Trust in government moderates the association between fear of COVID-19 as well as empathic concern and preventive behaviour

With the COVID-19 pandemic, behavioural scientists aimed to illuminate reasons why people comply with (or not) large-scale cooperative activities. Here we investigated the motives that underlie support for COVID-19 preventive behaviours in a sample of 12,758 individuals from 34 countries. We hypothesized that the associations of empathic prosocial concern and fear of disease with support towards preventive COVID-19 behaviours would be moderated by trust in the government. Results suggest that the association between fear of disease and support for COVID-19 preventive behaviours was strongest when trust in the government was weak (both at individual- and country-level). Conversely, the association with empathic prosocial concern was strongest when trust in the government was high, but this moderation was only found at individual-level scores of governmental trust. We discuss how motivations may be shaped by socio-cultural context, and outline how findings may contribute to a better understanding of collective action during global crises.

citations.pdf</a>.At a minimum the Data availability statement must explain how the data can be obtained and whether there are any restrictions on data sharing.Communications Psychology strongly endorses open sharing of data.If you do make your data openly available, please include in the statement: -Unique identifiers (such as DOIs and hyperlinks for datasets in public repositories) -Accession codes where appropriate -If applicable, a statement regarding data available with restrictions -If a dataset has a Digital Object Identifier (DOI) as its unique identifier, we strongly encourage including this in the Reference list and citing the dataset in the Data Availability Statement.
We recommend submitting the data to discipline-specific, community-recognized repositories, where possible and a list of recommended repositories is provided at <a href="http://www.nature.com/sdata/policies/repositories">http://www.nature.com/sdata/policies/repositories</a>.
If a community resource is unavailable, data can be submitted to generalist repositories such as <a href="https://figshare.com/">figshare</a> or <a href="http://datadryad.org/">Dryad Digital Repository</a>.Please provide a unique identifier for the data (for example a DOI or a permanent URL) in the data availability statement, if possible.If the repository does not provide identifiers, we encourage authors to supply the search terms that will return the data.For data that have been obtained from publicly available sources, please provide a URL and the specific data product name in the data availability statement.Data with a DOI should be further cited in the methods reference section.

REVIEWERS' EXPERTISE:
Reviewer #1: compliance with CVOID-19 measures Reviewer #2: compliance with CVOID-19 measures Reviewer #3: World Value Study, statistics REVIEWERS' COMMENTS: Reviewer #1 (Remarks to the Author): The manuscript is very well written and it presents a unique data set.It is not only reporting a multinational study, but also on that involves a wider variety of countries than most similar datasets.Here, most of the data does clearly not stem from Western industrialized countries, which is clearly an asset.Another strength is that the hypotheses have been preregisters.Unfortunately, the quality of the preregistration is lower regarding the data analysis procedure.Testing the moderation of the relations between COVID-19 containment behaviors and fear as well as empathy by governmental trust is to the best of my knowledge innovative.At the same time, the number of studies testing predictors of COVID-19 containment behaviors is almost countless and the main effects of empathy of fear, which have long been known, are much larger than the interactions with trust.In my opinion, standard effect sizes (e.g., Cohen's d) should be added for all reported effects.This would help to understand whether the B = .002for the TG x FoC interaction is meaningful in size.My sense is, that it might be not.Also, tables could profit from some additional information (e.g., Table 5 does not indicate, which type of indices are reported).Finally, numbers should be reported with 2 meaningful digits (e.g., in table 5 some SE = .00which is not informative).As acknowledged by the authors the reported data is cross-sectional, which limits the conclusions that can be drawn about causality.A longitudinal design would be hard to implement in a multinational study.Still it is hard to draw implications from cross-sectional data.Overall, I like the idea, the way the paper is written, the data, and the findings and I strongly believe it should be published in Communications Psychology or somewhere else.
Reviewer #2 (Remarks to the Author): Reviewer report: Empathy, Fear of Disease and Support for COVID-19 Containment Behaviors: Evidence from 34 Countries on the Moderating Role of Governmental Trust The paper investigates the association between empathy and fear on one hand and support for Covid containment measures on the other in a large data collection exercise spanning 34 countries.The article moreover investigates and uncovers the moderating effect of governmental trust on the relationship under investigation.
