, vol. 66, pp. 774-
807.

248
Piliavin, I., Gartner, R., Gartner, R., Thornton, C. & Matseuda, R. L. 1986, ‘Crime
deterrence and rational choice’, American Sociological Review, vol. 51, no. 1,
pp. 101-119.
Piquero, A. & Paternoster, R. 1998, ‘An application of Stafford & Warr's
reconceptualisation of deterrence to drinking and driving’, Journal of
Research in Crime and Delinquency, vol. 35, no. 1, pp. 3-37.
Piquero, A. & Pogarsky, G. 2002, ‘Beyond Stafford and Warr's reconceptualization
of deterrence: Personal and vicarious experiences, impulsivity, and offending
behavior’, Journal of Research in Crime and Delinquency, vol. 39, no. 2, pp.
153-186.
Piquero, A. & Rengert, G.F. 1999, ‘Studying deterrence with active residential
burglars’, Justice Quarterly, vol. 16, no. 2, pp. 451-471.
Pogarsky, G. 2002, ‘Identifying "deterrable" offenders: Implications for research on
deterrence’, Justice Quarterly, vol. 19, no. 3, pp. 431-452.
Pogarsky, G. & Piquero, A. 2003, ‘Can punishment encourage offending?
Investigating the "resetting" effect’, Journal of Research in Crime and
Delinquency, vol. 40, no. 1, pp. 95-120.
Pogarsky, G. & Piquero, A. 2004, ‘Studying the reach of deterrence: Can deterrence
theory help explain police misconduct?’, Journal of Criminal Justice, vol. 32,
pp. 371-386.
Pogarsky, G., Piquero, A. & Paternoster, R. 2004, ‘Modeling change in perceptions
about sanction threats: The neglected linkage in deterrence theory’, Journal
of Quantitative Criminology, vol. 20, no. 4, pp. 343-369.
Robinson, P. H. & Darley, J.M. 2004, ‘Does criminal law deter? A behavioural
science investigation’, Oxford Journal of Legal Studies, vol. 24, no. 2, pp.
173-205.
Rogers, P. N. & Schoenig, S.E. 1994, ‘A time-series evaluation of California's 1982
driving-under-the-influence legislative reforms’, Accident, Analysis and
Prevention, vol. 26, no. 1, pp. 63-78.
Ross, H. L. 1984, Deterring the Drinking Driver: Legal Policy and Social Control
(revised and updated edition), Lexington Books, Lexington.
Ross, H. L. & Blumenthal, M. 1975, ‘Some problems in experimentation in a legal
setting’, The American Sociologist, vol. 10, August Issue, pp. 150-155.

249
Ross, H. L. & Gonzales, P. 1988, ‘Effects of license revocation on drunk-driving
offenders’, Accident, Analysis and Prevention, vol. 20, no. 5, pp. 379-391.
Ross, H. L. & LaFree, G.D. 1986, ‘Deterrence in criminology and social policy’, in
N. J. Smesler & D. R. Gerstein (eds), Behavioral and Social Science: Fifty
Years of Discovery, National Academy Press, Washington, DC, pp 129-152.
Ross, H. L. & McCleary R. 1983, ‘Methods for studying the impact of drunk driving
laws’, Accident, Analysis and Prevention, vol. 15, no. 6, pp. 415-428.
Ross, H. L. & Voas, R.B. 1990, ‘The New Philadelphia story: The effects of severe
punishment for drunk driving’, Law and Policy, vol. 12, no. 1, pp. 51-77.
Ryan, G. A., Ferrante, A.M., Loh, N. & Cercarelli, L. R. 1996, Repeat Drink-Driving
Offenders in Western Australia, 1984 to 1994, Federal Office of Road Safety,
Canberra, ACT.
Saffer, H. & Chaloupka, F.J. 1989, ‘Breath testing and highway fatalities’, Applied
Economics, vol. 21, no. 7, pp. 901-912.
Saffer, H. & Grossman, M. 1987, ‘Beer taxes, the legal drinking age and youth motor
vehicle fatalities’, Journal of Legal Studies, vol. 16, pp 351-374.
Sanson, A., Montgomery, B., Gault, U., Gridley, H. & Thomson, D. 1996,
‘Punishment and behaviour change: An Australian Psychological Society
Position Paper’, Australian Psychologist, vol. 31, pp. 157-165.
Sherman, L. W. 1990, ‘Police Crackdowns: Initial and Residual Deterrence’, in M.
Tonry & N. Morris (eds), Crime and Justice: A Review of Research, vol. 12,
University of Chicago Press, Chicago, pp. 1-48.
Sherman, L. W., Gottfredson, D., MacKenzie, D., Eck, J., Reuter, P. & Bushway, S.
1998, Preventing Crime: What Works, What Doesn't, What's Promising?,
National Institute of Justice, Washington, DC.
Sherman, L. W., Smith, D.A., Schmidt, J. D. & Rogan, D. P. 1992, ‘Crime,
punishment and stake in conformity: Legal and informal control of domestic
violence’, American Sociological Review, vol. 57, pp. 680-690.
Shults, R. A., Elder, R.W., Shults, R. A., Elder, R. W., Sleet, D. A., Nichols, J. L.,
Alao, M. O., Carande-Kulis, V. G., Zaza, S., Sosin, D. M., Thompson, R. S.
& Task Force on Community Preventive Services 2001, ‘Reviews of
evidence regarding intervention to reduce alcohol-impaired driving’,
American Journal of Preventive Medicine, vol. 21, supp. 4, pp. S66-S88.

