1 Introduction

The novel Corona virus was first reported in December 2019 in the Chinese Wuhan city and within short duration spread across numerous countries. The Covid-19 has emerged as a global pandemic and challenged the economic, social and psychological well-being of human beings. Historically, human beings have shown immense resistance against disasters and for achieving shared prosperity during crises. The scale of destruction and challenges posed by the Covid-19 are however unprecedented and require special efforts to ensure the multi-facet damage is curtailed quickly. The existing preliminary research shows that pro-social preferences and trust can play an important role in fighting the challenges of the Covid-19 and in the enhancement of socio-economic and psychological well-being of people (Van Bavel et al 2020). Both pro-sociality and trust can enable individuals to cope up with difficult situations caused by pandemic, extend help towards others and cooperate with each other for repairing the damage caused to the societal setup, political institutions and the economic development (Bjørnskov and Méon 2013; Brañas-Garza et al. 2022; Brück et al. 2020). Moreover, recent evidence also suggests that prosocial people are more likely to take preventive measures (such as social distancing, staying home, wearing masks and intent of vaccination) for the containment of Covid-19 and hence make individual-level contributions for the overall safety and improvement of the society (Campos-Mercade et al. 2021; Ospina et al. 2021; Yu et al. 2021).

The potentially important role of pro-sociality and trust in mediating the adverse effects of pandemic necessitates an investigation how Covid-19 has influenced these preferences. There is a possibility that adverse effects of the Covid-19 on the economic activity and co-occurrence of fear of virus contraction can produce psychological distress and make people less prosocial and less trusting (Pan et al. 2022). Similarly, uncertainty in the labor markets and fear of job loss can lead to reduced prosocial behavior specifically in the form of monetary donations. Also, enhanced economic inequalities due to disruptions in the labor market can cause social unrest and ultimately lead to a trust deficit in the society. Furthermore, as Covid-19 mandates social distancing and isolation, reduced interpersonal interactions due to such preventive measures can also decrease general trust in the society. Contrarily, ‘posttraumatic growth’ theory advocates the transformative influence of suffering and predicts a positive change in the personal and interpersonal behavior after experiencing a disaster or a stressful event (Tedeschi and Calhoun 2004). In the literature several other psychological theories also predict similar outcomes as suggested by the ‘posttraumatic growth’. Some of these include positive psychological changes (Yalom and Lieberman 1991), stress related growth (Park et al. 1996), positive by-products of the struggle (McMillen et al. 2001), stress and positive emotions (Folkman and Moskowitz 2000) and others (Tedeschi and Calhoun 2004). Similarly, positive psychology also suggests the possibility that stressful events can be a source of growth. Therefore, both posttraumatic growth theory and positive psychology would predict an enhancement in prosocial behavior and trust due to adverse effects of direct confrontation with the Covid-19.

The prosocial behavior can also increase because donors understand the need and efficacy of donating money is higher during the pandemic when a vast majority of people is suffering.Footnote 1 Similarly, to tackle the stressful environment created by the Covid-19, people might become more prosocial because often these acts provide personal satisfaction and enhance well-being.Footnote 2 Some of the existing studies also report that exposure to challenges and stress can improve prosocial behavior and trust (Buchanan and Preston 2014; Von Dawans et al. 2012).Footnote 3 From an economic perspective, the policy studies and welfare analyses are performed based on the premises that preferences are stable over time (Drichoutis and Nayga 2021). Therefore, the traditional economic theory would predict no significant influence of pandemic on the prosocial behavior and trust.

