Social implications of Covid-19: Its impact on general trust, political trust, and trust in physicians in China

Abstract

Motivated by current debates over the relationship between epidemic and trust, this paper estimates the short-term effects of the Covid-19 pandemic on general trust, political trust, and trust in physicians in China. Using an individual-level national longitude dataset, results from the Difference-in-Difference estimation show that greater exposure to Covid-19 risks significantly decreased general and political trust among the Chinese population, except for the younger generation (age 8–22). Higher exposure to Covid-19 in malleable ages of trust formation (age 8–22) may worsen individuals’ general trust but improve their trust in local officials and physicians. Results from heterogeneity tests reveal that Covid-19 exacerbated general trust among the vulnerable groups, whereas their political trust was stable.

1. Introduction

The global Covid-19 pandemic has brought unprecedented challenges to human society. The latest statistical data show that the global cases of Covid-19 have reached 500 million and have resulted in 6.19 million deaths (WHO, 2022). As one of the most deadly pandemics in human history, Covid-19 wreaks havoc on societies, economies, communities, and individuals and challenges national governance (Cingolani, 2022). Motivated by recent scholarly contradictions regarding the effects of Covid-19 on political trust (Gozgor, 2021; Oude Groeniger et al., 2021; Schraff, 2021) and the examination of the relationship between epidemics and trust, this paper explores the impact of Covid-19 on trust in China. Given their socially acknowledged roles in mitigating Covid-19 and potential contributions to national governance and collective recovery, we specify trust as general trust, political trust, and trust in physicians.

Trust has been theoretically and empirically evidenced as the glue of societies, economics, and politics (Nikolakis and Nelson, 2019). Greater general trust helps generate higher cooperation, a practical contributing factor to social, institutional, and political stability and economic efficiency (Abascal and Baldassarri, 2015; Nikolakis and Nelson, 2019; Uslaner). These outcomes reciprocally boost citizens' general, specific, and political trust (Nikolakis and Nelson, 2019; Watkins, 2021). Besides general trust, political trust is a vital constituent of national governance and the foundation of legitimate governance (Dellmuth and Tallberg, 2020; Watkins, 2021; Zhang et al., 2022); trust in physicians, given their role in mitigating Covid-19, is a validated form of special interpersonal trust that facilitates health production, support of the government, and social order (Baicker and Finkelstein, 2019; Gilson, 2003).

Historical events offer considerable insights into the deterioration of trust due to their ripple effects. The slave trade in Africa (Nunn and Wantchekon, 2011), the 9/11 terror attack in the United States (Devine et al., 2021), Cultural Revolution in China (Bai and Wu, 2020), and the Spanish flu epidemic (Aassve et al., 2021; Cohn Jr, 2018) all have demonstrated the long-term adverse effects on trust. Notably, abundant literature keeps high engagement in investigating how pandemics affect trust because pandemic outbreaks provide a unique setting with arguably exogenous shock (Devine et al., 2021). The long-term consequences of historical pandemics like the Black Death and the Spanish Flu have thus been documented (Aassve et al., 2021; Alfani and Murphy, 2017; Alfani and Percoco, 2019).

However, despite the essentially variant features of recent epidemics, their consequences are less known. Scholars have estimated how HIV, Ebola, and the BSE crises affect societies, most notably the erosion of trust (Balog-Way and McComas, 2020; Hayden, 2014; Young, 2005). However, as Aassve et al. (2021) have implied, the consequences of a specific pandemic cannot be easily generalized. Moreover, social media, medical interventions like new drugs, vaccines, and information technology may also intervene in pandemic control and risk perception. For instance, Flückiger et al. (2019) found that effective measures to mitigate Ebola outbreaks increased government trust, not general trust. We argue that a similar dispute is quite possible for Covid-19 because available studies have generated contradictory results on the effects of Covid-19 on political trust (Gozgor, 2021; Oude Groeniger et al., 2021; Schraff, 2021).

Integrating data from a national longitudinal survey with China's official data on the spread of Covid-19, we explored how Covid-19 influenced general trust, trust in local officials, and trust in physicians in China. The latest data was collected in June 2020, two months after the first outbreak of Covid-19 was contained (i.e., cases of infection were close to zero in some cities) in China, providing an excellent natural experiment setting. The proxies for an individual's exposure to Covid-19 risks are confirmed cases of Covid-19 infection and death at the regional level, following the specification of Schraff (2021). All provinces had Covid-19 cases, so we used different strategies to construct the control groups. We adopted the classic Difference-in-Difference (DID) design, the continuous DID, and the cohort DID approach following Nunn and Qian (2011), Chen et al. (2020), and Duflo (2001).

Our main results from the 2018–2020 data and the 2010–2020 data suggest that Covid-19 adversely affected general trust and trust in local officials. However, the cohort DID analysis shows that higher exposure to Covid-19 in malleable ages of trust formation significantly impacted trust, specifically worsened individuals’ general trust but improved their trust in local officials and physicians. Robustness checks validate the main results. The adverse effects of Covid-19 are greater among vulnerable groups (women, rural population, and the educationally deprived) in terms of general trust but are weaker concerning trust in local officials and physicians. In other words, Covid-19 may generate different short-term consequences for general trust, political trust, and trust in physicians across populations.