250
Simon, H.A. 1983, Reason in Human Affairs, Blackwell, Oxford.
Siskind, V. 1996, ‘Does license disqualification reduce reoffence rates’, Accident,
Analysis and Prevention, vol. 28, no. 4, pp. 519-524.
Sloan, F. A., Reilly, B.A., & Schenzler, C. 1994, ‘Effects of prices, civil and criminal
sanctions, and law enforcement on alcohol-related mortality’, Journal of
Studies on Alcohol, vol. 55, pp. 454-465.
Smith, D. I. 1988, ‘Effect on traffic accidents of introducing Sunday alcohol sales in
Brisbane, Australia’, International Journal of the Addictions, vol. 23, no. 10,
pp. 1091-1099.
Snortum, J. R. & Berger, D.E. 1989, ‘Drinking-driving compliance in the United
States: Perceptions and behavior in 1983 and 1986’, Journal of Studies on
Alcohol, vol. 50, vol. 4, pp. 306-319.
South, D. 1998, ‘General deterrence and behaviour change: A comment on the
Australian Psychological Society position paper on punishment and
behaviour change’, Australian Psychologist, vol. 33, no. 1, pp. 76-78.
Span, D. 1995, Research on Knowledge, Attitudes and Reported Behaviour on Drink-
Driving in New South Wales, Proceedings of the 13th International
Conference on Alcohol, Drugs and Traffic Safety, Aug. 13-18, Adelaide, SA.
Span, D. & Stanislaw, H. 1995, Evaluation of the Long Term Impact of a
Deterrence-Based Random Breath Testing Program in New South Wales,
Proceedings of the 13th International Conference on Alcohol, Drugs and
Traffic Safety, Aug. 13-18, Adelaide, SA.
Spegilman, J.J. 2005, ‘A new way to sentence for serious crime’, Judicial Officers’
Bulletin, vol. 17, no. 1, pp. 1-4.
Spelman, W. 2000, ‘What Recent Studies Do (and Don't) Tell Us about
Imprisonment and Crime’, in M. Tonry (ed), Crime and Justice: A Review of
Research, vol. 27, University of Chicago Press, Chicago.
Spohn, C. & Holleran, D. 2002, ‘The effect of imprisonment on recidivism rates of
felony offenders: A focus on drug offenders’, Criminology, vol. 40, no. 2, pp.
329-357.
Stafford, M. C. & Warr, M. 1993, ‘A reconceptualization of general and specific
deterrence’, Journal of Research in Crime and Delinquency, vol. 30, no. 2,
pp. 123-135.