The existing literature that examines the influence of pandemic on pro-sociality and trust is relatively sparse and offers divergent results. For example, Shachat et al. (2021) and Zhang and Liu (2021) report that pandemic has increased prosocial behavior in China while Brañas-Garza et al. (2022) report opposite results from Spain. Brück et al. (2020) find that the general trust does not differ across the Covid-19 infected and uninfected people while Sibley et al. (2020) find increased trust post-pandemic among New Zealanders. This expanding pandemic literature only partially informs our understanding regarding the possible effects of the Covid-19 on prosocial preferences and trust because most of these studies use data collected at different times after the emergence of the Covid-19 (for example: Brañas-Garza et al. 2022) or use pre-Covid-19 data from a sample of same population (for example: Sibley et al. 2020). To precisely identify the impacts of the Covid-19 on pro-social preferences and trust, a panel analysis based on pre and post Covid-19 data is an optimal choice. Furthermore, most of the existing studies (very few exceptions such as Brück et al. 2020) per se do not examine how pro-sociality and trust vary across people who unfortunately contracted virus and those who remain safe (‘specific’ or ‘causal’ effects of pandemic). They examine general influence of pandemic on behavior. While this general analysis might capture some of the effects of ongoing pandemic (such as fear of virus contraction, mental health and psychological challenges, possible economic, social and political spillover effects), it does not reveal how pro-social behavior and trust levels of people who unfortunately contracted virus is affected in comparison to those who remained safe. This analysis is important because it can provide us profounder insights regarding the possible disproportionate effects of pandemic on preferences of infected and non-infected people, which subsequently can have implications for the public policies directed at mediating the effects of pandemic to support the socio-economic development.

This paper addresses the aforementioned gaps in the rapidly growing literature and disentangles the ‘general’ and ‘causal effects’ of pandemic with the help of two studies. In the first study panel data from the Netherlands (LISS panel 2019–2020 by CentERdata) is used to examine the proportion of people donating to different causes as well as trust in other individuals (general trust) in a pre-and-post Covid-19 situation. The high-quality balanced panel data enables to perform a relatively precise analysis for the pre-and-post effects of the Covid-19 and draw conclusions based on a relatively large number of observations. The analysis reveals no significant changes in the proportion of donors as well as general trust level in a pre-and-post Covid-19 situation. The second study dwells further and examines whether the proportion of donors and general trust levels vary across the Covid-19 infected and uninfected people using the LISS panel data from the Netherlands. Unlike study 1, study 2 draws causal inferences for the effect of Covid-19 infections on preferences by using the difference-in-differences (DD) method. The DD analysis leads to insignificant variation in pro-sociality and trust levels among infected and uninfected people in a post-pandemic situation. However, sub-group analysis shows a positive causal impact of infections on trust for respondents above 60 years. Overall, both studies offer evidence that pro-sociality and trust are largely stable.

The study contributes to the literature through several important avenues. First, it provides evidence for the effects of pandemic on prosocial behavior and trust based on a rich and balanced panel data. The high-quality nature of panel data helps to precisely identify whether people modified their prosocial behavior and trust in response to the pandemic. Second, the study performs inter-group comparison of pro-social behavior and trust across people infected and uninfected with virus. This comparison provides initial evidence on the extrinsic effects of the pandemic on preferences and places this study in the broader existing literature that examined the direct effects of exposure to violence and recession on pro-sociality (Voors et al. 2012; Fisman et al. 2015), and the effect of tsunami on trust (Cassar et al. 2017). The findings also contribute to the rapidly growing literature on the effect of Covid-19 on economic preferences (Angrisani et al. 2020; Binder 2020; Brañas-Garza et al. 2022; Brück et al. 2020; Fetzer et al. 2020; Sibley et al. 2020; Shachat et al. 2021; Zhang and Liu 2021 and others). However, the current work is different from these aforementioned studies in two aspects. (a) Several previous studies have examined the effect of extent of exposure to the Covid-19 (such as lockdown duration, local deaths due to virus, previous interaction with an infected person) while the current work (Study 2) focusses on the causal effect of already being infected by virus. (b) The time analyzed in this study is longer (2019–2020) than some of the existing studies (for example, Buso et al. 2020) examined preferences over six weeks after the start of pandemic while Brañas-Garza et al. 2022 examined generosity based on the data collected over a week). Therefore, the methodology and nature of data in this study are not completely comparable to the previous studies.

Third, this work contributes to the ongoing debate on the stability of social preferences across time and their resistance to change due to the real-world shocks. The existing literature on the stability of preferences is polarized with studies arguing in favor (please see review article by Chuang and Schechter 2015) and psychological (for example: Hertwig et al. 2004; Filipski et al. 2019) and empirical studies (for example: Choi et al. 2009; Malmendier and Nagel 2011; Andersen et al. 2019) arguing against preference stability across time and natural shocks. The evidence in this study supports the literature that social preferences are stable over time and amid natural disasters.

