Governance, Nature’s Contributions to People, and Investing in Conservation Influence the Valuation of Urban Green Areas
Round 1
Reviewer 1 Report
The manuscript titled “Perceived governance and investing in conservation influences the valuation of urban green areas” aims to fill critical research gaps regarding (i) how people perceive urban benefits and costs of urban green areas (UGA) in Colombia, (ii) how these perceptions are influenced by socioeconomic factors, and (iii) how the way people perceive the governance of UGA affects their willingness to pay for their conservation. To do so, the authors have surveyed a total of 500 respondents in 10 different sites, representing different classes of age, gender, education level and socioeconomic stratum. These surveys were then analyzed through quantitative methods in order to explore, amongst other, how people’s perceptions of UGA-related benefits are influenced by their environmental awareness, sense of security, education level and socioeconomic attributes (Eq. 1; L.214), and how people’s unwillingness to pay for conservation was influenced by their perception of local governance and socioeconomic attributes (Eq. 2; L.232).
While the research question addressed in this ms is important and needs to be tackled for more effective, inclusive urban planning and sustainable management, I think this ms suffers from important methodological flaws and would therefore not recommend it for publication. As I further explain below, a first major issue concerns the dependency of response and explanatory variables in Eq.1 and Eq.2, a second issue is related to the origin and calculation of the Weak_governance variable used in Eq.2, and a third issue is linked to the (over)reductionist approach used on certain variables prior to quantitative analyses. I would conclude by highlighting the mismatches between the objectives of the paper, the survey design, and the statistical methods used to analyze the collected data.
Variable dependency in Eq.1 and Eq.2:
The statistical method used in this ms – the ordinary least squares (OLS) method – has several assumptions, including homoscedasticity, normality, and independency between the response variable and the explanatory variables. The present ms does not give any guarantee that the first two assumptions are met, but beyond this lack of information, there is an auto-correlation issue in Eq.1 and Eq.2.
The equation Eq.1 aims to model the response variable “Benefits” (calculated from the number of benefits identified by people) according to four explanatory variables: environmental awareness (EA), sense of security (SS), education level (AE), and socioeconomic variables (X). Yet, the EA variable is a variable calculated, among other, from the answer respondents gave to the question: “Is this place offering any benefits?” (Appendix A). Thus, EA is not, per nature, independent from the response variable “Benefits”. Indeed, if people answered “no” to the first question, then they might logically cite no benefit in the following question, which makes the variables “Benefits” and EA auto-correlated.
In the equation Eq.2, the Boolean variable “Weak_governance” was attributed the value “1” if respondents indicated they there were not willing to pay for UGA conservation because (i) they perceive corruption, (ii) they consider their taxes already allow to pay for conservation, (iii) they consider they pay too much taxes, or (iv) they consider that paying for UGA conservation is the role of the government and not theirs. In other cases, “Weak_governance” takes the value “0”. Yet, once again, the answer used for calculating this variable is not independent from the answer used to assess people’s unwillingness to pay for UGA conservation (the response variable in Eq.2). Indeed, the two questions are not formulated independently: if people answered that they were willing to pay for UGA conservation in the first question, then the options that remain available to them in the second question are reduced.
Thus, these two points suggest a mismatch between the survey design and the analytical methods used. The successive questions asked to respondents are not independent from each other, whereas the variables derived from these questions must be independent to not violate OLS method assumptions. If the questions asked to respondents had been something like “how much are you willing to pay for UGA conservation?” and “do you think there is a corruption issue regarding the use of public money?” or “do you trust the government on the good distribution of public money?”, variable dependency would have been respected.
Beyond this variable dependency issue, the ms does not provide any rationale about why Eq.1 uses certain variables derived from the surveys but not other variables. For example, why the question “Is this place safe” was used in Eq.1 (for the variable SS), whereas the question “What problems does this place present” was not used? This is very appealing, in particular considering that the authors wrote that “respondent’s perceptions regarding costs from urban ecosystems, do affect their identification of benefits” (L.297). So, why people’s perceptions of costs were not used as an explanatory variable of their perceptions of benefits?
