ARTICLE IN PRESS

Journal of Environmental Psychology 23 (2003) 225–236

www.elsevier.com/locate/jep

Spillover of environment-friendly consumer behaviour

.
John Th^gersen*, Folke Olander
Department of Marketing, Aarhus School of Business, Haslegaardsvej 10, 8210 Aarhus V, Denmark
Received 22 June 2001; received in revised form 7 January 2003; accepted 10 February 2003

Abstract

This paper contributes empirical input to current reasoning about consumers’ propensity to behave in an environment-friendly
way in different domains. Specifically, we investigate whether environmental-friendly behaviours spread to more and more areas of
the consumption pattern in a virtuous circle and whether it is a necessary prerequisite for a virtuous circle to emerge and to continue
to work that the individual possesses certain general values or ethical norms. The data material is a three-wave panel study with a
large random sample of Danish consumers. The data were analysed by means of structural equation modelling and other techniques.
Cases of transfer of environment-friendly conduct between behavioural categories are found, but only in a few of the possible
instances and only of a modest size. The panel analysis also identifies a few negative cross-lagged effects. Such effects may indicate
that the performance of an environment-friendly behaviour reduces the propensity to behave environmentally friendly in other
areas. However, they may also be the result of the two involved behaviours correlating positively from the outset. Multigroup SEM
analyses indicate that the likelihood of spillover is marginally but significantly higher when respondents give high priority to the
value domain that Shalom H. Schwartz termed universalism or hold strong personal norms for environment-friendly behaviour.
r 2003 Elsevier Ltd. All rights reserved.

1. Introduction contains studies that report positive correlations be-

tween people’s propensities to behave in an environ-
For each attempt at changing consumer activities and ment-friendly way across domains (e.g. Berger, 1997)
lifestyles there is a need to better understand the as well as studies that fail to find such correlations
determinants of consumer behaviour. Until now, most (e.g. Stern & Oskamp, 1987). We are only aware of a
attempts to change consumer behaviour in an environ- single, preliminary study that investigates this issue on
ment-friendly direction have focused on isolated habits Danish data (Th^gersen, 1999b).
or groups of habits (for instance on persuading There are good arguments both for and against a
consumers to participate in a recycling programme, hypothesis that there are interrelated propensities to
lower their thermostats, or save water) rather than on behave in an environment-friendly way in different
achieving sweeping changes in lifestyles. As a conse- situations.
quence, most research aimed at improving the basis for It is possible that environment-friendly behaviour
environmental policy has focused on providing insights often requires knowledge or a willingness to seek
into consumer behaviour—and its causes—within nar- information that is more prevalent in some people than
row sectors (waste, water, organic food products, etc.). in others. If this is true, these people would more often
This is apparent from Danish as well as international than others, everything else being equal, behave in an
.
reviews of the literature (e.g. Gray, 1985; Olander & environment-friendly way in any specific domain. A
Th^gersen, 1995; Th^gersen, 1995). positive relationship across domains may also have
We even lack scientific knowledge about how causes of a psychological nature. The hypothesis that
individuals’ propensities to behave in an environment- environment-friendly behaviour has a tendency to ‘‘spill
friendly way in different domains or situations are over’’ into other behavioural domains (Frey, 1993;
related to each other. The international literature Th^gersen, 1999b) is backed by a group of psychological
theories (especially Balance Theory and Dissonance
*Corresponding author. Tel.: +45-89-48-666-88; fax: +45-86-15-
Theory) which claim that we have a need to avoid
39-88. inconsistencies in our beliefs, attitudes, and behaviours
E-mail address: [email protected] (J. Th^gersen). (for an extensive review, see Eagly & Chaiken, 1993).

0272-4944/03/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0272-4944(03)00018-5
 ARTICLE IN PRESS
226 .
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236

Bem’s (1972) Self-Perception Theory of attitude forma- reverse the case, i.e. that the performance of an
tion gives additional support to this hypothesis. This environment-friendly behaviour reduces the pro-
theory predicts that if a person starts behaving pensity to behave environmentally friendly in other
environmentally friendly in one area (e.g. separating areas?
one’s waste for recycling), that person’s attitudes and
self-image are likely to change in a way that increases his Of course, the answer to this question may be
or her preparedness to behave environmentally friendly contingent on certain background characteristics and,
in other areas. hence, vary between (groups of) consumers. In parti-
However, other arguments support the view that cular, it stands to reason that the spreading, or spillover,
correlations between behaviours in different situations of environment-friendly behaviour depends on the
or domains are small, nonexisting, or perhaps even person’s value priorities, in particular the priority given
negative. Some psychological theories suggest mechan- to self-transcendent values (Schwartz, 1994), and/or on
isms that may block the spillover between behavioural the strength of his or her ethical norms regarding the
domains, for example Schwartz’s (1977) Norm Activa- environment (Heberlein, 1972; Th^gersen, 1996). Hence,
tion Theory. This theory predicts that when the personal the second objective of this paper is to answer the
costs of behaving in a way that primarily benefits others following question:
or the society at large are perceived to be too high,
people tend, as a defence reaction, to post-rationalize Q2 Is it a necessary prerequisite for a virtuous circle to
the situation. People may neutralize the moral attitude emerge and to continue to work that the individual
or norm dictating pro-social behaviour by denying that possesses certain general values or ethical norms?
continuing their current behaviour has any serious
consequences or by denying their own responsibility Elsewhere we have argued that the likelihood of
for solving the problems produced by their current environment-friendly conduct spreading from one beha-
behaviour (Schwartz, 1968, 1973, 1977; Schwartz & viour to another depends on how closely the behaviours
Howard, 1980). Negative correlations are predicted by are associated in the person’s mind and that the
those that interpret environment-friendly behaviour in association is related to their perceived similarity
specific settings as an attempt at making amends (e.g. .
(Olander & Th^gersen, 2000; Th^gersen & Olander,.
Wenke, 1993). It is claimed that people to some degree 1999). Ajzen and Fishbein (1980) use the term ‘‘beha-
undertake environment-friendly behaviours that are vioural category’’ for clusters of behaviours that reflect
relatively easy to perform (such as participation in a the same underlying disposition. In fact, environment-
municipal recycling program) in order to make it easier friendly behaviour is usually discussed at the level of
to avoid performing more demanding or costly beha- behavioural categories (e.g. recycling, energy conserva-
viours (such as biking to work instead of going by car) tion, purchase of organic products) rather than at the
(e.g. Halkier, 1997). In a less judgmental vein, one may level of specific behaviours (e.g. delivering glass packa-
reason that it is unlikely that most individuals expect of ging to the municipal collector, turning down the
themselves, or perceive an expectation from society, to thermostat, buying organic milk). It may be debatable
contribute to environmental protection and resource whether transfer between different behaviours in the
conservation in all spheres of their lives. Environmental same behavioural category (for instance, when a con-
and resource problems are the result of, and can only be sumer who has bought organic milk for a while starts
solved by, collective action. Based on this premise, it buying other types of organic food) is spillover or just
seems rational to base decisions about environment- performing the same behaviour more consistently. There-
sensitive behaviours on a contribution ethics, rather fore, we focus here on possible spillover of environment-
than on an individual responsibility ethics (Guagnano, friendly conduct across behavioural categories.
Dietz, & Stern, 1994). A contribution ethics allows the
individual a choice of what to contribute, as long as it is
a fair contribution. 2. Method
Due to the unsatisfying speculative nature of current
reasoning about the relationship between consumers’ 2.1. Overview
propensity to behave in an environment-friendly way in
different domains, the first objective of this paper is to Since our research questions concern individual-level
provide an empirical answer to the following question: evolution through time (behaviour changes), we con-
ducted a panel study consisting of three waves of
Q1 Are environment-friendly behaviours spreading to interviews, each separated by a year. Subjects were a
more and more areas of the consumption pattern random sample of Danish consumers. A panel study
in a virtuous circle, possibly due to psychological allows us to investigate whether consumers who behave
mechanisms like those mentioned above? Or is the environmentally friendly in area A today are more likely
 ARTICLE IN PRESS
.
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236 227