These results contribute to studies documenting the importance of trust and the social and political consequences of Covid-19. Given the crucial role of trust summarized by Nunn and Wantchekon (2011), the current research contributes to understanding the origin of trust and the consequences of the global Covid-19 pandemic. Our results supplement current studies explaining the origin of trust through cultural norms and institutional changes (Abascal and Baldassarri, 2015; Mishler and Rose, 2001; North, 1990; Uslaner, 2002; van Ingen and Bekkers, 2015) during similar pandemics like the Spanish Flu (Aassve et al., 2021). Our results also complement the debate on how Covid-19 affects political trust with new evidence explaining the effects of Covid-19 on general trust and trust in physicians. Finally, we document the positive attitude towards government and physicians among the younger generation in China.

The rest of this paper is organized as follows. Section 2 briefly reviews related literature. Section 3 introduces the method, including the dataset and empirical models. Section 4 reports the main results. Section 5 tests the robustness of our results, and Section 6 is the concluding discussion.

2. Literature review

As more and more studies explore how trust helps mitigate the spread of Covid-19, scholars begin to investigate a potentially reverse causal link (Brodeur et al., 2021). Recent studies pay much attention to the effects of Covid-19 and pandemic policies on political trust, including trust in government, institutions, and officials (Devine et al., 2021). These studies can be further categorized by their focus: (1) policies to combat Covid-19 and their impact on political trust; (2) Covid-19 spread and political trust; and (3) social demographic factors and political trust.

2.1. Policies to combat Covid-19 and their impact on political trust

Though the decline of trust in government has been commonly acknowledged in many countries recently (Devine et al., 2021), the outbreak of Covid-19 did increase trust in government in many countries (Esaiasson et al., 2021; Goldfinch et al., 2021). Some scholars ascribe this increase to such policies as lockdown measures (Schraff, 2021). A potential explanation is the ‘rally-round-the-flag (or rally effects)’ theory (Kritzinger et al., 2021; Mueller, 1970), which states that trust is established through policy salience and performance (Hetherington and Husser, 2012).

Such evidence is mainly from European countries. Bol et al. (2021) used online survey data from 15 Western European countries, noting that lockdown policies increased trust in governments. Studies in the Netherlands and Denmark also found that lockdown measures improved government trust by 18% (Oude Groeniger et al., 2021). A similar increase in government trust was sustained throughout Denmark's whole lockdown period (Baekgaard et al., 2020). Outside Europe, one study from New Zealand reports similar results via the propensity score matching approach (Sibley et al., 2020). However, findings have not been consistent across studies. For example, a study in the Netherlands showed no effect of lockdown policies on political trust (Schraff, 2021).

2.2. Covid-19 spread and political trust

The second strand of literature investigates the increase in political trust as Covid-19 cases arose. Supporting evidence is reported from the Netherlands (Schraff, 2021), Spain (Amat et al., 2020), South Korea (Kye and Hwang, 2020), along with a cross-country study based on data from 750,000 participants in 142 countries (Aksoy et al., 2020). Utilizing panel data from the Netherlands, Schraff (2021) finds that other than lockdown policies, the rising cases of Covid-19 lead to the rally effects. Amat et al. (2020) reported a case in Spain where individuals who have close contacts infected by Covid-19 express less political trust. Aksoy et al. (2020) found that the negative relationship between Covid-19 risks and political trust is more significant among young people aged between 18 and 25. Another study on trust in medical personnel also produced similar results (Thoresen et al., 2021).

Schraff (2021) theorizes this phenomenon as rally effects, i.e., “the intensity of the pandemic rallied people around political institutions.” The author argues that the rising cases of Covid-19 bring collective anxiety, damage routine cognitive judgment, and finally lead to people's rally around governmental institutions. In this case, the author argues that common origins of political trust lose efficacy and that trust may primarily depend on emotional and cognitive evaluations. The author further states that the severity of Covid-19 pushes for increasing political trust. Finally, the author concludes that (1) policies like lockdowns do not account for the increase in people's political trust, and (2) the fast spread of Covid-19 drives the growth of political trust.

2.3. Social demographic factors and political trust

Another strand of studies explores how social demographic factors affect political trust. In that regard, relevant studies did not explore the causal link between Covid-19 and political trust but instead treated Covid-19 as a contextual factor. For example, a global survey from 57 countries found that media freedom reduced political trust overall, but trust levels varied with people's education (Rieger and Wang, 2021). Another study based on data from the U.S., Norway, Australia, and the U.K. investigated whether social demographic factors such as age, gender, education, employment, place of residence, and social media usage influenced political trust during Covid-19 (Price et al., 2021). However, another global study with respondents from 178 countries revealed different results (Gozgor, 2021). These controversies inspired us to select a series of covariates in our empirical models and implement the heterogeneous analysis.

