A Comprehensive Model to Explain Europeans’ Environmental Behaviors

Reviewer 1 comments

Answers

I consider the article interesting and methodically well prepared. The article contains hypotheses. Define clearly the research assumption, question, hypothesis. The hypothesis is an exakt assumption. The proposed hypotheses are rather assumptions. Exactly improve hypotheses, designate Ho and H1. In the results clearly respond to the hypotheses. In the discussion the authors do not discuss sufficiently. I recommend comparing the findings with the current state of knowledge and theoretical bases.

Thanks to the reviewer for the positive appraisal.

In the methodological section, we have added a section related to the analysis where we explain how to interpret the null hypotheses and alternative hypotheses related to the effects that we expect. We have further developed the discussion sections and even have added the implications for policymakers

Reviewer 2 comments

Answers

I would like to stress out that I support the potential publication of this paper due to its scientific interest. On the other hand, many aspects of the manuscript need to be extensively improved; otherwise, I will not be able to support the final publication. As such, I strongly advise the authors to take into serious consideration all of the following major and minor remarks to improve the quality of the presentation of their work. The quality of the manuscript must be strengthened, or else it would be extremely difficult to be finally accepted.

Thanks to the reviewer for the positive appraisal.

We have modified the manuscript taking into account your suggestions. See below.

1.    Title

After carefully reading your manuscript, I believe that the current title is inappropriate. “Responsible environmental behaviour theory” implies a single theory tested, not an abstract conceptual model that combines the existing theoretical frameworks. I recommend a more relevant title, for example, “A comprehensive model to explain Europeans’ environmental behaviours.”

We have changed the title, thank you for the suggestion.

2.    Abstract

The manuscript’s abstract needs to be thoroughly revised. First, the authors are advised to revise English carefully. Second, it is less confusing to say “Theoretical framework of responsible environmental behaviour” instead of “Responsible environmental behaviour theory.” Also, the authors should clearly state the purpose of the study, the methods used, and the main findings. Eventually, what is your main goal? Please, be more specific (e.g., explore the predictive power of cognitive, personality, and social factors concerning Europeans’ environmental behaviours at the individual or country level).

At the end of the Abstract, you should add 2-3 lines to explain the potential contribution of this work and possible policy implications.

We have revised the abstract following your suggestions and have had  the paper reviewed by a native English-speaking professional editor.

3.    Introduction and theoretical framework

a) The “Introduction” section of the manuscript requires extensive revision. The aim of the study needs to be appropriately highlighted and justified. I would suggest that the authors attempt to present the key objectives of their study with regards to what is currently known (i.e., literature), thus highlighting the added value of the paper. At the end of the Introduction, I highly recommend to clearly state the main objective of this study, secondary objectives- if any-, and the contribution of this research.

b) It would be nice if the authors added some more information in the introduction. For example, focus on the necessity of fostering environmental behavior due to climate change or maybe justify why environmental behavior is a prerequisite to a transition to a low carbon future (possibly refer to the ICCP reports).

c) The authors need to expand the review of literature that is relevant to their study: Section 2. Theoretical framework must be extended; the authors omit an essential piece of environmental behavior literature (i.e., take a look at the following studies, Gkargkavouzi, Halkos & Matsiori, 2019a DOI https://doi.org/10.1016/j.resconrec.2019.01.039, or Li et al., 2019 DOI https://doi.org/10.1016/j.resconrec.2019.03.024 ).

d) Line 66: Please add another definition of “environmental behavior,” more inclusive; for example, see Alisat & Riemer (2015) DOI https://doi.org/10.1016/j.jenvp.2015.05.006 .

e) Lines 103-104: Revise this definition; it is not clear. For example, see Farrow et al. (2017) DOI: 10.1016/j.ecolecon.2017.04.017.

f) Lines 205-106: Personal norms, as a psychological construct, are distinguished from “awareness of consequences” and “ascription of responsibilities.” Personal norms refer to a sense of moral obligation to act appropriately. Please revise it accordingly.