to behave environmentally friendly in area B tomorrow, ber–December 1998, and again 1 and 2 years later, for a
which would indicate that a virtuous circle is operating. telephone interview.3 Two years is a short period if one
Structural equation modelling (SEM) with a cross- wants to study changes in the consumption pattern,
lagged panel design was used to analyse the relationship which is generally expected to be fairly stable in the
between behaviours (Bagozzi, 1980; Bentler & Speckart, short run. In an attempt to ensure that changes took
1981; Bynner, 1994; Engel & Meyer, 1996). Cross-lagged place in at least some behaviours, the sample consisted
panel analysis is often considered a quasi-experimental of two subgroups: One drawn from telephone registers
procedure for ascertaining the direction of causality covering the whole country (1112 interviews in the first
between two variables measured at two or more points wave) and one drawn from the telephone register of a
in time (Bagozzi, 1980). The classical use of cross-lagged particular, mixed urban–rural, municipality (408 inter-
correlation analysis for this purpose has been heavily views in the first wave). In this municipality a radical
criticised, however (Rogosa, 1980). Acknowledging the change in the waste-collection system was implemented
criticism, modern cross-lagged panel analysis based on between the two first waves, which improved the
SEM is less mechanical and more theory-driven, and it opportunities for source separation. Hence, by over-
utilizes the capability of SEM to evaluate and compare sampling from this municipality we hoped to increase
the overall fit of alternative specifications of the causal the likelihood of changes in waste-related behaviours. In
relationships among variables to the data at hand (e.g. contacted households the interviewee was picked ran-
Bentler & Speckart, 1981; Bynner, 1994; Deary, 1995; domly among individuals 18 years of age or older (next
Lorenz, Conger, Simons, & Whitbeck, 1995; Engel & birthday method) for the first interview. In 58% of those
Meyer, 1996). cases where a contact with the right person was
It is only meaningful to study spillover across established, the subject accepted to participate.
behavioural categories if behavioural categories are When the first and second interview ended, we asked for
reasonably shared and stable. People may differ as to permission to call back a year later, ‘‘should we need to do
which behavioural categories are salient, as to which so.’’4 Ninety-eight per cent accepted in the first wave and
behaviours they classify in a certain behavioural 99% in the second wave. Hence, they were contacted again
category, and as to the category in which they classify 1 year later. Since the interview took about 30 min, on
a particular behaviour.1 Hence, we carried out a average, we decided to offer an incentive for participating
prestudy based on an independent sample with the a second and a third time: the possibility to win one of 10
objective to identify possible behavioural categories that prizes worth approximately 100 Euro. Still, 11% in the
Danish consumers use to classify environmentally first and another 11% in the second wave refused to be re-
sensitive behaviours (reported more thoroughly in interviewed, and 15% in the first and 13% in the second
.
Olander & Th^gersen, 2000, and Th^gersen & wave could not be contacted after 35 trials, reducing the
.
Olander, 1999). In the present paper, confirmatory final sample to 828. Based on an analysis of the answers in
factor analysis (CFA) is used to examine whether the the first interview, the dropout seems only slightly biased
categories of behaviours identified in the prestudy also regarding variables included in the present study and not
appear in the behavioural data of the main study. An in a way that is likely to have any impact on the results.
important advantage of using SEM in the analyses at With regard to six of 17 behaviour items is the difference in
the behavioural category level is that it allows us to mean scores between dropouts and re-interviewed statis-
control for measurement error2 in the behavioural data, tically significant (po0:05). However, even in these cases,
provided that we have more than one observed variable the differences are of a small magnitude.5
representing a behavioural category.

2.2. Subjects 3
The interviews were carried out by Jysk Analyseinstitut, a
professional marketing research company.
4
The prestudy was carried out in May 1998 as a mall- It is mandatory in Denmark to obtain permission if a market
intercept survey with 300 consumers, randomly selected research company wants to save personal information for re-contact.
In order to minimize possible interviewer effects from such a request,
among the customers in two shopping malls in the city
we gave the impression that only a minor, random sample of those
of Aarhus, Denmark. For the main study, a random contacted in the first (and second) round would be contacted again.
sample of Danish consumers was contacted in Novem- 5
The behaviour of dropouts was more environment-friendly with
regard to one and less with regard to five of the six behaviours. On a
1
Complicating matters even further, people may classify behaviours five-point scale dropouts bought 0.22 units less organic milk (t-value:
differently in different situations and their implicit classification system 2.68, po0:01), took the bus or train for shopping 0.12 units more (t-
may change over time, thus increasing the uncertainty of any value: 2.51, po0:05), cycled to work 0.21 units less (t-value: 2.58,
hypothetical taxonomy of behaviour. po0:01) and to shopping 0.20 units less (t-value: 2.68, po0:01),
2
Strictly spoken, what is accounted for is unique variance in the source-separated green kitchen waste for composting 0.27 units less (t-
manifest variables, i.e. variance not reflecting the hypothesized latent value: 2.80, po0:01), and delivered 0.29 units less hazardous waste to
category, but something else, including measurement error. special collection (t-value: 3.51, po0:001).
 ARTICLE IN PRESS
228 .
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236