The rest of the paper is structured as follows. The second section reports details about panel data and the statistical analysis that examines the general effects of pandemic on pro-sociality and trust. The third section reports details about panel data, examines endogeneity, performs baseline checkups and subsequently examines the prosocial behavior and trust across infected and uninfected people (causal effects) using DD analysis. Last section concludes the paper by discussing important implications.

2 Study 1: general effects of the Covid-19 on pro-sociality and trust

The first study uses data from the LISS (Longitudinal Internet studies for the Social Sciences) panel administered by CentERdata (Tilburg University, Netherlands). The panel is administered to a sample of around 5000 households that are drawn from the population based on a true probability sampling method. The respondents are incentivized for completing the questionnaire that elicits information on a broad range of life aspects.

The first Corona virus case in the Netherlands was reported on February 27th, 2020, the first virus related death occurred on March 6th, 2020 and the first lockdown was imposed on March 15th, 2020.Footnote 4 To perform a pre-and-post Covid-19 analysis for pro-social behavior, data from the survey module “Social Integration and Leisure” implemented in 2019 (wave 12; implemented in October–November) and 2020 (wave 13; implemented in October–November) is used respectively. To perform a pre-and-post Covid-19 comparison for trust, data from questionnaire on ‘Personality’ implemented in 2019 (wave 11; implemented in May–June) and 2020 (wave 12; implemented in May–June) is used respectively. The aforementioned data waves also elicited information for a rich set of demographics, economic, social and religious variables that are used as controls in the regressions. The post-Covid-19 data used in this study is collected well after the emergence of the Covid-19 virus in the Netherlands. The 2019 and 2020 data sets are merged and the analysis is restricted to only those respondents who participated in both waves (balanced panel) leading to a total of 4502 observations for two years. The two dependent variables that measure pro-social behavior and trust are briefly reported in Table 1.

Table 1 Description of dependent variables
Full size table

The dependent variable ‘Donated’ examines monetary donations to a wide range of organizations (complete list of organizations reported in Appendix 1).Footnote 5 The donation space is not restricted to a typical classification (such as humanitarian aid and religious giving) because such a restriction can underestimate the proportion of individuals who donated in the last 12 months. As information regarding the exact amount donated by individuals is unavailable in the survey, possible changes in the amount donated (intensive margins) in a pre-and-post pandemic situation cannot be analyzed. The binary variable ‘Donated’ nevertheless captures the proportion of donors that is more appropriate proxy for prosocial behavior in the Dutch context which is a welfare state and people are simply not used to donatingFootnote 6 as opposed to a country like the US where people donate more frequently.Footnote 7 Another limitation of data for pro-social behavior is linked to the timeline of data collection; the information for pro-social behavior in the last 12 months is collected in October–November in 2020 and therefore partially overlaps with pre-Covid-19 time. As monthly data related to donations is unavailable in the survey, it is impossible to control for the overlapping months through data. One possible confounding effect due to the time overlap is the impact of religious activities in December 2019 that can attract more donors in comparison to other months and hence cause an upward bias in the variable ‘Donated’ measured in 2020. However, Bekkers et al. (2017) report that religiosity has significantly decreased in the Netherlands and this decline explains 40% decrease in the overall Dutch generosity.Footnote 8 Therefore, donations driven by religious element in December are unlikely to have a large impact on donor proportion. Furthermore, the religiosity of respondents in the current study is also very low with almost 40% of the respondents certainly not religious and only about 12% reporting certainly religious (further details in Appendix 2). Therefore, the impact of time overlap is expected to be trivial in the current data.

The dependent variable ‘Trust’ elicits information on general trust in other people. A higher value means people have higher trust towards other individuals. There are 52 respondents that did not answer this question (Trust = − 9 or missing value) and these are excluded from pre-and-post Covid-19 data. Two different sets of control variables are used in the regression analysis. Main controls include essential demographic and economic variables while extended controls include information about religiosity, health condition and general relation with family. Further details about dependent variables are reported in Table 2 while complete data details are reported in Appendix 2.