251
Stewart, K., Gruenewald, P.J., & Parker, R. N. 1992, ‘Assessing legal change:
Recidivism and administrative per se laws’, Journal of Quantitative
Criminology, vol. 8, no. 4, pp. 375-394.
Stoduto, G., Mann, R.E., Smart, R. G., Adlaf, E. M., Vingilis, E. R., Beirness, D. J.
& Lamble, R. 2000, The Impact of the Administrative Driver's Licence
Suspension Law in Ontario, paper presented at the 15th International
Conference on Alcohol, Drugs and Traffic Safety, May 22-26, Stockholm,
Sweden.
Stover, R.V. & Brown, D. W. 1975, ‘Understanding compliance and non-compliance
with the law’, Social Science Quarterly, vol. 56, pp. 361-375.
Tibbetts, S. G. 1997, ‘Shame and rational choice in offending decisions’, Criminal
Justice and Behavior, vol. 24, no. 2, pp. 234-255.
Tittle, C. R. 1977, ‘Sanction fear and the maintenance of social order’, Social Forces,
vol. 55, no. 3, pp. 579-596.
Tittle, C. R. 1980, ‘Evaluating the deterrent effects of criminal sanctions’, in M. W.
Klein & K. S. Teilman (eds), Handbook of Criminal Justice Evaluation, Sage
Publications, Beverley Hills.
Tonry, M. 1996, Sentencing Matters, Oxford University Press, New York.
Tversky, A. & Kahneman, D. 1981, ‘The framing of decisions and the psychology of
choice’, Science, vol. 211, January Issue, pp. 453-458.
Tyler, T. R. 1990, Why People Obey the Law, Yale University Press, New Haven.
Voas, R. B. 2001, ‘Have the courts and the motor vehicle departments adequate
power to control the hard-core drunk driver?’, Addiction, vol. 96, pp. 1701-
1707.
Voas, R. B. & Hause, J.M. 1987, ‘Deterring the drinking driver: The Stockton
experience’, Accident, Analysis and Prevention, vol. 19, no. 2, pp. 81-90.
Voas, R. B., Holder, H. & Gruenewald, P. 1997, ‘The effect of drinking and driving
interventions on alcohol-involved traffic crashes within a comprehensive
community trial’, Addiction, vol. 92, supp. 2, pp. S221-S236.
Voas, R. B., Marques, P.R., Tippetts, A. S. & Beirness, D. J. 1999, ‘The Alberta
interlock program: The evaluation of a province-wide program on DUI
recidivism’, Addiction, vol. 94, no. 12, pp. 1849-1859.

252
Voas, R. B. & Tippetts, A.S. 1999, The Relationship of Alcohol Safety Laws to
Drinking Drivers in Fatal Crashes, National Highway Traffic Safety
Administration, Washington, DC.
von Hirsch, A., Bottoms, A.E., Burney, E. & Wikstrom, P. O. 1999, Criminal
Deterrence and Sentence Severity, Hart Publishing, Oxford.
Votey, H. L. & Shapiro, P. 1983, ‘Highway accidents in Sweden: Modelling the
process of drunken driving behaviour and control’, Accident, Analysis and
Prevention, vol. 15, no. 6, pp. 523-533.
Waldo, G. P. & Chiricos, T.G. 1972, ‘Perceived penal sanctions and self-reported
criminality: A neglected approach to deterrence research’, Social Problems,
vol. 19, no. 4, pp. 522-540.
Weatherburn, D., Lind, B. & Hua, J. 2003, ‘Contact with the New South Wales court
and prison systems: The influence of age Indigenous status and gender’,
Contemporary Issues in Crime and Criminal Justice, no. 78, NSW Bureau of
Crime Statistics and Research, Sydney, NSW.
Weatherburn, D., Topp, L., Midford, R. & Allsopp, S. 2000, Drug Crime Prevention
and Mitigation: A Literature Review and Research Agenda, NSW Bureau of
Crime Statistics and Research, Sydney, NSW.
Weinwrath, M. & Gartrell, J. 2001, ‘Specific deterrence and sentence length’,
Journal of Contemporary Criminal Justice, vol. 17, no. 2, pp. 105-122.
Wells-Parker, E. 2000, Assessment and screening of impaired driving offenders: An
analysis of underlying hypotheses as a guide for development of validation
strategies, paper presented at the 15th International Conference on Alcohol,
Drugs and Traffic Safety, May 22 – 26, Stockholm, Sweden.
Wells-Parker, E., Bangert-Downs, R., McMillen, R. & Williams, M. 1995, ‘Final
results from a meta-analysis of remedial interventions with drink/drive
offenders’, Addiction, vol. 90, pp. 907-926.
Wilkinson, J. T. 1987, ‘Reducing drunken driving: Which policies are most
effective?’, Southern Economic Journal, vol. 54, pp. 322-334.
Williams, A. F., Ferguson, S.A., & Cammisa, M. X. 2000, Self-reported Drinking
and Driving Practices and Attitudes in Four Countries and Receptions of
Enforcement, paper presented at the 15th International Conference on Alcohol
Drugs and Traffic Safety, May 22 – 26, Stockholm, Sweden.