2.4. Trust other than political trust

As probably the most concerned topic in current literature, political trust is understudied. Several studies have discussed the impact of Covid-19 on trust in science or scientists (Oude Groeniger et al., 2021; Sibley et al., 2020), health systems or professionals (Aksoy et al., 2020; Thoresen et al., 2021), religious organizations (Kye and Hwang, 2020), and general trust (Esaiasson et al., 2021; Li et al., 2021). Sibley et al. (2020) found that the group that accepted lockdown had more trust in science and scientists than the control group in New Zealand. Data from the Netherlands showed a 6% increase in trust in science after the lockdown (Oude Groeniger et al., 2021). Accordingly, experiencing a pandemic like Covid-19 may increase trust in the health system or medical personnel (Aksoy et al., 2020; Thoresen et al., 2021). In general, scholars found that political trust and general and interpersonal trust declined with the spread of Covid-19 (Esaiasson et al., 2021; Li et al., 2021). These studies show interesting results of Covid-19 on trust; however, the impact of Covid-19 on political trust needs to be further scrutinized, along with other types of trust.

3. Method

3.1. Data

The individual-level data are from the 2012–2020 China Family Panel Study (CFPS). CFPS is a national representative biennial longitudinal survey covering about 16,000 households in 25 of 31 provinces in China; a detailed description can be found in Xie and Hu (2014). CFPS surveys individual, family, and community-level longitudinal data, using multi-stage probability sampling based on implicit stratification to obtain random information (Xie and Lu, 2015). The design of questions and the implementation of the survey follow the highest academic requirement, and the interview is random and voluntary (Xie and Hu, 2014). Since 2010, the data quality of CFPS has been academically evaluated (Wu and Zhang, 2020; Xie and Jin, 2015; Zhang et al., 2014).

We utilized five waves of data from 2012 to 2020 in this research. The dataset provided us with a potential sample of 165,011 respondents. Finally, the valid observations are from 153,169 to 153,579, depending on specific variables. CFPS routinely asks respondents whether they trust others and how much they trust local officials and physicians on a scale from 1 to 10.1 indicates the least amount of trust, and 10 indicates the highest amount of trust. The variable for general trust is binary: 0 means no trust, and 1 means absolute trust. Given its high academic quality and reputation, this dataset has been widely used in studies on trust in China (Bai and Wu, 2020; Dou et al., 2019; Fan, 2019), including political trust (measured by the trust in local officials) (Bai and Wu, 2020; Yang and Shen, 2021).

The distribution of different types of trust is reported in Fig. 1 (Panel I). The data show a middle level of trust in local officials, with an average score of 5.1. The mean of general trust is 0.6, above the natural average (0.5), and suggests relatively higher social trust. The trust in physicians is higher than the trust in local officials. This data contradicts the statement that trust in physicians in China has collapsed (Tam, 2012; Tucker et al., 2016; Zhu et al., 2018).

Fig. 1.

Quartile distribution of trust and the confirmed cases of Covid-19.

The time variations of trusts are shown in Fig. 2 . We used the polynomial fitting strategy and found that trust trends increased in the past eight years. The only difference is the slopes and magnitudes of increase for different types of trust. Studies have shown that trust is the glue of societies, economics, and politics, and these outcomes can reciprocally boost the establishment of trust (Nikolakis and Nelson, 2019; Watkins, 2021; Zhang et al., 2022). Thereby, the trend shown in Fig. 2 may be primarily associated with better economic performance and social stability in the past decade in China (Liu et al., 2020), particularly the national launch of social, health, and poverty alleviation policies (Li and Wu, 2018; Yang and Shen, 2021; Zhang, 2020).

Fig. 2.

Time variation of trust (2012–2020).

The risk of Covid-19 is measured following the methods by Liu et al. (2021a) and Bartscher et al. (2021). Trust is usually built within administrative regions in China (Bai and Wu, 2020; Li and Wu, 2018; Shi, 2001), where data statistics and implementation of control policies for Covid-19 are implemented (Fang et al., 2020; Jia et al., 2020; Liu et al., 2021a). Because of the uncertain nature of Covid-19, the risk of getting infected (represented by confirmed cases in a community) is the primary threat brought by Covid-19 to every individual, whose safety depends on contact with others and governmental control efforts. Within the same province in China, anti-Covid policies are usually implemented top-down with little room for negotiation, yet the heterogeneity of policy stringency within a province does exist.

Given this background and the cross-regional model of contagious disease proposed by Adda (2016), we assume that Covid-19 risks mainly come from the number of Covid-19 cases within a region. However, as individuals’ risk perception can be subjective and volatile when facing an unknown virus without effective drugs, we measured Covid-19 risk equivalently for everyone from an objective perspective. Consequently, we used the number of confirmed cases of Covid-19 infection and death at the provincial level to approximate the risk that Covid-19 individuals face. We collected data on confirmed Covid-19 infection and death cases from the Chinese national and regional health authorities. Because the 2020 wave of the CFPS survey began in July 2020, the final time point for this data was June 30, 2020.