g) Additionally, VBN theory (Stern et al., 1999) is missing. VNB theory is an extended version of NAM but includes an important psychological construct, “personal norms” (you didn’t even include it in the proposed model!). Other studies that present extended models that combine VBN or NAM and TPB are also missing. Therefore, I do not agree with the proposed conceptual model. It relies on the Hines et al. (1987) model, a quite old one. The theories you reviewed in the introduction suggest more complex relationships compared to those of the proposed model.

h) The structure of the paper is confusing. Cognitive, personality, situational, and social factors should be analyzed and presented right after section 2.2. Maybe the authors could rename 2.3 section to “Additional factors influencing environmental behavior” and then add the following subsections: 2.3.1. Cognitive factors 2.3.2. Personality factors 2.3.3. Social factors and 2.3.4. Situational factors. Research Hypotheses can be placed at the end of each section, respectively. Thus, delete the current section 2.3. Research model and hypothesis, and add “2.3.5. The conceptual model” at the end of section 2. Theoretical framework and hypotheses (Figure 1 fits to section 2.3.5.). I also suggest to revise Figure 1 and link each arrow with the corresponding Hypotheses.

i) However, I must disagree with the authors that “attitudes, perceived behavioural control and subjective environmental norms” are personality factors! Environmental attitudes are a psychological tendency to evaluate the natural and built environments, and factors affecting their quality, with some degree of favor or disfavor (Milfont, 2007, 2012). I strongly suggest changing the word “personality” throughout the text.

j) Line 132: Please pay attention, locus of control, and perceived behavioral control are not the same psychological construct.

k) The authors should consider redefining their REB model and maybe delete “Environmental attitudes.” The Environmental Psychology literature suggests that general environmental attitudes are only moderate or weak predictors of behavior (i.e., see Farjam et al., 2019 DOI https://doi.org/10.1016/j.ecolecon.2019.106434 , Claudy et al., 2013 DOI https://doi.org/10.1177%2F0276146713481605, Eilam & Trop, 2012 DOI 10.3390/su4092210). At least, consider adding some lines to refer to the attitude-behavior gap in the literature review. Similarly, the literature suggests that environmental knowledge is a somewhat questionable direct predictor of behavior (for example, see Gkargkavouzi, Halkos & Matsiori, 2019b  DOI https://doi.org/10.1016/j.ecolecon.2019.106394 ).

l) I recommend keeping the layout of the paper as simple as possible. For example, Research Hypotheses should be placed ONLY at the end of the literature review.

m) In general, use either “behavior” or “behaviour,” use either American or British English throughout the paper.

a)      We have revised the introduction following your indications.

b)      We have added some additional information regarding the  need to foster environmentally friendly behaviors due to climate change and have justified why changing these behaviors is a prerequisite for  a transition to a low-carbon future.

c)      We have expanded the theoretical framework within the space limits and linked it to the recent research you mention, thanks for your suggestions.

d)      We have added another  more inclusive definition of “environmental behavior,” as in Alisat & Riemer (2015).

e)      We have revised the definitions taking into account the reference you suggested. Thanks.

f)       Personal norms, as a psychological construct, are distinguished from “awareness of consequences” and “ascription of responsibilities.” Personal norms refer to a sense of moral obligation to act appropriately. We have revised the text accordingly.

g)      We have included the VBN theory (Stern et al., 1999), but we did not expand on it as values, beliefs and norms are an antecedent to attitudes, and a SEM model was not the aim of this study.  

h)      We have revised the structure of the Theoretical framework following your indications, and hypotheses are included in the graphical summary. Thanks.

i)       We  agree with you and have revised this part of the text.

j)       We agree and have rewritten it.

k)      There is no general agreement on that. The REB model, as Kurisu describes it, takes into account not only environmental attitudes but also attitudes towards environmental behaviors, and it is proposed that the latter could have a greater impact on behaviors. As our findings suggest, the effect of these factors may  differ according to the behavior considered.  We have added some lines to refer to the attitude-behavior gap in the literature review.  We have also have expanded  the literature review  regarding environmental knowledge.

l)       We have simplified the structure of the paper and put all hypotheses at the end of the literature reviewin the graphical model.

m)    We have revised the paper and now  use US English consistently.

a) Based on the previous comments, I suggest deleting section 3.1. Research model and hypotheses, it does not add any value to the paper (the proposed model has been already described). Lines 213-216 are probably the secondary objectives of the study, so to the Introduction section.