Table 2 Summary and statistics of panel data
Full size table

The data has relatively balanced proportion of male and female respondents. Average age of respondents is a little bit on the higher side (approximately 59 years) and almost half of the respondents are married. The average family size is 2, a major proportion has college degree (approximately 52%) and a relatively high proportion of respondents has paid work (approximately 44%). A large number of respondents in the sample either do not known their exact monthly gross income or did not share this information. To avoid losing significant number of observations, the income categories reported by respondents are used as a substitute for the exact income. Overall, the data set has diverse representation of people from different backgrounds.

2.1 Statistical analysis

The primary objective is to analyze the general effects of Covid-19 on donor proportion and general trust level. To achieve these goals, panel regressions of the following type are estimated:

$$Y_{it } = \beta_{i} + \beta_{1} D_{t} + \beta_{i} X_{it} + \epsilon_{it}$$
(1)

Where \({Y}_{it}\) is the dependent variable (Donated or Trust) for individual i at time t. The main independent variable of interest is dummy variable \({D}_{t}\) and it is equal to 1 for year 2020 (post Covid-19) and 0 for pre-Covid-19-time frame. A positive and significant coefficient for the main explanatory variable would mean that post-Covid-19 pro-social behavior and trust have increased. The vector \({X}_{it}\) is a set of controls (both main and extended controls) while \({\epsilon }_{it}\) represents the random error term.

Even though dummy variable \({D}_{t}\) intends to capture the effect of Covid-19, it captures other time bound micro and macro effects as well. One possible solution is to use an interaction of \({D}_{t}\) with regional variation in the Covid-19 severity. However, as regional details of the respondents are unavailable, it is not possible to perform the aforementioned exercise. Therefore, \({D}_{t}\) might have contamination caused by other variables and that is why its coefficient captures the ‘general’ effect of the Covid-19 on preferences. The methodology used here however is similar to the rapidly growing literature that uses time dummies to perform a pre-and-post Covid-19 analysis or pre-and-post lockdown analysis to capture the impact of the Covid-19 on preferences (for example: Angrisani et al. 2020; Brañas-Garza et al. 2022; Shachat et al. 2021).

To estimate Eq. 1, either random effects or fixed effects specification can be used. The Hausman and Breusch-Pagan tests are employed to identify whether fixed or random effects specification is appropriate. Both tests suggest that fixed specification is appropriate. The analysis is performed using STATA 16. For brevity reasons, main result is reported in Table 3 while complete results for main and extended controls are provided in Appendix 3.

Table 3 Fixed effects panel regression: general effects of the Covid-19
Full size table

The coefficient for the primary variable \({D}_{t}\) has a positive magnitude in regressions with ‘Donation’ and ‘Trust’ as the dependent variables irrespective of whether primary controls are added (regressions 1 and 3) or both primary and extended controls are incorporated (regressions 2 and 4). However, the coefficients are insignificant in all regressions and hence, no significant ‘general effect’ or ‘time effect’ of pandemic on the proportion of donors and trust levels is observed.

As balanced panel data is used for the current analysis, the results offer relatively precise insights into the general effects of pandemic on behavior and stand in contrast to most existing studies (Brañas-Garza et al. 2022; Brück et al. 2020; Sibley et al. 2020; Shachat et al. 2021) that do not rely on panel data. These results also support the predominant stream of literature that argues preferences are stable over time.Footnote 9

As a last step of stability analysis, a correlation and OLS regression (preferences in 2020 as dependent and those in 2019 as independent variable) is also performed to examine the relation between prosocial behavior and trust in 2019 and 2020. The Pearson correlation coefficient for pro-sociality is 0.413 (p < 0.01) while for trust is 0.722 (p < 0.01). Both coefficients indicate the presence of strong correlation between preferences measured in 2019 and 2020. The regression coefficients are almost similar as well. The complete outcome of this exercise is reported in Appendix 4.

