Extending Planned Environmentalism
Anticipated Guilt and Embarrassment Across Cultures
Abstract
This paper cross-culturally tests an extended version of the planned behavior theory. Using cross-sectional surveys of 801 university students from four different cultures (high vs. low individualism, and English- vs. Spanish-speaking), we expected anticipated feelings of guilt to predict behavioral intention in cultures high on individualism, whereas anticipated feelings of embarrassment would be predictive of intention in cultures low on individualism. Results from a series of structural equation models showed that anticipated embarrassment had virtually the same effect as guilt across all four cultures. Although technically distinct, anticipated guilt and embarrassment were nearly indistinguishable from an individual perspective so that either concept is able to increase the explanatory power of the planned behavior theory for environmental conservation.
A high-quality communal life almost inevitably requires protecting collective interests against individuals’ self-interests. Affordable public transportation, access to an efficient health-care system, clean drinking water, and recreational public parks (to name just a few) all require a balance between the interests of the collective and the wants of the individual. Philosophically speaking, defending these collective interests can be achieved with reference either to moral standards or to conventions (see Tugendhat, 1994). While individual moral standards are grounded in moral concepts such as the welfare of others and justice, conventions relate to social standards and established traditions (see Turiel, 1985). One way in which moral standards and conventions are maintained is through the emotional responses that result when they are violated. Emotionally, embarrassment is felt when conventions are violated; guilt is experienced for moral transgressions (e.g., Keltner & Buswell, 1996).
Persons who value individual freedom tend to ignore conventions (see Triandis, 1994), and we would expect these individuals to respond to social praise and condemnation only weakly. Among such individualists (i.e., persons high on the individualism-collectivism dimension; see Schimmack, Oishi, & Diener, 2005), protecting collective interests against self-interests is achieved through personal moral self-restrictions. By contrast, persons who prioritize interdependence and collective interests are more sensitive to social sanctions. Expectedly, for these collectivists, protecting collective interests against self-interests is a matter of following social convention (see Triandis, 1994).
In other words, we anticipate that different people might be differentially motivated to adhere to moral standards or to social conventions. So, while individualists are expected to respond to anticipated guilt, collectivists are predicted to be more sensitive to possible embarrassment (see Triandis, 1994). Although anticipated future guilt has already been studied in the context of environmental conservation (Kaiser, 2006), anticipated embarrassment has not been explored. In the present research, we test an extended planned behavior model (e.g., Ajzen, 1991) in which anticipated guilt and embarrassment are differentially effective in promoting people’s intention to act more environmentally, depending on how individualistic their cultural background is (see Hofstede, 2001). To our knowledge, such a differential hypothesis has not yet been tested.
Self- and Collective Interests
Considerations of both collective and self-interests shape an individual’s motivation to act. Surprisingly, protecting one’s self-interest and the interest of others is usually not an either/or decision as it appears in Hardin’s (1968) “tragedy of the commons.” In the case of environmental conservation, the motivation to protect collective interests overlaps with the motivation to protect self-interests, yet both can lead to environmental protection (e.g., Schultz, 2001). While rational-choice-based models, such as the theory of planned behavior (e.g., Ajzen, 1991), allow us to address a wide array of utilitarian, self-interest-oriented reasons for action, they are often criticized for neglecting moral considerations (see Manstead, 2000). This is particularly problematic in morally relevant domains, such as the domain of environmental conservation (Thøgersen, 1996). In our review of the literature, we have found no reference to anticipated future social sanctions – as an alternative motivational source of behavior in the collective interest – being addressed in the planned behavior model beyond self-acknowledged compliance to social conventions, assessed with subjective norms.
The Planned Behavior Theory
According to the planned behavior theory (see Figure 1), intention is the ultimate and strongest determinant of behavior. Intention, in turn, is predicted by attitude toward performing a particular act, subjective norms, and perceived behavioral control. While attitude represents the expected utility of a given behavior, subjective norms stand for the motivation to comply with the behavior-relevant social conventions. The more a behavior depends on the presence of appropriate circumstances that are external to a person, the less a behavior is intentionally controllable. Thus, in addition to the relationship between perceived control and intention, the planned behavior theory also models a direct influence of perceived control on behavior (e.g., Ajzen & Madden, 1986). 