Conception of the Weak_governance variable in Eq.2:
As the authors explain in introduction, governance is a multidimensional concept that involves legitimacy, transparency, accountability, inclusiveness, fairness, resilience, connectivity and performance issues (to name the principles identified in Lockwood et al. 2010 that is cited in the ms). Thus, asking only one question to assess how people perceive governance seems very reductive, especially for a study that aims to analyze the role of perceived governance on the valuation of UGA. We could have expected several questions that would have aimed to assess how people perceive the different governance principles identified in Lockwood et al. 2010, in particular because the authors say they have designed their Eq.2 based on the Lockwood’s paper (L.234). I am therefore not convinced by the relevance of this Weak_governance variable, which is a Boolean variable, to assess perceived governance. This oversimplification of people’s perception of a given situation leads me to my third point.
The (over)reductionist approach:
With the surveys, a relatively rich qualitative data has been collected, but then a drastic “degradation” or “simplification” is operated prior to the quantitative analyses. While degrading qualitative data to be able to perform quantitative analyses is sometimes necessary, such operations have important epistemic implications and so should be properly justified and balanced, and their potential influences on the results properly addressed, which is not the case in the present ms. For example, the authors had five categories for assessing people’s willingness to pay, but they transform this categorical variable into a Boolean one (loosing the information on how much people were willing to pay for only keeping a Yes/No information). But can we really consider that someone who would agree to pay up to 4000 COP has the same willingness than someone who agrees to pay more than 30000 COP? What are the implications on the results? Why such a simplification of the data necessary? Unfortunately, the ms does not provide any answer to these questions. Such unexplained (and potentially problematic) variable degradation also concerns (i) the number of wetlands named by people, (ii) the number of trees, palms or plants identified by people (which are both used to calculate the environmental awareness variable used in Eq.1), and (iii) the weak governance variable.
To conclude, I think the data that supports this ms could be interesting and could result in valuable results on people’s perceptions of the benefits and costs associated to UGAs. Yet, the analyses conducted in this present ms do not seem to be suited to the type of data collected because the questions were not formulated in an independent manner, but also because OLS method have been designed for analyzing continuous quantitative data and not discrete pseudo-quantitative data (derived from qualitative data). I further consider that, with only one question referring to perceived governance, the data collected does not allow a robust and serious assessment of how people perceive governance. My main recommendation to the authors would therefore be to refocus their ms toward an objective that would better fit with the data they collected, and ideally to use analytical methods that would not only allow to account for the number of benefits (and costs) cited by people, but also for the nature of cited benefits and costs (see for example free association techniques, consensus analyses, or multi-factorial analyses for qualitative data).
Author Response
While the research question addressed in this ms is important and needs to be tackled for more effective, inclusive urban planning and sustainable management, I think this ms suffers from important methodological flaws and would therefore not recommend it for publication. As I further explain below, a first major issue concerns the dependency of response and explanatory variables in Eq.1 and Eq.2, a second issue is related to the origin and calculation of the Weak_governance variable used in Eq.2, and a third issue is linked to the (over)reductionist approach used on certain variables prior to quantitative analyses. I would conclude by highlighting the mismatches between the objectives of the paper, the survey design, and the statistical methods used to analyze the collected data.
Issue 1: Variable dependency in Eq.1 and Eq.2:
The statistical method used in this ms – the ordinary least squares (OLS) method – has several assumptions, including homoscedasticity, normality, and independency between the response variable and the explanatory variables. The present ms does not give any guarantee that the first two assumptions are met, but beyond this lack of information, there is an auto-correlation issue in Eq.1 and Eq.2.
RESPONSE: Yes, you are correct in that OLS regressions do require several assumptions. However, those assumptions can be met in several ways. In this manuscript, we’re interested in obtaining unbiased estimators. Thus, we could use the five Gauss-Markov assumptions in order to get those unbiased estimators. That said, according to Gauss-Markov assumptions, the normality assumption plays no role in showing whether the OLS estimators are indeed the best linear unbiased estimators. In fact, an OLS does not need for an error term to be normally distributed to yield unbiased estimates. So, given the basic statistical nature of these aspects of OLS, we feel that delving into this detail and including additional text explaining normality is unnecessary as it would interrupt the flow of our manuscript. However to address you comment, we now add revised text in Line 210-211 that states, “Then an ordinary least squares (OLS) regression with robust standard errors….”
