Trust as a Driver of Pro-Ecological Behaviour: The Power of Experts and Interpersonal Networks in a Low-Trust Context

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Topic is timely. The literature review and rationale for the study are excellent. The topic is likely to be of interest to a large audience. The following are my suggestions to improve the manuscript:

-Introduction is well done, with good use of sources, explanations, and good foundations for the study. 

-The sample is large (N = 1008) but readers are not likely to be convinced it was representative, as no sampling strategy is described. How exactly were participants chosen and then selected, what was the response rate if a national probabilistic sample was used? Was the sampling stratified, quota, probabilistic? 35-40 minutes per interview seems too long for close-ended questions (rated on Likert scales). Were narrative summaries or transcriptions also collected?  It is stated in the discussion the sample was cross-sectional. Were the participants demographics matched to national percentages? More description is needed on sampling. If the sample was collected through convenience, that should be stated--it is not a disqualifying characteristic as long as it is stated that was the approach, if indeed the case, then noting generalizations limitations, in addition to the fact these were self-reports (as acknowledged in the discussion of the manuscript). 

-On the different referent groups, I recommend that a ONEWAY analysis of variance be done, with perhaps a Tukey procedure for cross-group comparisons. Perhaps the only significant difference is  among the highest and lowest means. It should be an easy computation in SPSS. 

-On the dependent measure or index, more needs to be said which items were used, how many. Perhaps put asterisks on those items listed in the appendix that reference pro-environment choices.  It seems possible those items could have been factor analyzed to see if there are multiple dimensions present, rather than assume that those items in the index are measuring only one dimension (was it a weighted score=average of their sum divided by number of items)? The authors can also compute the internal consistency of the index (Cronbach's alpha). 

For the regression table, perhaps the eta square for each predictor can be added as well (percent of variance accounted by each item). 

The discussion can elaborate more on implications on alternative explanations and limitations.  The authors do a great job tying social dilemmas to individualism in regards to pro-ecological behavior. But there are more explanations can be use more discussion, such as a possible role of Diffusion of Responsibility for not endorsing those pro-ecological behaviors.

Also, separating national identity from "community identity," which the authors mention. The former having a negative connotation, the latter a higher value for social proof based on immediacy of social comparison based on similar others...which facilitates norm creation as the authors state in the introduction.   

Framing effects can also be involved, with ecological issues often framed in terms of losses, which increases pro-action. Perhaps some of these explanations can be ruled out with the existing data, but if now, those should be mentioned as viable explanations also. 

Overall the manuscript has a good focus by focusing in Bulgaria, a society that due to its history has had low institutional trust. In case the authors have not seen the case of Bangladesh, where institutional trust is also very low, but yet community groups clean their waterways in what is really an impressive feat (https://www.youtube.com/watch?v=xCLobD-ZkE0) . Exactly what these authors are describing in their paper.

Well wishes on revisions. 

 

Author Response

Reviewer 1:

Topic is timely. The literature review and rationale for the study are excellent. The topic is likely to be of interest to a large audience. The following are my suggestions to improve the manuscript:

Introduction is well done, with good use of sources, explanations, and good foundations for the study. 

Comment:

The sample is large (N = 1008) but readers are not likely to be convinced it was representative, as no sampling strategy is described. How exactly were participants chosen and then selected, what was the response rate if a national probabilistic sample was used? Was the sampling stratified, quota, probabilistic? 35-40 minutes per interview seems too long for close-ended questions (rated on Likert scales). Were narrative summaries or transcriptions also collected?  It is stated in the discussion the sample was cross-sectional. Were the participants demographics matched to national percentages? More description is needed on sampling. If the sample was collected through convenience, that should be stated--it is not a disqualifying characteristic as long as it is stated that was the approach, if indeed the case, then noting generalizations limitations, in addition to the fact these were self-reports (as acknowledged in the discussion of the manuscript). 

Thank you for this important comment. We agree that additional clarification regarding the sampling procedure and data collection method is necessary. The survey was conducted by a professional research agency using a nationally representative stratified multistage sampling design covering the adult population of Bulgaria.

