Professional Documents
Culture Documents
Crime Reporting As Institutional Development
Crime Reporting As Institutional Development
Development*
Rodrigo R. Soares
University of Maryland and IBMEC Business School, São Paulo, Brazil
I. Introduction
This article investigates the determinants of the rate of crime reporting. The
rate of crime reporting is the fraction of the total number of crimes that is
actually reported to the police. The article constructs this variable by crossing
data from official crime records with data from victimization surveys. The
results show that the variation of rates of crime reporting across countries is
strongly related to measures of institutional stability, to police presence, and,
most important, to a subjective index of corruption. This evidence uncovers
the underlying forces determining the correlation between reporting rates and
income per capita noticed by Soares (2004). In addition, it supports the view
that subjective indexes of governance and institutional development indeed
capture relevant dimensions of the performance and efficiency of the public
sector.
The fraction of the total number of crimes reported to the police varies
widely across countries and across different types of crimes, from virtually
zero (as for thefts in Egypt or India) to virtually one (as for burglaries in
Austria and Finland).1 Soares (2004) shows that this variation is strongly
related to income per capita. Figures 1, 2, and 3 reproduce this result and
illustrate the strong positive correlation between reporting rates of, respec-
tively, thefts, burglaries, and contact crimes and income. Income per capita
alone explains 65% of the cross-country variation in reporting rates for thefts,
54% for burglaries, and 45% for contact crimes. Soares also shows that this
correlation is responsible for the criminologists’ erroneous conclusion that
development and crime rates are positively related.
Nevertheless, it is unlikely that income per se is the variable driving the
behavior of reporting rates. In reality, income is probably capturing effects
related to development that are difficult to measure precisely, such as insti-
tutional maturity, degree of law enforcement, and community cohesiveness.
The goal of this article is to uncover the mechanism behind the correlation
between reporting rates and income per capita.
The appeal of this topic to the crime literature is obvious, but we believe
that its relevance is much broader. The role of institutions in the process of
economic development has received increased attention in recent years (see,
e.g., North 1991; Banerjee 1997; La Porta et al. 1999; and Laffont and Meleu
2001). Yet, objective data on institutional development are rare, if not inex-
istent. Virtually all the empirical work in the area has focused on subjective
surveys of perceived levels of corruption and efficiency of the public sector
(Mauro 1995; Treisman 2000; Lederman, Loayza, and Soares 2001), and there
is no objective evidence of the adequacy of such indicators.
The rate of crime reporting is an objective measure of institutional de-
velopment that is generated by individuals’ actual behavior in a real setting.
In principle, it should be correlated with several dimensions of institutional
development that researchers are usually interested in, such as confidence of
the citizens in the system, efficiency of the public service, and sense of civic
duty.
This article uses data from several different international sources to an-
alyze the determinants of the rate of crime reporting. I show that crime re-
porting is strongly related to institutional stability, police presence, and per-
ceived corruption. This evidence supports the view that subjective measures
of institutional development do capture relevant dimensions of governance
practices within the public sector.
The remainder of the article is structured as follows. Section II presents
an informal theory of the determinants of the reporting rate and discusses the
variables selected for the empirical analysis. Section III describes the defi-
nitions and sources of the data. Section IV discusses the statistical methods
used in the analysis and the results, and Section V concludes the article.
A. Police Presence
Police presence is likely to be related to both the costs and benefits of crime
reporting. It indicates the potential efficiency of the police force and the
strictness of the justice system, and it also affects the cost of reaching an
authority in order to report a crime. I measure police presence by number of
policemen as percentage of the total population.
B. Institutional Development
Institutional development affects the efficiency of the police and the justice
system as well as community cohesiveness and sense of civic duty. The
variable used to capture institutional development is the time of democratic
stability. This variable has been shown to be positively related to a series of
measures of institutional performance and good governance (see Treisman
2000 and Lederman et al. 2001). I assume that time of democratic stability
allows for gradual development of institutions, improving the efficiency of
the public service, and community cohesiveness. This may happen via natural
learning by doing or via the presence of checks and balances mechanisms in
the political system.
C. Degree of Urbanization
The degree of urbanization within society affects the costs of reaching a state
agency where crimes can be reported. It also affects the ability of the infor-
mation to navigate the system, from where the crime was first registered to
the central office where statistics are computed. Degree of urbanization is
measured by the percentage of the population living in urban areas.
D. Level of Education
Community cohesiveness and the sense of civic duty are likely to depend also
on the educational level of the population. Along the same lines, education
may capture the population’s degree of knowledge of individual rights and
its capacity to demand services from the relevant branches of government.
