1 Introduction

As of 2022, the COVID-19 pandemic still poses a catastrophic threat to the world. Aside from potentially causing millions of deaths and severe illness, the pandemic is placing the global economy in peril. While there are a few existing studies on the effect of the COVID-19 pandemic on preferences, this study differs from the existing literature by attempting to identify the effect of prevalence rate of COVID-19 and expected income drop due to COVID-19.Footnote ¹

The current research compares a wide range of preferences, including trust and trustworthiness, risk aversion, and patience (temporal discounting) before and after the onset of the COVID-19 outbreak in China. To this end, we compare the choices of subjects in two sets of incentivized online experiments using (mainly cross-sectional, but some longitudinal) samples drawn from residents across provinces in mainland China. The experiments were conducted before (Study 1, 2019/9–2019/12) and after (Study 2, early 2020/3) the COVID-19 outbreak. We measured trust and trustworthiness using a simplified trust game (Berg, et al., Reference Berg, Dickhaut and McCabe1995). In addition, we assessed the subjects’ risk attitudes and time preferences. We also collected a range of demographic information on factors such as gender, income, occupation, and location, which enable us to investigate whether the effects differ across groups with different socioeconomic status.

We first report the preferences elicited in study 1 and study 2. Then, we report within-subject analysis using a sub-sample of subjects who participated in both Study 1 and Study 2. To attempt to identify the effect of the COVID-19 outbreak on preferences, in Study 2, we use the variation in prevalence rate of COVID-19 across provinces, and we investigate its relationship with the preferences, and the effect of expected income drop due to COVID-19. Since the infection of COVID-19 can be viewed as random, the variation of prevalence rate and expected income change will allow us to identify the possible effects of COVID-19 on the preferences.

Trust and Trustworthiness. In the current study, we test two competing hypotheses on trust and trustworthiness. First, the COVID-19 can be viewed as a collective threat and the threat may enhance co-operation (Boyd & Richerson, Reference Boyd and Richerson2009; Brewer, Reference Brewer2004), and as such may enhance people’s trust and trustworthiness. We may consider the prevalence rate as a proxy of the magnitude of the threat. Accordingly, we hypothesize that subjects in provinces with higher prevalence rate would be more likely to be trusting and trustworthy. Second, the COVID-19 pandemic will induce a negative income shock whereby most people will expect their future income to decline, which may evoke more self-protective than prosocial behaviors, and hence leading to lower trust and trustworthiness. Thus, it is possible that subjects expecting a negative income shock (i.e., those expecting a lower income level or a drop in income) would be less likely to be trusting and trustworthy.Footnote ² Note that trust and trustworthiness are two distinct types of social preference that differ from other measures of social preference such as altruism (which can be measured via a dictator game where the dictator decides the allocation of a fixed amount of money between himself/herself and the receiver); trusting another person involves risk that the trustee may take all profit and not reciprocate any. By contrast, trustworthiness measures the reciprocity tendency if one plays the role of the trustee.

Some studies have found that collective threats enhance in-group cohesion and cooperation, presumably because these behaviors can help groups to weather adversity and enhance their survival (Boyd & Richerson, Reference Boyd and Richerson2009; Brewer, Reference Brewer2004; for a review, see Cheon & Hong, Reference Cheon, Hong, Chiao, Li, Seligman and Turner2015). Consistently, a longitudinal study (Chan et al., Reference Chan, Wang, Zuo, Chiu, Liu, Yiu and Hong2021) has found that increases in national identification during COVID-19 outbreak in China brought about greater compliance to disease preventive behaviors. As such, we expect that the pandemic as a collective threat induces more pro-social behavior by enhancing in-group cohesion and cooperation in China. Therefore, this stream of research suggests that the participants in this study will display greater trust and trustworthiness after the onset of the COVID-19 pandemic in comparison to before. We use prevalence rate as a proxy on the magnitude of the collective threat. We hypothesize that the higher is the prevalence rate, the more likely subjects will be trusting and trustworthy.

In an experiment conducted in rural Thailand, Cassar et al. (Reference Cassar, Andrew and Carl2017) found that there were high levels of trust and trustworthiness after the 2004 tsunami.Footnote ³ They used satellite imagery to identify the tsunami-affected villages and then recruited participants based on their responses to questions about whether they or their household had suffered financial damage due to the tsunami, and whether any of their family members were injured or killed in the tsunami. However, they did not have data on the participants’ attitudes before the tsunami.

As noted, there could be an opposing influence as well. The COVID-19 pandemic will likely induce a negative income shock. In study 2, we measure people’s belief on the effect of COVID-19 on their income. We hypothesize that people who expect their future income to decline will be less prosocial, and hence less trusting and trustworthy.Footnote ⁴

In short, if the outbreak induces in-group cohesion, and this effect is higher than the effect of negative income shock, then we may observe more trust and trustworthiness.