Prior research has shown that models of specific behavior can yield equivocal results for distinct conservation actions (e.g., McKenzie-Mohr, Nemiroff, Beers, & Desmarais, 1995) or when exploring people living in different countries (e.g., Lévy-Leboyer, Bonnes, Chase, Ferreira-Marques, & Pawlik, 1996). But a more general version of the planned behavior theory, aiming to explain an entire class of behavior and, thus, based on multi-item measures, has proven to be less sensitive to sample and/or behavior specifics and is expected to produce less ambiguous and more generalizable results (see Kaiser, Schultz, & Scheuthle, 2007). If the theory is tested using multi-item measures, the direct influence of perceived behavioral control upon behavior must be omitted, as was shown by Kaiser and Gutscher (2003).
Conventional and Moral Obligations to Act Altruistically
Although relatively successful in predicting a wide range of behavior, the rational-choice-based planned behavior model has been criticized for being overly utility and self-interest-oriented (Manstead, 2000). Several attempts to extend the theory into a more collective-interest-oriented (i.e., altruistic) direction have to date only considered moral concepts. Yet existing research on environmental conservation has failed to show a net gain in the explanatory power of a moral-norms-extended theory of planned behavior (e.g., Harland, Staats, & Wilke, 1999; Heath & Gifford, 2002). In three systematic explorations with the general application of the planned behavior theory, based on multi-item measures, moral norms did not significantly improve the explanatory power (see Kaiser, 2006; Kaiser, Hübner, & Bogner, 2005; Kaiser & Scheuthle, 2003). Instead, the evidence to date shows a substantial overlap between attitude and moral norms, which resulted in suppressor effects in two of the three studies of Kaiser and his colleagues (see also Heath & Gifford, 2002). It seems as though moral norms are already represented in people’s environmental attitudes, either as its powerful antecedent or as its evaluative essence (see Berenguer, Corraliza, & Martín, 2005). By contrast, and in line with theoretical expectations, Kaiser (2006) nevertheless found anticipated guilt feelings to be a supplementary determinant of people’s intention to act environmentally.
The current study explores anticipated feelings of embarrassment as an additional source of collective-interest-oriented (i.e., altruistic) behavior. Such a link is supported by data from a cross-cultural comparison of U.S.-American and Chinese consumers (Chan & Lau, 2001). In this study, the researchers applied a traditional planned behavior model to explain purchasing of green products. Results showed that Chinese customers (the collectivists) were more sensitive to social expectations and pressure compared to the individualistic counterparts from the United States, as subjective norms were significantly more intention-relevant in the Chinese sample (but see Ando, Ohnuma, & Chang, 2007).
Research Goals
In the current study, we apply a general version (based on multi-item measures) of the planned behavior theory to environmental conservation. Our primary goal was to test a model, in which anticipated guilt and embarrassment are included as additional determinants of intention (see Figure 1). A secondary goal was to explore the differential nature of this expansion, by testing the moderating role of culture. So, depending on people’s socio-cultural background, we expect them to show differential sensitivity for guilt and embarrassment. While individualists are expected to be relatively more sensitive to anticipated guilt, collectivists are predicted to more strongly respond to anticipated embarrassment.
Methods
To test our hypotheses, survey data were collected from samples of students in the United States, India, Spain, and Mexico.
Design and Participants
Countries were sampled systematically, based on a 2 × 2 language-individualism index design. Two samples were drawn from English (United States, India) and two from Spanish-speaking nations (Mexico, Spain); two of these countries (United States, Spain) are considered to be high on individualism based on Hofstede’s (2001) individualism index, while two (India, Mexico) are low – if not in an absolute sense then at least comparatively.
To control for as many demographic differences as possible, we exclusively sampled university students (see Van de Vijver & Leung, 1997): 200 in each of the four countries (201 in Mexico). Of the 801 college students, 75.2% (n = 602) were female and 24.8% (n = 198) male. Their mean age was 21.8 (SD = 3.5). Despite our attempt to match samples, we also found some differences in our four groups. Females were slightly overrepresented in the United States (164 instead of 150) and underrepresented in India (132 instead of 150): χ2(3) = 17.0, p < .001. Participants from Spain (M = 20.7; SD = 2.3) and India (M = 21.1; SD = 1.7) were found to be somewhat younger compared to those from the United States (M = 22.9; SD = 5.0) and those from Mexico (M = 22.6; SD = 3.7): F(3, 796) = 20.5, p < .001, η2 = 7.2%.