The findings report a strong and positive correlation between support for containment measures and fear and empathy with a strong moderating role of trust in government, whereby strong trust in government is associated with a strong empathy-support correlation and low trust in government is associated with a strong fear-support correlation.I believe this evidence, though I do not agree with the mechanism the author propose as explanation for it.
I struggle with the logic behind the proposed relationship between trust in government and empathydriven cooperation (page 13).This logic is key in the paper as the interpretation of the results the authors obtain rests entirely on it.The authors argue that "Only when trust in the government is high, individuals would perceive their individual efforts to contain the virus as positive for the community at large, because preventing the spread of the COVID-19 pandemic would be perceived as a concerted effort and a shared goal by many.On the other hand, when there is no trust or little trust in the government, individuals would not expect that their individual compliance with public health measures will produce any public, other-oriented benefit, as they would not expect their contribution to be reciprocated by their government's actions and regulations."If this logic holds, then the interpretation of the slopes in the interacted regression presented holds (with the caveat that the causal relationship remains unobservable and can only be taken with a pinch of salt and only to the extent with which we believe in the logic presented above).However, as I hinted at, the logic is not fully convincing.If regulations are in place, why would I not comply in fear of the government not implementing regulations?(this question paraphrases the authors' words, if I understand them correctly) Here's an alternative mechanism which to my understanding would produce the same observations reported in this study.Trust in government and generalised social trust are suspected to be (to the least) correlates (see for instance Putnam or Sønderskov et al. 2016).If I state I trust the government, it is likely I trust others in society too.If I am empathic, I will this support covid 19 measures because my belief is that others in society will comply with those (enacted) regulations (Hypothesis 1a is confirmed).If I don't trust the government and hence neither I trust others, the only reason I would want more stringent regulation is my fear that people won't comply (because I don't trust them) and hence stricter rules are required to get a minimum amount of compliance (Hypothesis 2a is confirmed).The same logic would produce arguments close to the above and would result in statements which would confirm Hypotheses 1b and 2b as well.
The authors are open about the correlational nature of their study, but only on page 24, and the entire paper reads as if trust in government is indeed the moderator between empathy/fear and support for governmental policies.That the study is correlational is not my concern here.What I am concerned with is that acceptance of their theory rests upon assumptions (the logic reported above) which in my view are not very convincing.Should the authors have individual data on social trust they could replicate their analyses to check whether models including social trust are more explanatory than those including governmental trust.Perhaps some statistical technique might be able to (perhaps imperfectly) disentangle the two (I fear no suggestion comes to my mind).Country level data on social trust is publicly available.Perhaps they could include analogous analyses to the mixed-level ones using this data.
Reviewer #3 (Remarks to the Author): Referee Report -Empathy, Fear of Disease and Support for COVID-19 Containment Behaviors: Evidence from 34 Countries on the Moderating Role of Governmental Trust

Summary of the findings
Using a large sample of 12,758 individuals from 34 countries, the authors show that empathic prosocial concern and fear of disease are positively and significantly related with support towards preventive COVID-19 behaviors.Further, the authors emphasize that these relationships are moderated by the level of trust in the countries' government.Finally, the authors discuss how both fear and empathy motivations to support preventive COVID-19 behaviors may be shaped by the sociocultural context.

Comments
I appreciated the opportunity to read the paper.Overall, the paper is well-written and presents some interesting findings.Yet, I have a couple of points (especially related to methodology) that I would like to elaborate on in more detail.
1.I like how the authors carry out their main analysis; however, I wonder why the authors do not control for the individual characteristics of the respondents (e.g., age and gender).What is the reason for that? 2. Similarly, I would suggest to the authors to include a control variable for the severeness of COVID-19 in the respective countries?For example, the number of cases/death per inhabitants.
3. Further, I would suggest to the authors to use a different regression specification for robustness purposes.Specifically, I would suggest using a specification including country fixed effects and clustering the standard errors across countries.This would help to address endogeneity concerns.4. I like that the authors use the country-level scores for trust in the government from the World Values Survey.However, Engelhardt et al. (2021) also investigated a related question in the finance literature and used scores from the OECD for robustness checks.I would suggest to the authors to check robustness by using these scores as well.