The Covid-19 confirmed case amount distribution is presented in Fig. 1 (Panel II). The box plot shows the quartile of cases for provinces where more than 50% had cases above 300. Additionally, we plotted the histogram of confirmed cases, as shown in Fig. 3 (Panel I). The data is far from the normal distribution of logarithmic transformation patterns. This result informed us to employ the logarithmic number of confirmed and death cases in our ongoing estimations.

Fig. 3.

Histogram of confirmed cases of Covid-19.

Appendix Table A1 gives the summary statistics for key variables. The surveyed population was averagely aged 44.8, ranging from 9 to 104; about 50% were male. We treated the trust variables as continuous variables and teased out provinces with confirmed cases below the first and second quartile as controls. Then, we calculated the difference between the treatment and control groups. Results in Table 1 show that the trust variables differ significantly between the treatment and control groups.

Table 1.

Differences between the treatment and control groups for trust.

Variable	Cutoff = P25					Cutoff = P50
	Control		Treatment		diff	Control		Treatment		diff
	Obs	Mean	Obs	Mean	t-test	Obs	Mean	Obs	Mean	t-test
General trust	39804	0.55	113555	0.56	0.00	59117	0.55	94242	0.57	0.00
Trust in local officials	39735	5.05	113296	5.18	0.00	59003	5.08	94028	5.18	0.00
Trust in physicians	39816	6.79	113707	6.84	0.00	59151	6.84	94372	6.81	0.01

Notes: 1. P₂₅ refers to the control group containing individuals who lived in provinces where the number of confirmed cases fell within the first quartile; P50 refers to those who lived in provinces where the number of confirmed cases fell within the second quartile; 2. Column diff (t-test) reports the p-value of the independent samples t-test for the means of the control and treatment groups.

3.2. Empirical model

The outbreak of Covid-19 provides an adequate setting for comparing data under natural circumstances via the Difference-in-Difference approach (DID). However, our independent variables are continuous. To identify the control groups, we used several strategies.

First, as elaborated above, we teased out provinces with confirmed cases below the first and second quartile ($P_{25}$ or $P_{50}$) as controls. The estimating equation is equation (1).

$$Trus{t}_{i,d,t}=\alpha +\beta I{\left(cases>P_{25}\right)}_d\times Y_t+X_{i,d,t}^{\prime }\Gamma +{\delta }_d+{\phi }_t+{\epsilon }_{i,d,t}$$ (1)

$i$ indexes individual, $d$ province, and $t$ year. $Trus{t}_{i,d,t}$ denotes one of the three trust variables varying across individuals and years. ${\delta }_d$ is provincial fixed effects adopted to capture province-specific factors that may affect trust. ${\phi }_t$ is the yearly fixed effects containing unobservable factors that might have affected trust but varied across years. $X_{i,d,t}^{\prime }$ is the vector denoting a set of individual-level covariates, such as age, gender, marriage status, years of education, and whether living in a rural area. $I{\left(\#cases>P_{25}\right)}_d$ is the indicative function, in which value 1 means the number of confirmed cases in province $d$ exceeded the $P_{25}$ or $P_{50}$ threshold; $Y_t$ for 2020 is also valued at 1. The coefficient of interest is $\beta$, indicating the estimated effects of Covid-19 on the individual level of trust.

Our second strategy follows Nunn and Qian (2011). We integrated the continuous measurement of the spread of Covid-19, confirmed cases, and death cases on the provincial level to denote the treatment intensity in need to capture more variation in the data. We replaced the variable $I{\left(\#cases>P_{25}\right)}_d$ in equation (1) with $Case{s}_d$ in equation (2). Other than this, all variables hold the same meanings with equation (1).

$$Trus{t}_{i,d,t}=\alpha +\beta Case{s}_d\times Y_t+X_{i,d,t}^{\prime }\Gamma +{\delta }_d+{\phi }_t+{\epsilon }_{i,d,t}$$ (2)

The cultural and institutional explanation of trust indicates that trust may not change dramatically in a short time (Abascal and Baldassarri, 2015; Mishler and Rose, 2001; Nunn and Wantchekon, 2011). Notably, some studies find that trust is the most malleable between ages 8 and 22. Beyond this age range, trust may not change too much (Bai and Wu, 2020; Sutter and Kocher, 2007). In this context, we introduced the classic cohort DID approach, as shown in equation (3), following similar practices by Chen et al. (2020) and Duflo (2001).