253
Williams, K. R. & Gibbs, J.P. 1981, ‘Deterrence and knowledge of statutory
penalties’, The Sociological Quarterly, vol. 22, Autumn Issue, pp. 591-606.
Williams, K. R., Gibbs, J.P., & Erickson, M. L. 1980, ‘Public knowledge of statutory
penalties: The extent and basis of accurate perception’, Pacific Sociological
Review, vol. 23, no. 1, pp. 105-128.
Williams, K. R. & Hawkins, R. 1986, ‘Perceptual research on general deterrence: A
critical review’, Law and Society Review, vol. 20, no. 4, pp. 545-572.
Wilson, J.Q. & Hernstein, R. 1985, Crime and Human Nature, Simon Schuster, New
York.
Wright, B. R. E., Caspi, A., Moffit, T. E. & Paternoster, R. 2004, ‘Does the
perceived risk of punishment deter criminally prone individuals? Rational
choice, self-control, and crime’, Journal of Research in Crime and
Delinquency, vol. 41, no. 2, pp. 180-213.
Yin, R. 1994, Case Study Research: Design and Methods, Sage, Thousand Oaks,
California.
Yu, J. 1994, ‘Punishment celerity and severity: Testing a specific deterrence model
on drunk driving recidivism’, Journal of Criminal Justice, vol. 22, no. 4, pp.
355-366.
Zaal, D. 1994, Traffic Law Enforcement: A Review of the Literature, Monash
University Accident Research Centre, Melbourne, Victoria.
Zador, P. L. 1991, ‘Alcohol-related relative risk of fatal driver injuries in relation to
driver age and sex’, Journal of Studies on Alcohol, vol. 52, no. 4, pp. 302-
310.
Zador, P. L., Krawchuk, S.A. & Voas, R. B. 2000, ‘Alcohol-related relative risk of
driver fatalities and driver involvement in fatal crashes in relation to driver
age and gender: An update using 1996 data’, Journal of Studies on Alcohol,
vol. 61, pp. 387-395.
Zimring, F. E. & Hawkins, G.J. 1973, Deterrence: The Legal Threat in Crime
Control, University of Chicago Press, Chicago.

254
Appendices
Appendix I – Summary of drink-driving studies reviewed in
Chapter 3
The tables below present summary data from each of the studies reviewed in Chapter
3. Included in these tables is information on the jurisdiction studied (i.e. State,
Territory or Province), the type of study design employed and the specific outcome
measures analysed (e.g. self-reported drink driving or road fatalities). In the case of
recidivism studies on penalty severity, Table A3 includes additional information on
the type of penalty examined. If the study described found a statistically significant
reduction, in the outcome examined, this is also indicated in the table.

Table A1: Aggregate-level studies on punishment certainty

Study Jurisdiction Design Outcome measure Statistically
significant
reduction?
Wilkinson 1987 USA Ecological Road fatalities No
Evans, Neville USA Ecological Road fatalities No
& Graham 1991
Sloan, Reilly & USA Ecological Road fatalities Yes
Schenzler 1994
Saffer & USA Ecological Road fatalities Yes
Chaloupka 1994
Benson, USA Ecological Driver involvement rate Yes
Rasmussen &
Mast 1999
Kenkel 1993 USA Ecological Self reported drink-driving Yes
Self reported drinking Yes
Chaloupka & USA Ecological Self-reported drinking Yes
Weschler 1996
Ross 1984 Britain Quasi- Road fatalities Yes
experimental Serious injury accidents
Voas & Hause Stockton, CA, Quasi- Night-time accidents Yes
1987 USA experimental Roadside breath tests
Voas, Holder & California, Quasi- SVNT accidents & positive Yes
Gruenewald USA experimental BACs Yes
1997 Self reported drink-driving
Henstridge, NSW, QLD, Quasi- Road fatalities Yes
Homel & WA & experimental SVNT accidents
Mackay 1997 Tasmania,
Australia
NB: SVNT accidents are those where only one vehicle is involved in an accident occurring at night