The sampling frame was based on the latest official population statistics from the National Statistical Institute. The sample was stratified by administrative region (the 28 administrative districts) and type of settlement (capital, regional centers, small towns, and villages). Within the selected strata, a cluster (nest) selection of settlements or urban districts was applied, followed by the selection of starting addresses. Interviewers followed a random route procedure, contacting households until the required number of interviews was achieved. Only one eligible adult respondent per household was interviewed, with quotas applied for gender and age to ensure demographic balance.

Data were collected through face-to-face semi-structured interviews using tablets (CAPI) conducted in respondents’ homes. The typical interview length of 35–40 minutes reflects the fact that the questionnaire contained several thematic modules (e.g., trust, pro-environmental behaviour, institutional attitudes, and socio-demographic variables), rather than only short Likert-scale batteries. The interview was structured and quantitative in nature; therefore, no narrative responses or qualitative transcripts were collected.

Quality control included multiple procedures implemented by the research agency: supervision of interviewers by regional supervisors, telephone verification of 10–15% of completed interviews, and software-based checks including GPS verification of interview locations and detection of anomalous response patterns.

The maximum stochastic sampling error for proportions around 50% is ±3.1%, which is standard for national surveys of approximately 1000 respondents. We have clarified these methodological details in the revised manuscript – the new text is highlighted.

Additionally, we have also revised and expanded Table 1 (Sample Characteristics). The updated table now presents not only the sample distribution but also the corresponding population proportions based on official statistics, together with the difference between sample and population shares and t-tests for statistical comparison.

This additional information allows readers to directly assess the degree to which the achieved sample corresponds to the demographic structure of the Bulgarian adult population. As shown in the revised table, the sample closely matches the population distribution for gender and most age groups, with statistically non-significant differences in nearly all categories. The only notable deviation appears in the 70+ age group, which is somewhat underrepresented in the sample—a common issue in face-to-face surveys due to accessibility and response limitations among the oldest population groups. Overall, the comparison confirms that the sample provides a good approximation of the population structure, supporting the representativeness of the survey data.

Comment:

On the different referent groups, I recommend that a ONEWAY analysis of variance be done, with perhaps a Tukey procedure for cross-group comparisons. Perhaps the only significant difference is among the highest and lowest means. It should be an easy computation in SPSS. 

Thank you for this suggestion. In the current study, the analysis does not aim to compare mean differences between referent groups. Instead, our research design focuses on examining the associations between variables, particularly the relationship between trust and pro-environmental behaviour. For this reason, we employed regression analysis, which allows us to assess the relative contribution of multiple predictors simultaneously.

Since the study does not formulate hypotheses about mean differences between referent groups, conducting a one-way ANOVA with post hoc comparisons (e.g., Tukey tests) would not directly address the research questions of the paper. Therefore, we have retained the regression-based analytical strategy, which is more appropriate for testing the relationships examined in this study.

At the same time, another reviewer suggested conducting group comparisons based on settlement type. In response, we performed an exploratory one-way ANOVA with Tukey post hoc tests using settlement type as the independent variable and both trust and pro-environmental behaviour as dependent variables. The results indicated no statistically significant differences in pro-environmental behaviour across settlement types. For trust, the overall ANOVA reached statistical significance (F(3, 824) = 2.72, p = 0.044), although Tukey post hoc comparisons did not reveal statistically significant pairwise differences, with the homogeneous subsets suggesting only a difference between respondents from the capital and those from rural areas.

These additional analyses did not change the overall conclusions of the study. Since the research questions focus on predictive relationships rather than group mean differences, we retained regression analysis as the primary analytical strategy and decided not the describe these results in the text.

Comment:

On the dependent measure or index, more needs to be said which items were used, how many. Perhaps put asterisks on those items listed in the appendix that reference pro-environment choices.  It seems possible those items could have been factor analyzed to see if there are multiple dimensions present, rather than assume that those items in the index are measuring only one dimension (was it a weighted score=average of their sum divided by number of items)? The authors can also compute the internal consistency of the index (Cronbach's alpha). 

Thank you for this helpful suggestion. We have clarified the construction of the dependent variable and added additional information about the measurement properties of the index in the revised manuscript.