As such, we intend to test the impact of Covid-19 on the 8–22 generation with malleable trust compared with age groups whose trust is less changeable. Therefore, we only used the wave 2020 data, and individuals younger than eight were excluded. $I\left(1998\le g\le 2012\right)$ denotes indicative function in which value 1 represents the cohort born between 1998 and 2012. ${\delta }_d$ and ${\lambda }_g$ are regional and cohort fixed effects absorbing provincial and cohort features. ${\Omega }_{d,g}$ is the province-cohort fixed effects to alleviate the potential interaction between cohort trends and Covid-19 intensity that may influence trust.

$$Trus{t}_{i,g,d}=\alpha +\beta Cas{e}_d\times I\left(1998\le g\le 2012\right)+X_{i,g,d}^{\prime }\Gamma +{\delta }_d+{\lambda }_g+{\Omega }_{d,g}+{\epsilon }_{i,g,d}$$ (3)

As shown in Fig. 3, the trend of confirmed cases of Covid-19 is highly skewed. We used the natural log of the confirmed and death cases to estimate equations (2), (3). We also clustered the standard errors at the provincial level. Our dependent variables are categorized, and several estimation strategies can be used. We use the strategy of Nunn and Wantchekon (2011). The OLS estimation is our primary strategy because the respondents’ trust values ranged from 0 to 10. It is categorical but also seems to be continuous.

4. Results

Equation (1). We first present the results from equation (1). Table 2 shows results only with the 2018 and 2020 samples. The classic DID results reveal that Covid-19 may have significantly reduced individual trust compared with 2018. When the control cutoff is the first quartile (P ₂₅), trust in local officials and general trust significantly decreases (Columns (1) and (2)). When the control cutoff is the second quartile (P ₅₀), the two types of trust in columns (1) and (2) also decrease significantly. Impressively, the magnitude of coefficients in Panel B is greater than in Panel A, except for the negative but insignificant coefficient for trust in physicians in column (3). Appendix Table A2 shows the results of equation (1) with the sample from 2012 to 2020. When extending the control groups to 4 waves, we achieved the same results as in Table 2.

Table 2.

The effects of Covid-19 on trusts in China, DID with 2018–2020 data.

	(1)	(2)	(3)
	General Trust	Trust in Local Officials	Trust in Physicians
Panel A: Cutoff=P25
Covid-19 × year	−0.0236* (0.0116)	−0.0664* (0.0357)	−0.0478 (0.0389)
Observations	56,890	56,771	57,019
R2	0.021	0.045	0.038
Panel B: Cutoff=P50
Covid-19 × year	−0.0441*** (0.0103)	−0.1140*** (0.0328)	−0.0269 (0.0412)
Observations	56,890	56,771	57,019
R2	0.021	0.045	0.038
Controls	✓	✓	✓
Regional FE	✓	✓	✓
Year FE	✓	✓	✓

Notes: 1. * = p < 0.1, ** = p < 0.05, *** = p < 0.01.2. The brackets under coefficients are the robust standard error. 3. P₂₅ indicates that the control group consists of individuals who lived in provinces where the number of confirmed cases was lower than the first quartile; P50 indicates that the control group consists of individuals who lived in provinces where the number of confirmed cases was lower than the second quartile.

Equation (2). Table 3 reports results from equation (2) only with the samples from waves 2018 and 2020. Panel A reports the effect of Covid-19 confirmed cases, and Panel B shows results regarding the effect of death cases. Column (1) shows that increased confirmed and death cases significantly reduced general trust. These results are consistent with what has been reported in Table 2. Column (2) reveals that the increase of confirmed cases only reduced trust in local officials, as reported in column (2) of Table 2. We find no significant correlation between the amount of Covid-19 cases and trust in physicians. Not surprisingly, the effect of confirmed cases on trust is greater than that in death cases. Similarly, we ran equation (2) with samples from 2012 to 2020. Results in Table A3 align with Table 3, except for trust in local officials. When the sample is extended to 4 waves, the adverse effects of Covid-19 on trust in local officials become insignificant.

Table 3.

The effects of Covid-19 on trusts in China, continuous DID with 2018–2020 data.

	(1)	(2)	(3)
	General Trust	Trust in Local Officials	Trust in Physicians
Panel A: Confirmed cases from January 2020 to June 2020
Confirmed cases × year	−0.0177*** (0.0050)	−0.0481*** (0.0135)	−0.0176 (0.0151)
Observations	56,890	56,771	57,019
R2	0.021	0.045	0.038
Mean, control group	0.568	5.192	6.804
Panel B: Death cases from January 2020 to June 2020
Death cases × year	−0.0121** (0.0046)	−0.0215 (0.0156)	−0.0177 (0.0145)
Observations	56,890	56,771	57,019
R2	0.021	0.045	0.038
Mean, control group	0.568	5.192	6.804
Controls	✓	✓	✓
Regional FE	✓	✓	✓
Year FE	✓	✓	✓

Notes: 1. * = p < 0.1, ** = p < 0.05, *** = p < 0.01.2. The brackets under coefficients are the robust standard error. 3. Mean of the control group is the mean for the dependent variables of the control group.

Equation (3). Table 4 shows equation (3) results with samples from wave 2020. In column (1), we documented a more significant negative impact of Covid-19 on general trust among the 8–22 cohort. These results are consistent with what has been found in Table 2, Table 3 but demonstrate a more negative effect of Covid-19 on trust in the malleable age of trust formation. However, coefficients for trust in local officials and physicians here differ from Table 2, Table 3 We find that the higher exposure to Covid-19, denoted by confirmed and death cases, significantly improved trust in local officials and physicians among the younger generation. Notably, the effects of death cases are much more significant than confirmed cases across all trust variables.