Furthermore, it is likely to be related to social changes that affect the psy-
chological costs of reporting certain types of crimes, such as sex-related in-
E. Corruption
Corruption is a specific dimension of institutional efficiency that is very likely
to affect reporting behavior. Corruption can reduce the gains from reporting
a crime through something like a tax on the recovered good. Additionally, it
reduces the efficiency of the police force, since it increases the probability
that the force is actually working together with the criminals. I use an index
of perceived level of corruption to account for these effects.
These five variables—police presence, institutional development, degree
of urbanization, level of education, and corruption—capture some of the most
relevant dimensions of the benefits and costs of reporting crimes. This set of
variables constitutes our baseline specification:
ln (re port crime) p b 0 ⫹ b1 ( police) ⫹ b 2 (time democ stab)
III. Data
A. Reporting Rates
Rates of crime reporting are constructed using two international sources of
crime data: the International Crime Victim Survey (ICVS) and the United
Nations Survey of Crime Trends and Operations of Criminal Justice Systems
(UNCS). The ICVS is a survey conducted by a group of international research
institutes under the coordination of the United Nations Interregional Crime
and Justice Research Institute (UNICRI). It contains data for selected countries,
irregularly distributed over the years 1989, 1992, and/or 1996–97. The project
was set up to obtain adequate crime statistics as opposed to those based on
traditional police or justice system crime data. That is, it sought to increase
the comparability of crime rates across countries by providing crime statistics
based on homogeneous definitions of the different crimes and not contaminated
by the underreporting problem that plagues official data.2
The ICVS is a standardized sample survey that looks at households’
experiences with crime in a large number of nations. There are a total of 92
surveys carried out in about 56 nations worldwide and involving a total of
135,465 people interviewed (ICVS International Working Group 2002). The
ICVS was carried out in three waves and also made limited use of some
independent national and local surveys. As with any other household survey,
it is likely to contain measurement error. But since it is an independent stan-
dardized victimization survey, the measurement error should not be in any
way related to the reporting problem common in official crime records. Even
if contaminated by measurement error, the ICVS should be free from the
systematic bias introduced by the problem of underreporting and, therefore,
it should give an unbiased estimate of the “true” crime rate and of its variation
across countries.
The other data set used for this study, the UNCS, is created for the United
B. Burglaries
Burglary Someone got into respondent’s house Burglary Unlawful entry into someone else’s
or flat without permission and stole or premises with an intention to commit
tried to steal something (does not in- crime.
clude thefts from garages, sheds, or
lock-ups).
C. Contact Crimes
Assault and threat Respondent was personally attacked, Assault Physical attack against the body of
or threatened in a frightening way; ei- another person, including battery but
ther at home or elsewhere, such as in excluding indecent assault. Some
a pub, in the street, at school, on pub- criminal or penal codes distinguish
lic transport, on the beach, or at between aggravated and simple as-
workplace. sault depending on the degree of re-
sulting injury. Some countries distin-
guish between assaults and major
assaults (both are included in these
cases).
Sexual incident Respondent was grabbed, touched or Rape Sexual intercourse without valid con-
assaulted for sexual reasons, in a really sent (may include statutory rape).
offensive way; either at home or else-
where, for instance in a pub, the
street, at school, on public transport,
in cinemas, on the beach, or at
workplace.
Robbery Someone took something from the re- Robbery Theft of property from a person,
spondent, or tried to do so, by using overcoming resistance by force or
force or threatening. threat of force.
Sources.—ICVS and UNCS questionnaires and codebooks.
Note.—The ICVS data we use here refer to victimization during the last year.
860
B. Explanatory Variables
The control variables are obtained from various sources. They are calculated
as averages for the period between 1990 and 1995, or years available within
this interval.
The variable for number of policemen per capita is built from the UNCS
data set. The variable for time of democratic stability (time democ stability)
is constructed from the database on political institutions put together by Beck
et al. (2001) and is defined as uninterrupted years of democratic stability,
counting from 1930. The inequality variable (inequality) is the ratio of the
income of the 20% richest fraction of the population to the income of the
20% poorest fraction. This variable, the percentage of the population living
in urban areas (urban), and a dummy for British legal tradition (British legal
tradit) are taken from the World Bank’s Development Database (World Bank
2000). The education variable (education) is the average schooling in the
population aged 15 and above, and this is taken from the Barro and Lee (2000)
data set. The freedom of press and the corruption indexes are constructed
from, respectively, the Freedom House and the International Country Risk
Guide indexes. I define these indexes such that higher values correspond to,
respectively, more freedom of press and more corruption. Precise definitions
of the variables and their sources are contained in the appendix.
C. Descriptive Analysis
The data on reporting rates are available only for a limited number of countries.
Those included in at least one of the regressions discussed in the next section
are listed in the appendix.
Table 2 presents some basic statistics for the core variables. Reporting
rates vary widely across crimes and countries. As should be expected, reporting
rates are higher for crimes that typically involve higher economic losses (e.g.,
burglaries, as opposed to thefts or contact crimes).4 Reporting rates of thefts
TABLE 2
Summary Statistics for Core Variables, Cross Section of Countries, 1990s
Variable Observations Mean SD Min Max
report theft 41 .084 .089 .001 .313
report burgl 36 .205 .371 .001 1.918
report cont 42 .035 .040 .001 .213
police 36 .75 2.73 .04 16.62
time democ stability 42 24 26 0. 62
urban 42 64 17 26 97
education 38 8.0 2.3 3.7 12.1
corruption 40 1.81 1.22 0. 4.00
Note.—Variables are reporting rates for, respectively, thefts, burglaries, and contact crimes,
number of policemen as percentage of population, time of democratic stability, percentage of
population living in urban areas, average schooling in the population aged 15 and above, and
ICRG corruption index.
in China, Colombia, India, and the Philippines are virtually zero, while re-
porting rates of burglaries in Austria, Finland, Malta, the Netherlands, Sweden,
Switzerland, and the United Kingdom are all above 30%.
In reality, one of the observations for the case of burglaries (Finland)
implies a reporting rate considerably higher than 100%, which is obviously
impossible. This is the reason why the maximum value for report burgl equals
1.92. This happens because the rate of burglaries in Finland, according to the
victimization survey, is extremely low (0.9%), and so any small measurement
error in the official records or in the victimization data can generate an absurdly
high reporting rate. Finland is the only case for which this happens, and so
we ignore the problem and proceed as if it did not exist.5 As a matter of fact,
Finland is not an outlier, and the results are virtually identical if it is removed
altogether from the sample.
The descriptive statistics for the independent variables imply that the
average country in the sample has a police force equal to 0.75% of the
population, 24 years of democratic stability, 64% of the population living in
urban areas, and 8 years of average schooling in the population aged 15 and
above. Number of policemen as a percentage of the population for Costa Rica
has the astonishingly large value of 16.62, which may reflect a measurement
error in the UNCS data set and may be artificially inflating the mean of the
police variable. Without Costa Rica, police has a mean of 0.33%. I deal with
this outlier a little more carefully in the regression analysis.
Table 3 presents pair-wise correlations between the same set of variables
included in table 2. Reporting rates for different crimes are strongly correlated
with each other (the correlation coefficient is significant and above 0.6 in all
cases). Time of democratic stability, degree of urbanization, and average
schooling are also positively and significantly related to the rate of crime
reporting for the three types of crimes. Number of policemen per capita is
positively correlated with reporting rates, but the coefficients are not signif-
icant. Finally, corruption has an extremely high negative and significant cor-
relation with the reporting rates of all types of crimes.
[41]
ln(report burgl) .72 1
(.00)
[35] [36]
ln(report cont) .72 .62 1
(.00) (.00)
[41] [36] [42]
ln(police) .30 .18 .25 1
(.08) (.34) (.13)
[35] [32] [36] [36]
863
TABLE 4
Reporting Rate Regressions, Cross Section of Countries, 1990s
ln(report theft) ln(report burgl) ln(report cont)
1 2 1 2 1 2
ln(police) 1.4439* 1.4856* 1.2543* 1.5436* .1431 .1559
(.2612) (.3114) (.5999) (.5053) (.1772) (.1869)
time democ stab .0389* .0196* .0279* .0045 .0280* .0066
(.0059) (.0076) (.0135) (.0129) (.0070) (.0094)
urban ⫺.0083 .0044 .0048 ⫺.0085 .0062 .0222
(.0114) (.0124) (.0252) (.0194) (.0130) (.0137)
education .0204 ⫺.0535 .0762 ⫺.2071 .0109 ⫺.0796
(.0792) (.1067) (.1761) (.1716) (.0965) (.1181)
corruption ⫺.5826* ⫺1.2849* ⫺.5020*
(.2150) (.3758) (.2367)
Constant ⫺1.7850⫹ ⫺.5721 ⫺2.6520 3.6299 ⫺4.9126* ⫺3.7837*
(.9647) (1.6521) (2.0824) (2.7438) (.7616) (1.2783)
F-value 26.25 20.85 4.45 8.64 6.99 6.09
N 33 33 30 30 35 35
Note.—N p number of observations. Standard errors are in parentheses. Robust regression
with iteratively reweighted least squares used to deal with outliers. Dependent variables are
natural logarithms of reporting rates for, respectively, thefts, burglaries, and contact crimes.
Independent variables are natural log of number of policemen as percentage of population, time
of democratic stability, percentage of population living in urban areas, average schooling in the
population aged 15 and above, and ICRG corruption index.
⫹
Significant at the 10% level.
* Significant at the 5% level.
IV. Results
Table 4 presents the results of the estimation with the baseline specification.
Results are presented for the three types of crimes with and without the
corruption variable, since time of democratic stability and corruption are very
highly correlated with each other.
Standard procedures for identifying outliers—such as plots of leverage
by normalized residual squared—indicated that Costa Rica was a strong outlier
for the case of thefts and burglaries and that India might be a concern in the
case of contact crimes.6 Therefore, I decided to use a robust regression method
that uses iteratively reweighted least squares to estimate both the coefficients
and standard errors. The procedure attributes to each observation a weight
that varies inversely with the observation’s leverage and influence. These are
the regressions presented in tables 4 and 5.7 Since I do not present the R2 for
the weighted least squares estimates, I include in the tables the F-statistics
for the joint significance test.8
Table 4 shows that police presence is positively and significantly related
to crime reporting for the cases of thefts and burglaries. In the case of contact
crimes, the results are not significant. The dependent variable in this case is
the logarithm of police, and so the coefficient can be interpreted as an elasticity.
Increasing the police force by 10% (for constant population) tends to increase
crime reporting for thefts and burglaries by roughly 14% of its initial value.
Institutional development, captured by time of democratic stability, is
positively related to reporting rates in the specifications that exclude corruption
(specification 1). But once the corruption index is introduced in the equation,
the coefficient on time democ stab becomes nonsignificant for burglaries and
contact crimes and is reduced by approximately 50% for the case of thefts.
It seems that institutional development is an important determinant of crime
reporting but that its effect works mostly through corruption. The average of
the time democ stab coefficients estimated in specification 1 implies that 10
additional years of democratic stability increase reporting rates by 32%.
However, once corruption is introduced in the regression, most of the
significance in the estimation shifts to the corruption index. The variable
corruption is negatively and significantly related to the reporting rate of the
three types of crimes. A reduction of one point in the ICRG corruption index
is associated with increases between 50% and 128% of the reporting rates
for, respectively, burglaries and contact crimes.
Table 5 checks the robustness of these results to the alternative hypothesis
discussed in Section II. The index of corruption is significant at 5% for all
the specifications, except for specifications 1 and 3 in the case of contact
crimes, where it is significant at 10%. In all cases, including these two ex-
ceptions, the F-test of joint significance for the coefficients of corruption and
time democ stab are significant at 5%. Also, in all cases, the coefficient on
time democ stability would be significant at 5% if corruption were excluded
from the regression. The result that institutional development is an important
determinant of the reporting rate is robust to the introduction of the inequality,
legal tradition, and freedom of press variables. The way institutional devel-
opment and good governance affects crime reporting seems to be mostly
1 2 3 1 2 3 1 2 3
ln(police) 1.4109* 1.6347* 1.4016* 1.8751* 1.5164* 1.5839* .1609 .1736 .1195
(.3000) (.3164) (.3183) (.4418) (.5204) (.5043) (.2056) (.1862) (.1895)
time democ stab .0210* .0140 .0185* .0087 .0032 .0029 .0057 .0022 .0061
(.0071) (.0093) (.0081) (.0108) (.0164) (.0132) (.0096) (.0111) (.0096)
urban .0130 .0070 .0023 ⫺.0150 ⫺.0081 ⫺.0107 .0275⫹ .0229 .0178
(.0123) (.0133) (.0134) (.0173) (.0210) (.0201) (.0153) (.0143) (.0142)
education ⫺.0597 ⫺.1136 ⫺.0645 ⫺.2892⫹ ⫺.2063 ⫺.2677 ⫺.1049 ⫺.1178 ⫺.0990
(.1082) (.1108) (.1145) (.1570) (.1823) (.1780) (.1303) (.1215) (.1215)
corruption ⫺.4503* ⫺.6674* ⫺.5801* ⫺1.2388* ⫺1.3123* ⫺1.1865* ⫺.4996⫹ ⫺.5793* ⫺.4379⫹
(.2022) (.2291) (.2214) (.3137) (.4152) (.3943) (.2432) (.2505) (.2373)
866
Note.—N p number of observations. Standard errors are in parentheses. Robust regression with iteratively reweighted least squares used to deal with outliers. Dependent variable is
natural log of reporting rate for thefts. Independent variables are natural log of number of policemen as percentage of population, time of democratic stability, percentage of population living
in urban areas, average schooling in the population aged 15 and above, ICRG corruption index, ratio between income per capita of 20% richest and 20% poorest, British legal tradition
dummy, and freedom of press index.