Measures
The questionnaire consisted of seven measures: attitude toward behavior, subjective norms, perceived behavioral control, anticipated feelings of guilt and of embarrassment, intention, and behavior. Note that all measures were identically assembled in each country. “Not applicable” was a response option with each item.
Conservation behavior was measured with 60 behaviors, 50 from a recent version of the General Ecological Behavior (GEB) scale (Kaiser & Wilson, 2004). Ten nonredundant behaviors are additionally taken from a 12-item instrument that has been used in prior multicountry comparisons (e.g., Schultz et al., 2005). Such an inclusion of new items into the GEB scale is possible, because the GEB measure refers to a particular measurement approach with any coincidental sample of conservation behaviors rather than to a standard set of behaviors (Kaiser & Wilson, 2000). Typical item examples are: (a) I ride a bicycle or take public transportation to work or school; (b) I reuse my shopping bags; (c) I use a clothes dryer [i.e., unecological behavior]; (d) I talk with friends about problems related to the environment; (e) In nearby areas (around 30 kilometers; around 20 miles), I use public transportation or ride a bike; (f) I buy products in refillable packages; (g) I put dead batteries in the garbage [i.e., unecological behavior]; and (h) I contribute financially to environmental organizations. Negative responses to unecological behaviors were recoded as ecological engagement responses and vice versa.
All 60 behaviors were calibrated as our GEB measures using the dichotomous Rasch model separately for each of the four country samples. Overall, the four behavior measures were all in line with prior scale explorations (e.g., Kaiser & Wilson, 2000, 2004) – except for the fact that 10 new behaviors were adopted – and generally corroborate the fact that the 60 conservation behaviors fall into a single class of behaviors. Their reliabilities were rel = .71 (United States), rel = .73 (India), rel = .62 (Mexico), and rel = .77 (Spain).
The planned behavior measures conform to common assessment practices in this field of research (e.g., Ajzen & Fishbein, 2005; Ajzen & Madden, 1986). Following previous studies (e.g., Kaiser, 2006; Kaiser & Scheuthle, 2003), we employed the same set of eight conservation behaviors (only four with anticipated embarrassment; see below) and presented them in two groups of four behaviors. The eight activities were all taken from our conservation behavior measure, in fact, they are the ones listed above.
Behavior intention was measured by rating each of four behavior items on (a) a 5-point likely/unlikely response scale with the opening phrase “In the future, I will .. .” and on (b) a 5-point decided/undecided response scale with the opening phrase “In the future, I intend to .. ..” The internal consistency of the eight items was α = .73, based on the combined sample (N = 801).
Attitude was measured by means of four behaviors on (a) a 5-point good/bad response scale and on (b) a 5-point appropriate/inappropriate scale. The eight attitude items were internally consistent with an α-coefficient of .75.
Subjective norms were measured by rating each of four behavior statements on (a) a 5-point likely/unlikely response scale with the opening phrase “Most people who are important to me think I should .. .” and on (b) an agree/disagree response scale with the opening phrase “Most people who are important to me .. ..” The eight subjective norms items were internally consistent with α = .78.
Perceived behavioral control was measured by rating each of four conservation behaviors on (a) a 5-point easy/difficult response scale and on (b) a 5-point simple/complicated scale. The internal consistency of the eight control items was α = .68.
Anticipated feelings of guilt were assessed by rating each of eight behaviors on a 5-point agree/disagree response scale, four with the opening statement (a) “I would feel guilty if I would not .. .” and four with (b) “My conscience would bother me if I would not .. ..” (Note that with unecological behavior a positive wording was used.) The internal consistencies of the eight anticipated feelings of guilt items was α = .86.
Anticipated embarrassment was measured by rating each of four behaviors on a 5-point agree/disagree response scale with the opening phrase “I would feel embarrassed, if others knew that I would not .. ..” (Again, a positive wording was used with unecological behavior.) Because of a mistake in the production process of the Spanish version of the questionnaire, four other statements were inadvertently omitted so that we had to settle with only four items. However, given their internal consistency, which was acceptable with α = .85, we do not believe that this oversight substantially affected the results.