Risk Aversion. Previous studies have shown that natural disasters led to increase in risk aversion. For example, Cameron and Sahah (2015) found that villagers who have experienced earthquakes or floods are more risk averse than those who have not had such an experience. Consistently, Cassar et al. (Reference Cassar, Andrew and Carl2017) also found high individual risk aversion among the victims after the 2004 tsunami. In an experiment conducted in Latin America, Van Den Berg et al. (Reference Van d, Marrit and Kees2009) found that experiencing natural shocks makes people more risk averse in both the short run and long run. Likewise, the COVID-19 pandemic may lead to an increase in risk aversion. The COVID-19 pandemic has clearly generated fear, and studies (e.g., Lerner & Keltner, Reference Lerner and Keltner2001) have shown that fear induces people to be more risk averse. Raghunathan and Pham (Reference Raghunathan and Pham1999) showed that anxious individuals were more likely to favor low-risk/low-reward options in gambles.

Alternatively, the COVID-19 outbreak may have induced people to become more risk taking. Bchir and Willinger (Reference Bchir and Willinger2013) found that in areas exposed to the risk of lahar flows, poor people were more risk seeking and more impatient than were those in unexposed areas. However, this effect was absent among high income individuals. Moreover, research has shown that bad moods can lead to more risk-taking behavior (Leith & Baumeister. Reference Leith and Baumeister1996). Given that the pandemic could depress people’s mood, people may become more risk seeking after the onset of the pandemic. Hanaoka et al. (Reference Hanaoka, Shigeoka and Watanabe2018) found that after the 2011 Great East Japan Earthquake, the subjects were more risk seeking. However, their survey did not cover social preferences and time preferences. In another study, Eckel et al. (Reference Eckel, Mahmoud and Rick2009) found that in a sample taken shortly after the event, female evacuees from Hurricane Katrina were more risk seeking, but the participants were mildly more risk averse a year later. In sum, the research on the effects of catastrophic events on risk aversion is inconclusive and seem hard to explain as they might be confounded with other variables such as government’s subsidies and aids.

In the current study, we argued that there are two possible aggravating effects of COVID-19 on risk aversion. First, the expected negative income shock (expecting a drop in income or a lower income level) may induce higher degree of risk aversion as these people need to prepare for the income drop. Thus, subjects with more pessimistic income expectations could be more risk averse. Second, the pandemic could induce fear, and focus on bad outcomes that may occur, and hence leading to more risk aversion. We may view the prevalence rate as a proxy for the fear. Thus, we hypothesize that subjects in provinces with higher prevalence rate will be more risk averse.

Temporal Discounting. Research suggests that negative affect, such as emotional distress and fear, can lead people to indulge their immediate impulses (Tice, et al., Reference Tice, Bratslavsky and Baumeister2001), while positive affect reduced people’s discount rates (Ifcher & Zarghamee, Reference Ifcher and Zarghamee2011). In a survey of Sri Lankan wage workers, Callen (Reference Callen2015) found that exposure to the Indian Ocean Earthquake tsunami increased the respondents’ patience, which was measured using consumption choice decisions. The author attributed the change to a change in preference, rather than to changes in the economic environment. Cassar et al. (Reference Cassar, Andrew and Carl2017) found that the subjects were less patient after the tsunami. Overall, the existing research on the effects of catastrophic events on patience is inconclusive and seem hard to explain.

In our view, the pandemic could induce two opposing tendencies on individuals’ time preference (Frederick et al. Reference Frederick, Loewenstein and O’donoghue2002), which refers to the preference for present utility over future utility. First, the COVID-19 outbreak may induce emotional distress and fear, and hence leading to lower patience (i.e., higher discount rate). Thus, subjects in provinces with higher prevalence rate could have higher discount rates. Second, the expected income drop due to COVID-19 may make people less impulsive (as they need to save more for the future) and more self-controlled (Ameriks et al., Reference Ameriks, Caplin, Leahy and Tyler2007; as they delay gratification of a smaller award sooner for a larger award later), hence leading to lower discount rate. Thus, subjects with an expected income drop or a lower expected income level could have lower discount rates.

Related literature on Effect of COVID-19 on Preferences. There is by now a small literature on the effect COVID-19 on preferences. A study closely related to our study is that of Hellman et al. (Reference Hellmann, Dorrough and Glöckner2021), which used dictator games to investigate prosocial behavior in Germany during the COVID-19 pandemic. They found that prosocial behavior increased when vulnerability of others to virus increased. This result suggests that trust and trustworthiness will likely be positively correlated with prevalence rate, which is consistent with our expectations.