255
Table A2: Aggregate-level studies on punishment severity
Study Jurisdiction Design Outcome measure Statistically
significant
reduction?
Rogers & California, Quasi- Night-time accidents Yes
Shoenig 1994 USA experimental SVNT accidents
Accidents designated by
police as alcohol-related
Peck 1991 California, Quasi- Recidivism Yes
USA experimental
Borschos 2000 Sweden Quasi- Road fatalities Yes
experimental Accidents w/ severe injury
Beirness, Canada Quasi- Road fatalities & arrests No
Simpson & experimental Self-reported drink-driving
Mayhew 1993 & positive BACs
Ross 1984 Finland Quasi- Road fatalities No
experimental Serious injury accidents
Single-vehicle fatalities
Hingson et al. Massachusetts Quasi- Road fatalities No
1987 & Maine, USA experimental Self-reported drink-driving
Homel 1986 NSW, Quasi- Self-reported drink-driving Yes
Australia experimental
Ross & Voas New Quasi- Roadside breath tests No
1990 Philadelphia, experimental Self-reported drink-driving
USA
Sloan, Reilly & USA Ecological Road fatalities No
Schenzler 1994
Chaloupka, USA Ecological Road fatalities Yes
Saffer &
Grossman 1993
Votey & Shapiro Sweden Ecological Road fatalities Yes
1983 Serious injury accidents

Table A3: Individual-level studies on punishment severity

Study Jurisdiction Design Outcome Penalty type Stat. sig.
measure reduction?
Holden 1983 USA Experimental Rearrest rates Supervised No
probation,
Education &
Therapy
Bloomberg Wisconsin, Quasi- Rearrest rates 3-month Yes
et al. 1987 USA experimental Road accident minimum
rates licence
disqualification
DeYoung California, Cross-sectional Reconviction 2-day gaol No (gaol)
1997 USA rates sentences
Licence Yes
sanctions
Homel 1980 NSW, Cross-sectional Reconviction Gaol sentences No
Australia rates Licence
sanctions
Fines
Probation
Martin, Minnesota, Cross-sectional Reconviction 2-day gaol No
Annan & USA rates sentences
Forst 1993

256
 Siskind 1996 Qld, Cross-sectional Rearrest rate Licence Yes
Australia Road accident sanctions
rate
Yu 1994 New York, Cross-sectional Reconviction Licence No
USA rate sanctions
Fine Yes (fines)
Mann et al Ontario, Cross-sectional Rearrest rate Gaol sentences No (gaol)
1991 Canada Road accident Licence Yes
rate sanctions
Fines No (fines)

Table A4: Aggregate and Individual-level studies on punishment celerity

Study Jurisdiction Design Outcome measure Statistically
significant
reduction?
Nagin & Pogarsky Students, Experimental Likelihood of committing No
2001 USA a drink-driving offence
Simon, Cleary & Minnesota, Cross-sectional Reconviction rates Yes
Storkamp 2000 USA
Yu 1994 New York, Cross-sectional Reconviction rates No
USA
Mann et. al. 2000 Ontario, Quasi- Proportion of fatalities w/ Yes
Canada experimental positive BACs
Self-reported drink-driving
Voas & Tippets USA Ecological Road fatalities Yes
1999
Stewart, Louisiana, Quasi- Recidivism rates Yes
Gruenewald & Mississippi & experimental
Parker 1992 North
Dakota, USA

257
Appendix II – Participant information sheet and drink-driving
questionnaire

258
 SCHOOL OF SOCIAL SCIENCE & POLICY

Approval No
HREC03163

THE UNIVERSITY OF NEW SOUTH WALES

PARTICIPANT INFORMATION STATEMENT

The deterrent effect of increased penalties for drink-driving

You are invited to participate in a study of drink-driving. From this study we hope
to learn what impact the penalties for drink-driving have on decisions to offend.
We are recruiting students from various disciplines at the University of New South
Wales to ensure that we have responses from a wide range of young people.

If you decide to participate, we will ask you to answer a series of questions related
to drink-driving, which will take you approximately five minutes to complete. To
ensure anonymity, we would ask that you do not write your name or any other
identifying information on any part of the questionnaire.

If you give us your permission by completing this questionnaire, we plan to publish

the results as part of a PhD thesis to be submitted to The University of New South
Wales. In any publication, information will be provided in such a way that you
cannot be identified.

Your participation in this study is voluntary and your decision whether or not to
participate will not prejudice your future relations with The University of New
South Wales. If you decide to participate, you are free to discontinue participation
at any time without prejudice. Your self-completion of this questionnaire indicates
that, having read the information provided above, you have decided to participate.

If you have any questions, please feel free to ask us. If you have any additional
questions later, Suzanne Briscoe (ph 9231 9151) will be happy to answer them.
Any complaints about this study may be directed to the Ethics Secretariat, The
University of New South Wales, SYDNEY 2052 AUSTRALIA (ph 9385 4234, fax
9385 6648, email ethics.sec@unsw.edu.au).