Procedure
The Spanish version represents a literal translation of the original English questionnaire. The translation was conducted by the third and fourth authors who are native Mexican-Spanish or Spanish-Spanish speakers respectively, with English as their second language. Such a straightforward procedure was reasonable, since we only sought construct and not item equivalence with our conventional planned behavior measures (see Van de Vijver & Leung, 1997). A few of the behavior items had to be adjusted to the specific local contexts. For example, air-conditioning instead of heating was used as one of the behavior items in Mexico. These adaptations were trivial in nature, as evidenced by the successful calibration of the four behavior scales. Lecturers at the local universities distributed the questionnaires to students. The surveys were completed during class sessions.
Statistical Analysis
Because of the statistical identifiability of the factor loadings, each of the planned behavior concepts needs at least two indicators. These two indicators represent two parcels. For each parcel, we averaged the four items that we had used with one of the two response formats in the assessment of the planned behavior components. In the case of anticipated embarrassment, we could only combine two items per parcel. Note that for situations like this one, when the primary goal is theory testing and when measurement procedures are well-established, parceling is a methodologically acceptable strategy (e.g., Little, Cunningham, Shahar, & Widaman, 2002).
All structural equation models were assessed using the maximum likelihood method. Except for the exploration of the structure of the two newly included measures, all models were tested confirmatorily (i.e., without allowing for any model modification). Fixing the behavior scales’ reliabilities and their corresponding error variances (see Figure 2) at the pre-estimated values from the Rasch scale calibration was only possible using a correlation rather than a covariance matrix. As is customary in reporting results from structural equation models, we supplement the χ2-statistic with two widely-used goodness of fit indices: the comparative fit index (CFI) and the root mean square error of approximation (RMSEA). Following traditional recommendations, cutoff values around .90 for the CFI, and around .08 for the RMSEA were adopted as indicators of a reasonable fit between model-implied and observed correlations (Bentler & Bonett, 1980; Browne & Cudeck, 1993). 
Results
We begin by exploring the discriminant validity of the two measures: anticipated feelings of guilt and embarrassment. Second, we test whether the planned behavior model can be extended with either anticipated guilt or with anticipated embarrassment feelings. Finally, we examine the differential impact of guilt and embarrassment as a function of culture. For these latter trials, we adopted a multisample comparison approach in which the data from the four samples are analyzed simultaneously with all parameters constrained to be equal, except for the behavior-related parameters. Such a multisample comparison is useful and recommended when one wishes to explore the structural invariance cross-culturally; that is equivalence in the structure of the measures and in the structure of the concepts (e.g., Steenkamp & Baumgartner, 1998).
From previous research, we already knew that, with conservation behavior, measurement equivalence across countries should not be assumed (e.g., Scheuthle, Carabias-Hütter, & Kaiser, 2005). For this reason, the behavior scales were calibrated separately by country using the Rasch approach. Accordingly, the reliability indices and the uniquenesses were fixed to the estimated values from the various Rasch model tests. As a consequence, the behavior variances and the β-weights of intention, because they depend on the behavior-related estimates, were allowed to vary in the four country-models (see Figure 2).
Anticipated Guilt and Embarrassment Feelings
While exploring the appropriate measurement model for anticipated guilt and embarrassment, we first tested a one-factor model and found considerable misfit: χ2(32) = 197.3; CFI = .90; RMSEA = .16; N = 801. A 2-factor model, by contrast, showed a significantly better, and reasonable fit (Δχ2(1) of 118.6; see Model 2a in Table 1). This improvement, in turn, settles concerns about the two concept measures’ discriminant validity. Despite the relatively high correlation between anticipated guilt and embarrassment feelings (i.e., r = .80), the two concepts are technically distinguishable. 1The two measures’ discriminant validity also shows in a conventional exploratory factor analysis of the eight anticipated guilt and four embarrassment items with a principal factors’ extraction and a subsequent varimax rotation. Two factors yielded Eigenvalues greater than one (i.e., 5.7 & 1.3), covering 58.2% of the common variance of the 12 guilt and embarrassment items. All communalities were substantial (i.e., above .40, with only one exception at .35); and all but one item had nontrivial loadings of a > .50 on the a priori anticipated factor, either anticipated guilt or embarrassment. Unexpectedly, one guilt item was found to have its highest loading (a = .58) on the embarrassment factor with a cross-loading of a = .28 on the a priori anticipated guilt factor. Overall, our items had nontrivial cross-loadings on the alternative second factor (ranging from a = .12 to a = .43), which confirms the conceptual overlap of the two measures. This conceptual overlap can also be addressed by using an oblique rotation (direct oblimin) instead of the orthogonal varimax rotation. Such a modification yields substantially reduced cross-loadings at the expense of two correlated factors (r = .61).