It would be interesting to add some more information in section 3.2. Sample and procedures.

b) Lines 226-227: Better describe data analysis methods in a separate section, after measures.

c) Lines 284-290: Factor Analysis and Principal Component Analysis are NOT the same. Please revise these lines and be more precise about the methodology used. EFA aims to explore the factorial structure of a set of items.

d) Lines 294-295: Univariate normality does not ensure multivariate normality; did you test for it?

e) Lines 306-307: Why did you apply Varimax rotation? Are you sure that the underlying factors are not related?

f) Why didn’t you test for common method bias before applying PCA?

a)     We have deleted section 3.1 and have provided more information in section 3.2

b)    We have added a final subsection to the methods section to explain the modeling strategy

c)     We explain that PCA is not (E)FA, but that EFA can use PCA when factors are considered as formative, forming a scale.

d)    We tested the MVN of the dataset, but did not pass it. However, this is not a problem when using PCA as the method for extracting the factor scores.

e)     We conducted varimax and oblimin rotations to explore the correlation among factors. The correlations when using oblimin  were below .32 so varimax rotation was considered acceptable (see Tabachnik and Fidell 2007: 646). We have added this information to the manuscript

f)     We tested for evidence of common method variance (Tehseen, Ramayah & Sajilan 2017), since behaviors as well as environmental factors indicators were provided by the same source. We conducted Harman’s single-factor test, an examination of cross-loadings and correlation among behaviors and environmental factors. The first and last tests did not provide any evidence of common method variance. The examination of cross-loadings for  public transport use produced the expected results, and, for eco-friendly purchasing and reduced resource consumption, the behavioral indicators never loaded on the environmental factors.  Results are now reported in the appendix.

Is exploratory factor analysis (EFA) suitable to obtain information on the discriminant validity of those groups? Convergent validity states that tests having the same or similar constructs should be highly correlated. Two methods are often applied to test convergent validity. One is to correlate the scores between two assessment tools or tools’ sub-domains that are considered to measure the same construct. The other method is the multitrait-multimethod matrix (MTMM) approach. It includes the correlations between multiple constructs and multiple measuring methods and is rich in information regarding reliability, convergent validity, and discriminant validity. MTMM analyzes the same construct with different measuring tools. The discovery of moderate to high correlation can support the existing of convergent validity (Campbell & Fiske, 1959). Moreover, discriminant validity can be adequately tested in the context of other- more complex- methods, like Structural Equation Modeling.

The authors provide all the necessary information concerning the multilevel regression analysis and how this analysis was performed (you only forgot to report which centering method was used).

We decided to use EFA with PCA as the method of extraction because we were interested in summarizing and reducing the data, that is, identifying the main environmental factors and reducing the original indicators so as to build a scale of the environmental factors we were interested in.  Additionally, our data did not pass the MNV test, so the PCA seemed to be the most robust method for extracting the factors (Tabachnik  and Fidell 2007; Hair, Black, Babin & Anderson 2014).

We are aware that, in using PCA as the EFA method, we are treating the factors as scales formed by the indicators, not the indicators as the observed manifestations of the theoretical construct.  We have indicated that in the text.

The MTMM approach is suitable for confirmatory factor analysis but not exploratory factor analysis with PCA; that is why we did not conduct that test

We have reported in the manuscript that numerical variables  have been scaled to make it easier to compare their effects.

5.    Results

This section is by far the most exciting and well-organized. The results are presents clearly and in great detail. This section might be slightly revised if the authors adopt previous comments (methods section).