The pro-environmental behaviour index consists of 23 items capturing a broad range of everyday ecological practices (e.g., consumption choices, waste management, energy and water conservation, transportation, and participation in environmental initiatives). The index was calculated as the mean score across all items (sum of responses divided by the number of items). Item 14 was reverse-coded prior to computing the index.

We also assessed the internal consistency of the scale. The Cronbach’s alpha for the pro-environmental behaviour index is 0.833, indicating good reliability. For the trust scale (6 items), Cronbach’s alpha is 0.828.

Following the reviewer’s suggestion, we additionally explored the dimensional structure of the behavioural items using exploratory factor analysis. The analysis suggested the presence of several behavioural domains, which is consistent with the literature indicating that pro-environmental behaviour includes multiple types of actions. However, since the aim of the study is to examine overall pro-environmental behaviour, we retained the composite index approach commonly used in survey research.

All changes are highlighted in the text.

Comment:

For the regression table, perhaps the eta square for each predictor can be added as well (percent of variance accounted by each item). 

Thank you for this suggestion. The analysis in the paper is based on multiple regression models rather than analysis of variance (ANOVA). Because eta squared is typically used as an effect size measure in ANOVA designs to indicate the proportion of variance explained by group differences, it is not directly applicable to the regression framework used in this study.

Based on your comment, to provide additional information about the strength of the relationships between predictors and the dependent variable, we have expanded the regression table by including partial correlation coefficients for each predictor.

Comment:

The discussion can elaborate more on implications on alternative explanations and limitations.  The authors do a great job tying social dilemmas to individualism in regards to pro-ecological behavior. But there are more explanations can be use more discussion, such as a possible role of Diffusion of Responsibility for not endorsing those pro-ecological behaviors.

Thank you for this valuable suggestion. Following your recommendation, we expanded the Discussion section to incorporate additional theoretical explanations that may account for the observed patterns. In particular, we included a discussion of the diffusion of responsibility as a potential mechanism that may reduce individual engagement in pro-ecological behaviour, especially in large-scale collective environmental problems such as climate change. In addition to this, we also improved the Practical Implications section.

The following text has been added to the manuscript:

Several additional mechanisms may help explain the observed patterns. One potential explanation is the diffusion of responsibility, a well-established phenomenon in social psychology whereby individuals are less likely to take action when responsibility is perceived to be shared among many actors (Darley & Latané, 1968). Environmental problems such as climate change are frequently perceived as large-scale collective challenges, which may encourage individuals to assume that governments, institutions, or other societal actors should take primary responsibility for addressing them. In low-trust institutional environments, such diffusion of responsibility may further weaken individual motivation to engage in pro-ecological behaviour if citizens perceive institutional actors as ineffective or unreliable.

Comment:

Also, separating national identity from "community identity," which the authors mention. The former having a negative connotation, the latter a higher value for social proof based on immediacy of social comparison based on similar others...which facilitates norm creation as the authors state in the introduction. 

Thank you for this helpful comment. We agree that distinguishing between national identity and community identity strengthens the theoretical explanation of how social norms influence pro-ecological behaviour. Accordingly, we expanded the Discussion section to clarify this distinction and to emphasize the role of local social comparison and normative influence within immediate social networks.

The following text has been added to the manuscript:

The results also highlight the importance of distinguishing between national identity and community identity when considering the role of social norms in environmental behaviour. While national identity may sometimes be associated with politicized narratives or institutional distrust in post-socialist contexts, community identity is grounded in more immediate social relationships and localized networks. Research on social norms and pro-environmental behaviour suggests that individuals are particularly responsive to behavioural expectations within their immediate social environment, where comparison with similar others is most salient (Goldstein, Cialdini & Griskevicius, 2008). Such localized normative influence can facilitate the emergence of collective behavioural standards that reinforce environmentally responsible practices.

Comment:

Framing effects can also be involved, with ecological issues often framed in terms of losses, which increases pro-action. Perhaps some of these explanations can be ruled out with the existing data, but if now, those should be mentioned as viable explanations also. 

Thank you for your suggestion. We also added a discussion of framing effects in environmental communication, noting that the way environmental issues are presented may influence behavioural engagement and interact with levels of institutional trust.