Since not all fit indicators in this 2-factor model are fully satisfactory (particularly the RMSEA value still seems a little too high at .09; see Table 1), we investigated the most obvious source of this misfit: the invariance of the guilt-embarrassment correlation. In other words, we wondered whether the strength of the relationship between guilt and embarrassment might be country-specific rather than universal. To further explore this hypothesis, we tested two more models. In Model 2c (see Table 1), we allowed the correlations between anticipated guilt and embarrassment to vary across all four countries. In Model 2b (see Table 1), only India was allowed to have a distinct correlation between the two latent variables. Table 1 gives the model fit indices, which generally corroborate both models superiority – obvious in a significant Δχ2(1) of 18.5 – relative to the universal model, which assumes the same guilt-embarrassment correlations in all four countries. From the nonsignificant Δχ2(2) of 1.9, indicating an absence of improvement in overall model fit, we learn that the Models 2c and 2b can be regarded as equivalent. In fact, while the guilt-embarrassment correlation in the United States, Spain, and Mexico seems to be uniform at about r = .84, for India the same figure is around r = .64.
Guilt and Embarrassment-Extended Planned Behavior
Without any model modification, the planned behavior model fits the data reasonably well from an empirical point of view: χ2(150) = 350.9, p < .001, CFI = .91, RMSEA = .082; N = 801. The fit statistics look even better if anticipated guilt and embarrassment are also included in the estimation procedure – not as determinants but as covariates of the determinants of intention (Model 1 in Table 2). Since the latter – the one we call the traditional extended planned behavior model – permits a more accurate comparison with the three subsequently tested guilt and embarrassment-extended models, we use it as our benchmark model.
In this traditional but extended planned behavior model, the determinants, attitude (β = .23), subjective norms (β = .69), and perceived behavioral control (β = –.14, p > .05; two-tailed), explain 57% of the variance of behavior intention. In turn, behavior intention (β = .65, .85, .41, .70) explains 43%, 73%, 17%, and 50% of the variance in a person’s conservation behavior in the United States, Spain, Mexico, and India, respectively. Table 3 gives the correlation matrix of the theoretically relevant latent concepts.
As Table 2 shows, the inclusion of either anticipated guilt (Model 2) or embarrassment (Model 3) as an additional determinant of intention in all four cultures, or using guilt with the United States and Spain (the two cultures high in individualism), and embarrassment with Mexico and India (the two cultures low in individualism; Model 4), leaves the overall fit statistics virtually unaffected.
Nevertheless, even if the overall fit of a model appears acceptable, the theoretically substantial paths may still prove to be poorly specified (McDonald & Ho, 2002). After subtracting the contribution of the measurement model from the overall fit statistics, we find both model parts, that is the theoretically substantial one and the measurement model, to acceptably fit the data (see Table 4). An RMSEA value of .07 looks reasonable – given the reference value close to .08. From a model fit perspective, it seems fair to conclude that the individualism-dependent guilt or embarrassment-extended planned behavior model is quite accurate in the way it identifies the relations among the concepts, as an exclusively guilt or embarrassment-extended model would be in all four cultures. This is, however, not enough to justify a model extension with anticipated guilt and embarrassment. For that, anticipated guilt and embarrassment would also need to contribute to the explained variance of intention.
In the individualism-dependent model, the determinants, attitude (β = .23), subjective norms (β = .65), perceived behavior control (β = –.12, p > .05; two-tailed), and anticipated guilt (β = .04, p > .05; two-tailed) explain 58% of the variance of behavior intention. Behavior intention (β = .66, .85), in turn, explains 43% and 73% of the variance in a person’s conservation behavior in the two individualistic cultures. In the two cultures low in individualism, attitude (β = .24), subjective norms (β = .66), perceived behavioral control (β = –.12, p > .05; two-tailed), and anticipated embarrassment (β = .00, p > .05; two-tailed) explain 56% of the variance of behavior intention. In turn, behavior intention (β = .41, .70) explains 17% and 49% of the variance in a person’s conservation behavior. Obviously and unexpectedly, perceived control, anticipated guilt and embarrassment do not seem to directly affect intention, despite their remarkable overlap with intention (see Table 3).