In addition, three other studies have investigated effect of the outbreak on preferences using Chinese subjects, namely Bu et al. (Reference Bu, Hanspal, Liao and Liu2020), Lohmann et al. (Reference Lohmann, Gsottbauer, You and Kontoleon2020), and Shachat et al. (Reference Shachat, Walker and Wei2021). Shachat et al. (Reference Shachat, Walker and Wei2021) compared the pro-social behavior and risk taking of Wuhan university students before and after the outbreak. They used a trust game to measure trust and trustworthiness and found a decrease in trust post lockdown, while there was no significant effect on trustworthiness. By contrast, Lohmann et al. (Reference Lohmann, Gsottbauer, You and Kontoleon2020) conducted survey experiments on students of Renmin University in China before and after the outbreak and found no significant difference in trust. Both Shachat et al. and Lohmann et al. used random pairing of subjects from the same university, who share the same social identity and were aware of it; the degree of trust and trustworthiness among Wuhan university students (in-group members) might be higher than among strangers (out-group members). Bu et al. (Reference Bu, Hanspal, Liao and Liu2020) conducted a survey with graduate students at the Wuhan University of Science and Technology before and after the outbreak. They found that planned risk-taking (which refers to whether the subject plans to take more or less risk in the coming year compared to the last year) decreased. Shachat et al. (Reference Shachat, Walker and Wei2021) found lesser risk aversion after the outbreak. Lohmann et al. (Reference Lohmann, Gsottbauer, You and Kontoleon2020) found no significant effect. In short, the findings from these studies were mixed and inconclusive. The current study attempts to examine the issue using a sample beyond college students.

For studies conducted in other countries, there is mixed evidence on the effect of the COVID-19 pandemic on risk taking and time preference. Angrisani et al. (Reference Angrisani, Cipriani, Guarino, Kendall and Ortiz de Zarate2020) found no significant effect of the pandemic on risk taking and time preference with university students and professional traders in London. Drichoutis and Nayga (Reference Drichoutis and Nayga2021) found no significant effect on risk and time preference with undergraduates in Greece, while Harrison et al. (2020) found higher risk aversion but stable time preference with US undergraduate students. Glöckner et al. (Reference Glöckner, Dorrough, Wingen and Dohle2020) investigated perception of COVID-19 inflection risk during the early and later outbreak of COVID-19 in Germany. They found that risk perception change over time (inversed U-shape) with the development of the outbreak.

The current study differs from the above literature on the effect of COVID-19 on social preferences. First, regarding trust and trustworthiness, we are not aware of other studies except ours, Lohmann et al. (Reference Lohmann, Gsottbauer, You and Kontoleon2020), and Shachat et al. (2021) that used the trust game.Footnote ⁵ In comparison with Lohmann et al. (Reference Shachat, Walker and Wei2020), and Shachat et al. (Reference Shachat, Walker and Wei2021), our design has the advantage that subjects do not belong to the same school. Second, we measured the subjects’ expectations about the change in their own income, which allows us to test whether the differences observed in preferences were associated with perceived changes in income. To the best of our knowledge, this is the first study to examine the effects of a shock on people’s preferences that takes their expectations about income into account. Third, we investigate the correlations between disease prevalence with social preferences. Fourth, we include within-subject analysis with 54 subjects who participated in both study 1 and study 2. Although the size of this subsample is small, the results can be compared with those found in the large cross-sectional samples. Finally, our study includes subjects residing in different provinces in China, and is thus unlikely to be driven by location-specific factors.

The main findings can be summarized as follows. Taking together the results of the analyses, we found that after the COVID-19 outbreak, the subjects were more trustworthy than before the outbreak. There were mixed results on trusting in the comparison of study 1 and 2, and in the within-subject analysis. We found that trust and trustworthiness were positively correlated with the prevalence rate, a result consistent with the contention that collective threats enhance in-group cohesion and cooperation (Boyd & Richerson, Reference Boyd and Richerson2009; Chan et al., Reference Chan, Wang, Zuo, Chiu, Liu, Yiu and Hong2021; Brewer, Reference Brewer2004; Cheon & Hong, Reference Cheon, Hong, Chiao, Li, Seligman and Turner2015). We only found a significant negative correlation between trust and expected income drop (in 5 years), but no significant correlations between trust or trustworthiness and rank of expected income change (i.e., the expected income change relative to other subjects such that the lower the rank, the more pessimistic the expected income change).

We found that the subjects were more risk averse after the outbreak. In study 2, we found a positive correlation between degree of risk aversion and prevalence rate, possibly because fear make people more risk averse, although we did not measure participants’ fear directly. Further, as predicted, subjects with more pessimistic income expectation showed lower valuation in the risky lottery task, and were thus more risk averse.

The subjects had higher average discount rates after the outbreak, possibly because emotional distress reduces people’s self-control (Tice et al., Reference Tice, Bratslavsky and Baumeister2001) although we did not measure participants’ self-control directly. We also found a positive correlation between average discount rate and rank of expected income change. These results support our hypothesis that prevalence rate and expected income decline are two important channels affecting time preference.

The rest of this report is organized as follows. Section 2 reports the experimental design, and Section 3 reports the experimental results. Section 4 concludes.