You can keep this form.

259
 THE UNIVERSITY OF NEW SOUTH WALES

The deterrent effect of increased penalties for drink-driving

QUESTIONNAIRE

___________________________________________________

260
Here are some questions about what the law can and cannot do when a person is
convicted of drink-driving in NSW. Please answer all questions in the order in
which they appear by ticking the most correct answer.

A. If a person was caught driving with a blood alcohol level just over the legal limit
(i.e. just over 0.05g/100ml) and it was their first drink-driving conviction……..

1. They can be fined.

 a) True ► The maximum fine allowed is:
 $110
 $220
 $550
 $1,100
 $2,200
 b) False

2. They can have their driver’s licence disqualified.

 a) True ► The minimum licence disqualification allowed is:
 there is no minimum
 1 month
 2 months
 3 months
 6 months
The maximum licence disqualification allowed is:
 3 months
 4 months
 6 months
 9 months
 there is no maximum
 b) False

3. They can be sentenced to imprisonment.

 a) True ► The maximum gaol term allowed is:
 1 month
 3 months
 6 months
 9 months
 12 months
 b) False

261
B. If a person was caught driving with a higher blood alcohol level (i.e. somewhere
between 0.08g/100ml and 0.15g/100ml), and it was their first drink-driving
conviction……

4. They can be fined.

 a) True ► The maximum fine allowed is:
 $110
 $220
 $550
 $1,100
 $2,200
 b) False

5. They can have their driver’s licence disqualified.

 a) True ► The minimum licence disqualification allowed is:
 there is no minimum
 1 month
 2 months
 3 months
 6 months
The maximum licence disqualification allowed is:
 3 months
 4 months
 6 months
 9 months
 there is no maximum
 b) False

6. They can be sentenced to imprisonment.

 a) True ► The maximum gaol term allowed is:
 1 month
 3 months
 6 months
 9 months
 12 months

 b) False

262
The previous questions asked about the minimum and maximum penalties, allowed
by NSW law, for drink-driving offences……

C. Suppose you were caught driving with a blood alcohol level just over the legal
limit (i.e. just over 0.05g/100ml). Give your best estimate as to what the actual
penalty (or penalties) would be if you were convicted of your first drink-driving
offence?

7. You would receive a fine.

 a) True ► The fine would be:
 Less than $220
 $220 - $350
 $351 - $550
 $551 - $1,100
 More than $1,100
 b) False

8. You would have your driver’s licence disqualified.

 a) True ► Your licence would be disqualified for:
 Less than 1 month
 1 – less than 3 months
 3 – less than 6 months
 6 – less than 9 months
 9 months or more
 b) False

9. You would be sentenced to imprisonment.

 a) True ► The gaol term would be:
 Less than 1 month
 1 – less than 2 months
 2 – less than 4 months
 4 – less than 6 months
 6 months or more

 b) False

263
D. Suppose you were caught driving with a higher blood alcohol level (i.e.
somewhere between 0.08g/100ml and 0.15g/100ml). Give your best estimate as to
what the actual penalty (or penalties) would be if you were convicted of your first
drink-driving offence?

10. You would receive a fine.

 a) True ► The fine would be:
 Less than $220
 $220 - $350
 $351 - $550
 $551 - $1,100
 More than $1,100
 b) False

11. You would have your driver’s licence disqualified.

 a) True ► Your licence would be disqualified for:
 Less than 1 month
 1 – less than 3 months
 3 – less than 6 months
 6 – less than 9 months
 9 months or more
 b) False

12. You would be sentenced to imprisonment.

 a) True ► The gaol term would be:
 Less than 1 month
 1 – less than 2 months
 2 – less than 4 months
 4 – less than 6 months
 6 months or more

 b) False

264
(Scenario 1)

Please read the following scenario and answer the questions below:

Suppose you drove by yourself one evening to meet some friends at a pub.
You didn’t intend to drink much alcohol but by the end of the evening,
you’ve had enough drinks so that you’re pretty sure your blood alcohol
level is above the legal limit. You haven’t seen police conducting Random
Breath Tests (RBT) in this area for some time. You live about 10 kms
away and you have to be at work early the next morning. You can either
drive home or find some other way home, but if you leave your car at the
pub, you will have to return early the next morning to pick it up.

13. How likely is it, on a scale of 0 (not at all) to 100 (definitely), that
you would drive home under the circumstances provided in the
scenario above?