Table 1
Stability and change in environment-friendly behaviour, Denmark 1998–2000

How ofteny Mean Difference between means, F -value Correlation, r

1998 1999 1999 1998–1999 1999–2000 1998–2000 1998–1999 1999–2000 1998–2000

Is the milk you buy organic? 2.75 2.85 2.88 7.82b 0.82 10.17c 0.79 0.83 0.73
Are the frozen peas you buy organic? 1.83 1.99 2.03 15.30c 0.98 20.47c 0.53 0.48 0.43
Is the minced beef you buy organic? 2.12 2.24 2.26 7.67b 0.11 8.39b 0.54 0.60 0.48
Do you take the bus or train to work? 1.58 1.59 1.54 0.03 1.70 1.04 0.67 0.69 0.62
Do you take the bus or train to 1.36 1.33 1.31 1.62 0.64 3.28 0.58 0.63 0.50
shopping?
Do you bike to work? 2.31 2.26 2.14 2.10 8.58b 16.10c 0.80 0.74 0.71
Do you bike to shopping? 2.54 2.50 2.53 0.74 0.63 0.01 0.72 0.74 0.69
Do you keep the temperature in the 3.19 3.24 3.40 0.74 9.43b 14.40c 0.59 0.53 0.52
living room below 21-degree Celsius in
the winter?
Do you turn of all lights when you leave 4.48 4.44 4.44 1.15 0.01 1.38 0.41 0.47 0.41
a room (as last person)?
Do you turn off the water when 3.71 3.84 3.97 9.87b 12.01c 37.08c 0.71 0.75 0.69
brushing your teeth?
Are the light bulbs you buy energy 2.66 2.82 3.02 15.23c 29.80c 65.36c 0.67 0.69 0.60
saving?
Is the paint you buy based on water or 3.72 3.73 3.78 0.04 0.75 1.00 0.43 0.44 0.40
linen oil?
Is the shampoo you buy of an 2.57 2.70 2.76 5.02a 0.95 9.52b 0.51 0.47 0.46
environmentally approved brand?
Do you deliver empty packaging glass 4.51 4.63 4.71 11.91c 6.44a 33.89c 0.39 0.46 0.39
and newspaper to recycling?
Is your green kitchen waste composted 2.88 3.40 3.47 73.42c 3.16 96.44c 0.53 0.79 0.54
in the garden or at a municipal facility?
Do you deliver paint residues and 4.09 4.19 4.33 3.58 8.57b 21.34c 0.48 0.50 0.46
similar waste to the ‘‘environment car’’
or a paint shop?
Do you wash your ‘‘boiling wash’’ at 60 3.67 3.74 3.86 2.28 5.38a 13.08c 0.64 0.59 0.57
instead of 90 degree Celsius?

Behavioural categories
Buying organic 2.23 2.36 2.39 25.63c 1.37 31.33c 0.77 0.79 0.72
Buying green nonfood 2.98 3.08 3.18 10.57b 12.01c 39.44c 0.60 0.58 0.55
Transport 1.95 1.92 1.88 2.38 4.16a 10.72c 0.79 0.77 0.73
Conservation 3.76 3.82 3.92 5.55a 20.45c 42.77c 0.66 0.65 0.63
Recycling 3.83 4.07 4.17 64.10c 15.95c 123.54c 0.59 0.73 0.57

Note: N ¼ 828: Behavioural items are measured on 5-point scales with the values 1: never, 2: rarely, 3: half the time, 4: often, and 5: every time/
always. Item non-response is coded as 1. All correlations in the three rightmost columns are significant at po0:001: F -values marked by letters are
significant at the following levels: a: po0:05; b: po0:01; c: po0:001:

2.3. Variables environment-friendly behaviours (see Table 1 below), and

using a five-point scale with the labels ‘‘never’’, ‘‘rarely’’,
In all three waves, questions were asked about the ‘‘half the time’’, ‘‘often’’, and ‘‘always/every time’’. In
actual performance of a number of environment- order to ensure that the analysed behaviours make a
friendly consumer behaviours, about the priority given difference for the sustainability of the consumption
to a selection of the many values a person may hold pattern we used a study on the environmental impact of
(Schwartz, 1994), and about their personal norms consumption, commissioned by the Danish Environment
regarding the performance of a number of environ- and Consumer Agencies (Forbrugerstyrelsen og Milj^-
ment-friendly behaviours (Schwartz, 1977), besides styrelsen, 1996), as the basis for selecting the behaviours.
questions not used for the present study. The variables The 17 behaviours cover all phases of the consump-
measured in the prestudy are described together with the tion cycle (Pieters, 1991), a wide range of environmental
results. and resource problems (direct and indirect consumption
of energy and water, product-related pollution during
2.3.1. Behaviour production and use, waste), and a range of settings
Behaviour was measured by 17 questions of the type (the home, the supermarket, waste facilities, transport).
‘‘How often do you X,’’ where X refers to each of 17 Due to differing requirements and prerequisites for
 ARTICLE IN PRESS
.
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236 229