Table 4.

The effects of Covid-19 on trusts in China, cohort DID with wave 2020 data.

	(1)	(2)	(3)
	General Trust	Trust in local Officials	Trust in Physicians
Panel A: Confirmed cases from January 2020 to June 2020 & cohort 1998–2012
Confirmed cases × treated cohort	−0.4310*** (0.0029)	0.2310*** (0.0157)	0.4050*** (0.0162)
Observations	24,541	24,512	24,634
R2	0.101	0.109	0.116
Mean, control group	0.589	5.726	7.104
Panel B: Death cases from January 2020 to June 2020 & cohort 1998–2012
Death cases × treated cohort	−0.9770*** (0.0065)	0.5240*** (0.0357)	0.9190*** (0.0368)
Observations	24,541	24,512	24,634
R2	0.101	0.109	0.116
Mean, control group	0.589	5.726	7.104
Controls	✓	✓	✓
Regional FE	✓	✓	✓
Cohort FE	✓	✓	✓
Regional × cohort FE	✓	✓	✓

Notes: 1. * = p < 0.1, ** = p < 0.05, *** = p < 0.01.2. The brackets under coefficients are the robust standard error. 3. Mean of the control group is the mean for the dependent variables of the control group. 4. The treated cohort was born from 1998 to 2012, and the control cohort was born from 1901 to 1997.

Heterogeneity. Fig. 4 reports the heterogeneous effect of Covid-19 on trust by gender, urban (living in the urban or rural area), age (above or below 60), and education (primary school and below, middle and high school, college and above). In this analysis, we mainly estimate equation (2). We find that the severity of Covid-19 decreases trust, but these effects vary among different groups. The spread of Covid-19 may have more significant adverse effects on men than women. The effects of Covid-19 spread on those who live in rural and urban areas are complex. Covid-19 has fewer effects on the rural population's trust in local officials and physicians. However, the effect on general trust is less for the urban population than for the rural population.

Fig. 4.

Results of heterogeneous analysis.

The similar complex effects of Covid-19 on different age groups deserve more attention. The findings suggest that the elderly are less affected by Covid-19 in terms of general trust but more affected regarding trust in local officials and physicians. These results align with previous studies that trust among the elderly may be less affected by current events (Mishler and Rose, 2001; Nunn and Wantchekon, 2011). For example, the institutional changes experienced by the elderly at an early age, such as the Cultural Revolution, and the market-oriented reform of the health system, may deteriorate political trust and trust in health care for a long time (Bai and Wu, 2020; Liu et al., 2021b).

In addition to variations in age, gender, and area of residence, it is also helpful to understand the heterogeneous effects of Covid-19 on trust among populations with different education levels. Fig. 4 reveals that general trust among those with higher education is less affected by the severity of Covid-19. However, trust in local officials and physicians is greater among the highly educated than the less. Covid-19 generates the most adverse effects on trust in physicians for the group holding college and higher degrees. The middle and high school education group shows worsened trust in local officials compared to college-and-above degree holders. If we measure education by continuous years, U-shape and reverse U-shape effects can be identified in terms of trust in local officials and physicians.

5. Robustness

5.1. Parallel-trend test

Our empirical strategy in equations (1), (2) has to meet the parallel-trend assumption: the trends in individual trust should not be associated with the severity of Covid-19 in the absence of this pandemic. We have several arguments to support this assumption.

First, the traditional theories of culture and institution on the origin of trust support our assumption. Theories indicate that trust may not change much in a short time without dramatic institutional or cultural change (Mishler and Rose, 2001; Nunn and Wantchekon, 2011; Shi, 2001). Since 2012 when the wave 2012 was surveyed, no dramatic institutional or cultural change like the Cultural Revolution, Great Famine, market-oriented health system reform, and SARS had happened in China before the Covid-19 pandemic.

Second, as we have discussed, the outbreak of Covid-19 is naturally exogenous, and the risks exposed to individuals are approximately random and seemingly identical. Although some studies argue that the elderly are more vulnerable to Covid-19, epidemiological evidence shows that everyone can be infected (Mueller et al., 2020). The severe symptoms of the elderly are associated mainly with their weak immunity. In that case, the threat is primarily from the intensity of exposure to Covid-19. Finally, we argue that the effects of Covid-19 on trust are also wholly exogenous.

Third, we also have empirical support for the parallel-trend assumption beyond the theoretical arguments. Our polynomial fit for trust with the time variation is shown in Fig. 2. It reports an upward trend for the three types of trust from 2012 to 2018. If the parallel trend is not satisfied, decline or fluctuation should be observed. Conclusively, we argue that these empirical results support the parallel-trend assumption.