______________________________

14. How likely is it, on a scale of 0 (not at all) to 100 (definitely), that
you would be apprehended and convicted for drink-driving if you
drove home?

_______________________________

15. Suppose you did drive home and were caught and convicted for
drink-driving. How big a problem, on a scale of 0 (no problem) to 100
(a very big problem), would the penalties for this offence create for
you?

_________________________________

16. If there was no possibility of you being caught and punished how
likely is it, on a scale of 0 (not at all) to 100 (definitely), that you
would drive home under the circumstances provided in the scenario
above?

__________________________________

265
(Scenario 2)

Please read the following scenario and answer the questions below:

Suppose you drove by yourself one evening to meet some friends at a pub.
You didn’t intend to drink much alcohol but by the end of the evening,
you’ve had enough drinks so that you’re pretty sure your blood alcohol
level is above the legal limit. You have often seen police conducting
Random Breath Tests (RBT) in this area. You live about 10 kms away and
you have to be at work early the next morning. You can either drive home
or find some other way home, but if you leave your car at the pub, you will
have to return early the next morning to pick it up.

13. How likely is it, on a scale of 0 (not at all) to 100 (definitely), that
you would drive home under the circumstances provided in the
scenario above?

______________________________

14. How likely is it, on a scale of 0 (not at all) to 100 (definitely), that
you would be apprehended and convicted for drink-driving if you
drove home?

_______________________________

15. Suppose you did drive home and were caught and convicted for
drink-driving. How big a problem, on a scale of 0 (no problem) to 100
(a very big problem), would the penalties for this offence create for
you?

_________________________________

16. If there was no possibility of you being caught and punished how
likely is it, on a scale of 0 (not at all) to 100 (definitely), that you
would drive home under the circumstances provided in the scenario
above?

__________________________________

266
(Scenario 3)

Please read the following scenario and answer the questions below:

Suppose you drove by yourself one evening to meet some friends at a pub.
You didn’t intend to drink much alcohol but by the end of the evening,
you’ve had enough drinks so that you’re pretty sure your blood alcohol
level is above the legal limit. You have often seen police conducting
Random Breath Tests (RBT) in this area and since it is a Friday night you
know that there will be more police on the roads. You live about 10 kms
away and you have to be at work early the next morning. You can either
drive home or find some other way home, but if you leave your car at the
pub, you will have to return early the next morning to pick it up.

13. How likely is it, on a scale of 0 (not at all) to 100 (definitely), that
you would drive home under the circumstances provided in the
scenario above?

______________________________

14. How likely is it, on a scale of 0 (not at all) to 100 (definitely), that
you would be apprehended and convicted for drink-driving if you
drove home?

_______________________________

15. Suppose you did drive home and were caught and convicted for
drink-driving. How big a problem, on a scale of 0 (no problem) to 100
(a very big problem), would the penalties for this offence create for
you?

_________________________________

16. If there was no possibility of you being caught and punished how
likely is it, on a scale of 0 (not at all) to 100 (definitely), that you
would drive home under the circumstances provided in the scenario
above?

__________________________________

267
Please answer the following questions:

17. Age: _______________

18. Gender:  Male

 Female
19. Degree currently enrolled in _____________

20. Have you ever driven a motor vehicle after drinking too much
alcohol?
 Yes ► If so, how many times _____________
 No

21. Have you ever been convicted of a drink-driving offence?

 Yes ► If so, how many times _____________
 No

22. Have you or any of your relatives ever been involved in an

alcohol-related accident?
 Yes ► If so, how many times ______________
 No

23A. (If you are MALE) In the last 12 months how often did you
drink more than 6 standard alcoholic drinks in one day? By a
standard drink, I mean the equivalent of a middy of full strength beer,
a schooner of light beer, a small glass of wine, a glass of port or a nip
of spirits.
 Every day
 4 to 6 days a week
 2 to 3 days a week
 About 1 day a week
 2 to 3 days a month
 About 1 day a month
 Less often
 Never

(continued over the page)

268
24B. (If you are FEMALE) In the last 12 months how often did you
drink more than 4 standard alcoholic drinks in one day? By a
standard drink, I mean the equivalent of a middy of full strength beer,
a schooner of light beer, a small glass of wine, a glass of port or a nip
of spirits.
 Every day
 4 to 6 days a week
 2 to 3 days a week
 About 1 day a week
 2 to 3 days a month
 About 1 day a month
 Less often
 Never