performing these behaviours, consumers may differ in behavioural categories, represented by averages of the
their abilities and opportunities as well as in their items assumed to belong to the category, were also
motivation to perform any of them. analysed in this step. Next, a sample of Danish
consumers’ mental categorization of the behaviours
2.3.2. Values was analysed by means of multidimensional scaling
When analysing the dependence of spillover on the using similarity judgments of pairs of behaviours as
possession of certain values we used an index based on input. CFA7 was used to investigate whether the
five items from Schwartz’s Value Survey6 that we identified mental categories are reflected in the beha-
measured in all three waves. Schwartz (1992, 1994) vioural structure. Next, the dynamic relationships
classifies these items in the motivational domain that he between different environment-friendly behaviours were
terms ‘‘universalism’’ and which belongs to the ‘‘self- tested by means of SEM with a cross-lagged panel
transcendence’’ corner of the value space. The items are design. Finally, multigroup SEM was used to test
questions of the type ‘‘How important or unimportant is whether interrelationships between behaviours are con-
X as a guiding principle in your life,’’ where X refers to tingent on certain values or ethical norms. SEM makes
the five items measured on a scale from 0 to 10 with the it possible to solve a series of hypothesized equations
end points ‘‘totally unimportant’’ and ‘‘of decisive simultaneously, and the fit of the estimated to the
importance.’’ In a previously published paper based on observed sample covariance matrix can be used to
an analysis of the two first waves of this data set we determine whether the hypothesized model gives an
found that the overall propensity to behave in an acceptable representation of the analysed data.
environment-friendly way was related more to univers- An important issue in cross-lagged analysis is the
.
alism than to other values (Th^gersen & Olander, 2002). optimal length of the lag. The lag between waves should
Other studies support this finding (e.g. Stern, Dietz, & mirror the time that it is assumed to take for the causal
Kalof, 1993; Stern & Dietz, 1994; Stern, Dietz, Kalof, & effect to evolve (Lorenz et al., 1995). The problem, of
Guagnano, 1995; Karp, 1996; Stern, Dietz, & Guagna- course, is that often we can only make a guess as to the
no, 1998; Schultz & Zelezny, 1999). length of that period. As already mentioned, we have
reasons to assume that broader behavioural patterns
2.3.3. Ethical norms change fairly slowly. Hence, if anything a lag of 1 year is
Previous studies have shown that internalized moral presumably too short, rather than too long.
norms, in psychological research termed personal norms
(e.g. Schwartz, 1977; Ajzen, 1991), are the ethical norms
with the strongest influence on the propensity to behave 3. Results
in an environment-friendly manner (e.g. Harland,
Staats, & Wilke, 1999; Th^gersen, 1999a, b). Hence, 3.1. Stability and change in manifest variables
when analysing the dependence of spillover on the
possession of certain ethical norms we used an index Before proceeding to the structural analysis, we take a
designed to capture personal norms regarding the look at the stability and change of the individual items
protection of the environment. Eight personal norm in order to assess the quality of the data, as reflected in
items were measured in all three waves. Four questions test–retest reliabilities, among other things. We also
were of the type ‘‘I feel that I should do X for the sake of investigate whether the analysed behaviours tend to be
the environment’’ and four were of the type ‘‘I get a bad stable or volatile at the sample means level. The top
consciousness if I don’t do X,’’ where X refers to the section of Table 1 reports analyses of the individual
following four behaviours: source-separate green kitch- behaviours and the bottom section contains the same
en waste for composting, choose organic instead of information with regard to the five behavioural cate-
conventional milk, use energy-saving instead of conven- gories in which we anticipated that Danish consumers
tional light-bulbs, and take the bus or train to work or might classify the 17 behaviours.
shopping instead of a private car. The answers were According to Table 1, Danes are fairly considerate in
measured on a five-point scale with the end points their everyday consumer behaviour when it comes to the
‘‘totally disagree’’ and ‘‘totally agree.’’ environment and resource use. However, the environ-
ment-responsible behaviours differ greatly in popularity,
2.4. Statistical analysis as indicated by means ranging from below two to over
four on a five-point scale.
The statistical analysis began with item-level studies The performance of these behaviours is fairly stable at
of stability and change in behaviour. Anticipated the individual level, all test–retest correlations (or
6 7
Equality, Unity with nature, A world of beauty, Social justice, The CFA and SEM were performed by means of LISREL 8.30
Protecting the environment. .
with WLS estimation (Joreskog, .
Sorbom, Toit, & Toit, 1999).
 ARTICLE IN PRESS
230 .
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236

stabilities) being highly significant. However, as one common’’ and 10 as ‘‘the two activities have very much
should expect, the strength of the correlation tends to in common.’’
decrease when the time-lag between test and retest Classical, metrical MDS was used to find the
increases. Further, the stabilities vary substantially. One dimensionality of the space that seems to account for
reason may be differences in the variability of external most of the differences in perceived similarity. A two-
conditions for the particular behaviours. Furthermore, dimensional solution was found to offer the best
there is a difference in the rate with which these compromise between fit and parsimony (Kruskal’s
behaviours are performed—in an environment-friendly stress=0.105, r2 ¼ 0:957).
way or not—which may influence the trustworthiness of Whereas we found it difficult to give a clear
an individual’s self-reports (and, hence, measurement interpretation of the two dimensions divulged by the
error). Notice that the stability of the behavioural perceptual map, it was clear that the various activities
category indexes generally is higher than that of any related to the purchasing of organic food remained
individual behaviour in the category, presumably closely related in the map. Likewise, the three activities
because aggregation cancels out random error (Epstein, related to waste handling were closely linked. Apart
1983). Behavioural stability is often assumed to be the from these, the various activities did not form intuitively
product of habit, i.e. that the behaviour is performed in meaningful clusters in the perceptual map, which was
a more or less automatic fashion (e.g. Landis, Triandis, rather fragmented. Hence, the findings support that
& Adamopoulos, 1978; Ouellette & Wood, 1998). consumers have formed two of the anticipated beha-
However, really all that we know is that, when vioural categories (purchase of organic food products
behaviour is stable, the balance of forces determining and source-separation activities). However, none of the
behaviour is unchanged (Ajzen, 1991, 2002). other anticipated behavioural categories were reflected
The mean values of 12 of the 17 behaviours and all in the perceptual map. In fact, the perceptual map
five behavioural categories changed significantly be- conveyed the impression that many activities with
tween Waves 1 and 3. Except for the change in environmental consequences do not ‘‘go together’’ in
transport-related behaviours all changes were in an people’s minds.
environment-friendly direction. Which of the beha-
viours that changed differed somewhat between the 3.2.2. Categorization of behaviours by means of
two time periods. In eight behaviours a statistically confirmatory factor analysis
significant change in the mean value occurred between It seems likely that a defining characteristic of the
the two first interviews, all of them in an environment- identified behavioural categories is a common goal. If
friendly direction. Between the last two interviews, the they share a common goal, we would expect the
mean values of 7 behaviours changed significantly, all performance of behaviours that consumers classify in
except transport-related behaviours in an environment- the same behavioural category to vary in concert, but
friendly direction. Hence, in the first period, the only under certain conditions. Their covariance may be
behaviour changes led to increased sustainability of reduced if conditions for performing the behaviours
the self-reported consumption pattern whereas in the vary widely, if they are competing ways to reach the
second period, the direction of changes was not same goal, or if people are unaware of the fact that the
unequivocal. behaviours lead to a common goal. Learning that
different behaviours lead to the same goal is probably
3.2. Behavioural categories in the environmental domain one of the most important mechanisms for transfer of
environment-friendly conduct.
3.2.1. Perceptual mapping of behaviours The perceptual mapping only identified two beha-
.
In a previous study (Th^gersen & Olander, 1999; vioural categories that seem to organise several environ-
.
Olander & Th^gersen, 2000) we investigated how ment-friendly behaviours: organic buying and recycling.
Danish consumers classify environment-friendly beha- However, with a larger set of behaviours we might have
viours in behavioural categories (Ajzen & Fishbein, identified additional behavioural categories. In the panel
1980) based on the perception of the similarity or study, we extended our list of behaviours with two items
dissimilarity of pairs of behaviour.8 The measures of referring to the use of each of the means of transporta-
perceived similarity were obtained by asking the tion covered in the prestudy (distinguished with regard
respondents to make a pairwise comparison of the to the purpose of the trip), implying that it might be
various activities and for each pair to report the degree possible to identify ‘‘means of transportation’’ beha-
of similarity on a scale from 0 to 10, where 0 was defined vioural categories in this material, if they exist. That we
as ‘‘the two activities have absolutely nothing in should expect to find such behavioural categories is
supported by research suggesting that people form
8
Except that we added two extra transport behaviours in the present general habits with regard to their use of means of
study, the sets of analysed behaviours are identical. transportation (e.g. Verplanken, Aarts, Knippenberg,
 ARTICLE IN PRESS
.
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236 231