5. I also recommend to the authors to check if there is a moderating effect when focusing the level of societal trust.In this respect, Engelhardt et al. (2021) note that "trust in governments might only be one side of the coin as trust in fellow citizens obeying the government's orders might also be of significant importance."The data on societal trust is also available from the World Values Survey.

RESPONSE LETTER Reviewer 1 Comments:
General comment: The manuscript is very well written and it presents a unique data set.It is not only reporting a multinational study, but also on that involves a wider variety of countries than most similar datasets.Here, most of the data does clearly not stem from Western industrialized countries, which is clearly an asset.Another strength is that the hypotheses have been preregisters.Unfortunately, the quality of the preregistration is lower regarding the data analysis procedure.Testing the moderation of the relations between COVID-19 containment behaviors and fear as well as empathy by governmental trust is to the best of my knowledge innovative.At the same time, the number of studies testing predictors of COVID-19 containment behaviors is almost countless and the main effects of empathy of fear, which have long been known, are much larger than the interactions with trust.
Response: We are happy that the Reviewer find merits in our work, and grateful for their compliments on the clarity of our presentation, our mindfulness to include data from lessrepresented countries, the innovative analytical approaches, and our open-science practices (with pre-registration of our hypotheses).We are using this opportunity to assure that we took care to address all remaining issues.In line with the reviewer's suggestion, we updated our preregistration by adding a detailed analysis plan (see https://osf.io/k2wjr).For the remainder of the issues, please refer to our detailed responses below.
1.In my opinion, standard effect sizes (e.g., Cohen's d) should be added for all reported effects.This would help to understand whether the B = .002for the TG x FoC interaction is meaningful in size.My sense is, that it might be not.
Response: We agree with the Reviewer and we added effect sizes (Cohen's f) for all estimates, when reporting the main effects and interactions in the main text (see highlighted parts in the results section on page 19-22).As already noted by the Reviewer, the effect sizes for the interaction were relatively small (especially for the individual level analyses).Therefore, we drew additional attention to this limitation in the discussion of the results on page 23/24 and page 28/29.
Page 23/24: It should be noted however, that even though trust in one's government altered the strength of the association between empathic concern and support for COVID-19 containment behaviors, the strength of the association under high versus low trust in the government was almost identical (Cohen's f = 2.01 vs. Cohen's f = 1.99).Thus, even though the association between empathy and support for large-scale cooperative action may become more pronounced when individuals perceive their governments as more (compared to less) trustworthy, such interplay with trust seems less salient for empathic concern (compared to fear of COVID-19).
Page 28/29: Last, we note that the moderation effects, though significant, are small in magnitude, especially the ones obtained from the individual level analysis (Cohen's f = 0.02).However, this is not in itself a concern due to the explorative nature of the study.It could become a concern if these results were taken at face-value without subsequent testing 2. Also, tables could profit from some additional information (e.g., Table 5 does not indicate, which type of indices are reported).Finally, numbers should be reported with 2 meaningful digits (e.g., in table 5 some SE = .00which is not informative).

Response: Following this recommendation we updated the tables to include: a) Legend with information regarding the reported indices; b) Brief description as regards the performed analyses and; c) Correct reporting of numbers (with 2 meaningful digits).
3. As acknowledged by the authors the reported data is cross-sectional, which limits the conclusions that can be drawn about causality.A longitudinal design would be hard to implement in a multinational study.Still it is hard to draw implications from cross-sectional data.Overall, I like the idea, the way the paper is written, the data, and the findings and I strongly believe it should be published in Communications Psychology or somewhere else.
We are very grateful for this encouraging feedback, the constructive approach with overall positive evaluation of our work, and the recommendation for publication of our results.In the revised version of the manuscript, we paid more attention to avoid language that would imply causality.Furthermore, we would like to reiterate, that the cross-sectional approach was highlighted as a possible limitation at the very begging of the corresponding "Limitation" section of the manuscript (on page 27), so as to acknowledge that caution is warranted when interpreting the results and to caution the audience against making overconfident conclusions.The statement reads as follows: Page 27: The present research allowed us to examine the links between trust and acceptance of imposed restrictions during times of COVID-19, albeit in a correlational rather than causal manner.Though correlational research is central for scientific progress 58 , it is but an intermediate step in the process.Correlational research such as the one presented in this study should ideally be supported by ensuing causal confirmations, or at least with corroborated results from independent samples (also see many labs project 59 ).This of course is not a feasible solution in the present situation, as, hopefully, the COVID-19 pandemic is nearing its end.It is, however, an observation that should make us interpret the present findings with due caution.General comment: The findings report a strong and positive correlation between support for containment measures and fear and empathy with a strong moderating role of trust in government, whereby strong trust in government is associated with a strong empathy-support correlation and low trust in government is associated with a strong fear-support correlation.I believe this evidence, though I do not agree with the mechanism the author propose as explanation for it.