We believe there is a methodological explanation for subjective norms’ excessive impact and the apparent negligible direct guilt/embarrassment and perceived control effect on intention: Highly correlated determinants of intention (see Table 3) together with a rather poor quality of the subjective norms’ measure.
Combined, nearly 70% to 85% of the variance of the two indicators of subjective norms represent error variance (see Figure 2). Necessarily, a correction for measurement error attenuation inflates subjective norms’ relevance for intention at the expense of the other determinants, which largely cover the same information.
To test our explanation, we removed subjective norms’ effect on intention and only considered subjective norms as a covariate of the other determinants of intention (see Figure 2). Overall, the statistical fit looks comparable to what we found previously with Model 4 in Table 2: χ2(317) = 668.0, p < .001, CFI = .93, RMSEA = .075. This time, however, attitude (β = .41), perceived behavioral control (β = .19), anticipated guilt (β = .34) or embarrassment (β = .24) affect behavior intention significantly. Note that all other estimates of the model stay essentially unchanged (i.e., ± .03) despite the omission of subjective norms’ direct path to intention.
Discussion
Our results support previous research and extend former conclusions cross-culturally (e.g., Kaiser, 2006). As before, a substantial amount of the variance in behavioral intention (≥ 55%) was explained by attitude, subjective norms, and – at least indirectly, mediated by subjective norms – by perceived behavioral control, anticipated feelings of guilt and embarrassment. Our findings are consistent with Manstead’s (2000) proposition that including anticipated guilt feelings significantly contributes to the explanation of people’s intention to act in a more altruistic manner. Inconclusive still is whether subjective norms necessarily mediate the anticipated guilt and embarrassment effect on intention or, as was found previously (see Kaiser, 2006), whether anticipated guilt and embarrassment uniquely add to the account of intention.
The present study shows evidence of a mediated effect, which seems plausible, given that subjective norms and anticipated feelings of embarrassment both probably derive from an underlying concern about what others might think. The excessive subjective norms effect (β ≈ .65) similarly suggests an enhanced sensitivity for social pressure, which is rather typical for students when inquired about environmental issues (see Kaiser, Ranney, Hartig, & Bowler, 1999). In other words, a more conclusive picture might require more population-representative samples as students’ general sensitivity for social pressure could have masked the more subtle anticipated embarrassment effect, which should be better recognizable with more mature and socially less sensitive samples (see Kaiser, 2006).
Unexpectedly, anticipated guilt and embarrassment were indistinguishable in their ability to predict intention (Figure 2). Although they are technically distinct (which confirms the two measures’ discriminant validity; see Table 1 and 1The two measures’ discriminant validity also shows in a conventional exploratory factor analysis of the eight anticipated guilt and four embarrassment items with a principal factors’ extraction and a subsequent varimax rotation. Two factors yielded Eigenvalues greater than one (i.e., 5.7 & 1.3), covering 58.2% of the common variance of the 12 guilt and embarrassment items. All communalities were substantial (i.e., above .40, with only one exception at .35); and all but one item had nontrivial loadings of a > .50 on the a priori anticipated factor, either anticipated guilt or embarrassment. Unexpectedly, one guilt item was found to have its highest loading (a = .58) on the embarrassment factor with a cross-loading of a = .28 on the a priori anticipated guilt factor. Overall, our items had nontrivial cross-loadings on the alternative second factor (ranging from a = .12 to a = .43), which confirms the conceptual overlap of the two measures. This conceptual overlap can also be addressed by using an oblique rotation (direct oblimin) instead of the orthogonal varimax rotation. Such a modification yields substantially reduced cross-loadings at the expense of two correlated factors (r = .61).
Contrary to our expectations, our results did not show a differential pattern of relationships for anticipated guilt and embarrassment in promoting people’s intention to act environmentally, as a function of culture. Even in India, where embarrassment and guilt seem to be somewhat more distinct, we found anticipated guilt and embarrassment to operate equally well as determinants of intention within a planned behavior framework (Figure 2). In other words, we have reason to believe in the universal nature of anticipated guilt/embarrassment’s supplementary significance for intention, at least in the domain of environmental conservation. Our four-sample comparison additionally speaks not only of an equivalent theoretical structure, but also of concept measures that operate equivalently cross-culturally (e.g., Steenkamp & Baumgartner, 1998). Thus, it seems sensible to assume that the proposed model holds universally.