3 Study 2: causal effects of Covid-19 on pro-sociality and trust

Study 1 analyzed the general influence of pandemic on prosocial behavior and trust. Study 2 analyzes a different aspect of pandemic by examination of prosocial behavior and trust across people who unfortunately contracted virus and those who remained safe, and therefore aims to establish a causal relation between Covid-19 infections and preferences using difference-in-differences method. The study 2 also uses data from the LISS (Longitudinal Internet studies for the Social Sciences) panel administered by CentERdata (Tilburg University, Netherlands). The information regarding people who contracted virus and general trust post-pandemic (2020) is extracted from the special questionnaire ‘Effect of the pandemic on social trust’ administered to the LISS panel members in July 2020. The pre-pandemic trust information is obtained from the survey module ‘Personality’ implemented in 2019 (wave 11; implemented in May–June). The information for prosocial behavior is extracted from the module “Social Integration and Leisure” collected in 2019 (wave 12: implemented in October–November) and 2020 (wave 13: implemented in October–November). The two outcome variables are same as those reported in Table 1. This study also uses balanced panel data however, the number of observations in the panel are 2744 and significantly fewer than those used in study 1. The primary reason for a smaller number of observations in study 2 is due to the smaller number of households invited for questionnaire designed specifically to study the impact of the Covid-19 on trust (Survey module: ‘Effect of the pandemic on social trust’).Footnote 10 The data summary and statistics for outcome variables and primary explanatory variables are in Table 4. As controls and extended controls used in study 2 are similar to those in study 1, their summary and statistics are reported in Appendix 5. The data characteristics are similar to the panel data reported in Table 2. The primary independent variable (people infected with Covid-19) indicates that almost 7% of the respondents self-reported they have been infected with virus.

Table 4 Summary and statistics of panel data
Full size table

3.1 Baseline checkups

Before proceeding to the DD analysis, necessary baseline checkups are performed by comparing the characteristics of infected and uninfected respondents at the baseline (year = 2019) and the output for important variables is reported in Table 5 while results for all controls are available in Appendix 6.

Table 5 Baseline checkups for main variables: uninfected versus infected respondents
Full size table

The generosity and trust levels of infected and uninfected respondents do not differ at the baseline. The age composition of infected and uninfected people however differs significantly; relatively younger people are part of the infected group at the baseline. Apart from it, other important characteristics of the sample are similar at the baseline. Overall, the results for baseline checkups in Table 5 show no stark differences among infected and uninfected respondents at the baseline.

3.2 Difference-in-differences (DD) analysis

As primary objective is to analyze the causal effects of the Covid-19 before and after the Covid-19 infections on prosocial behavior and general trust level, a difference-in-differences specification is used to estimate OLS regressions of the following type:

$$Y_{it } = \beta_{0} + \beta_{1} D_{t} + \beta_{2} \;{\text{Infected}} _{i} + \beta_{3} \left( { D_{t} \times {\text{Infected}} _{i} } \right) + \beta_{i} X_{it} + \epsilon_{it}$$

where \({Y}_{it}\) is the dependent variable (Donated or Trust) for individual i at time t. The variable \({D}_{t}\) is binary in nature and equal to 1 for year 2020 (post Covid-19) and 0 for pre-Covid-19-time frame. The variable \({Infected}_{ i}\) is also binary and equal to 1 if the respondent i contracted virus and 0 if the respondent remained safe. The coefficient \({\beta }_{3}\) is the DD coefficient and outcome of interest here. It measures the differential effect on pro-sociality or trust among the infected and uninfected respondents post Covid-19. The vector \({X}_{it}\) is a set of controls (both main and extended controls) while \({\epsilon }_{it}\) represents the random error term. The analysis is performed using STATA 16, output for primary independent variable is reported in Table 6 while results with discussion on controls is reported in Appendix 7 .

Table 6 DID results: infected versus non-infected
Full size table

The regression results show that the coefficient for DD variable is insignificant and indicates that pro-sociality and trust are stable among the infected and uninfected people pre-and-post-pandemic. The results also provide an early evidence regarding the stability of pro-sociality and trust despite exposure to unfavorable Covid-19 infections.