In terms of meeting the assumption of homoscedasticity and heteroskedacity, you are also correct in that we did not specifically provide any stated guarantee in our Methods that these two assumptions were met. We note, that stating whether you met or not these assumptions for OLS analyses is rarely done in manuscripts in journals with this level of impact factor. However, we explicitly state in Table 5 of our manuscript and int he revised methods that we are using OLS with robust standard errors; an approach that is used in the social sciences exactly because variation structure is typically not known (which is generally not the case in the biophysical sciences). As such, OLS with robust standard errors generally accounts for these assumptions. Also, we did not explicitly state that our goal was to address the presence of heteroskedasticity, therefore again, robust standard errors are generally appropriate even under homoskedasticity. So, again we do not feel it necessary to specify the technical aspects of these assumptions in additional text. In terms of auto-correlation and independence, please see the following revisions and responses. We hope this addresses your concern.
The equation Eq.1 aims to model the response variable “Benefits” (calculated from the number of benefits identified by people) according to four explanatory variables: environmental awareness (EA), sense of security (SS), education level (AE), and socioeconomic variables (X). Yet, the EA variable is a variable calculated, among other, from the answer respondents gave to the question: “Is this place offering any benefits?” (Appendix A). Thus, EA is not, per nature, independent from the response variable “Benefits”. Indeed, if people answered “no” to the first question, then they might logically cite no benefit in the following question, which makes the variables “Benefits” and EA auto-correlated.
RESPONSE: Yes, you are correct, these were the 4 factors used for our benefits model. And yes, if the respondents answered no to this question, then no benefits were listed. But we are not sure we understand the logic behind this comment, specifically in regard to independence and particularly auto-correlation. Again, this question was used to simply filter out respondents (n=15) who did not recognize the beneficial ecosystem functions of the site.
However, your confusion likely resulted from an oversight and translation error on our part in Appendix A and the methods- as written. Primarily the first question and variable (Row 1) in Appendix A was incorrectly described as “Environmental Awareness” when it should have been “Awareness of ecosystem functions”. We note that for this variable (Column 4) we clearly specify that we did not use this question/variable to calculate the “Environmental Awareness” variable in Equation 1. Rather we used 3 other questions/variables (Clearly designated in Column 4 and in Footnote ‘a’ in Table 6) to calculate the Environmental Awareness Variable in question:
- Can you identify 3 trees, palms or plants in this place?
- Are you aware of Reserva Forestal Protectora Bosque Oriental de Bogotá?
- Can you name wetlands that are located in Bogota?
Again in the 4th column describing these variables (“How data were analyzed”) in Appendix A and in the footnote in Table 6, we specify that these were the variables used for the Environmental Awareness variable.
However, to address your concern and avoid future confusion, In Appendix A for the question “Is this place offering any benefits”, we have revised the description of this variable from “Environmental Awareness” to “Awareness of ecosystem functions” to avoid further confusion. Further in the footnote in Table 6 we now clearly list the 3 questions used for the Environmental Awareness variable. The revised text for the footnote in Table 6 now reads, “The Environmental awareness variable consisted of 3 questions: Can you identify 3 trees, palms or plants in this place? Are you aware of the Reserva Forestal Protectora Bosque Oriental de Bogotá (the adjacent large extensive forest reserve to the east of the city)?, and Can you name wetlands that are located in Bogota?” We hope this addresses your concern.
In the equation Eq.2, the Boolean variable “Weak_governance” was attributed the value “1” if respondents indicated they there were not willing to pay for UGA conservation because (i) they perceive corruption, (ii) they consider their taxes already allow to pay for conservation, (iii) they consider they pay too much taxes, or (iv) they consider that paying for UGA conservation is the role of the government and not theirs. In other cases, “Weak_governance” takes the value “0”. Yet, once again, the answer used for calculating this variable is not independent from the answer used to assess people’s unwillingness to pay for UGA conservation (the response variable in Eq.2). Indeed, the two questions are not formulated independently: if people answered that they were willing to pay for UGA conservation in the first question, then the options that remain available to them in the second question are reduced.
RESPONSE: Again, we are not sure what you mean by independence here. In standard stated preference instruments and their analysis, the series of responses given by a person are used to estimate their demand or choice for a given suite of goods or services and to determine covariates that can infer why they would be willing to sacrifice something (or not) for such goods or services. The respondent can also provide their gender or income and a suite of other responses and yes these will be used to calculate a model to explain their demand schedule. Put another way, the Left Hand Side (LhS) Variable indicates a choice made (i.e., they are unwilling to invest), while the Right Hand Side (RhS) variables can have two characteristics: 1. choice varying attributes in the utility functions of respondents and 2. choice invariant variables. As such, our “weak governance” variable belongs to the first set of explanatory variables (choice varying attributes). Therefore, we do expect both sets of variables to be correlated with our dichotomous response (dependent variable), so variables such as gender, strata, etc are “Choice Invariant variables” that can also be correlated with our LhS variable.