Table 7: Please consider another title, maybe “Hypotheses testing results” (it is “Supported,” not “accepted” hypothesis).

Thanks to the reviewer for this positive appraisal.   

We have slightly revised this section because we now treat the effect of age as non-linear.

We agree with your suggestion and have changed the title of Table 7.

6.    Discussion

a) Lines 465-468: I disagree that your findings are in line with most of the existing literature- previous studies. In fact, this is not the case. I already highlighted in previous comments (Introduction) that knowledge is an intellectual prerequisite for behavior, but not as a significant predictor as norms or perceived behavioral control (i.e., Morren & Grinstein, 2016). Why is this? You only report two studies to support your findings! I think these inconsistencies with existing research should be highlighted as potential limitations of the study.

b) Lines 479-481: This is surprising again, please explain in more detail why you think this happened.

c) Are your findings useful? What is the main contribution of this work? Can policymakers use your results, and how? Please revise your discussion accordingly to answer these questions, and maybe make some recommendations for future research.

a)     The effect of environmental factors does not seem to be universal. Environmental knowledge is a great predictor of public transport use, but not of reduced resource consumption or of eco-friendly purchases. This means that researchers have to keep in mind that the effect may not be universal. We have rewritten this in the discussion

b)     Perceived behavioral control has its main effect on reducing the consumption of resources, meaning that everybody feels able to reduce their consumption of resources; however, this has to do with the fact that the indicators loading on this factor are related to cutting back on water, energy consumption and replacing old appliances.  But the effect of perceived behavioral control  on buying eco-friendly goods interacts with the income proxy, meaning that only people that do not have problems getting to the end of the month can afford this behavior.

c)     We have addressed the issue of how policymakers can use the results of this research.

Reviewer 3 comments

Answers

 The Authors should embed their work more in the current literature. What is new about the study? In lines 42-47, the Authors claimed the aim of the paper. However, I find it really strange to declare the objectives of the paper, which are based on REB theory, before that the same theory is introduced. Moreover, the Authors should discuss what is really new about the paper beyond the study area? By embedding more in the current literature, the Authors should argue what their new angle adds to the literature and why it is important

We have revised the abstract and the introduction section.

Several scholars pointed out that values directly and/or indirectly affect behaviour. Moreover, values embodied by institutions are usually the main determinants of individual values and behaviours (Terpstra and David, 1987; Agirdag et al., 2016). Although Eurobarometer did not detect personal values, it assessed ‘perceived values’ (e.g., peace, equality, freedom, tolerance, respect for environment and history) that are expected to relate to environmental behaviour (see Steg and de Groot, 2012; Gifford and Nilsson, 2014; Rajapaksa et al., 2018). In fact, European identity can be enmeshed with the different national identities, either reinforcing or lessening supranational values, thus leading to diverse national outcomes. Therefore, perceived values can be considered as similar to personal values in that they are internal predispositions that could presumably relate to attitudes and actions (Punzo et al., 2019). Taking this approach, the Authors should consider the possibility to broadening their approach by also modelling ‘perceived values’, with an effort to exploring their relationship with the environmental behaviour.

In this research, we are not interested in a sequential causality model, from values to behavior, but on the environmental factors that influence behavior. We measure the indirect effect of values through the effect of attitudes on behavior. Even the possible influence of environmental subjective norms on attitudes is not that strong according to the variance inflation factor, which is below 5. The SEM model has its strengths but also its limitations. It is great for measuring the effect of several factors in a sequential causal model of direct and indirect effects, but is not so good for taking into account individuals clustered into countries. We further explain this when explaining the decision about the modeling strategy. 

The same methodological choice was not justified enough. I think REB theory as a complex framework that makes it hard to be solved through the multilevel regression. Please, discuss your methodological choice more in detail with its strengths and limitations.

Alternatively, I suggest the Authors to reconsider their methodological choice by performing a more appropriate and advanced modelling. In this context, for example, structural equation modelling (SEM) – which allows testing complex hypotheses – should enable the Authors to explore the relationship among the different constructs in which the theories are expressed.