& Knippenberg, 1994; Aarts, Verplanken, & Knippen- Table 2

berg, 1997). Confirmatory factor analyses of items concerning the purchase of
In a CFA the tendency of behaviours in a behavioural organic food products, source separation of various waste fractions,
and transport for 3 consecutive years
category to vary in concert is reflected in significant
factor loadings (paths) from a common latent factor (the 1998 1999 2000
behavioural category) to the individual behaviours. If L1: buying organic 0.75 0.75 0.76
consumers have established a clear categorization of the Milk 0.78 0.74 0.80
analysed behaviours, the behavioural categories should Frozen peas 0.75 0.67 0.65
be distinguishable, i.e. show less than perfect correla- Minced beef 0.59 0.71 0.71
L2: recycling 0.60 0.65 0.63
tion. Hence, we used CFA to test whether at least part of
Paper and glass 0.66 0.68 0.58
Danish consumers’ environment-friendly behaviour is Compostable kitchen waste 0.46 0.45 0.44
organized in behavioural categories, not only men- Hazardous waste 0.61 0.71 0.76
tally, but also in a behavioural sense. We tested for L3: alternative transport 0.73 0.67 0.73
three behavioural categories: the two proposed by the Work 0.61 0.78 1.00
Shopping 0.88 0.63 0.46
perceptual map (organic buying and recycling) and the
third proposed by transportation research (using non- Correlations between behavioural categories
car based means of transportation9). With three waves BO-recycling 0.38a 0.20a 0.17a
we were able to perform three consecutive confirmatory BO-AT 0.18a 0.17a 0.19a
factor analyses of self-reported behaviour in these three Recycling-AT 0.05 0.14a
0.04
areas (see Table 2), which makes it possible to assess the
Test statistics, overall fit
reliability of the solution. We assumed a simple RMSEA 0.04 0.03 0.03
structure factor pattern, in addition to the usual CFI 0.97 0.98 0.97
assumptions, for instance about the unique (or error)
Note: N ¼ 828: The numbers in the rows of individual items are factor
variances of the individual items being mutually loadings and those in the rows of latent variables (behavioural
uncorrelated and uncorrelated with the latent variables. categories) are composite reliability indexes (r). The wording of the
The test statistics show that the proposed factor pertinent behavioural items can be found in Table 1. The two items
structure produces a good fit to the data in all concerning transport to work are summed and so are the two items
concerning transport to shopping. Correlations marked by ‘a’ are
three waves. All individual factor loadings surpass the
significant at po0:05:
conventionally accepted threshold and the same is true
for the composite reliability index r: Finally, all
correlations between latent constructs are low. Hence,
the proposed behavioural categories have acceptable modification indices calculated by Lisrel. In the basic
construct and discriminant validity. For one parameter model, we only included the stabilities (effects of a latent
an inadmissible value was found: a negative error variable measured at tn on itself at tnþ1 ), assuming a
variance10 for the use of alternative means of transpor- first-order Markov chain relationship. In order to avoid
tation for work trips. Since no obvious reason for this the possibly biasing influence of omitted variables, we
result could be detected, and since the error variance was allowed the errors of endogeneous (dependent) variables
not significantly different from 0, we judged it to be a measured at the same point in time to correlate.11
statistical artefact and fixed it to 0. Further, as is customary in panel analysis, unique
(error) variances of the same item measured at different
3.3. The spreading of environmental-friendly behaviour times were allowed to correlate, this because there may
be nonrandom sources of unique variance, which are
Next, the three data sets were combined in a three- likely to persist over time. In the following steps, we
wave panel analysis of the stability of and cross-lagged inspected the modification indices concerning the cross-
influences between the identified behavioural categories; lagged paths in order to see whether the addition of a
significant cross-lagged effects being interpreted as cross-lagged path would improve the overall fit of
indicators of spillover. Since we had no a priori the model. If the modification indices suggested that
hypotheses about which spillover effects to expect, we there were significant cross-lagged effects, we added the
adopted an exploratory stepwise approach based on one with the highest modification index and repeated the
analysis. This procedure was repeated as long as
9
Since bicycles and public transportation are competing alternatives, significant modification indices for cross-lagged paths
besides both being environment-friendly alternatives to the private car, were identified. Fig. 1 shows the structural part of the
we use a summed index of the items referring to using the two means of
transportation to the same destination in the following calculations.
10
The error variances and many other details from the Lisrel output
11
are not shown, but the full output as well as the analysed covariance Only one of six correlations between the errors of endogeneous
matrices can be acquired from the authors. variables is significant.
 ARTICLE IN PRESS
232 .
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236

0.06 0.14

0.90 0.93
Organic1 Organic2 Organic3
0.13 0.06

0.46 -0.10 -0.09 0.07

0.07

0.95 0.95
0.24 Recycling1 Recycling2 Recycling3

0.09
0.04 0.06 0.08

0.87 0.96
Transport1 Transport2 Transport3

Fig. 1. Stability and spillover at the behavioural category level: A three-wave panel analysis. N ¼ 828:
.
Note: RMSEA=0.031, CFI=0.99. For Wave 3, reduced form equation error variances are reported (see Joreskog, 1999a, b). A dashed line means
that the path is nonsignificant.