The DD results for pro-social behavior do not support recent findings by Brañas-Garza et al. (2022), Shachat et al. (2021) and Zhang and Liu (2021) who find prosocial behavior changed post-pandemic. The data used in this study however is fundamentally different from the previous studies because the current study compares prosocial tendencies across the infected and uninfected people using DD analysis and essentially uncovers the causal impact of Covid-19 infections on preferences. The DD results for trust are consistent with insignificant general effects of pandemic on trust obtained from panel analysis in study 1. These findings also reinforce results by Brück et al. (2020) that general trust does not differ across Covid-19 infected and uninfected people. In a broader perspective, the current findings are consistent with those by Fleming et al. (2014) that exposure to unfavorable natural disaster (comparison of earthquake affected and unaffected people in Chile) does not change trust levels.

3.3 Robustness checks

Randomized experiments or trials are generally used to analyze causal relations. However, the Covid-19 infections cannot be examined by designing a randomized experiment and panel data can serve as a natural experiment to draw causal inferences about the effects of the Covid-19 on preferences. One possible issue in the examination of causal influence of virus infections on pro-sociality and trust is the endogeneity problem. The endogeneity in the current analysis can be caused through several avenues.

First, as baseline checkups in Table 5 show, there is a significant difference in the age of infected and uninfected respondents. Therefore, as a robustness test an interaction term of infection with age is added to the DD analysis to examine if the results change. For brevity reasons, the primary output is reported in Table 7 (Panel A for Donations and Panel B for Trust) while complete output is reported in Appendix 8 and 9. The coefficient for DD as well as the interaction of infected with age are however insignificant for both prosocial behavior and trust (Table 7: Panel A and B, Regressions 1 and 2). To further examine the impact of virus infection on preferences, a sub-group analysis is performed by dividing the sample into two groups (Group 1: age <  = 60 years; Group 2: Age > 60 years). The DD coefficient remains insignificant in all regressions for donations (Panel A, Regressions 3–6). Similarly, DD coefficient for trust is also insignificant for respondents less than or equal to 60 years old (Panel B, Regressions 3–4). However, there is a positive and significant causal impact of virus infections on people elder than 60 years (Panel B, Regressions 5–6). As a further check based on the median age (62 years) I divided the sample into two groups (Group 1: Age <  = 62 years; Group 2: Age > 62 years). The findings for the DD term however do not change and hence are not reported here.

Table 7 Robustness checks for DD results
Full size table
Table 8 Propensity score matching (PSM) outcomes
Full size table

The sub-group analysis based on the respondent’s age shows that infections have a positive influence on trust. The findings suggest that relatively younger infected people are more resilient to change in preferences while elder infected people have malleable preferences. Overall, the positive causal impact of infections on trust observed for elder people (age > 60 years) is consistent with positive influence of Covid-19 priming on social trust reported by Daniele et al. (2020), positive effect of infections on interpersonal trust reported by Gambetta and Morisi (2020) and the positive impact of lockdown on the institutional trust reported by Sibley et al. (2020). The increased trust of elder people after exposure to the virus can be due several reasons. First, infected elder people would have most likely obtained medical assistance from other people. This interaction with generally stranger medical personnel can cause an upward revision in their trust. Second, the fear of pandemic and the possibility of perishing due to the virus can make infected elder people rely on strangers and ultimately increase their trust levels (Gambetta and Morisi 2020).

Several recent studies (for example: Campos-Mercade et al. 2021; Bargain and Aminjonov 2020) suggest that prosocial behavior and trust play an important role in predicting the preventive behavior against the Covid-19. A recent work of Umer (2022) based on the LISS panel data also shows that prosocial behavior and trust predict preventive behavior in the Netherlands. Importantly, Umer (2022) finds a robust relation between preventive behavior and trust measured in the short-run as well as long-run. Therefore, there is a possibility that pro-social people or those with higher trust in the current study observe preventive measures more frequently, are less likely to contract virus and hence cause endogeneity problem. Moreover, as Covid-19 infections are a natural experiment, the number of observations in the control and treatment groups are very uneven in the current data and this can also cause bias in estimates.