Response: We thank the Reviewer for the concise summary of our work, and correct interpretation of our findings.Also, we acknowledge the issues raised in the general comment and we made sure to address them in our responses below.
I struggle with the logic behind the proposed relationship between trust in government and empathy-driven cooperation (page 13).This logic is key in the paper as the interpretation of the results the authors obtain rests entirely on it.The authors argue that "Only when trust in the government is high, individuals would perceive their individual efforts to contain the virus as positive for the community at large, because preventing the spread of the COVID-19 pandemic would be perceived as a concerted effort and a shared goal by many.On the other hand, when there is no trust or little trust in the government, individuals would not expect that their individual compliance with public health measures will produce any public, other-oriented benefit, as they would not expect their contribution to be reciprocated by their government's actions and regulations."If this logic holds, then the interpretation of the slopes in the interacted regression presented holds (with the caveat that the causal relationship remains unobservable and can only be taken with a pinch of salt and only to the extent with which we believe in the logic presented above).
Response: We thank the Reviewer for their observation that the rationale for our hypotheses seemed less convincing in the first version of the manuscript.We believe that this was partly related to the fact that some parts of the justification did not sufficiently outline the unique role of trust in the government (as opposed to generalized trust).We extensively revised both the introduction and the discussion of the paper to reduce such confusion, and to strengthen the rationale for our hypotheses.Additionally, we ran supplemental analyses to address concerns raised by the Reviewer.We will outline these changes in detail below, and hope that the current version of the manuscript addresses the reviewer's concerns in a manner that is both clear, concise and comprehensive.
1.However, as I hinted at, the logic is not fully convincing.If regulations are in place, why would I not comply in fear of the government not implementing regulations?(this question paraphrases the authors' words, if I understand them correctly) Response: We totally agree with the issue raised by the Reviewer here.Actually, we did not hypothesize that fear would not operate as a motivator if regulations are in place.However, we agree that the way the argument was developed (in the original version) opened space for such an understanding.To clarify that our hypotheses would not imply that fear will become irrelevant (or less relevant than empathy) when trust is high, we added the following clarification to our manuscript on page 25 (discussion): Page 25: Yet, it should be noted that our results do not indicate that empathy would not operate in low trust settings, or that fear would not operate in high trust settings.Instead, our results support the view that both motivational pathways seem relevant for mobilizing cooperative action across individuals and contexts (i.e., as main effects were consistently recorded across various levels of governmental trust), but suggest that empathy-driven cooperation is likely to unfold optimally when trust in the government is perceived as high, while fear-driven cooperation is likely to become most pronounced in contexts where trust in the government is generally low, or when trust in the government is individually perceived as low.Response: We thank the Reviewer for this insightful observation, and for employing additional efforts to provide alternative interpretation of (and explanation for) our findings.And we agree that in the previous version of the manuscript the unique role of trust in the government (compared to trust in general) was not sufficiently outlined.We addressed this issue by revising the introduction extensively (p.14-16) and hope that these revisions address the concerns raised by the reviewer.