269
Appendix III – Supplementary analyses

Table A.5: Linear Regression Model for breath tests administered by NSW Police
each month, 1994-2002
Variable Coefficient Standard t p value
error
Intercept 197932.0 15878.85 12.47 < 0.01
Underlying trend -834.38 367.46 -2.27 0.03
Change in underlying 2376.63 544.44 4.37 < 0.01
trend from Oct. ‘98
Change in level of -30913.71 15282.67 -2.02 0.05
series from Oct. ‘98
Seasonal terms Jan. 77195.47 13392.44 5.76 < 0.01
Feb. -19155.00 14359.30 -1.33 0.19
Mar. 48652.79 13421.56 3.62 < 0.01
Apr. 67358.11 14018.71 4.80 < 0.01
May -9320.95 12975.26 -0.72 0.47
Jun. 24328.01 13958.85 1.74 0.08
Jul. -11143.15 11196.71 -1.00 0.32
Aug. -5474.15 13940.52 -0.39 0.70
Oct. 9886.39 14018.28 0.71 0.48
Nov. 18891.79 11262.79 1.68 0.10
Dec. 150095.3 13985.19 10.73 < 0.01
AR(2) 0.4 0.11 3.48 < 0.01
Notes: R2 = 0.807 & Durbin-Watson = 1.929; no autocorrelation or partial autocorrelation in
residuals

270
Table A.6: Linear Regression Model for PCA incidents recorded by NSW Police
each month, 1995-2002
Variable Coefficient Standard t p value
error
Intercept 1095.51 71.03 15.42 < 0.01
Underlying trend 7.00 1.45 4.81 < 0.01
Change in underlying -1.26 1.87 -0.67 0.50
trend from Oct. ‘98
Change in level of 297.33 51.93 5.73 < 0.01
series from Oct. ‘98
Seasonal terms Jan. 116.71 63.29 1.84 0.07
Feb. -250.66 61.18 -3.97 < 0.01
Mar. 311.22 63.09 4.93 < 0.01
Apr. -0.65 63.01 -0.01 0.99
May -58.52 62.95 -0.93 0.36
Jun. -18.14 62.90 -0.29 0.77
Jul. -146.26 62.86 -2.33 0.02
Aug. 2.12 62.84 0.03 0.97
Oct. -75.25 63.09 -1.19 0.24
Nov. 177.91 63.03 2.82 < 0.01
Dec. 250.07 62.98 3.97 < 0.01
Notes: R2 = 0.904 & Durbin-Watson = 1.902; no autocorrelation or partial autocorrelation in
residuals

271
Table A.7: Linear regression models for stated likelihood of offending specified for
prior offending groups
Model 1 Model 2 Model 3
(no priors n=257) (1-5 priors n=88) (>5 priors n=21)
b p b p b p
CERTAINTY
High v. low -6.23 0.11 -10.57 0.14 -28.45 0.08
(3.92) (7.00) (15.22)
KNOWLEDGE OF 1.30 <0.01 -0.53 0.51 -1.01 0.54
PENALTIES (0.35) (0.80) (1.59)
(Statutory penalties)
RISKY ALCOHOL
CONSUMPTION
Monthly v. rarely/never 2.79 0.52 4.67 0.65 39.74 0.19
(4.35) (10.10) (29.13)
Weekly v. rarely/never -5.46 0.31 -8.76 0.38 24.27 0.25
(5.31) (9.92) (20.36)
0.07 0.07 0.38
Notes: Standard error of estimates appear in parentheses.

Table A.8: Linear regression models for stated likelihood of offending specified for
prior offending groups
Model 1 Model 2 Model 3
(no priors n=257) (1-5 priors n=88) (>5 priors n=21)
b p b p b p
CERTAINTY
High v. low -6.74 0.09 -11.30 0.11 -23.80 0.09
(3.91) (7.03) (12.97)
KNOWLEDGE OF 2.47 <0.01 0.60 0.57 5.08 0.02
PENALTIES (0.63) (1.05) (1.95)
(Actual penalties)
RISKY ALCOHOL
CONSUMPTION
Monthly v. rarely/never 2.70 0.53 5.45 0.59 37.12 0.14
(4.33) (10.05) (24.13)
Weekly v. rarely/never -3.08 0.57 -7.50 0.44 33.97 0.06
(5.42) (9.76) (16.57)
0.08 0.07 0.56
Notes: Standard error of estimates appear in parentheses.

272