final model.12 According to the test statistics the model Two of the four significant cross-lagged path coeffi-
produces an excellent fit to the data. cients have a positive and two have a negative sign. The
Notice first the high stabilities and low residual positive path coefficients provide support for the
variances. It is obvious that in this case, with correction hypothesis that environment-friendly conduct may spill
for measurement error, the widely held belief that the over from one behavioural domain to another. In
best predictor of future behaviour is past behaviour substantive terms, they show that heavy recyclers in
(Ouellette & Wood, 1998) is confirmed. As noted earlier, 1998 tended to increase their purchase of organic food
high stabilities may reflect the force of habit, but they products more than average between 1998 and 1999 and
may also just be the product of the motives and other to decrease their use of public transport and/or bicycle
individual conditions determining these behaviours less than average between 1998 and 2000. On the other
remaining (practically) unchanged. With this high level hand, the negative cross-lagged path coefficients seem to
of behavioural inertia it is obvious that if there is any provide support for the hypothesis that the performance
propensity for environment-friendly behaviour to spill of some environment-friendly behaviour may reduce the
over from one behavioural category to another, the propensity to act in an environment-friendly way in
transfer is bound to be gradual and fairly slow. That this other areas. An alternative explanation exists for the
is indeed the case is reflected both in the small size of the negative cross-lagged effects, however. In substantive
path coefficients and in the fact that only four of the 18 terms, they show that heavy users of organic food
possible cross-lagged paths with a lag of 1 or 2 years13 products tended to increase their source separation of
reach statistical significance. waste less than average both between 1998 and 1999 and
Because we have controlled for the stabilities, a between 1999 and 2000. Notice, however, that there was
significant cross-lagged effect indicates that the level of a moderately strong positive correlation between recy-
the origin behaviour accounts for some of the change in cling and buying organic food from the outset. Hence,
the destination behaviour. A positive sign means that the negative cross-lagged effects may simply reflect that
individuals who perform the origin behaviour to a high heavy users of organic food products had less room for
(low) degree increase their performance of the destina- increasing their waste source separation than light users
tion behaviour more (less) than average or decrease their (i.e. a ceiling effect).
performance of the destination behaviour less (more)
than average. If the sign is negative, the reverse is true. 3.4. The moderating effects of general values and ethical
(From Table 1 we know that source separation of waste norms
and the purchase of organic products increased and that
the use of public transport decreased in the analysed The results reported above disregard any heterogene-
period, on average.) ity in the population, which could influence the like-
lihood of spillover. For instance, we suggested above
that spillover might be contingent on the individual
12
The measurement model and the analysed covariance matrix can holding certain value priorities or ethical norms. We
be acquired from the authors.
13 tested this proposition by means of multigroup SEM,
Due to the high correlation between consecutive measures of the
same latent variable, multicollinearity is inevitable if both are included using the universalism and personal norm indexes that
as predictors of the same dependent variable. Hence, in practice a we introduced earlier (each split on its mean) as
maximum of 12 cross-lagged paths can be included in the model. grouping variables. As a preparation for the multigroup
 ARTICLE IN PRESS
.
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236 233

analysis, we first followed the same procedure as Table 3

described above to identify all cross-lagged effects that Path coefficients, explained variances (R2 ), and test statistics for the
two multigroup analyses
are significant in at least one of the compared groups. In
order to test whether differences between groups are To From
statistically significant, this set of cross-lagged effects Organic Recycling Transport R2
was included and a model where the cross-lagged effects
were assumed to be identical between groups was High universalism
Recycling1 0.29
compared to one where they were allowed to vary Transport1 0.10
0.03
between groups. Organic2 0.91 0.08 0.85
The multigroup SEM analysis reveals that when the Recycling2 0.77
0.09 0.68
cross-lagged effects are restricted to be equal between Transport2 0.85 0.91
groups a significant drop in the overall fit is produced, Organic3 0.93 0.75
Recycling3
0.10 1.12 0.63
both when the universalism and when the personal norm Transport3 0.13 1.11 0.83
index is used as the grouping variable
(Dw2universalism ¼ 10:5; 4 df., po0:05; Dw2personalnorm ¼ 57:8; Low universalism
6 df., po0:001). Recycling1 0.26
We also checked whether the measurement model (i.e. Transport1 0.14 0.11
Organic2 1.02 0.99
the perception of the behavioural categories) and the Recycling2 1.00 1.03
stabilities vary between the groups. In fact, they do. Transport2 0.94 0.86
When either (a) the factor loadings of the latent Organic3 0.96 0.83
variables or (b) the error variances of the individual Recycling3 0.91 0.90
measures are set equal between groups, a significant Transport3 0.08 0.85 0.81
change in w2 is produced ((a) Dw2universalism ¼ 56:9; 15 df., RMSEA 0.05
po0:001; Dw2personalnorm ¼ 40:4; 15 df., po0:001; (b) CFI 0.99
Dw2universalism ¼ 134:3; 24 df., po0:001; Dw2personalnorm ¼
85:9; 24 df., po0:001), thus indicating that how the Strong personal norm
behavioural categories are perceived depends on value Recycling1 0.12
Transport1 0.13
0.03
priorities and personal norms. The same result was Organic2 0.92 0.19 0.98
found when the stabilities of the behavioural categories Recycling2 1.05 0.96
were fixed to be equal between groups Transport2 0.16 0.84 0.86
(Dw2universalism ¼ 42:1; 6 df., po0:001; Dw2personalnorm ¼ Organic3 0.95
0.12 0.76
63:9; 6 df., po0:001), showing that also the stability of Recycling3
0.10 1.00 0.93
Transport3 0.09 0.95 0.76
the behavioural categories depends on value priorities
and personal norms. Unfortunately, we have not been Weak personal norm
able to find any systematic pattern in—nor a simple Recycling1 0.13
explanation for—these differences between groups. The Transport1
0.05 0.00
path coefficients that are significant in at least one of the Organic2 0.94 0.87
Recycling2 0.90
0.29 0.77
compared groups, explained variances, and overall fit Transport2 0.84 1.00
statistics are reported in Table 3.14 Organic3 0.93 0.85
As anticipated, there are more significant cross-lagged Recycling3
0.09 0.83 0.83
effects among those who give high priority to univers- Transport3 1.22 0.86
alism values or hold strong personal norms about
RMSEA 0.06
environment-friendly behaviour than among those who CFI 0.99
score low on these background characteristics.15 In the
low universalism group only one cross-lagged effect Note: All coefficients have a one-period lag, except those going to
behavioural categories at time 1, which are correlations between
reaches significance. In the weak personal norm group (exogeneous) categories measured at the same point in time. For Wave
there are two, but one of them is dubious (the one 3 dependent variables, reduced form equation R2 is reported.
linking Organic2 and Recycling3) since it turns out not