To account for the aforementioned possible endogeneity problems, I use propensity score matching (PSM) to analyze the causal impact of virus infections on preferences by using data for 2020.Footnote 11 I use three different sets of covariates that might be related to the treatment variable infected to approximate a randomized experiment and identify the causal impact of virus contraction on preferences. The first two sets of covariates are identical controls and extended controls used in the DD analysis. The third set of covariates has an additional variable. For the outcome variable donation, I add trust as a covariate. Similarly, for trust as outcome variable, I add donation as a covariate. The addition of these preferences as covariates is to control for the possible endogeneity caused by the influence of these preferences on the preventive behavior and the Covid-19 infections. As donation variable is binary, I use logit while for trust use probit regressions.

The average treatment effect (ATE) based on the PSM analysis shows no causal effect of infections on both prosocial behavior and trust irrespective of the variables used as covariates (Table 8). Even when trust (Regression 3) and prosocial behavior (Regression 6) are added as covariates, ATE remains insignificant. Furthermore, when both trust and donation are added together as covariates, ATE remains insignificant and for brevity reasons not reported here. Overall, PSM outcomes support the DD findings.

4 Conclusions and discussions

This article examined the general effects of the pandemic on prosocial behavior and trust using balanced panel data for 2019–2020 from the Netherlands. There is no significant change in either prosocial behavior or trust pre-and-post Covid-19. The article also analyzed the causal effects of the pandemic on prosocial behavior and trust by examining these preferences among Covid-19 infected and uninfected people using panel data and difference-in-differences (DD) method. The DD estimates also indicate that pro-sociality and general trust levels do not differ across the two groups post Covid-19. Overall, the study disentangles the ‘general’ and ‘causal’ effects of pandemic on prosocial behavior and trust and offers foremost evidence based on the aforementioned insightful comparisons.

Both fixed effects panel regressions (study 1) and DD analysis (study 2) offer robust support that either indirect or direct exposure to the Covid-19 does not significantly alter pro-sociality and general trust and supports the notion that preferences do not change over time and with exposure to disasters. Only for the sub-group of respondents older than 60 years, there is evidence that virus infections lead to a higher trust and suggest that older people might have more malleable trust levels. As the results are obtained from a balanced panel data and from a comparison based on the infected and uninfected people, these findings are relatively robust and precise in comparison to most of the existing studies that analyzed pro-sociality or trust during the pandemic. The results also concord with the recent findings on the stability of risk and time preferences examined by Drichoutis and Nayga (2021) using data collected in a pre-and-post Covid-19 scenario.

The study has some important social and economic implications. The results suggest that despite unprecedented disruptions caused by the Covid-19 to the socio-economic fabric, pro-social tendencies and general trust in the Netherlands did not change. The stability of these preferences advocates that countries such as Netherlands might be able to withstand and quickly fight the negative effects of virus infections on the socio-economic development, leading to a resilient association among people and perhaps a quicker socio-economic recovery. Current findings also provide support to the stability of preferences assumption that underlies most policy analyses as well as welfare examinations performed by the economists (Drichoutis and Nayga 2021).

Finally, some of the caveats of data are discussed here. The information about pro-social behavior for last 12 months for post-Covid-19 analysis is elicited in October–November, 2020. The post-Covid-19 pro-sociality data therefore overlaps for around two to three months with pre-Covid-19 time period, and this overlap might influence results in both study 1 and study 2. However, as the overlap period is relatively narrow, it’s influence on results is expected to be trivial. Second, the data on virus infections is based on self-reports. There is a possibility that either respondents incorrectly reported an illness as Covid-19 or did not report virus infection due to the social stigma. Therefore, an upward or downward bias in self-reported virus infections cannot be ruled out. Lastly, the outcome variables (prosocial behavior and trust) are elicited through unincentivized survey questions and therefore, are not completely comparable to the incentivized outcome variables mostly used in experimental literature. Therefore, one of the reasons for null effect of pandemic on the outcome variables could be their unincentivized elicitation mechanism. Furthermore, as these unincentivized outcomes variables might be susceptible to the social desirability bias, an upward bias in both prosocial behavior and trust used in the current study is possible.