The SEM strategy is not the proper modeling strategy for the purpose of this research: discriminating between individual- level factors that influence the three groups of behaviors and the fact that respondents live in different countries.

We will consider your suggestions for another study that will use a modeling strategy that is a mixture of a causal model (PLS-SEM) and an exploratory strategy to account for heterogeneity in the effect of causal factors on a general measure of environmental behavior. However, the aim of that manuscript is different to the aim of this one. 

Your suggestion to measure the effect of a sequential causality model is interesting. However, in this research, we were interested in the last two levels, behavior and intention (measured by attitudes and knowledge) and perceived behavioral control, but not in the values that may affect attitudes.

As the Authors indirectly claimed, the results are likely not generalizable worldwide, being valid in the context of each country because different country structures may lead to different behaviours. In lines 196-198, the Authors claim that “among other factors, country differences in individuals’ REBs can be related to environmental policies, economic development and social and cultural specificities. What does this sentence mean in practical terms? It is meaningless if nothing is added in terms of policy implications (please, see the last comment).

 This (lines 196-198 of the first draft) means that to test the effect of environmental factors on behaviors, we have to take into account a multilevel modeling strategy that accounts for country differences in behaviors, once we have taken into account the environmental factors and social properties of individuals.  

However, identifying the country-specific socioeconomic and environmental factors that explain differences in the average behavior of Europeans is not the aim of this research. Here we want to measure the effect of environmental factors and social properties of Europeans on three particular behaviors.  Explaining what causes country-specific differences in general environmental behavior is another study that we have just started recently. In this paper we wanted to separate the contribution of the environmental factors that contribute to responsible environmental behavior, taking into account that individuals live in different contexts.

•                  Moreover, the study was performed with reference to all European countries, most of them are developed but some others are less developed countries. The Authors noted that environmental knowledge and personality constructs may differ between the 28 EU Member States due to situational factors (lines 199-205) and they also claimed that “comparative studies have explored differences in terms of developing/developed” countries (lines 194-195). However, the Authors remained unclear whether conclusions of their study particularly differ between these two subsets of countries (developed vs. developing countries). A wide literature discussed the more environmentally-friendly practices of citizens from developed countries than those living in developing countries (see, for example, Punzo et al., 2019). Empirical results would be interpreted in light of these issues, in order to make this matter clearer to the readers.

In this paper, we are interested in taking into account nation-states differences in the behavior of individuals. Instead of studying differences between two sets of countries, we wanted to control for all differences. That’s why we chose a multilevel framework.  The SEM modeling strategy that Punzo et al. (2019) use makes it almost impossible to contrast differences among the 28 countries in a way that can be clearly communicated. That is why they probably chose to study just 4 developed countries.

The focus on comparing environmentally friendly practices of citizens from different types of countries (developed countries versus less developed countries) is interesting but first you need an objective criterion to classify countries. In another paper that we are currently writing, we follow a more exploratory strategy regarding the properties of countries, and group countries together according to the behavior of their citizens and country-specific properties. But that is another research strategy that will provide complementary information.    In the methods sections, we have added a subsection to explain the choices made regarding the modeling strategy.

•                  In my opinion, the selection of variables sometimes lacks rationality. I suggest the Authors to provide more details to explain why these variables were chosen. For example, some studies demonstrated the role of political identity (this information is available in the Eurobarometer dataset) in affecting environmental behaviour. Moreover, in order to overcome the lacking of information about income in the Eurobarometer dataset, the Authors selected the variable concerning the difficulties in paying bills as a proxy of income, using it as a control variable. Really, I am not convinced that it is a good idea. Please, discuss this choice with references to the literature supporting it. I believe this proxy can capture only partially the overlapping between monetary and supplementary poverty, which two distinct concepts remain. I would invite the Authors to revise their choice.