Research with cultures rather than individuals as units of analysis commonly yields different results (Schimmack et al., 2005). Thus, one reason why our hypothesis was not substantiated could be that we portray cultures instead of individuals in terms of individualism-collectivism (see Schwartz, 1994). In other words, we could have designated the wrong samples as individualistic or collectivistic. The individualism-collectivism dichotomy indeed inverts if classification of our four samples is based on individual level data (i.e., by using Schwartz’ approach instead of the cultural taxonomy provided by Hofstede, 2001). However, in additional analyses not reported here, we find that our conclusions remain the same even if the originally anticipated guilt and embarrassment-effect pattern is inverted and if we define individualism-collectivism individually – using personal values – instead of culturally. (Details of this analysis are available upon request.)
Since our research has resulted in some surprising and unanticipated findings, our matched student samples could be criticized as being the source of it. This is, because homogeneous samples commonly mean reduced variability. Restricted variability in turn is suspected to dilute correlations (e.g., Tabachnick & Fidell, 2006). The restricted variability could have been additionally aggravated with researching comparatively social-pressure-sensitive students. Students’ propensity to respond in a socially desirable manner (i.e., social desirability effect; see Kaiser et al., 1999) could, for instance as a ceiling effect, have constrained the variability of variables. Moreover, as student samples are certainly not representative, the generalizability of the findings to the entire population becomes another issue. Despite all shortcomings, student samples are nevertheless a sensible and recommended means to control for extraneous factors, for example demographic differences, and, thus, possibly confounding variables in cross-cultural research (e.g., Van de Vijver & Leung, 1997).
Since our research employs some nonstandard methods, we would like to address two other possible shortcomings. First, parceling can be expected to result in more replicable findings on the conceptual level, as the procedure normally results in more sound concept indicators (Little et al., 2002). Second, common method variance can never be ruled out entirely as a nuisance effect with any cross-sectional survey research. For example, the comparatively high correlations between the planned behavior concepts (see Table 3), could derive from employing a compulsory measurement paradigm (i.e., the compatibility principle; e.g., Ajzen & Fishbein, 2005), which has already been suspected to artificially inflate the planned behavior theory-implied relationships (Kaiser et al., 2007).
Our research is consistent with prior findings showing that anticipated guilt significantly enhances the explanatory power of people’s intention to act in a certain way, at least with collective interest-relevant behavior like environmental conservation (e.g., Kaiser, 2006). Moreover, our results show that anticipated embarrassment had virtually the same effect on intention as anticipated guilt. Unexpectedly, this was the case across all four English or Spanish-speaking samples, representing varying levels of individualism (this was true for both culture-based or person-based analyses). While including either anticipated guilt or embarrassment increased the explanatory power of people’s intention to act environmentally, the effect was not moderated by culture.
Florian G. Kaiser is Professor of Personality and Social Psychology at the Otto von Guericke University, Magdeburg, Germany. His research interests include theory and measurement of individual behavior and attitudes, particularly with respect to nature conservation, health, and psychological restoration.
P. Wesley Schultz is Professor of Social Psychology at California State University, San Marcos. His research interests lie in applied social psychology, particularly in the area of sustainable behavior. His current work focuses on social norms, and the importance of social norms in fostering sustainable behavior.
Jaime Berenguer is Assistant Professor at the Universidad Autonoma de Madrid, Spain. His research interests are focused on the relationship between empathy and environmental attitudes and actions.
Victor Corral-Verdugo is Professor of Environmental Psychology at the Universidad de Sonora, Hermosillo, Mexico. His research focuses on the determinants of sustainable behavior.
Geetika Tankha, Ph.D., is a freelance psychologist and visiting faculty member at Rajasthan University of Health Sciences, Jaipur, India. At present, she works as a project fellow on a university grants commission project on positive psychology. Her areas of interest are environmental, positive, and clinical psychology.
References
(1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179–211.
(2005). The influence of attitudes on behavior. In , The handbook of attitudes (pp. 173–221). Mahwah, NJ: Erlbaum.
(1986). Prediction of goal-directed behavior: Attitudes, intentions, and perceived behavioral control. Journal of Experimental Social Psychology, 22, 453–474.
(2007). Comparing normative influences as determinants of environmentally conscious behaviours between the USA and Japan. Asian Journal of Social Psychology, 10, 171–178.
(1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88, 588–606.
(2005). Rural-urban differences in environmental concern, attitudes, and actions. European Journal of Psychological Assessment, 21, 128–138.