14
The measurement model and the analysed covariance matrices can to be significant when the calculation for the group is
be acquired from the authors. made outside the multigroup analysis framework. The
15
The difference in factor loadings between groups means that the positive cross-lagged paths in the high universalism and
metric of the latent variables is not identical between groups. Hence,
strong personal norm groups connect the same beha-
the size of the path coefficients is not directly comparable between
groups. However, by fixing the measurement models to be equal we vioural categories as in the aggregate analysis presented
establish a common metric for testing the significance of differences in Fig. 1, and they are still of modest size. Only one of
between structural parameters. the negative paths is recurrent from the earlier analysis,
 ARTICLE IN PRESS
234 .
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236

i.e. from organic to recycling. In the strong personal argued above, they may also be the result of a ‘‘ceiling
norm group there are no negative cross-lagged effects in effect’’ stemming from the fact that two involved
the first period. However, in the second period the behaviours correlated positively from the outset. Notice
negative path from organic to recycling is matched by a that this or a similar phenomenon cannot explain the
negative path from recycling to organic. In the high positive cross-lagged effects.
universalism group a negative path from transport to The second research question is answered affirma-
recycling has substituted that from organic to recycling tively by our study. The multigroup SEM analyses
in the first period. The lack of a consistent pattern with indicate that the likelihood of spillover is higher if the
regard to the negative paths makes them difficult to value domain that Schwartz (1994) termed universalism
interpret. is given high priority or if a person possesses strong
personal norms for environmental-friendly behaviour.
Of course, the practical relevance of this conclusion
4. Discussion depends on how important a phenomenon spillover is.
This study confirms the existence of transfer of
In this study, we wanted to investigate whether environment-friendly conduct between behavioural ca-
environment-friendly behaviours spread to more and tegories, but also indicates that the process is a slow one
more areas of the consumption pattern in a virtuous and that it depends on various conditions. We have only
circle, possibly due to psychological mechanisms like taken the first small step towards uncovering the
those discussed in the introduction, and whether it is a contingencies of such spillover.
necessary prerequisite for a virtuous circle to emerge and The remarkable stability of behaviour found in this
to continue to work that the individual possesses certain study indicates that habits deserve particular attention
general values or ethical norms. in future research on the contingencies of spillover.
Unfortunately, the study gives a somewhat ambig- Substantively, the high stabilities indicate that the
uous answer to the first question. The CFA identified individual and societal conditions for performing
some clusters of behaviour that seem to ‘‘go together’’ in the analysed behaviours changed very little during the
what we conceive as behavioural categories. The analysed period. Many of the analysed behaviours
existence of such behavioural categories may in itself are performed frequently and in a stable context, which
be taken as an indication that some transfer of are the ideal conditions for developing habits (e.g.
environment-friendly conduct goes on between beha- Ouellette & Wood, 1998). From research on habitual
viours that are closely associated in a person’s mind. On decision-making it can be inferred that the likelihood
the other hand, our analyses also show that many that environment-friendly behaviour makes a person
environment-friendly behaviours are not at all closely reflect on behaviours in other domains is lower the more
related in people’s minds. habitually these other behaviours are performed (e.g.
By means of the panel analysis of behavioural Ronis, Yates, & Kirscht, 1989; Verplanken & Aarts,
categories we attempted to capture aspects of the 1999). We leave it to future research to address this issue
process of evolution of the behavioural pattern, and empirically.
specifically whether the performance of an environment-
friendly behaviour increases (or decreases) the like-
lihood of performing other environment-friendly beha- Acknowledgements
viours later in time. Realizing that only fairly weak
spillover effects can be expected under normal circum- The study reported was part of a research programme
stances, our research design was specifically tailored to within the Centre for Social Science Research on the
make it possible to identify weak, but meaningful Environment (CeSaM), a multidisciplinary centre based
correlations between variables: (a) it is based on a fairly on collaboration among scholars from a number of
large sample, (b) it covers a time-span of 2 years, and (c) Danish universities and other research institutions, and
it relies on a data analysis technique that allows was financed by a grant from the Danish Strategic
correction for measurement error. Indeed, the panel Environment Research Programme. We are grateful to
analysis did find signs of transfer of environment- Joachim Scholderer and two anonymous reviewers for
friendly conduct between behavioural categories, but valuable comments on an earlier version.
only in a few of the possible instances and, as expected,
only transfers of a modest size. The panel analysis also
identified a few negative cross-lagged effects (between
References
buying organic food and recycling). Such effects may
indicate that the performance of an environmental- Aarts, H., Verplanken, B., & Knippenberg, A. v. (1997). Habit and
friendly behaviour reduces the propensity to behave information use in travel mode choices. Acta Psychologica, 96,
environmentally friendly in other areas. However, as 1–14.
 ARTICLE IN PRESS
.
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236 235

Ajzen, I. (1991). The theory of planned behavior. Organizational .