As following in the text:

Another possible explanatory mechanism relates to framing effects in environmental communication. Research in behavioural science and climate communication has shown that the way environmental problems are framed can significantly influence public engagement and behavioural responses (Nisbet, 2009; Spence & Pidgeon, 2010; van der Linden et al., 2015). However, framing effects may interact with trust and perceived efficacy. When individuals perceive environmental risks as severe but have limited trust in the institutions responsible for addressing them, concern may not translate into behavioural engagement. Although the present study did not directly measure framing effects or perceived efficacy, these factors may shape how environmental information is interpreted and acted upon.

Comment:

Overall the manuscript has a good focus by focusing in Bulgaria, a society that due to its history has had low institutional trust. In case the authors have not seen the case of Bangladesh, where institutional trust is also very low, but yet community groups clean their waterways in what is really an impressive feat (https://www.youtube.com/watch?v=xCLobD-ZkE0) . Exactly what these authors are describing in their paper.

 

Thank you for this insightful suggestion. We found the example of Bangladesh very informative, inspiring and relevant to our argument regarding the role of community-based environmental action in low-trust contexts. Following your recommendation, we incorporated this example into the Discussion section and complemented it with additional examples of grassroots environmental initiatives.

In text:

In this respect, the findings are consistent with evidence from other low-trust contexts where community-based initiatives have successfully mobilized environmental action. For example, grassroots community groups in Bangladesh have organized collective efforts to clean and restore local waterways despite limited institutional support (BD Clean, n.d.). Similar community-driven environmental initiatives have been documented in other parts of the world (Global Citizen, n.d.) where strong local networks and shared community identities can compensate for weak institutional capacity and enable collective environmental stewardship (Pretty & Ward, 2001; Ostrom, 2000). These examples further support the argument that interpersonal trust and localized social norms can function as key mechanisms for promoting pro-ecological behaviour in societies characterized by low institutional trust.

Well wishes on revisions. 

Thank you.

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

Firstly, I must acknowledge that the influence of social trust on pro-environmental behavior is a consistent theme.

The authors have described very well the sample of 1,008 respondents and the method used to develop it. What is missing, however, is information about geographical and administrative areas other than the capital Sofia (which covers 16.4% of the sample). The authors should justify why only 26% of respondents were selected from rural areas. Personally, I believe that half of the sample could have been selected from rural areas and half from urban areas (there is no need for demographic representation here). This would have made for an interesting comparative analysis between urban and rural areas. In fact, this analysis could be done even under the given conditions.

Table 1 comes after the sample is explained in the text. I believe table 1 is unnecessary and could be omitted, especially since no comparative interpretations are made regarding gender, education, age, etc. I suggest that the authors introduce this type of interpretation and, based on them, introduce tables or graphs.

Regarding the measurement scales, I noticed that the authors detail the structure and even the items of the scales appropriately. However, I did not see any details about how they constructed these measurement scales, given that there is a bibliography in this field. For example, we have the pro-environmental behavior scale (Larson, Stedman, Cooper & Decker; or Abraham, Pane & Chairiyani).

Author Response

Answers:

Firstly, I must acknowledge that the influence of social trust on pro-environmental behavior is a consistent theme.

Comment:

The authors have described very well the sample of 1,008 respondents and the method used to develop it. What is missing, however, is information about geographical and administrative areas other than the capital Sofia (which covers 16.4% of the sample). The authors should justify why only 26% of respondents were selected from rural areas. Personally, I believe that half of the sample could have been selected from rural areas and half from urban areas (there is no need for demographic representation here). This would have made for an interesting comparative analysis between urban and rural areas. In fact, this analysis could be done even under the given conditions.

Thank you for this observation. Following your comment and the one of the other reviewer, we have revised and improved Table 1 (Sample Characteristics) to provide a clearer comparison between the sample and the population structure.

First, the updated figures differ from those reported previously. In the corrected table, respondents from Sofia account for 21.5% of the sample, while respondents from rural areas represent 23.7%.

Second, the comparison between the sample distribution and the population distribution by type of settlement indicates that the differences are not statistically significant. This means that the sample reproduces the structure of the adult population of Bulgaria reasonably well with respect to settlement type.