Fourth, we ran the regular parallel-trend test for DID with the total sample from 2012 to 2020. We primarily run the test based on equation (1). Fig. 5 presents the results where the cutoff of control groups is P ₂₅ (Panels I, II, and III) and P ₅₀ (Panels IV, V, and VI). Panels I, II, and III suggest the parallel-trend tests are passed when the cutoff of control groups is P ₂₅. The tests with the cutoff of control groups as P ₅₀ also are passed, according to Panels IV, V, and VI. The insignificant post-treatment effects shown in Panels III and VI align with the results in Column (3) of Table 2, Table 3

Fig. 5.

The parallel-trend test for DID with the cutoff of the controls being P₂₅ and P₅₀.

Finally, as another argument, we supplement an additional estimation with the synthetic differences in differences (SDID) proposed by Arkhangelsky et al. (2021). SDID combines the advantages of DID and Synthetic Control (SC) methods but overcomes their defects by adjusted unit- and time-specific weight strategy. Notably, the parallel trends can be graphically visualized with the adjusted data (Arkhangelsky et al., 2021). Following Arkhangelsky et al. (2021), we trim our data as a balanced panel dataset and run the SDID estimation. The trends shown in Figure A1 plus the results in Table A7 also support the parallel-trend assumption of our analysis.

The test of parallel-trend assumption for the cohort DID approach shown in equation (3) follows the by-cohort specifications strategy of Chen et al. (2020). The results are presented in Fig. 6 showing the approximate support of the parallel-trend assumption. Because of the long birth range, we categorize the age into 9 groups, where the three groups in the gray region right of the vertical line are treatment cohorts, and the rest are the control cohorts. The coefficients approximately fluctuate or present a stable trend around zero for adult cohorts, implying that nearly no heterogeneous cohort trends in trust are related to COVID-19 in the absence of this epidemic. Individuals older than 22 (born before 1998) may report stable and close to zero effects on trust from high and low regional exposure to COVID-19, while individuals aged 8–22 from regions with more COVID-19 cases have significantly different levels of trust than those from regions with low exposure.

Fig. 6.

Effects of Covid-19 on the trust of different cohorts (Wave, 2020).

5.2. Reverse causality

Another potential challenge of our empirical estimation is reverse causality. Several studies have documented the effects of existing trust on the spread of Covid-19. Trust as social capital may help maintain social distancing and compliance with public health policies (Oronce and Tsugawa, 2021). Existing trust has been evidenced to effectively help control pandemics by curbing human mobility (Bargain and Aminjonov, 2020) and encouraging Covid-19 vaccine acceptance (Latkin et al., 2021; Opel et al., 2020; Petersen et al., 2021). However, we must emphasize that such trust is pre-determined. The trust we used in this study was investigated after the outbreak of Covid-19, just as what has been done by other scholars (Schraff (2021); Robinson et al. (2021); Devine et al. (2021)). Their causal relationship is one-way because of the antecedent outbreak of Covid-19.

5.3. Robustness check

We conducted several analyses to secure the robustness of our results. From the distribution of the confirmed cases, we can learn that the cases in Hubei province, whose capital (Wuhan) is the ground zero of the outbreak, are outliers. Therefore, we regressed equations (1) - (2) by excluding observations from Hubei province to test whether outliers can affect our results. Appendix Table A4 and A5 present results for equations (1) - (2). No difference can be found except for magnitude variation. Our main results in Table 2, Table 3 used 2020 and 2018 data. To test the effects of trust over more years, we extended the sample to the data from 2012 to 2020. Relevant results reported in Tables A2 and A3 also support the 2018 and 2020 data results. What should be highlighted here is that we also regressed the whole sample except for observations from Hubei for reasons as above. Results are consistent with our main results. Finally, as an alternative to measuring Covid-19 spread, we weighted the confirmed cases and deaths by provincial population (per 10,000 population). The results from equation (2) shown in Tables A6 have also validated the robustness of our main results.

6. Concluding discussion

This paper adds to a new and growing literature that seeks to understand the social consequences of Covid-19. Relevant empirical studies focused on how Covid-19 affects political trust via pandemic control policies or the severity of the pandemic. However, current studies generate contradicting results using data usually collected during the outbreak of Covid-19 and pay less attention to general or interpersonal trust. If the pandemic deteriorates the explanatory power of the typical determinant of trust, as Schraff (2021) has argued, the negative link between pandemic and trust may be challenged (Aassve et al., 2021). The next urgent and natural step is to comprehensively investigate the causal effects of Covid-19 on social and political trust.

Our results suggest that the spread of Covid-19 may adversely affect both social and political trust, at least for the population whose trust barely fluctuates, whereas the exception is the younger generation whose trust is malleable (Bai and Wu, 2020; Sutter and Kocher, 2007). High exposure to Covid-19 in the malleable age of trust formation worsened social trust but improved trust in local officials and physicians. Our heterogeneous analysis reveals that the vulnerable groups are more negatively affected by Covid-19 in general trust but less in political trust. Conversely, men, groups with higher education, people aged below 60, and those who live in urban areas may have lower political trust and higher general trust due to Covid-19.