Olander, F., & Th^gersen, J. (1995). Understanding of consumer
Behavior and Decision Processes, 50, 179–211. behaviour as a prerequisite for environmental protection. Journal
Ajzen, I. (2002). Residual effects of past on later behavior: Habituation of Consumer Policy, 18, 317–357.
and reasoned action perspectives. Personality and Social Psychol- .
Olander, F., & Th^gersen, J. (2000). Perceived similarities and
ogy Review, 6, 107–122. dissimilarities among activities with environmental consequences.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting In E. Kirchler, C. Rodler, K. Meier, & E. Holzl . (Eds.), IAREP/
social behavior. Englewood Cliffs, NJ: Prentice-Hall. SABE 2000 conference. Fairness and cooperation (pp. 319–323).
Bagozzi, R. P. (1980). Causal models in marketing. Pittsburgh: Wiley. Vienna: University of Vienna.
Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Ouellette, J. A., & Wood, W. (1998). Habit and intention in everyday
Advances in experimental social psychology,, Vol. 6 (pp. 1–62). life: The multiple processes by which past behavior predicts future
New York: Academic Press. behavior. Psychological Bulletin, 124, 54–74.
Bentler, P. M., & Speckart, G. (1981). Attitudes ‘cause’ behaviors: A Pieters, R. G. M. (1991). Changing garbage disposal patterns of
structural equation analysis. Journal of Personality and Social consumers: Motivation, ability, and performance. Journal of Public
Psychology, 40, 226–238. Policy and Marketing, 10, 59–76.
Berger, I. E. (1997). The demographics of recycling and the structure Rogosa, D. (1980). A critique of cross-lagged correlation. Psycholo-
of environmental behavior. Environment and Behavior, 29, gical Bulletin, 88, 245–258.
515–531. Ronis, D. L., Yates, J. F., & Kirscht, J. P. (1989). Attitudes, decisions,
Bynner, J. (1994). Analyzing change over time using LISREL: Social and habits as determinants of repeated behavior. In A. R.
life and delinguency. In A. Dale, & R. B. Davies (Eds.), Analyzing Pratkanis, S. J. Breckler, & A. G. Greenwald (Eds.), Attitude
social and political change (pp. 99–117). London: Sage. structure and function (pp. 213–239). Hillsdale, NJ: Lawrence
Deary, I. J. (1995). Auditory inspection time and intelligence: Erlbaum.
What is the direction of causation? Developmental Psychology, Schultz, P. W., & Zelezny, L. (1999). Value as predictors of
31, 237–250. environmental attitudes: Evidence for consistency across 14
Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort countries. Journal of Environmental Psychology, 19, 255–265.
Worth: Harcourt Brace Jovanovich. Schwartz, S. H. (1968). Words, deeds, and the perception of
Engel, U., & Meyer, W. (1996). Structural analysis in the study of consequences and responsibility in action situations. Journal of
social change. In U. Engel, & J. Reinecke (Eds.), Analysis of Personality and Social Psychology, 10, 232–242.
change. Advanced techniques in panel data analysis (pp. 221–252). Schwartz, S. H. (1973). Normative explanations of helping behavior: A
Berlin: Walter de Gruyter. critique, proposal, and empirical test. Journal of Experimental
Epstein, S. (1983). Aggregation and beyond: Some basic issues on the Social Psychology, 9, 349–364.
prediction of behavior. Journal of Personality, 51, 360–392. Schwartz, S. H. (1977). Normative influence on altruism. In L.
Forbrugerstyrelsen og Milj^styrelsen (1996). Milj^belastningen ved Berkowitz (Ed.), Advances in experimental social psychology, Vol.
familiens aktiviteter. Report 1996:1, The National Consumer 10 (pp. 221–279). New York: Academic Press.
Agency, Copenhagen. Schwartz, S. H. (1992). Universals in the content and structure of
Frey, B. S. (1993). Motivation as a limit to pricing. Journal of values: Theoretical advances and empirical tests in 20 countries.
Economic Psychology, 14, 635–664. In M. P. Zanna (Ed.), Advances in experimental social psychology,
Gray, D. B. (1985). Ecological beliefs and behaviors: Assessment and Vol. 25 (pp. 1–65). San Diego: Academic Press.
change. Westport, CT: Greenwood Press. Schwartz, S. H. (1994). Are there universal aspects in the structure
Guagnano, G. A., Dietz, T., & Stern, P. C. (1994). Willingness to pay and content of human values? Journal of Social Issues, 50(4),
for public goods: A test of the contribution model. Psychological 19–45.
Science, 5, 411–415. Schwartz, S. H., & Howard, J. A. (1980). Explanations of the
Halkier, B. (1997). Nemt at t^rre ansvaret af p(a forbrugerne. moderating effect of responsibility denial on the personal
Information, January 28, 3. norm-behavior relationship. Social Psychology Quarterly, 43,
Harland, P., Staats, H., & Wilke, H. A. M. (1999). Explaining pro- 441–446.
environmental intention and behavior by personal norms and the Stern, P. C., & Dietz, T. (1994). The value basis of environmental
theory of planned behavior. Journal of Applied Social Psychology, concern. Journal of Social Issues, 50(3), 65–84.
29, 2505–2528. Stern, P. C., Dietz, T., & Guagnano, G. A. (1998). A brief inventory of
Heberlein, T. A. (1972). The land ethic realized: Some social values. Educational and Psychological Measurement, 58, 984–1001.
psychological explanations for changing environmental attitudes. Stern, P. C., Dietz, T., & Kalof, L. (1993). Value orientations,
Journal of Social Issues, 28(4), 79–87. gender, and environmental concern. Environment and Behavior, 25,
.
Joreskog, K. G. (1999a). Interpretation of R2 revisited. http:// 322–348.
ssicentral.com/lisrel/column5.htm Stern, P. C., Dietz, T., Kalof, L., & Guagnano, G. (1995). Values,
.
Joreskog, K. G. (1999b). What is the interpretation of R2? http:// beliefs, and proenvironmental action: Attitude formation toward
ssicentral.com/lisrel/column3.htm emergent attitude objects. Journal of Applied Social Psychology, 25,
.
Joreskog, .
K. G., Sorbom, D., Toit, S. d., & Toit, M. d. (1999). 1611–1636.
LISREL8: New statistical features. Chicago: Scientific Software Stern, P. C., & Oskamp, S. (1987). Managing scarce environmental
International. resources. In D. Stokols, & I. Altman (Eds.), Handbook of
Karp, D. G. (1996). Values and their effect on pro-environmental environmental psychology, Vol. 2 (pp. 1043–1088). New York:
behavior. Environment and Behavior, 28, 111–133. Wiley.
Landis, D., Triandis, H. C., & Adamopoulos, J. (1978). Habit and Th^gersen, J. (1995). Forbrugeradf!crdsunders^gelser med milj^m!cssigt
behavioral intentions as predictors of social behavior. Journal of sigte. Erfaringer og perspektiver. Arbejdsrapport 1, The Danish
Social Psychology, 106, 227–237. Environmental Protection Agency, Copenhagen.
Lorenz, F. O., Conger, R. D., Simons, R. L., & Whitbeck, L. B. (1995). Th^gersen, J. (1996). Recycling and morality. A critical review of the
The effects of unequal covariances and reliabilities on contempora- literature. Environment and Behavior, 28, 536–558.
neous inference: The case of hostility and marital happiness. Th^gersen, J. (1999a). The ethical consumer. Moral norms and
Journal of Marriage and the Family, 57, 1049–1064. packaging choice. Journal of Consumer Policy, 22, 439–460.
 ARTICLE IN PRESS
236 .
J. Thøgersen, F. Olander / Journal of Environmental Psychology 23 (2003) 225–236

Th^gersen, J. (1999b). Spillover processes in the development of a Verplanken, B., & Aarts, H. (1999). Habit, attitude, and planned
sustainable consumption pattern. Journal of Economic Psychology, behaviour: Is habit an empty construct or an interesting case of
20, 53–81. goal-directed automaticity? European Review of Social Psychology,
.
Th^gersen, J., & Olander, F. (1999). Danske forbrugeres oplevelse af 10, 101–134.
forskelle og ligheder mellem milj^relevante aktiviteter. Working Verplanken, B., Aarts, H., Knippenberg, A. v., & Knippenberg, C. v.
Paper 1999-3. The Aarhus School of Business, Department of (1994). Attitude versus general habit: Antecedents of travel mode
Marketing, Aarhus. choice. Journal of Applied Social Psychology, 24, 285–300.
.
Th^gersen, J., & Olander, F. (2002). Human values and the emergence Wenke, M. (1993). Konsumstruktur, UmweltbewuX tsein und Umwelt-
of a sustainable consumption pattern: A panel study. Journal of .
politik. Eine makrookonomische Analyse des Zusammenhanges in
Economic Psychology, 23, 605–630. .
ausgewahlten Konsumbereichen. Berlin: Duncker & Humblot.