Our aim in constructing the sample was to achieve national demographic representativeness, rather than to create analytically balanced groups (e.g., equal urban and rural subsamples). Therefore, the proportions of respondents from urban and rural areas reflect the actual population structure rather than a deliberately balanced design.

Following your suggestion, we additionally conducted an exploratory one-way ANOVA with settlement type as the independent variable and both pro-environmental behaviour and trust as dependent variables. The analysis indicated no statistically significant differences in pro-environmental behaviour across settlement types. For trust, the overall ANOVA reached statistical significance (F(3, 824) = 2.72, p = 0.044). However, post hoc Tukey tests did not identify statistically significant pairwise differences, although the homogeneous subsets analysis suggested a difference primarily between respondents from the capital and those from rural areas.

Since these differences are small and not central to the research questions of the study, we have not incorporated a detailed analysis of settlement-type differences in the main results. Nevertheless, we appreciate the reviewer’s suggestion and explored the possibility empirically.

Comment:

Table 1 comes after the sample is explained in the text. I believe table 1 is unnecessary and could be omitted, especially since no comparative interpretations are made regarding gender, education, age, etc. I suggest that the authors introduce this type of interpretation and, based on them, introduce tables or graphs.

Thank you for this comment. In the revised manuscript, Table 1 has been expanded and improved to provide additional information about the structure of the sample. The table now includes not only the sample distribution but also population proportions, differences between the sample and the population, and statistical tests assessing these differences. For this reason, we believe that retaining Table 1 is important, as it allows readers to directly evaluate the representativeness of the sample.

The primary purpose of the table is therefore descriptive, providing transparency about the demographic composition of the survey sample and its correspondence to the population structure. The study does not aim to test hypotheses about differences between demographic groups (e.g., gender, age, or education), but rather to examine the relationships between trust and pro-environmental behaviour at the population level. For this reason, the analysis focuses on regression models, which are more appropriate for addressing the research questions.

Comment:

Regarding the measurement scales, I noticed that the authors detail the structure and even the items of the scales appropriately. However, I did not see any details about how they constructed these measurement scales, given that there is a bibliography in this field. For example, we have the pro-environmental behavior scale (Larson, Stedman, Cooper & Decker; or Abraham, Pane & Chairiyani).

Thank you for this helpful comment. We agree that it is important to clarify how the measurement scale for pro-environmental behaviour was constructed. The scale used in the study was not adopted from a single existing instrument but was developed by combining and adapting items from established classifications of pro-environmental behaviour. In particular, the selection of behavioural domains was guided by the framework proposed by Kaiser et al. (2003), which identifies key categories such as energy conservation, mobility and transportation, waste reduction, consumer behaviour, recycling, and social environmental actions. We also drew on the synthesis provided by Kurisu (2015), which summarizes common domains of pro-environmental behaviour including household energy efficiency, waste and recycling, water consumption, transportation, sustainable consumption, and environmental activism.

Based on these conceptual frameworks, we selected and adapted items reflecting everyday ecological behaviours relevant to the study context. The revised manuscript now clarifies the theoretical basis and development of the scale in the methodology section as follows:

The scale was constructed by adapting and combining items from established classifications of pro-environmental behaviour. In particular, the selection of behavioural domains was informed by the typology proposed by Kaiser and colleagues (2003), who identify several categories of ecological behaviour including energy conservation, mobility and transportation, waste reduction, consumer behaviour, recycling, and social environmental actions. The conceptual structure of the scale was also guided by the synthesis of pro-environmental behavioural domains presented by Kurisu (2015), which highlights key behavioural areas such as household energy efficiency, waste and recycling, water use, sustainable consumption, transportation, and environmental activism.

Based on these frameworks, items were selected, adapted, and contextualized to reflect common everyday behaviours relevant to the Bulgarian context. The final instrument included behaviours of varying environmental impact, ranging from routine household practices (e.g., saving energy, reducing waste) to more involved civic actions (e.g., participating in environmental campaigns or initiatives). All 23 statements used in the analysis are presented in Table 1 (Appendix).

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I have found the revisions and explanations for introduced changes to the manuscript to be appropriate and sound. Therefore, I recommend publications as revised.