As outlined above, mobilizing people for large-scale and long-term cooperative acts (as in times of the pandemic) is challenging; mainly due to the high level of uncertainty it involves.Some individuals (especially those who feel less vulnerable to the virus) might even experience a social dilemma whereby cooperating would require them to sacrifice their concrete, short-term, and selfish interests (e.g., by avoiding social events, by face masking, etc.) over the more abstract, less traceable, and long-term goal of protecting other (mostly vulnerable) people from infection.We argue that such a dilemma should become especially salient, and thus hinder empathy-driven cooperation, under highly unpredictable conditions where one is doubtful about whether or not individual "sacrifices" will promote any long-term public benefit.The way in which people perceive their governments, namely whether they view them as competent, protective, and caring -simply, whether they trust in their government or not -may represent one of the factors that guides people's perceptions of (un)predictability.Namely, when people anticipate their governments to be supportive and to take the needed measures for protecting their citizens from any threat (including the COVID-19 pandemic), individuals would become more confident that their selfless and empathy-driven activities will truly result in positive public health outcomes.On the contrary, when there is no trust or little trust in the government, unpredictability will increase, and individuals will doubt whether their individual compliance with public health measures will produce any public, other-oriented benefit.Even if respective measures are implemented by the government, those with little trust in the government would be propelled to suspect that they may be implemented in a superficial and non-transparent manner.For instance, the value of face masking (individually and even collectively) to protect others from infection would become less obvious and more questionable when governments do not offer the needed framing conditions and (are perceived to) act in pandemic-ignorant ways.
However, under such volatile conditions cooperation may still occur, yet for different sets of reasons.As outlined above, the compliance with safe and preventive COVID-19 behaviors is not only an act that may be performed for the sake of protecting others from infection, but also has direct (beneficial) implications on personal health, and may thus be considered as a self-protective act.In circumstances where trust in the government is low, people would experience higher uncertainty, feel more vulnerable and susceptible to the virus, and their cooperation would be more strongly driven by selfish concerns such as fear of COVID-19.Therefore, we propose that the relationship between empathic concern and support for COVID-19 containment behaviors will become stronger when the government is (generally) perceived as more (compared to less) trustworthy, and that conversely the association between fear of disease and support for COVID-19 containment behaviors will become stronger when the government is (generally) perceived as less (compared to more) trustworthy.
1.The authors are open about the correlational nature of their study, but only on page 24, and the entire paper reads as if trust in government is indeed the moderator between empathy/fear and support for governmental policies.That the study is correlational is not my concern here.What I am concerned with is that acceptance of their theory rests upon assumptions (the logic reported above) which in my view are not very convincing.Should the authors have individual data on social trust they could replicate their analyses to check whether models including social trust are more explanatory than those including governmental trust.Perhaps some statistical technique might be able to (perhaps imperfectly) disentangle the two (I fear no suggestion comes to my mind).Country level data on social trust is publicly available.Perhaps they could include analogous analyses to the mixed-level ones using this data.
Response: We thank the Reviewer for this excellent suggestion, which actually helped us to strengthen our argument.Accordingly, we conducted additional analyses with generalized trust (both at the individual and country level) and discussed their results in the discussion on p 25/26 The excerpt is copied as follows: Page 25/26: Another question to be addressed concerns the role of generalized trust (i.e., how much people trust each other in general), as much of our argumentation in regard to governmental trust seem also applicable to generalized trust, and since generalized trust has also been found to be a relevant predictor of compliance during the COVID-19 pandemic 52 .Yet -while we acknowledge the crucial role of generalized trust-the present research deemed important to focus on trust in the government, since pandemic management obviously takes more than just citizens' goodwill, and largely depends on the governments' activities and on how much people find them trustworthy 53 .This is also in line with previous research, confirming the central and unique role that "systemic trust" has played for cooperative action during the COVID-19 pandemic, by showing that not only interpersonal trust (r = 0.14, p < 0.05) but also -and more strongly -trust in politicians was associated with COVID-19 vaccination (r = 0.24, p < 0.01) 47,54,55 .Nevertheless, we conducted additional analyses to examine (1) whether our findings remain robust when generalized trust is added as another covariate to the analyses, and (2) whether the proposed moderation effects can also be obtained with generalized (instead of governmental) trust, using the same analytical procedures as described above.