Our results of general trust well follow the cultural origin of trust. Natural disasters may encourage individuals to cooperate in collectivistic countries like China, which fosters stronger social trust. However, the pandemic is not simply a natural disaster. Contagious diseases with higher mortality rates may naturally deteriorate social touch and enlarge distance among people (Aassve et al., 2021). Severe outbreaks bring panic and suspicion but not cooperation (Cohn Jr, 2018). Pandemics have been evidenced to decrease social trust in both the short and long run (Aassve et al., 2021; Alfani and Murphy, 2017; Alfani and Percoco, 2019). Following the theory in cultural anthropology, Nunn and Wantchekon (2011) argue that individuals exposed to an environment with imperfect risk information usually make decisions using the rule of thumb or heuristic strategy. The outbreak of Covid-19 provides a similar environment where an individual cannot screen who is a threat before new information and vaccine are introduced. In early 2020 in China, it was especially so when information on Covid-19 was uncertain. Wearing masks even became a token to eliminate suspicion from others.

Our results of political trust do not support the increase of political trust documented in European countries, except for the younger generation. The 2022 Edelman Trust Barometer reports China as having the highest political trust globally, but we argue that the “rally effects” may emerge for the central government, but not local government (Fang and Zhang, 2021; You et al., 2022). Our results focusing on the local government may be more consistent with the classic origins of political trust: institutions, government performance, and culture. As an exogenous shock, Covid-19 may change institutional arrangements and cultural traditions, but not quickly. The only reasonable explanation for trust variation is government performance. However, previous studies on political trust in China show that the central government enjoys more trust, but the local government does not (Li and Wu, 2018; Shi, 2001). The mitigation of Covid-19 is the same. The central government's ironfisted conduct drives the local government to perform better. The performance of local governments such as Wuhan had been heavily criticized, and several local officials were removed from office.

Our results for trust in physicians are pretty interesting. Trust in physicians depends on the competence and willingness of the physicians to serve the patient's best interests (Newcomer, 1997). It is a special trust developed through interaction between individuals and physicians from three dimensions: cognitive, knowledge, and identity (Pérez-Chiqués and Meza, 2021). In countries where the public health system serves the people, most physicians are street-level bureaucrats and professionals as the “gatekeepers” who provide health services and implement health policies (Gilson, 2003; Spink et al., 2021). Given that the dominator of the health system in China is the public hospitals where physicians are employed as government employees, previous studies argue that the trust in physicians in China has deteriorated in the past decades due to institutional changes (Tam, 2012; Tucker et al., 2016; Zhu et al., 2018). However, our data shows an upward trend of trust in physicians in the past years and no significant adverse effects of Covid-19 on trust in physicians. We argue that the classic theories may explain the unexpected results that trust is more stable for adults. Their experience with the failed health care reform in the 1980–1990s may last for decades in the adult group. The outstanding efforts and sacrifices of medical personnel in the combat against Covid-19 were not enough to touch them.

The most unintended results are lower general trust and higher political trust and trust in physicians for the younger generation. Two channels can explain the outliers. One is that they are in the age of shaping trust toward others, and their ideology is susceptible to change. The other is that they were born around 2000 and grew up with China's booming economic development in the recent two decades. They also have not experienced the Cultural Revolution, the great famine in the 1960s, and the failed health care reform in the 1980–1990s. These particularities might have induced their patriotism and pride in the country, government, governmental employees, physicians, and hospitals. The performances of governments and medical personnel in mitigating Covid-19 trigger their patriotism and pride. These results also echo the learning model of trust, where culture and institutions make more sense for the young (Mishler and Rose, 2001; Schoon and Cheng, 2011).

In conclusion, the short-term effects of Covid-19 on trust may be associated mainly with government performance and culture. The causal relationship between government performance and political trust has been validated (Zhang et al., 2022). Our data collected after the excellent control of the first outbreak may support the results of Kritzinger et al. (2021) but not those of Schraff (2021). The ‘rally effects’ may be more related to government performance, though these effects are usually given to China's central government. The culture here is not the long-standing rules established in the development of human society but the natural human instinct of risk aversion. We further translate it to an information barrier for individuals' decision-making following Nunn and Qian (2011), on account of the unknown origin and severe mortality rate of Covid-19 during the first outbreak. What makes it different from other pandemics is the struggling but great success of health care institutions in pharmaceutical intervention, which might have increased trust in physicians and eliminated perception of risks.

Credit author statement

Ning Liu: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft preparation, Writing- Reviewing and Editing. Guoxian Bao: Conceptualization, Supervision. Shaolong Wu: Conceptualization, Writing – original draft preparation, Writing- Reviewing and Editing.

Funding statement

This work was supported by Humanities and Social Science Foundation of Ministry of Education in China [21YJC630082], Key Research and Development Plan of Gansu Province, China [20YF8GA068], and the Fundamental Research Funds for the Central Universities, China [21lzujbkydx030].

Ethical approval

Not related.

Conflict of interest disclosure

No.

Patient consent statement

N/A.

Handling Editor: W Yip

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

The authors do not have permission to share data.