(1993). Alternative ways of assessing model fit. In , Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage.
(2001). Explaining green purchasing behavior: A cross-cultural study on American and Chinese consumers. Journal of International Consumer Marketing, 14(2/3), 9–40.
(1968). The tragedy of the commons. Science, 162, 1243–1248.
(1999). Explaining proenvironmental intention and behavior by personal norms and the theory of planned behavior. Journal of Applied Social Psychology, 29, 2505–2528.
(2002). Extending the theory of planned behavior: Predicting the use of public transportation. Journal of Applied Social Psychology, 32, 2154–2189.
(2001). Culture’s consequences (2nd ed.). Beverly Hills, CA: Sage.
(2006). A moral extension of the theory of planned behavior: Norms and anticipated feelings of regret in conservationism. Personality and Individual Differences, 41, 71–81.
(2003). The proposition of a general version of the theory of planned behavior: Predicting ecological behavior. Journal of Applied Social Psychology, 33, 586–603.
(2005). Contrasting the theory of planned behavior with the value-belief-norm model in explaining conservation behavior. Journal of Applied Social Psychology, 35, 2150–2170.
(1999). Ecological behavior, environmental attitude, and feelings of responsibility for the environment. European Psychologist, 4, 59–74.
(2003). Two challenges to a moral extension of the theory of planned behavior: Moral norms and just world beliefs in conservationism. Personality and Individual Differences, 35, 1033–1048.
(2007). The theory of planned behavior without compatibility? Beyond method bias and past trivial associations. Journal of Applied Social Psychology, 37, 1522–1544.
(2000). Assessing people’s general ecological behavior: A cross-cultural measure. Journal of Applied Social Psychology, 30, 952–978.
(2004). Goal-directed conservation behavior: The specific composition of a general performance. Personality and Individual Differences, 36, 1531–1544.
(1996). Evidence for the distinctness of embarrassment, shame, and guilt: A study of recalled antecedents and facial expressions of emotion. Cognition and Emotion, 10, 155–171.
(1996). Determinants of pro-environmental behaviors: A five countries comparison. European Psychologist, 1, 123–129.
(2002). To parcel or not to parcel: Exploring the question, weighing the merits. Structural Equation Modeling, 9, 151–173.
(2000). The role of moral norm in the attitude-behavior relation. In , Attitudes, behavior and social context: The role of norms and group membership (pp. 11–30). Mahwah, NJ: Erlbaum.
(2002). Principles and practice in reporting structural equation analysis. Psychological Methods, 7, 64–82.
(1995). Determinants of responsible environmental behavior. Journal of Social Issues, 51(4), 139–156.
(2005). The motivational and instantaneous behavior effects of contexts: Steps toward a theory of goal-directed behavior. Journal of Applied Social Psychology, 35, 2076–2093.
(2005). Individualism: A valid and important dimension of cultural differences between nations. Personality and Social Psychology Review, 9, 17–31.
(2001). The structure of environmental concern: Concern for self, other people, and the biosphere. Journal of Environmental Psychology, 21, 327–339.
(2005). Values and their relationship to environmental concern and conservation behavior. Journal of Cross-Cultural Psychology, 36, 457–475.
(1994). Are there universal aspects in the structure and contents of human values? Journal of Social Issues, 50(4), 19–45.
(1998). Assessing measurement invariance in cross-national consumer research. Journal of Consumer Research, 25, 78–90.
(2006). Using multivariate statistics (5th ed.). Boston, MA: Allyn and Bacon.
(1996). Recycling and morality: A critical review of the literature. Environment and Behavior, 28, 536–558.
(1994). Culture and social behavior. New York: McGraw-Hill.
(1994). Vorlesungen über Ethik [
Lectures on ethics ] (2nd ed.). Frankfurt a. M., Germany: Suhrkamp.(1985). The development of social knowledge. Morality and convention (2nd ed.). Cambridge, MA: University Press.
(1997). Methods and data analysis for cross-cultural research. Thousand Oaks, CA: Sage.
The present research was supported by the Department of Technology Management at Eindhoven University of Technology, Eindhoven, The Netherlands. We gratefully acknowledge Katarzyne Byrka for her assistance in analyzing the data, Steven Ralston for his language support, and Katarzyne Byrka, Cees Midden, Miguel Ruiz, and three anonymous reviewers for their comments on earlier drafts of this paper.