The scores for individual-level generalized trust were obtained by averaging the scores that respondents provided to the following two questions: "In general, most people in our community can be trusted" and "Most people in our community are fair and do not take advantage of you" (1 = totally disagree, 5 = totally agree, r = 0.78).The scores for country-level generalized trust, were extracted from the most recent wave of the World Values Survey (WVS).First, our analyses revealed that the results did not change when generalized trust was added as a covariate into the regression models.Second, when analyses were conducted with generalized instead of governmental trust, no interaction with fear and empathy emerged, neither on the individual-level nor the country-level analyses (expect for a marginal significant interaction with fear of COVID-19 in the proposed direction with p = 0.089).Interestingly, generalized trust was also not associated directly with support for COVID-19 containment measures, neither on an individual level (p = 0.073) nor on a country level (p = 0.204).This supports the view that trust in the government seems conceptually distinct from generalized trust (in our research they were correlated with r = 0.58 on the individual-level; and r = 0.10 on at the country-level), exhibiting unique associations with support for COVID-19 containment behaviors (detailed results are available at https://osf.io/kws9x/files/under supplementary analyses).

Reviewer 3 Comments:
General comment I appreciated the opportunity to read the paper.Overall, the paper is well-written and presents some interesting findings.Yet, I have a couple of points (especially related to methodology) that I would like to elaborate on in more detail.
Response: Thank you, we truly appreciate your kind words and very concrete instructions that helped us improve the presentation of our findings.
1.I like how the authors carry out their main analysis; however, I wonder why the authors do not control for the individual characteristics of the respondents (e.g., age and gender).What is the reason for that?

Response: Thank you for drawing attention to this issue. The reason for this choice was simply due to parsimony. In the original version of the paper, these analyses were actually conducted and reported in a footnote (the output was made available via the OSF). As the results did not change substantially, we decided to present results without covariate effects. However, in the revised version of the paper, we present the analyses and results by controlling for the effects of gender, age, HDI, hospital beds per 1000, data collection month, government stringency level, COVID-19 cases and death rates by the time of data collection (as kindly suggested by you and the Editor). In short, our findings remain even after controlling for the above mentioned socio-demographic variables. This is noted inside the manuscript (on page 19 and in Table 4 & 5) as follows:
Page 19: Additionally, to account for possible socio-demographic, country-and pandemic-specific effects, we entered respondents' age and their gender, and country-level scores of the Human Development Index (HDI, a composite score reflecting the level of a country's overall development in the domains of economy, education, and health) 43 , number of hospital beds per 1000, month of data collection, government stringency level, and the number of new daily COVID-19 cases and deaths by the time of data collection as covariates to the analyses.
2. Similarly, I would suggest to the authors to include a control variable for the severeness of COVID-19 in the respective countries?For example, the number of cases/death per inhabitants.
Response: As noted above, we added rates for COVID-19 cases and deaths as covariates into our regression model.For more details, please refer to the copied excerpt (in response to the above comment).
3. Further, I would suggest to the authors to use a different regression specification for robustness purposes.Specifically, I would suggest using a specification including country fixed effects and clustering the standard errors across countries.This would help to address endogeneity concerns.

Response: We thank the reviewer for this suggestion which made us search for analytical methodologies to address endogeneity concerns in mixed-level regressions. However, we could not find out how to incorporate this suggestion into our analyses that we conducted with jamovi (which also produces a command that can be used in R). Some of the coauthors suggested to add 33 dummy variables for the 34 countries into the model; others suggested to cluster the countries into meaningful groups based on the study variables (e.g., via k-means), while a third group suggested to conduct analyses with only fixed effects. In the current version, we tested our predictions with a regression model where we specified fixed effects for the predictors' main effects and the two-way interactions, plus random effects for the predictors' main effects (country, trust, empathy and fear). Hence, due to my unfamiliarity with such kind of analyses (as it is rarely used in our discipline), we had trouble in clearly understanding what to do. If the Reviewer deems this analysis important, we would be happy to run these additional analyses and
present them in the OSF files.However, for doing so, we would appreciate if the Reviewer can provide us with some clarification, assistance and guidance in how to do this (e.g., is there a command that can be added to the R for running these analyses?).
4. I like that the authors use the country-level scores for trust in the government from the World Values Survey.However, Engelhardt et al. ( 2021) also investigated a related question in the finance literature and used scores from the OECD for robustness checks.I would suggest to the authors to check robustness by using these scores as well.
