The Meaning and Measurement of Political Trust

The study of public attitudes towards governmental institutions has occupied a central role in survey research. The wealth of research on the topic has provided valuable theoretical and empirical insights into the nature of public trust and confidence in government. David Easton’s (Reference Easton1965) conceptualization of specific and diffuse regime support has occupied a central place in the study of political trust. The former embodies attitudes towards political authorities and public policy whereas the latter represents the political system or regime more broadly. Specific support concerns the politics of the day, who is in office, and what they are doing; it is mercurial and apt to change with new leadership and shifting policy priorities. Diffuse support, on the other hand, is a long-standing orientation, what Easton refers to as a “reservoir” of goodwill towards the regime, that is relatively stable and rooted in early political socialization.

In one of the most influential scholarly exchanges about political trust, Miller (Reference Miller1974) and Citrin (Reference Citrin1974) offered different interpretations of declining trust in the United States government drawing on Easton’s conceptualization. Miller (Reference Miller1974) interpreted the decline as a loss of confidence in the political regime, an erosion of diffuse support, whereas Citrin (Reference Citrin1974) interpreted it to be dissatisfaction with incumbent political leaders, the waning of specific support. The emphasis on theory and measurement showcased in the Miller-Citrin exchange was a forerunner to the rich literature addressing these issues in the study of political trust (e.g., Cook and Gronke Reference Cook and Gronke2005; Denk and Christensen Reference Denk and Christensen2016; Doorenspleet Reference Doorenspleet2012; Fisher, Van Heerde, and Tucker Reference Fisher, Van Heerde and Tucker2010; Hibbing and Theiss-Morse Reference Hibbing and Theiss-Morse2002; Hetherington Reference Hetherington2005; Marien Reference Marien, Zmerli and Hooghe2011; Mishler and Rose Reference Mishler and Rose1997; Norris Reference Norris1999; Schneider Reference Schneider2017).

Despite the wealth of knowledge produced by self-reported measures, they can be limited since people are sometimes unwilling (e.g., social desirability) or unable (lack of self-awareness) to answer survey questions. Although Citrin (Reference Citrin1974) focused primarily on how the decline of political trust in the United States represented a loss of confidence in leaders, he also suggested that political distrust might be a relatively harmless form of venting frustration at the political system. In the aftermath of Watergate and Vietnam, admitting that one trusts the United States government might make one appear naïve so responses to questions about political trust may, in part, be biased by social desirability.

Whereas Americans may be reluctant to express trust in government so as not to appear naive, the Chinese people may be reluctant to report distrust because they fear punishment by the government. Many Chinese criticize the government on the Internet (King, Pan, and Roberts Reference King, Pan and Roberts2013), but the extent to which such criticism is allowed varies across time and domain (Gueorguiev and Malesky Reference Gueorguiev and Malesky2019). In the past decade, the Chinese government has further tightened the scope for expression, banning “improper discussions,” insisting “media make the party their surname,” and detaining government critics and rights lawyers (Shirk Reference Shirk2018; Yuan Reference Yuan2021; Zhao Reference Zhao2016). The ever present—and in fact increasing—repression in Chinese society thus undermines popular and scholarly perceptions about the high levels of trust in government. To be clear, the skepticism is not due to expectations that trust should be low but rather that it might be artificially high.

The nature of trust in government in China has motivated a good deal of research into whether social desirability bias, particularly fear of retribution, inflates expressed regime support. The findings are somewhat mixed. Many studies have been unable to find evidence of substantial overreporting or the claim that fear is a primary motivation (e.g., Lei and Lu Reference Lei and Lu2017; Shi Reference Shi2001; Tang Reference Tang2016). For instance, Shi (Reference Shi2001) found anemic correlations between trust in government and fear of political persecution. Tang (Reference Tang2016) offers a key comparison by reporting that responses given to a question about whether Chinese respondents would support the government even if it is wrong are roughly comparable to the responses found in other countries, including democracies. Another key comparison is between levels of trust in the local and central governments wherein scholars have routinely found lower trust in the local government (Dickson Reference Dickson2016; Li Reference Li2016; Tang Reference Tang2016). Scholars have also used experiments to test whether responses to political trust questions are biased by fear of retribution. Lei and Lu (Reference Lei and Lu2017) randomly assigned participants to either a standard face-to-face interview (control) or a treatment wherein participants were told the survey was sponsored by the CCP, the expectation being that people would be less forthcoming in expressing distrust in the latter case. Yet participants in the CCP treatment did not exhibit a significantly higher non-response rate nor report higher trust in China’s political system than participants in the (normal interviewer) control condition. Drawing on affect transfer theory, Stockmann, