Bamberg 2013

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Journal of Environmental Psychology 33 (2013) 68e75

Contents lists available at SciVerse ScienceDirect

Journal of Environmental Psychology


journal homepage: www.elsevier.com/locate/jep

Applying the stage model of self-regulated behavioral change in a car use


reduction intervention
Sebastian Bamberg
Department of Social Work, University of Applied Science Bielefeld, Kurt-Schumacher-Str. 6, 33615 Bielefeld, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The stage model of behavioral change (SSCB, Bamberg, 2011, submitted for publication) conceptualizes
Available online 15 November 2012 behavioral change as a transition through the time-ordered sequence of four stages: predecision, pre-
action, action, and postaction. This study uses a phone-based social marketing campaign aiming to
Keywords: promote motor car use reduction to demonstrate how SSCB can be used to develop intervention modules
Motor car use targeting the proposed stages. It further reports on a randomized control trial evaluating the effective-
Intervention
ness of the SSBC-based intervention and testing hypotheses derived from the SSBC on the mechanisms
Behavioral change
that mediate the intervention effects. Results showed that the stage-based intervention significantly
Stage model
reduced motor car use (d ¼ 0.51). They also confirmed the hypothesis that the intervention triggers
participants’ transition to later, more action-oriented stages of behavioral change, and that this stage
progression mediates the intervention effect on behavioral change. Limitations to the study’s evaluation
design are discussed along with implications for future research.
Ó 2012 Elsevier Ltd. All rights reserved.

The development of effective intervention strategies that will same time, it stresses its self-regulative nature: Behavioral
motivate people to voluntarily change environmentally damaging change is viewed as a process in which people actively invest
lifestyles is a central applied research field in environmental effort in setting/activating goals, developing and enacting
psychology. However, it requires detailed insights into the factors strategies to achieve these goals, appraising progress, and
or processes underlying behavioral change. Recently, Bamberg revising goals and strategies accordingly (Baumeister, 2005; De
(submitted for publication) has questioned whether the theoret- Ridder & De Wit, 2006).
ical frameworks currently prevailing in environmental psychology
provide a sufficient basis for effective intervention development.
1. The present research
These frameworks not only struggle to bridge the intentione
behavior gap but also neglect the time-related and self-regulatory
The present article aims to demonstrate how the stage model of
aspects of behavioral change. These deficits may well explain the
self-regulated behavioral change proposed by Bamberg (submitted
inconsistencies in the available evidence on whether interventions
for publication) could be used as a theoretical framework for
based explicitly on these frameworks are more effective than ones
systematic intervention development. After presenting the SSBC
that are not explicitly theory-based (e.g., Hardeman, Johnston,
within the context of a phone-based social marketing campaign
Bonetti, Wareham, & Kinmonth, 2002; Michie, Abraham,
aiming to motivate citizens of Berlin to reduce their motor-car use
Whittington, & McAteer, 2009).
for everyday trips, it reports the development of dialog modules
As a consequence, Bamberg (2011, submitted for publication)
explicitly targeting the four stages proposed by this model. Because
has proposed the stage model of self-regulated behavioral
Berlin offers a high-quality public transportation (PT) system and
change (SSBC) as a new theoretical framework for systematic
a growing number of cycling and walking trails, PT use, cycling, or
intervention development. By assuming that behavioral change
walking are objectively acceptable motor-car-use alternatives.
is best conceptualized as a transition through a temporally
Therefore, a social marketing campaign was expected to be an
ordered sequence of qualitatively different stages, the SSBC
effective intervention approach. Two meta-analyses (Fujii,
explicitly takes the time-related dimension into account. At the
Bamberg, Friedman, & Gärling, 2009; Möser & Bamberg, 2008) of
quasi-experimental studies evaluating this type of intervention on
motor-car use in urban areas indicate that it has a significant but
E-mail address: sebastian.bamberg@fh-bielefeld.de. small behavioral effect (Cohen’s h ¼ 0.15, Cohen’s d ¼ 0.17).

0272-4944/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jenvp.2012.10.001
S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75 69

The empirical part reports results from an experimental field Fig. 1 also includes sets of stage-specific affective and social-
study evaluating the behavioral effects of the newly developed cognitive constructs assumed by the SSBC to influence the forma-
stage-tailored intervention. Comparing the effect of this interven- tion of the critical three intention types: goal intention, behavioral
tion with the reported small effects of traditional, non-stage- intention, and implementation intention. In the first, predecisional
tailored interventions could provide first information on the stage, constructs derived from the norm-activation model
effectiveness of this new approach. (Schwartz & Howard, 1981) are used to explain the formation of
Because it includes information about preintervention stage a goal intention. More precisely, the SSBC assumes that when an
membership and travel behavior, the intervention study also allows individual becomes aware that her or his current behavior has
a test of the theoretical assumptions underlying the SSBC, in harmful consequences and accepts personal responsibility for
particular, the assumption that it is the intervention effect on stage causing this harm, a state of self-awareness arises. Self-awareness
transition that is the cause of observable behavioral change. The automatically activates a tendency to compare current behavior
last section discusses the consequences of these results for future with important personal standards (e.g., “My frequent motor-car
research. use contributes to climate change” vs. “I am a person who cares
about the environment”). The perceived idealeactual self-
2. The stage model of self-regulated behavioral change discrepancy elicits bad feelings such as anxiety or guilt (e.g.,
Higgins, 1989) that raise the felt obligation to behave more in line
Core of the SSBC is the idea that behavioral change is best with the standard (personal norm). Simultaneously, the internal
viewed as a person’s transition through a temporally ordered attribution of responsibility may give rise to concerns about what
sequence of qualitatively different stages. Drawing on the model of important reference persons might expect her or him to do
action phases (e.g., Gollwitzer, 1990; Heckhausen & Gollwitzer, (perceived social norms). Fear of social disapproval may addition-
1987), the SSBC assumes that the temporal path of behavioral ally contribute to the activation of the personal norm. The felt
change can be broken down into four independent, qualitatively obligation to fulfill a personal norm leads to the anticipation of
different stages. In each of these four stages, a person is confronted positive feelings (pride, satisfaction) associated with its fulfillment.
with solving a specific task in order to successfully change her or his The strength of the activated personal norm and the anticipated
behavior. In the first (predecisional) stage, this task is to reflect positive feelings provide the incentive to form a goal intention.
deliberately on competing wishes and turn some of these into However, whether a person actually commits to a behavioral
binding goals. Self-commitment to a specific behavioral goal is then change goal depends on its perceived feasibility.
reflected in a respective goal intention. The formation of this goal In the second, preactional stage, the SSBC predicts that goal
intention simultaneously marks the person’s transition to the intention is joined by two additional factors influencing the
second (preactional) stage. Because several actions could normally formation of a behavioral intention. These are the constructs atti-
be used as a means to achieve the intended goal, the task in this tudes toward behavioral alternatives and perceived behavioral
second stage is to select the most suitable behavioral strategy. control over performing these alternatives derived from the theory
Weighing up the pros and cons of different alternative behavioral of planned behavior (Ajzen, 1991).
strategies results in a behavioral intention reflecting an individual’s In the third, actional stage the SSBC views a strong behavioral
self-commitment to one strategy. The formation of an explicit intention as one important prerequisite for forming an imple-
behavioral intention marks the individual’s transition into the third mentation intention (Gollwitzer, 1999). Therefore, the SSBC inte-
(actional) stage. In this stage, the task is to actually implement the grates additional factors promoting the formation of an
chosen behavioral strategy that will initiate the necessary actions. implementation intention. These are the constructs coping plan-
Gollwitzer (1999) assumes that the initiation of a new behavioral ning, action planning, and coping self-efficacy proposed by
strategy is facilitated by forming an implementation intention. The Schwarzer (2008). Coping planning refers to the ability to imagine
formation of an implementation intention marks the transition to scenarios that may hinder the performance of an intended behavior
the fourth (postactional) stage. In this stage, an individual’s task is and to develop one or more plans to cope with such a challenging
to evaluate what she or he has achieved and to decide whether situation. Action planning refers to the specific situation parame-
further action is necessary. This is done by comparing desired with ters and the sequence of actions necessary to implement the
actually achieved outcomes. A second task in the postactional stage intended new behavior. Coping self-efficacy refers to a person’s
is to struggle with temptation, that is, to prevent a relapse into the confidence in being capable of maintaining a difficult behavior.
old behavior. Fig. 1 presents the proposed four different stages of In the fourth, postactional stage, the SSBC views a well-formed
behavioral change. implementation intention as one prerequisite for successfully

Fig. 1. The stage model of self-regulated behavioral change.


70 S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75

maintaining the implemented new behavior. Schwarzer (2008) change but also elements designed to reduce reactance. Research
proposes a person’s confidence in being capable of resuming (e.g., Knowles & Riner, 2007) has indicated that one of the most
a difficult behavior after a relapse (so-called recovery self-efficacy) effective ways to reduce reactance is to merely acknowledge the
as an additional factor promoting the maintenance of the new fact that the person might feel some resistance. A second strategy
behavior. for reducing reactance is to minimize the request; that is, instead of
requesting complete behavioral change, ask for only a small change
3. Development of the four stage-based intervention modules (“even a penny will help”). A third technique to prevent reactance is
the “power of yes”; that is, asking people whether they agree with
The following sections describe the stage-tailored intervention an assumption/conclusion. Providing choice between multiple
designed to promote motor car use reduction and the theoretical behavioral alternatives is another way to prevent reactance. The
assumptions derived from the SSBC underlying its development. predecisional module developed for this study is documented in
Appendix A.
3.1. The general intervention approach
3.3. The dialog module targeting participants in the preactional
As well as tailoring the intervention to a person’s current stage stage
of behavioral change, the general features of the intervention were
to combine personal contact with supportive print materials and to The SSBC assumes that in the preactional stage of behavioral
repeat the intervention twice. More precisely, the intervention change, people already have the general goal of changing their
consisted of the following three elements: (a) The first was current behavior (high goal intention). However, because several
a contact letter describing the purpose of the campaign and inviting actions could normally be used as a means to achieve this goal, the
citizens to participate in it. This letter was accompanied by an task confronting them is to select the personally most suitable
answer postcard that participants could send back, giving their behavioral strategy. Consequently, the main intervention goal in
consent to participate and a contact phone number. The postcard this stage is to provide more knowledge about the pros and cons as
also contained questions on mobility behavior (see below) and well as the personal feasibility of behavioral alternatives, and then
asked participants to complete a newly developed stage diagnosis help people to select the one they find personally most suitable. In
tool (also see below). (b) The main intervention element consisted the present context, this can be done by analyzing mobility needs
of four standardized stage-tailored dialog modules that were used and preferences together with a participant.
during phone contact with participants in the predecisional, pre- Besides promoting arguments, the preactional stage dialog
actional, actional, and postactional stages. The development of module should also include arguments targeting potential sources
these stage-tailored phone dialog modules was based on the SSBC of reactance/resistance: According to Knowles and Riner (2007),
assumptions about the specific tasks a person has to solve at each of one effective strategy to deal with concerns about an offer consists
the four stages. (c) The third intervention element consisted of four in giving guarantees. In the mobility case, for example, participants
stage-specific sets of supportive information materials. Participants could be offered the opportunity to test whether PT services fit
could select from one set the materials that interested them most. their purposes with a free one-week travel pass. The preactional
module formulated for this is documented in Appendix B.
3.2. The dialog module targeting participants in the predecisional
stage 3.4. The dialog module targeting participants in the actional stage

The SSBC assumes that in the first predecisional stage, people The SSBC assumes that in the actional stage, people not only
are performing the problem behavior on a regular, habitual basis. possess a strong goal intention but also have formed a strong
They are not fully aware of the negative consequences associated behavioral intention; that is, they have made a decision on which
with this behavior, and thus see no reasons for behavioral change. It new behavioral strategy they want to test instead of the old one.
further assumes that confronting people in this stage with a direct The main task confronting them in this stage is to actually initiate
request to change their behavior runs the risk of triggering reac- the actions necessary to implement this new behavioral strategy.
tance. People who are psychologically not ready for behavioral Consequently, the focus of the dialog module targeting participants
change may perceive such a request as an attempt to limit their in this stage is to help them plan the initiation of these actions as
freedom of choice by social pressure. Furthermore, in this stage, concretely as possible. In this stage, people may also benefit from
people may be convinced that behavioral change is not feasible for recalling the perceived positive consequences associated with the
them personally. As a consequence, the main intervention task in new behavior strategy as well as receiving information that would
this stage is not to request behavioral change directly, but to give help them to overcome potential barriers. This actional module is
the individual a push toward thinking about the negative conse- documented in Appendix C.
quences of her or his current behavior, thereby raising the aware-
ness for the discrepancy between this behavior and important 3.5. The dialog module targeting participants in the postactional
personal standards and increasing her or his insight that a change stage
of her or his current behavior is both necessary and possible
(formation of a goal intention). The SSBC assumes that in the postactional stage, persons actu-
On the promotion side, the SSBC provides clear guidelines ally have been engaging in the selected new behavior strategy for
regarding which intervention strategies the predecisional stage some time. Thus, people in this stage are reflecting on the experi-
dialog module should focus on: (a) enhancing problem awareness; ences they have gained with the new behavior and are comparing it
(b) increasing acceptance of personal responsibility; (c) making critically with the old behavior. As a result, the main intervention
social norms salient; (d) strengthening perceived ability to change task in this stage is twofold: to provide feedback on how success-
current behavior; and (e) promoting the formation of a clear and fully they have achieved their personal change goal and to help
challenging, but not excessive personal change goal. However, them to cope with the temptation to relapse into the old behavior.
because of the aforementioned risk of eliciting reactance, the dialog As a consequence, besides repeating and strengthening the positive
module should include not only arguments promoting behavioral consequences associated with the new behavior and helping them
S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75 71

to overcome barriers, the focus of this dialog module is on 1. The stage-tailored intervention based on the SSBC will cause
providing support. For this reason, it starts by explicitly thanking a significant reduction in participants’ postintervention motor-
and praising people for their good decision and offering them car use.
a small gift in recognition (a free one-day regional family ticket). 2. This effect of the stage-tailored intervention will be signifi-
Besides asking them whether they need additional information, cantly stronger than the effect of the pure information
another intervention element for this stage group consists in intervention.
motivating them to think about buying a permanent monthly or 3. With the help of the information obtained through the newly
annual PT pass. The idea behind this intervention element is to developed stage measure, it will be possible to identify four
increase participants’ commitment to their current behavior and to homogeneous subgroups in the total sample representing the
strengthen its habitual nature. At the end of each dialog module, four stage groups of behavioral change proposed by the SSBC.
participants are offered the respective list of stage-specific infor- 4. The stronger effect of the stage-tailored intervention will be
mation materials. The selected materials are then sent to them via reflected in a significantly higher proportion of participants in
mail. The postactional module is documented in Appendix D. this experimental condition who progress from earlier to later
stages of behavioral change.
3.6. Intervention delivery 5. Stage progression will be the mediator of the intervention
effect on postintervention travel behavior. In other words, after
As reported above, the postcard presented questions on partic- controlling for the intervention effect on stage progression, the
ipants’ current travel mode choice as well as a short stage diagnosis direct effect of the stage-tailored intervention condition on
instrument. This information provided the input for a statistical travel behavior will no longer be significant.
diagnosis (latent class cluster analysis) of participants’ current
stage of behavioral change. In the next step, participants with the 5. Method
same stage diagnosis were assigned to the respective stage-specific
intervention modules. The phone numbers posted in by partici- 5.1. Participants and procedure
pants were then used to call them personally. Calls were conducted
by agents working for a professional marketing company. To Based on statistical power analyses and estimated dropout rates,
guarantee that all clients received a similar intervention, agents the plan was to start with four equal-sized stage groups (n ¼ 180)
were trained intensively in how to use the standardized dialog that is, a total sample of N ¼ 720 participants. Participants were
protocol provided on a computer screen during the talks. The whole recruited from a random sample of 12,000 addresses of Berlin
process was steered and controlled by a CATI system. citizens of German nationality obtained from the city administra-
Two weeks later, participants with the preintervention diag- tion. These prospective participants were contacted through
nosis “predecisional,” “preactional,” and “actional stage” were a letter inviting them to participate in a climate protection
contacted for a second time. The purpose of this second call was to campaign carried out under the patronage of the Berlin Senate on
support participants in these stages who were showing stage Environment, Health, and Consumer Protection. Of the 12,000
progression following the first intervention in their further tran- contacted addresses, 877 persons (7.3%) sent back the answer
sition through the stage sequence. Participants showing a stage postcard. However, 252 persons had to be excluded because they
progression received the dialog module matched to the respective were captive PT users; that is, they did not own or have access to
stage. At the end, they were once again offered the respective stage- a car. Of the remaining 625 participants, 317 (50.1%) were diag-
specific information material. nosed as in the postactional stage; 72 (11.5%), in the actional stage;
150 (24%), in the preactional stage; and 86 (13.8%), in the pre-
4. The evaluation study decisional stage. In a second step, participants within each stage
group were randomly assigned to either the control group
Six weeks after the second call, the effect of the two dialog (n ¼ 209) or one of the two experimental groups (n ¼ 208 each). Of
marketing campaigns was evaluated with a randomized control the 208 persons assigned to the stage-tailored phone intervention,
trial design. Before the start of the intervention, participants within 148 could be reached by phone and received at least one dialog
each of the four stage groups were randomly assigned to one of the intervention. Of the 565 persons (209 control group, 208 infor-
following experimental conditions: (a) a control group without any mation intervention, 148 stage-tailored intervention group) who
treatment participating only in the postintervention evaluation participated in the study as planned, 291 also participated in the
interview; (b) an experimental group participating in the phone- evaluation study. Table 1 presents the distribution of these 291
based dialog marketing campaign; and (c) a second experimental persons study across (preintervention) stages and experimental
group receiving only a standardized information package without conditions. Information from the evaluation study about the
a personal talk. The information package consisted of professionally socioeconomic background of these 291 participants indicated that
designed small brochures/leaflets used by federal institutions, they represented the better educated, more affluent part of the
consumer agencies, and companies in campaigns to promote the
use of non-motor-car transportation means. The brochures/leaflets Table 1
contained information about PT use (route map, pocket time tables Distribution of participants contacted after the intervention across (Preintervention)
of nearby PT lines, tariffs), bike and walking trails, and motor-car- stages and experimental conditions.
sharing services in Berlin. Participants assigned to the standard- Stage Experimental condition
ized information campaign received the information package only
Control Standardized Stage-based dialog
once via mail. group information campaign
Predecisional 18 15 5 38
4.1. Hypotheses Preactional 28 26 14 68
Actional 9 12 8 29
Postactional 63 52 41 156
The aim of the evaluation study was to test the following 118 105 68 291
hypotheses empirically:
72 S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75

Berlin population: 72.0% were female, the mean age was 36.2 years, protocol, participants were asked by phone how many trips they
the mean household size was 2.7 persons, and 43.6% had an had taken and with which travel modes on three consecutive days.
academic degree. Two-thirds (67%) were full or part-time Across all reported trips, the proportion of trips taken by motor
employed. The average household net income was V 2500e3000 car, PT, cycling, and walking was then calculated (so-called modal
per month. split). Data analysis indicate a satisfactory internal consistency of
Because of the high rate of participants who did not participate the calculated modal split variable and the answers provided to
in the evaluation (n ¼ 274, 48.5% of 565), it was tested whether this the two self-rating scales (motor-car use a ¼ 0.86, PT use a ¼ 0.85,
dropout rate reflected a systematic self-selection process. A logistic walking a ¼ 0.79; cycling a ¼ 0.91). Hence, the behavioral effects
regression taking “participation in the evaluation” as binary of the intervention were assessed by four sum indices combining
dependent variable and the information on preintervention the information obtained from the three measures. Each of the
behavior from the postcard as predictor did not indicate any four transportation specific indices could theoretically range from
significant relation between preintervention behavior and study 0 to 12.
dropout. A MANOVA was also calculated with preintervention
transportation means use as dependent variable and assignment to 6. Results
the three experimental conditions as an independent variable.
Results did not indicate any significant difference between exper- 6.1. Behavioral effect of the intervention
imental conditions in preintervention behavior. These results
provided no evidence that the high dropout rate was due to It was assumed (Hypothesis 1) that the dialog marketing
a systematic self-selection process. intervention would not just have a significant effect on post-
intervention travel behavior, but that this effect would be stronger
5.2. Measurement instruments than that of the standardized information package (Hypothesis 2).
Table 2 presents the postintervention means and standard devia-
When trying to empirically test assumptions of stage models, tions of the four travel behavior indices. For an overall test of
one methodological challenge is to develop a reliable and valid intervention effects, a MANCOVA was calculated with post-
measure for assessing an individual’s current stage membership. intervention motor-car use, PT use, walking, and cycling as
Bamberg (submitted for publication) has presented a first stage dependent variables and experimental condition as independent
diagnosis measure and provided some evidence that it works. variable. To increase statistical power, preintervention use of these
However, he strongly recommended developing alternative transportation means reported on the consensus postcard was
measurement approaches in order to check the construct validity entered as covariates. Using Pillai’s trace, the MANCOVA revealed
of the latent stage construct. Therefore, the present study applied significant differences between experimental conditions, V ¼ 0.09,
a newly developed alternative measure for assessing pre- F(14, 454) ¼ 2.68, p < 0.01.
(answer postcard) and postintervention stage membership Because a MANCOVA provides only an overall test of interven-
(evaluation study) that is based explicitly on the SSBC. The tion effects, the second step was to calculate linear regression
measure consisted of the introductory sentence “Which of the analyses for postintervention motor-car use, PT use, walking, and
following statements best describes how you feel about your cycling separately. Linear regression was used because SPSS 18
current motor-car use?” It was followed by five statements, and provides a bootstrap option that allows correcting for the poten-
the participant had to select the one that best fitted her or his tially biasing effect of the dependent variables’ nonnormal distri-
personal situation. For example, the statement “At the moment, I butions on the estimated standard errors of the regression
use the motor car for most of my trips. I am happy with my coefficients. In the regression analyses, the independent effects of
current level of motor-car use and see no reason why I should the two interventions (standardized information package and
reduce it” is assumed to reflect the mind-set of participants in dialog campaign) were represented by two dummy variables with
the predecisional stage (see Appendix E for the other stage- control group as reference category. In each analysis, the respective
related statements). One additional statement was included to preintervention behavior was used as an additional predictor. The
identify captive non-motor-car users. As recommended by regression analyses revealed a significant negative effect of the
Bamberg (submitted for publication), the new measure also dialog marketing intervention on motor-car use and a significant
included information about participants’ current travel behavior positive effect on PT use (see Table 3). Compared with controls, the
preferences by asking how frequently they had used a motor car, standardized effect size of the dialog intervention on motor-car use
PT, and bicycle or walking for everyday trips in Berlin within the was d ¼ 0.51, and on PT use d ¼ 0.39. The standardized infor-
last month. All four items used 5-point rating scales ranging from mation package did not have a significant effect on either motor-car
0 (never) to 4 (very often). or PT use. Neither the dialog marketing intervention nor the
Besides the assessment of postintervention stage diagnosis, the information package had a significant effect on postintervention
main focus of the evaluation study was on a reliable measurement cycling and walking.
of participants’ postintervention travel behavior. One approach to
measure this behavior consisted in asking the following two self-
rating items at the beginning and end of the first phone contact Table 2
during the evaluation study: (a) How frequently have you used the Postintervention means (Theoretical value range 0e12) and standard deviations in
the four travel behavior indices (N ¼ 291).
motor car (PT, cycling, walking) for everyday trips in Berlin within
the last month (5-point rating scale ranging from never [0] to very Postintervention behavior Control Info. package Dialog
often [4]); (b) How many days have you used the motor car (PT, group n ¼ 105 marketing
n ¼ 118 n ¼ 68
cycling, walking) within the last week for everyday trips in Berlin
(7-point rating scale ranging from never [0] to every day [7]). The M SD M SD M SD
second approach to measure travel behavior consisted of Motor-car use 7.13 2.70 6.71 2.31 5.80 1.98
a detailed mobility diary for the day of the first phone contact (all Cycling 4.75 2.91 4.52 2.86 4.85 3.06
Walking 8.40 2.14 8.51 1.94 8.31 2.10
interviews were conducted in the evening) as well as the two PT use 5.37 2.66 5.53 2.70 6.46 2.72
following days. Thus with the help of a standardized interview
S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75 73

Table 3
Results of regression analyses on the effects of the two intervention types on postintervention travel behavior.

Mode 1 Model 2 Model 3 Model 4


postintervention motor-car use postintervention PT use postintervention cycling postintervention walking
R2 ¼ 0.54 R2 ¼ 0.57 R2 ¼ 0.54 R2 ¼ 0.44

b SEa p b b SEa p b b SEa p b b SEa p b


Constant 2.29 0.36 *** 0.77 0.30 * 1.20 0.27 *** 3.33 0.48 ***
Information package 0.14 0.25 0.03 0.07 0.26 0.01 0.07 0.28 0.01 0.07 0.23 0.00
Dialog 0.80 0.26 ** 0.14 0.00 0.30 ** 0.16 0.26 0.38 0.04 0.19 0.25 0.04
Past motor-car use 1.39 0.09 *** 0.71
Past PT use 1.51 0.09 *** 0.74
Past cycling 1.53 0.09 *** 0.73
Past walking 1.29 0.11 *** 0.66

Note. PT ¼ public transport, b ¼ unstandardized regression coefficient, b ¼ standardized regression coefficient.


***p < 0.001, **p < 0.01, *p < 0.05.
a
SE ¼ bootstrapped standard error.

6.2. Identification of stage groups membership. In the dialog intervention group, the number of
participants changing from the preintervention predecisional and
It was assumed (Hypothesis 3) that the information obtained preactional to the postintervention actional and postactional stages
from the stage measure in the total sample could be used to identify should be significantly higher than that in the other two experi-
four homogeneous subgroups representing the predecisional, pre- mental conditions.
actional, actional, and postactional stages of behavioral change. To gain an overall impression on how frequently stage
Although cluster analysis is widely used for such analyses, latent progression and stage regression occurred, participants’ pre- and
class cluster analysis (LCCA, Vermunt & Magidson, 2002) provides postintervention stage diagnoses were cross-tabulated (see
a methodologically more advanced analysis tool. An important Table 5). There were high levels of both stage progression and stage
difference between standard cluster analysis techniques and LCCA is regression. For example, 22.2% of the participants with the pre-
that the latter is a model-based clustering approach. This means that intervention diagnosis “preactional stage” progressed to the later
a statistical model is postulated for the population from which the actional/postactional stages, whereas 15.9% regressed to the earlier
sample under study is drawn. More precisely, it is assumed that the predecisional stage. The predecisional and the postactional stage
data is generated by a mixture of underlying probability distribu- seemed to be quite stable, whereas the actional stage seemed to be
tions. One advantage of LCCA is that formal statistical tests could be unstable: After the intervention, only 17.6% of participants with
used to decide which model (number of clusters) fits the data best. a preintervention actional stage diagnosis remained in this stage,
In the LCCA, the following nine variables served as indicators of whereas just as many participants progressed to a later or regressed
the latent ordered categorical stage variable: the five statements on to an earlier stage.
the stage measure and the four rating scales assessing travel But did the interventions influence stage progression? To test
behavior. Table 4 presents the fit statistics of the one to five LCCA this question formally, a KruskaleWallis H test was conducted with
cluster solutions estimated for the preintervention (consensus pre- and postintervention stage membership as dependent variable
postcard) as well as postintervention (evaluation interview) stage and experimental conditions as independent variable. Because
data. Because it takes model parsimony into account, the BIC assignment was random, it was assumed that preintervention stage
statistic was used as the criterion for deciding which cluster solu- membership would not be associated significantly with the
tion fitted the data best. It showed that the theoretically expected experimental conditions. Results supported this assumption,
four-cluster solution provided the best fit to the data (lowest BIC H(2) ¼ 2.11, p ¼ 0.38. However, postintervention stage membership
value) in both datasets (see Table 4). Further inspection of the four- was associated significantly with the experimental conditions,
cluster LCCA solution indicated that the classification was deter- H(2) ¼ 7.42, p < 0.05. Jockheere’s test revealed a significant trend in
mined mainly by the chosen stage statement. Using only these five the data: The proportion of participants in higher stages increased
statements as stage indicators resulted in a quasiidentical classifi- from the control, over the standardized information, to the dialog
cation (rs ¼ 0.98, respectively rs ¼ 0.97). intervention group, J ¼ 10,586.5, z ¼ 2.56, p < 0.01. ManneWhitney
tests were used to determine which of the two intervention forms
contributed to these findings. Results showed that compared to
6.3. Intervention effect on stage transition controls, the standardized information intervention had no signif-
icant effect on postintervention stage membership, U ¼ 6478.5,
Hypothesis 4 predicted that the dialog marketing intervention z ¼ 0.50, p ¼ 0.62, whereas the dialog campaign had a significant
would have a significant effect on postintervention stage positive effect, U ¼ 4008, z ¼ 2.62, p < 0.01.
Table 4
Model fit of LCCA solutions for the pre- and postintervention stage measure
(N ¼ 291).291. Table 5
Percentage pre- and postintervention stage progression and regression (N ¼ 291).
Cluster Preintervention LCCA Postintervention LCCA
results results Postintervention Preintervention stage

LL BIC (LL) LL BIC (LL) Predecisional Preactional Actional Postactional Total post
1 1157.34 2374.92 1191.29 244,336 Predecisional 43.2 15.9 23.5 3.5 15.5
2 967.92 2045.37 978.89 2068.28 Preactional 29.5 61.9 17.6 6.1 26.1
3 908.51 1975.84 919.11 1998-49 Actional 13.6 6.3 17.6 4.4 7.6
4 871.74 1951.58 881.69 1973.34 Postactional 13.6 15.9 41.2 86.0 50.8
5 866.87 1991.13 876.88 2013.46 Total pre 100.0 100.0 100.0 100.0 100-0
74 S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75

6.4. Stage progression as mediator of the intervention effect that it makes no reference to the social-cognitive constructs
underlying stage transition proposed by the SSBS. The reason for
Hypothesis 5 predicted that the stage progression elicited by the this was that in the context of an evaluation study with “normal”
intervention should be the mechanism mediating the intervention citizens, there was a danger that asking them the huge number of
effect on behavior. Table 6 presents the results of linear regression items necessary for a reliable measurement of these constructs
analyses testing this hypothesis. These analyses used the pre- and would further increase the dropout rate. Thus one task of future,
postintervention stage diagnoses based only on the five statements more laboratory-oriented studies would be to test whether the
as predictors. The behavioral indicators were excluded from the effects of stage-specific interventions on stage transition are indeed
stage diagnosis to prevent a confounding of stage diagnosis and mediated by the social-cognitive variables postulated by the model.
postintervention behavior. As expected, the analyses (Table 6, If the model is correct, the stage transition elicited by the inter-
Model 2) indicated that after controlling for the effect of post- vention should be reflected in the formation of the three intention
intervention stage diagnosis, the effect of the dummy variable types viewed as the critical transition points between the four
“dialog intervention” on behavior was no longer significant for stages. The formation of the intention types themselves should be
either motor-car or PT use. associated with respective changes in the stage-specific sets of
social-cognitive variables. Although Bamberg (submitted for
7. Discussion publication) provides some correlational support for these model
assumptions, stronger experimental evidence is needed. Future
One goal of the present paper was to demonstrate how the studies should also test whether the SSBC provides a general
stage model of self-regulated behavioral change (SSBC) could be framework for understanding the processes underlying change in
used as a theoretical framework for systematic intervention environmentally relevant behaviors. If this claim holds, model-
developmentdin this case, a social marketing campaign aiming to based stage-specific social marketing campaigns should also be
motivate participants to reduce their motor-car use for daily trips in effective in other important environmentally related consumption
Berlin. A second goal was to test empirically the effectiveness of this fields such as household energy use or food consumption. In this
intervention as well as hypotheses on the mechanisms underlying context, future intervention studies should also think about
the intervention effect derived from SSBC. switching to other forms of intervention delivery. The low response
In summary, the evaluation results provided support for all rate (7% of the 12,000) as well as the huge dropout rate indicates
hypotheses: Whereas the delivery of a standardized information that the acceptance of mail and phone-based interventions is
package had no significant effect on participants’ postintervention decreasing. Whereas the low response rate questions the practical
travel behavior, participation in the stage-based dialog intervention effectiveness of these delivery channels, the huge dropout rate
resulted in a significant motor-car use reduction. The likewise threatens the validity of the results. Although analyses revealed no
significantly higher PT use in this group indicated that car trips evidence for a systematic self-selection process, they could not
were mostly substituted by PT use, whereas cycling and walking completely rule out the possibility that the dropout caused some
remained unchanged. Compared with the small motor-car use bias. Web-based interventions should be considered as an attrac-
reduction effect (0.15e0.17) indicated by two meta-analyses for tive alternative delivery channel. For participants, Web-based
traditional social marketing, the medium motor car use reduction interventions would offer the advantage that they could use them
effect (0.51) found in the present study provides some evidence 24 h a day. For researchers and practitioners, Web-based inter-
that this approach may be more effective. The analyses also provide ventions have the advantage of automatizing intervention delivery
support for the construct validity of the new stage measure: Using and data collection. Perhaps Web-based interventions are even
another theory-based approach to measure this construct also a better delivery channel from a theoretical viewpoint, because
resulted in the identification of four subgroups corresponding to they make it possible to combine the attributes of interpersonal
the four postulated stages. The theoretically most relevant finding and mass communication (e.g., Cassell, Jackson, & Cheuvront, 1998).
is that the stage-based dialog intervention had a significant effect Furthermore, future intervention studies should use a different
not only on behavior but also on postintervention stage member- research design than the present one, which compares the effects of
ship. Furthermore, when controlling the intervention effect on the stage-matched intervention to those of a traditional non-stage-
stage membership, the direct effect of the intervention on behavior based information package. The reason for choosing a standardized
was no longer significant, indicating that stage transition seems to information package as second experimental condition was the
be the mechanism mediating between the intervention and intervention partner’s interest in comparing the effectiveness of
behavioral change as assumed by the SSBS. the more expensive phone-based dialog marketing campaign with
However, the present study also has limitations that will need to that of a cheaper information package without personal contact.
be targeted by future research. One limitation consists in the fact Thus, the significant behavioral effect of the dialog marketing

Table 6
Results of regression analyses testing whether stage progression mediates the intervention effect on behavior (N ¼ 291).

Variable Motor-car use as dependent variable PT use as dependent variable

Model 1 Model 2 Model 1 Model 2


(R2 ¼ 0.04) (R2 ¼ 0.55) (R2 ¼ 0.03) (R2 ¼ 0.21)
DR2 ¼ 0.51 DR2 ¼ 0.19
F(1, 290) ¼ 280.9, F(1, 290) ¼ 55.8,
p < 0.001 p < 0.001

b SEa b SEa b SEa b SEa


Constant 6.87*** 0.18 11.15*** 0.28 5.52*** 0.20 2.70*** 0.42
Dialogue intervention 1.11** 0.32 0.36 0.25 1.03* 0.40 0.53 0.37
Postintervention stage diagnosis e e 1.52*** 0.09 e e 1.00*** 0.13

*p < 0.05, **p < 0.01, ***p < 0.001.


a
Bootstrapped SEs.
S. Bamberg / Journal of Environmental Psychology 33 (2013) 68e75 75

campaign per se does not demonstrate the superiority of the stage- Baumeister, R. F. (2005). The cultural animal: Human nature, meaning, and social life.
New York, NY: Oxford University Press.
matched intervention approach. It only indicates that combining
Cassell, M. M., Jackson, C., & Cheuvront, B. (1998). Health communication on the
a personal talk with information delivery is a more effective way to Internet: An effective channel for health behavior change? Journal of Health
trigger behavioral change than the delivery of information alone. In Communication, 3, 71e79.
the literature (e.g., Weinstein, Rothman, & Sutton, 1998), a design De Ridder, D. T. D., & de Wit, J. B. F. (Eds.), (2006). Self-regulation in health behavior.
London, England: Wiley.
comparing the effectiveness of the same-stage specific intervention Fujii, S., Bamberg, S., Friman, M., & Gärling, T. (2009). Are effects of travel feedback
modules delivered one time in a stage-matched way and the other programs correctly assessed? Transportmetrica, 5, 43e57.
time randomly is discussed as a more convincing way to test the Gollwitzer, P. M. (1990). Action phases and mind-sets. In E. T. Higgins, &
R. M. Sorrentino (Eds.). Handbook of motivation and cognition: Foundations of
assumed higher effectiveness of stage-based interventions. Future social behavior, Vol. 2 (pp. 53e92). New York, NY: Guilford.
studies should also apply such a design. Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans.
American Psychologist, 54, 493e503.
Hardeman, W., Johnston, D., Bonetti, D., Wareham, N. J., & Kinmonth, A. L. (2002).
Acknowledgments Application of the theory of planned behavior in behavior-change interven-
tions: A systematic review. Psychology and Health, 17, 123e158.
This study was supported by a grant from the German Ministry Heckhausen, H., & Gollwitzer, P. M. (1987). Thought contents and cognitive func-
tioning in motivational versus volitional states of mind. Motivation and Emotion,
of Research and Education (No 01UVO0809B) to Sebastian Bam- 11, 101e120.
berg. The intervention was developed and conducted in collabo- Higgins, E. T. (1989). Self-discrepancy: A theory relating self and affect. Psychological
ration with Martina Schäfer and Melanie Jäger-Erben (TU-Berlin) Review, 94, 319e340.
Knowles, E. S., & Riner, D. D. (2007). Omega approaches to persuasion: Overcoming
and Omniphon (Leipzig). I thank Christian Glöckner for his helpful
resistance. In A. R. Pratkanis (Ed.), The science of social influence (pp. 83e114).
comments on an earlier version of this paper. New York, NY: Psychology Press.
Michie, S., Whittington, C., Abraham, C., & McAteer, J. (2009). Effective Techniques in
Appendix A. Supplementary data Healthy Eating and Physical Activity Interventions: A Meta-Regression. Health
Psychology, 28, 690e701.
Möser, G., & Bamberg, S. (2008). The effectiveness of soft transport policy measures:
Supplementary data related to this article can be found at A critical assessment and meta-analysis of empirical evidence. Journal of Envi-
http://dx.doi.org/10.1016/j.jenvp.2012.10.001. ronmental Psychology, 28, 10e26.
Schwartz, S. H., & Howard, J. A. (1981). A normative decision-making model of
altruism. In J. P. Rushton, & R. M. Sorrentino (Eds.), Altruism and helping behavior
References (pp. 189e211). Hillsdale, NJ: Erlbaum.
Schwarzer, R. (2008). Modeling health behavior change: How to predict and modify
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human the adoption and maintenance of health behaviors. Applied Psychology: An
Decision Processes, 50, 179e211. International Review, 57, 1e29.
Bamberg, S. (2011). Processes of change. In L. Steg, A. E. van den Berg, & J. I. M. de Vermunt, J. K., & Magidson, J. (2002). Latent class cluster analysis. In J. Hagenaars, &
Groot (Eds.), Environmental psychology: An introduction (pp. 267e280). Chi- A. McCutcheon (Eds.), Applied latent class analysis (pp. 89e106). Cambridge,
chester, UK: Wiley. England: Cambridge University Press.
Bamberg, S. When and how do people change environmentally harmful behaviors? Weinstein, N. D., Rothman, A. J., & Sutton, S. R. (1998). Stage theories of
A stage model of self-regulated behavioral change, Manuscript submitted for health behavior: Conceptual and methodological issues. Health Psychology, 17,
publication. 290e299.

You might also like