Response: We thank the reviewer for this excellent suggestion which we have followed faithfully.When running the country-level analyses with the OECD trust scores (instead of WVS scores), we find a significant interaction in the hypothesized direction with empathy, but not with fear.We copied the analysis output towards the end of this letter.We believe that there may be several reasons for this.First, it should be noted that the sample decreases from 29 countries to 13 countries (N = 5541) when the OECD data is used, as much of the countries represented in our paper are not represented in the OECD.As such, the OECD countries may represent a more selective and less comprehensive sample than the WVS countries which are more inclusive.Second, the OECD data on trust in the government more strongly reflects trust in times of COVID-19 while the WVS trust has been obtained from pre-pandemic times and would thus be more reflective of trust in government in a way that is more independent of the pandemic.Moreover, while the WVS trust and OECD trust scores were moderately correlated (r = .29),it is likely that they are substantially different.Correlational analyses showed that the OECD governmental trust score was more strongly associated with interpersonal trust form the WVS; also it was positively associated with a country's HDI while the trust score from the WVS was negatively associated with it.All this suggests that both types of assessment may measure different conceptualizations of trust in the government and thus not show parallel effects.While we believe that this additional robustness check may be valuable, we also thought that adding this analysis along with the discussion of the inconsistent results would spread the scope of the present paper.Hence, for now we decided not to include the OECD robustness check into the paper.However, if the Reviewer and the Editor consider this important, we may add this additional analysis along with its discussion into the paper (as we did with generalized trust).
5. I also recommend to the authors to check if there is a moderating effect when focusing the level of societal trust.In this respect, Engelhardt et al. (2021) note that "trust in governments might only be one side of the coin as trust in fellow citizens obeying the government's orders might also be of significant importance."The data on societal trust is also available from the World Values Survey.
Response: We thank the Reviewer for this excellent suggestion.As already outlined above (in our response to the third comment 3 by Reviewer 2), we conducted additional analyses with generalized trust (both at the individual and country level), and reasoned about the obtained results in the discussion of the revised manuscript on page 25/26.
Page 25/26: Another question to be addressed concerns the role of generalized trust (i.e., how much people trust each other in general), as much of our argumentation in regard to governmental trust seem also applicable to generalized trust, and since generalized trust has also been found to be a relevant predictor of compliance during the COVID-19 pandemic 52 .Yet -while we acknowledge the crucial role of generalized trust-the present research deemed important to focus on trust in the government, since pandemic management obviously takes more than just citizens' goodwill, and largely depends on the governments' activities and on how much people find them trustworthy 53 .This is also in line with previous research, confirming the central and unique role that "systemic trust" has played for cooperative action during the COVID-19 pandemic, by showing that not only interpersonal trust (r = 0.14, p < 0.05) but also -and more strongly -trust in politicians was associated with COVID-19 vaccination (r = 0.24, p < 0.01) 47,54,55 .Nevertheless, we conducted additional analyses to examine (1) whether our findings remain robust when generalized trust is added as another covariate to the analyses, and (2) whether the proposed moderation effects can also be obtained with generalized (instead of governmental) trust, using the same analytical procedures as described above.
The scores for individual-level generalized trust were obtained by averaging the scores that respondents provided to the following two questions: "In general, most people in our community can be trusted" and "Most people in our community are fair and do not take advantage of you" (1 = totally disagree, 5 = totally agree, r = 0.78).The scores for country-level generalized trust, were extracted from the most recent wave of the World Values Survey (WVS).First, our analyses revealed that the results did not change when generalized trust was added as a covariate into the regression models.Second, when analyses were conducted with generalized instead of governmental trust, no interaction with fear and empathy emerged, neither on the individual-level nor the country-level analyses (expect for a marginal significant interaction with fear of COVID-19 in the proposed direction with p = 0.089).Interestingly, generalized trust was also not associated directly with support for COVID-19 containment measures, neither on an individual level (p = 0.073) nor on a country level (p = 0.204).This supports the view that trust in the government seems conceptually distinct from generalized trust (in our research they were correlated with r = 0.58 on the individual-level; and r = 0.10 on at the country-level), exhibiting unique associations with support for COVID-19 containment behaviors (detailed results are available at https://osf.io/kws9x/files/under supplementary analyses).We apologise for the delay in processing your manuscript and thank you for your patience during the peer-review process.Your revised manuscript titled "Empathy, Fear of Disease and Support for COVID-19 Containment Behaviors: Evidence from 34 Countries on the Moderating Role of Governmental Trust" has now been seen by 2 reviewers, and I include their comments at the end of this message.The third reviewer (Reviewer #3) was unable to re-review your manuscript, so we asked Reviewer #2 to assess your responses to points 3 and 5 of Reviewer #3's report.Based on this assessment, you will see there are some outstanding points we would like you to address before making a final decision.