⫹
Significant at the 10% level.
* Significant at the 5% level.
V. Conclusion
This article shows that rates of crime reporting are strongly related to dem-
ocratic stability and perceived corruption. I believe that this result reflects the
fact that longer periods of political stability allow for institutional improve-
ments when they are combined with the checks and balances mechanisms
present in democratic regimes. One of the main implications of this institu-
tional development, which greatly affects reporting behavior, is the reduction
in the perceived level of corruption. Together with police presence, I found
perceived corruption to be the most important factor associated with rates of
crime reporting.
My results support the view that subjective indexes of institutional de-
velopment—such as the ones used in the recent governance literature—do
capture relevant features of the performance of the public sector. In reality,
I present the first piece of evidence relating one of these indexes to a directly
observable manifestation of institutional quality.
Appendix
TABLE A1
Data Description
Variable Source Definition
Independent variable:
police UNCS Number of policemen as a percent-
age of the total population.
time democ stab Beck et al. (2001) Years of continuous democratic sta-
bility (democp1 uninterruptedly)
since 1930.
urban World Bank Development Percentage of population living in
Indicators (World Bank urban areas.
2000)
education Barro and Lee (2000) Average years of schooling in the
population aged 15 and above.
corruption International Country Risk Indicator related to financial risk as-
Guide Detailed (ICRG sociated with corruption; based
1999) on the analysis of worldwide net-
work of experts.
inequality World Bank Development Share of income (or consumption)
Indicators (World Bank of the 20% richest of the popula-
2000) tion divided by the share of in-
come (or consumption) of the
20% poorest.
British legal tradit World Bank Global Devel- Dummy for British legal tradition.
opment Network Growth
Database (World Bank
2000)
freedom of press Freedom House (http:// Freedom of Press Index obtained
www.freedomhouse.org/) from the Human Development
Index (HDI); based on academic
advisors, in-house experts,
publications, and local
correspondents.
income per capita World Bank Development GDP per capita, PPP (current inter-
Indicators (World Bank national $).
2000)
TABLE A1 (Continued )
TABLE A2
Countries Included in the Sample
Australia Czech Republic Latvia Spain
Austria Egypt, Arab Republic Lithuania Sweden
Belarus Estonia Malta Switzerland
Belgium Finland Netherlands Ukraine
Bolivia France Norway United Kingdom
Bulgaria Georgia Philippines United States
Canada Hungary Poland West Germany
China India Romania Former Yugoslavia
Colombia Indonesia Russian Federation Zimbabwe
Costa Rica Italy Slovak Republic
Croatia Kyrgyz Republic Slovenia
References
Banerjee, Abhijit. 1997. “A Theory of Misgovernance.” Quarterly Journal of Eco-
nomics 112, no. 4 (November): 1289–1332.
Barro, Robert J., and Jong-Wha Lee. 2000. “International Data on Education Attain-
ment: Updates and Implications.” Working Paper no. 42, Harvard University, Center
for International Development.
Beck, Thorsten, George Clark, Alberto Groff, Philip Keefer, and Patrick Walsh. 2001.
“New Tools in Comparative Political Economy: The Database of Political Institu-
tions.” World Bank Economic Review 15, no. 1:165–76.
Biderman, Albert, and James Lynch. 1991. Understanding Crime Incidence Statistics:
Why the UCS Diverges from the NCS. New York: Springer-Verlag.
Cohen, Lawrence E., and Kenneth C. Land. 1984. “Discrepancies between Crime
Reports and Crime Surveys: Urban and Structural Determinants.” Criminology 22,
no. 4:499–530.