Accordingly, this does not nullify their use in these types of logistic models that are regularly used for contingent valuation analyses nor do they violate assumptions of statistical independence. In fact this type of regression approach and use of this type of variable analysis is regularly used for benefits-cost analyses and hedonic price models as it is a regression approach that “..provides a theoretical basis for statistically isolating the independent effects of various characteristics of a product on price (i.e., value; Cost–Benefit Analysis by R.H. Haveman, D.L. Weimer, in International Encyclopedia of the Social & Behavioral Sciences, 2001)”.
Furthermore, you state “…the two questions are not formulated independently: if people answered that they were willing to pay for UGA conservation in the first question, then the options that remain available to them in the second question are reduced.” However, as shown in Appendix A if they answered yes, then they were presented with these following reasons for doing so: Air purification, I use these green areas, Improves citizens’ welfare, Urban biodiversity conservation, I do not know, and No answer. However, these were not included in the weak governance variable. Hence, beyond what we have tried to explain here, we apologize but we are not sure what else or how to respond to this comment.
Thus, these two points suggest a mismatch between the survey design and the analytical methods used. The successive questions asked to respondents are not independent from each other, whereas the variables derived from these questions must be independent to not violate OLS method assumptions. If the questions asked to respondents had been something like “how much are you willing to pay for UGA conservation?” and “do you think there is a corruption issue regarding the use of public money?” or “do you trust the government on the good distribution of public money?”, variable dependency would have been respected.
RESPONSE: We disagree and we have already responded to your OLS concerns. Having asked these specific questions in the way described by the Reviewer would have biased responses particularly if mentioning corruption and trust in government. As laid out in our introduction and discussion, these are serious and complex issues in places like Colombia and many other places, and directly specifying such value-laden terms in the questions would have induced biased responses.
However, we do admit that governance is a very difficult concept to measure. So, to meet you halfway we changed the terms “measure….governance” in our third objective (last paragraph of our Introduction) to “explore…governance” as a means to convey the limitations to our approach and the difficulty in measuring it. In addition, our responses to your previous 2 comments - we feel - should address your observations and concerns.
Beyond this variable dependency issue, the ms does not provide any rationale about why Eq.1 uses certain variables derived from the surveys but not other variables. For example, why the question “Is this place safe” was used in Eq.1 (for the variable SS), whereas the question “What problems does this place present” was not used? This is very appealing, in particular considering that the authors wrote that “respondent’s perceptions regarding costs from urban ecosystems, do affect their identification of benefits” (L.297). So, why people’s perceptions of costs were not used as an explanatory variable of their perceptions of benefits?
RESPONSE: These are interesting points, but as in any manuscript that uses this type of approach, one has to limit the number of variables used based on the study objectives, the available literature, and if they make socioeconomic and ecological sense given the study context. Accordingly, if we are trying to model “benefits”, we feel it does not make sense to include “What problems does this place present”. We could perhaps include this, but then this raises the question, why not include other questions on for example climate change, policy awareness, etc.
However, to address your point we have now added revised text to the future research (2nd sentence, 8th paragraph) section of our Discussion summarizing your observations. “As indicated by a reviewer, the opportunity exists for also including other questions as explanatory variables in or models for better understanding citizen opinions of these functions such as perceptions regarding costs, different dimensions of governance or other environmental and societal factors and how they affect the valuation of benefits.” As an added plus, this might spark interest in other researchers to undertake these questions you pose.
Issue 2: Conception of the Weak_governance variable in Eq.2:
As the authors explain in introduction, governance is a multidimensional concept that involves legitimacy, transparency, accountability, inclusiveness, fairness, resilience, connectivity and performance issues (to name the principles identified in Lockwood et al. 2010 that is cited in the ms). Thus, asking only one question to assess how people perceive governance seems very reductive, especially for a study that aims to analyze the role of perceived governance on the valuation of UGA. We could have expected several questions that would have aimed to assess how people perceive the different governance principles identified in Lockwood et al. 2010, in particular because the authors say they have designed their Eq.2 based on the Lockwood’s paper (L.234). I am therefore not convinced by the relevance of this Weak_governance variable, which is a Boolean variable, to assess perceived governance. This oversimplification of people’s perception of a given situation leads me to my third point.