Info
In particular, we ask you to conduct and analysis to check the robustness of your estimates to clustering (see second to last paragraph of Reviewer #3's additional comments).
We therefore invite you to revise and resubmit your manuscript, along with a point-by-point response to the reviewers.Please highlight all changes in the manuscript text file.
Editorially, we also ask you to use appropriate language to describe the null results.Statements such as 'Findings from the regression analysis suggest that country-level scores of trust in government did not interact with empathic prosocial concern (p = 0.422), which disconfirms Hypothesis 1b.' must be revised to read 'We found [no/little] credible evidence of X'.
We are committed to providing a fair and constructive peer-review process.Please don't hesitate to contact us if you wish to discuss the revision in more detail.
Please use the following link to submit your revised manuscript, point-by-point response to the referees' comments (which should be in a separate document to any cover letter) and the completed checklist: [Link redacted] ** This url links to your confidential home page and associated information about manuscripts you may have submitted or be reviewing for us.If you wish to forward this email to co-authors, please delete the link to your homepage first ** We hope to receive your revised paper within 8 weeks; please let us know if you aren't able to submit it within this time so that we can discuss how best to proceed.If we don't hear from you, and the revision process takes significantly longer, we may close your file.In this event, we will still be happy to reconsider your paper at a later date, provided it still presents a significant contribution to the literature at that stage.
We would appreciate it if you could keep us informed about an estimated timescale for resubmission, to facilitate our planning.
Please do not hesitate to contact me if you have any questions or would like to discuss these revisions further.We look forward to seeing the revised manuscript and thank you for the opportunity to review your work.

Best regards,
Decision letter and referee reports: second round Editor Comments: 1.In particular, we ask you to conduct and analysis to check the robustness of your estimates to clustering (see second to last paragraph of Reviewer #3's additional comments).
We thank the Editor and Reviewer 3 for their comment regarding the robustness of our results to a different regression approach.In the current version of the manuscript, the additional analyses are referred to on page 27/28; and details regarding the results of these analyses and a discussion of the results is presented in the Supplementary Materials.
Page 27/28: Finally, we tested the robustness of our results when using a different methodological approach, namely a fixed effects regression model with cluster-robust standard errors.For this purpose, two additional analyses were performed: first, a simple OLS regression; and second, an OLS regression with the clustered standard errors correction.A table comparing the results obtained from these approaches, and a brief discussion concerning the differences in the results can be found in the Supplementary Materials (SM).The comparison of the results obtained under the three different methodological approaches revealed that results from the OLS regression approach (Model 1) confirmed the results that we obtained when using a random-slopes MLM approach (Model 1).In both types of regression models, the interaction between trust in the government on the country-level and fear of COVID-19 was found significant, while the statistical analyses did not provide evidence for a significant interaction between country-level trust in the government and empathic prosocial concern.However, when applying the cluster-robust standard errors correction to the OLS regression (Model 3), the results obtained from this analysis did no longer support the existence of a significant interaction between trust in the government and fear of COVID-19.Hence, while Hypothesis 2b seems supported under Model 1 and Model 2, the analysis results obtained under Model 3, where the size of standard errors is typically larger, do not support this hypothesis.While this difference in results between Model 1 and Model 3 may be interpreted as an indicator for a less robust interaction effect, it should also be noted that there is an ongoing discussion concerning the use of these two regression approaches (MLM vs. OLS with cluster-robust errors).Whether one or the other approach should be preferred pretty much depends on the research setting and the type of research question that is to be tested.While the OLS with cluster-robust standard errors seems more appropriate for single-level models where the researcher theorizes variation across different units (i.e., countries) more like a nuisance factor that must be controlled for, the MLM approach seems more appropriate in settings that assume a multi-level structure where the different units are more independent from each other and can produce different effects (for a comparison of these two approaches, see 1 ).We thus believe that the random-effects MLM approach (Model 1) that we reported in the main analyses is more suitable for testing the hypotheses in the current research setting.