We were interested in how the social (not political) position of individuals and their social categories moderate the effect of the environmental factors on behavior. The Special Eurobarometer survey does not ask directly about the respondents’ income because many individuals refuse to answer. Instead, they use a  proxy.  Question D60 asks respondents whether they have difficulties to pay their bills or not and is measured with three levels: “Most of the time”, “From time to time”, and “Almost Never/Never”. Even though this measure is not perfect, at least it captures whether an individual’s income influences their behavior. It is not so much a control but a moderator. We mention this in the paper.

•                  Then, how were the moderating and control variables treated before being included into the model? And the age variable? Why was not its quadratic version considered?

Control variables were standardized to make it easy to compare the magnitude of the effects.

The age indicator has now been dichotomized to make it easy to show that its moderating effect is not constant along the life span of individuals.

•                  The Authors called “Environmental attitudes” the construct composed of items that most literature classified as concerns in the framework of Schwartz’s models. In particular, I refer to the items from QD5.4 to QD5.7. How do the Authors intend to justify this choice? What is the difference with the next construct “Attitudes toward environmental behaviours”?

The literature on responsible pro-environmental behavior makes the distinction between environmental attitudes and attitudes towards the environmental behaviors.

The first captures a general environmental attitude that represents an individual's concern about environmental problems (Davies et al., 2002). We propose to measure environmental attitude as Europeans worries about environmental problems. According to Hayward (1990) worries about environmental problems are related to environmental attitude.

The second construct is Kurisu’s (2015) ATEB (attitudes toward environmental behavior), who considered that ATEB may be more influential on behavior than a general environmental attitude.

This information was in a previous draft but was omitteddue to space limits.

•                  Why did the Author choose not to control for other variables than those used? I guess the Authors selected the socio-demographic variables as moderating or control variables according to the prominent literature in the field. Did they run simulations in which they took a wider set of control variables into account? And then, did they decide to exclude from the model those variables that never demonstrated significance in the data? In brief, more explanations of the reasons for variables’ selection are expected.

Yes, we selected the social indicators as moderating variables, according to the literature review. We use other indicators as control variables (household composition and community type). Summing the moderating plus the control variables we have social indicators. We left out occupation because it is usually correlated with income and education.  We have further explained the selection of these moderating  and control variables.

•                  What are the main hypotheses of the research? I find the insertion of a large number of hypotheses extremely dispersive and misleading. I strongly suggest the Authors to focus in the “Theoretical framework” on the main hypotheses of the work (H1-H5). The remaining hypotheses are strictly linked to control variables (H6-H9) and could be directly discussed in the results.

We want to differentiate between control variables and moderating variables. We agree that the main effect of social indicators could be interpreted as control variables, but their role as moderating variables of the main effect of theoretical constructs must be hypothesized according to the literature review. For that reason we will maintain hypotheses H6-H9 that states the interaction, not the main effects, but we present them all at the section ends and integrate them in the graphical model at the end second section.

•                  I find the excessive fragmentation of the sections very dispersive, each of them even of a few lines (see, for example, sub-section 3.1-3.3, and sub-section 3.3.1).

We agree with your and reviewer 3 regarding this problem. We have reordered these sections.

•                  Empirical results were merely discussed from the statistical perspective. The Authors made no efforts to discuss the role of the national Governments and/or European institutions. The empirical results should be interpreted in light of the ongoing institutional settings. In other words, the results imply policies and measures of which no suggestions were given in the paper. I believe the Authors should extend Section 5 with the discussion of the most important statistical results in light of policy implications and not only comparing them with those of previous studies. Which strategies for Governments? Do best practices already exist? Discussing best practices in some specific environmental issues, as well as further practical policy proposals aimed at enhancing certain environmental behaviours, should allow the paper to gain in originality.

At the end of the discussion section, we discuss how the results can be used to improve the policies of national governments and European institutions. The fact that the findings discriminate between the individual nation-state levels can be used to think about European as well as national levels of policymaking.

Thanks for directing us to consider the practical value of our findings.