RESPONSE: We did not use Lockwood et al 2010 to define governance, rather as clearly stated in Lines 226-230, “ Per [31- Colombian program evaluation criteria for governance], a response of not willing to invest (i.e., UTI) included the following reasons for not doing so: 1. Perception of corruption or funds will not be used appropriately, 2. Conservation of green areas and wetlands is already paid for in taxes, and 3. The respondents already pay too much tax, and 4. It is the government´s responsibility to conserve green areas.” Then we elaborate this in our discussion section by stating in Line 397, “Based on Colombian program evaluation criteria [31] for principles and performance outcomes for good governance [34], our survey and modelling found…” Thus, the Colombian governance program evaluation criteria is reductionist, however it is more in-line with real world application of the concept than more theoretical works mentioned in our literature review.
But to address you concern we have added this revised text to our statement on the use of this crietera. The revised text in lines 226-227 now reads, “Per Colombian program evaluation standards [31], a response of not willing to invest (i.e., UTI) included the following reasons for not doing so..”
I believe we have addressed the second point in the previous comments. The third point (i.e., boolean variables), is explained in the following response.
Issue 3: The (over)reductionist approach:
With the surveys, a relatively rich qualitative data has been collected, but then a drastic “degradation” or “simplification” is operated prior to the quantitative analyses. While degrading qualitative data to be able to perform quantitative analyses is sometimes necessary, such operations have important epistemic implications and so should be properly justified and balanced, and their potential influences on the results properly addressed, which is not the case in the present ms. For example, the authors had five categories for assessing people’s willingness to pay, but they transform this categorical variable into a Boolean one (loosing the information on how much people were willing to pay for only keeping a Yes/No information). But can we really consider that someone who would agree to pay up to 4000 COP has the same willingness than someone who agrees to pay more than 30000 COP? What are the implications on the results? Why such a simplification of the data necessary? Unfortunately, the ms does not provide any answer to these questions. Such unexplained (and potentially problematic) variable degradation also concerns (i) the number of wetlands named by people, (ii) the number of trees, palms or plants identified by people (which are both used to calculate the environmental awareness variable used in Eq.1), and (iii) the weak governance variable.
RESPONSE: This is an interesting point you bring out. However, we would like to emphasize that as stated in our objective, we are trying to get at how governance influences the discrete choice decision to invest or not for the benefits provided by these urban green spaces. No where do we state that the focus of our study is to elicit the “maximum willingness to pay” for conservation and that would require an entirely different approach in both our instrument and econometric analyses. Therefore, we do not intend to account for non-linearities that may arise from the differing levels or amounts of potential contributions. We are simply interested in exploring how weak governance affects the decision of participating (or not) in a decision involving investing in “something”.
So yes, we used a boolean explanatory variable. And indeed you could apply this to a number of variables like gender and we could have gone beyond non-binary responses (male/female) and explored the role of bi or transgender respondents. And as we previously responded to you, we could have included other explanatory variables to better understand the role of costs, different dimensions of governance and other environmental and societal factors.
However, this is a finite study with specific objectives and we doubt that using a suite of potential variables and advanced quantitative/mixed analyses would have yielded different findings. Hence, as we state in our objectives, we are “assessing” and “Exploring” the effects of the variables on the valuation of green spaces. We also dedicate a paragraph to report our study limitations, among them we admit “Similarly, concepts such as governance and costs as pointed out by our literature review are complex metaphors to define and measure. Environmental values, resource conflicts, power relationships, green infrastructure types and structure will also affect how people perceive benefits-cost bundles”. However, we later state “That said, the approach used in our study to measure benefits and effects of weak governance and institutional transparency, can be used in … Latin American socio-ecological systems and their contexts [20].”
To conclude, I think the data that supports this ms could be interesting and could result in valuable results on people’s perceptions of the benefits and costs associated to UGAs. Yet, the analyses conducted in this present ms do not seem to be suited to the type of data collected because the questions were not formulated in an independent manner, but also because OLS method have been designed for analyzing continuous quantitative data and not discrete pseudo-quantitative data (derived from qualitative data).