Figlio, Robert M. 1994. “Self-Reported and Officially Defined Offenses in the 1958
Philadelphia Birth Cohort.” In Cross-national Longitudinal Research on Human
Development and Criminal Behavior, ed. Elmar G. Weitekamp and Hans-Jergen
Kerner, pp. 267–80. London: Khemer Academic.
Glaeser, Edward L., and Andrei Shleifer. 2002. “Legal Origins.” Quarterly Journal of
Economics 117, no. 4 (November): 1193–1229.
ICRG (International Country Risk Guide). 1999. Brief Guide to the Rating System.
http://www.prsgroup.com/icrg/riskdata.html.
ICVS International Working Group. 2002. International Crime Victimization Survey,
1989–2000. Computer file. ICPSR version. Leiden, Netherlands: University of Lei-
den; Turin, Italy: United Nations Interregional Crime and Justice Research Institute
(UNICRI).
Kitsuse, John, and Aaron V. Cicourel. 1963. “A Note on the Use of Official Statistics.”
Social Problems 1:131–39.
Laffont, Jean-Jacques, and Mathieu Meleu. 2001. “Separation of Powers and Devel-
opment.” Journal of Development Economics 64:129–45.
La Porta, Rafael, Florencio López-de-Silanes, Andrei Shleifer, and Robert Vishny.
1999. “The Quality of Government.” Journal of Law, Economics, and Organization
15:229–79.
Lederman, Daniel, Norman Loayza, and Rodrigo R. Soares. 2001. “Accountability
and Corruption: Political Institutions Matter.” World Bank Policy Research Working
Paper no. 2708, World Bank, Washington, DC.
Levitt, Steven D. 1998. “The Relationship between Crime Reporting and Police: Im-
plications for the Use of Uniform Crime Reports.” Journal of Quantitative Crim-
inology 14, no. 1:61–82.
Mauro, Paolo. 1995. “Corruption and Growth.” Quarterly Journal of Economics 110,
no. 3 (August): 681–712.
North, Douglas. 1991. “Institutions.” Journal of Economic Perspectives 5, no. 1 (Win-
ter): 97–112.
O’Brien, Robert M. 1996. “Police Productivity and Crime Rates, 1973–1992.” Crim-
inology 34:83–207.
Skogan, Wesley. 1976. “Citizen Reporting of Crime: Some National Panel Data.”
Criminology 13:535–49.
Soares, Rodrigo R. 2004. “Development, Crime, and Punishment: Accounting for the
International Differences in Crime Rates.” Journal of Development Economics 73,
no. 1 (February): 155–84.
Treisman, Daniel. 2000. “The Causes of Corruption: A Cross-national Study.” Journal
of Public Economics 76:399–457.
World Bank. 2000. World Bank Development Indicators. Washington, DC: World Bank.
———. 2000. “World Bank Global Development Network Growth Database.” http:/
/www.worldbank.org/research/growth/GDNdata.htm.
Notes
* This article benefited from conversations with Steven Levitt and Ming-Jen Lin
and from comments from the late D. Gale Johnson and John Strauss (the editor). I
thank Anna del Frate and the United Nations Interregional Crime and Justice Research
Institute for the data from the International Crime Victim Survey.
1. The numbers refer to the beginning of the 1990s. The data are described in
detail in Sec. III.
2. Official statistics only present the number of crimes actually reported to the
police. Since a significant fraction of crimes are not reported and this fraction seems
to vary with countries’ characteristics, the use of official data to compare crime rates
across countries can be very misleading (see Soares 2004).
3. The analysis of the differences between data from official records and from
victimization surveys is a recurrent subject in applied criminology research. References
in this area include Kitsuse and Cicourel (1963), Skogan (1976), Cohen and Land
(1984), Biderman and Lynch (1991), Figlio (1994), O’Brien (1996), and Levitt (1998).
Nevertheless, most cases concentrate on national data (e.g., U.S.), and no study deals
with the topic proposed here.
4. Notice that, in our data set, contact crimes do not include homicides. The
statement would not be true if homicides were included, since their reporting rates
are very high.
5. The rate of burglary reporting in Austria is measured to be 105%, which, even
though impossible, is extremely close to 100%.
6. I owe an anonymous referee thanks for the suggestion that outliers might be
an issue in such a small data set.
7. The procedure is the rreg command in Stata 7.0. The results for thefts and
burglaries are very similar to the ones obtained when Costa Rica is removed from the
sample, and standard OLS techniques are applied. The results for contact crimes are
similar to the ones obtained with standard OLS techniques in the full sample.
8. All of the F-statistics presented reject the null hypothesis of all coefficients
equal to zero at the 1% significance level.