RESPONSE: As explained in our first response to you, we consider OLS models as used in the study to be informative, they do complement our descriptive results, and it can be used explore the relationships between environmental, cultural and provision benefits identified in the El Salitre watershed and socioeconomic variables. Nonetheless, this was not the only analysis used. For added depth and completeness in our analyses , later in our methods we explore these aspects using a more robust discrete choice model where the response variable was a binary choice that was estimated using a logistic model, which is more than sufficiently suitable for binary responses.
I further consider that, with only one question referring to perceived governance, the data collected does not allow a robust and serious assessment of how people perceive governance. My main recommendation to the authors would therefore be to refocus their ms toward an objective that would better fit with the data they collected, and ideally to use analytical methods that would not only allow to account for the number of benefits (and costs) cited by people, but also for the nature of cited benefits and costs (see for example free association techniques, consensus analyses, or multi-factorial analyses for qualitative data).
RESPONSE: Thank you for these suggestions but what this comment suggests would require a completely different study and manuscript. That said, we hope that the above revisions we have laid out and our responses, justify our manuscript to the Editor, to you, and other two reviewers. Indeed, we are encouraged by the other 2 reviewers who found no issues with our econometric approach to exploring out stated objectives. As stated earlier, we feel this revised manuscript is now more transparent in reporting our study limitations and, where we could, we have tried to incorporate your comments as much as possible.
However to address your comment about refocusing our manuscript, we now include “Nature’s Contributions to People” in the revised title and “Unwillingness to invest” in the keywords so as to better fit the data collected and make more relevant and highlight the implications of our study.
Reviewer 2 Report
I had little issue with the original manuscript and found it suitable for publication, and thus have little to offer regarding this resubmission. However, I still feel that it reads very well and presents important research by offering a different geographic perspective from the excessive focus in much related literature on the global north.
Author Response
I had little issue with the original manuscript and found it suitable for publication, and thus have little to offer regarding this resubmission. However, I still feel that it reads very well and presents important research by offering a different geographic perspective from the excessive focus in much related literature on the global north.
RESPONSE: Thank you for your comment. We hope the revisions made in response to another reviewer have further improved the manuscript and that you find them acceptable.
Reviewer 3 Report
The ms is very interesting and it deals with the green areas through a very transdisciplinary approach. The inclusion of the perception in the environmental analyses gives an additional value to the traditional approach to environmental assessments.
The green areas play an important role in terms of ecosystem services' provision. In this context I suggest Authors to better discuss the results in the context of the MEA in terms of the contribution of green areas to urban well-being and quality of life. In this context, I suggest Authors to have a look of a recent paper Valente et al. 2020 published on Ecological Indicators.
For what concern the perception, some of the results of the present paper could be discussed in terms of the contribution of green areas to the maintenance of social capital.
Author Response
The ms is very interesting and it deals with the green areas through a very transdisciplinary approach. The inclusion of the perception in the environmental analyses gives an additional value to the traditional approach to environmental assessments.
The green areas play an important role in terms of ecosystem services' provision. In this context I suggest Authors to better discuss the results in the context of the MEA in terms of the contribution of green areas to urban well-being and quality of life. In this context, I suggest Authors to have a look of a recent paper Valente et al. 2020 published on Ecological Indicators.
RESPONSE: This is an interesting point in that one of the things we found was that the MEA might not be the most relevant ecosystem service framework for places such as Bogota and other Global South contexts. In fact, one of our findings is that many of the benefits and costs identified did not fit nicely into the MEA approach but rather the more controversial Nature’s contributions to people seemed more appropriate. We clearly state this in our Abstract, Introduction, Discussion and now in our revised title.
So, we did look at the article “The role of green infrastructures in Italian cities by linking natural and social capital”. And after reading it, we replaced one of our references [2] with this one as it was more appropriate. And to address your comment we have also added the following text to reflect your suggestion in our study limitations paragraph (8th paragraph of the Discussion section), “Indeed, other biophysical factors such as size and density of green areas per neighborhood could also affect human well-being [2].”
For what concern the perception, some of the results of the present paper could be discussed in terms of the contribution of green areas to the maintenance of social capital.
RESPONSE: Thank you for this comment. In our limitations section (8th paragraph of the Discussion section) we have added this statement, “As indicated by a reviewer, the opportunity exists for also including other questions as explanatory variables in or models for better understanding citizen opinions of these functions such as perceptions regarding ….social capital factors and how they affect the valuation of benefits.” Again thank you for your comments and suggestions.
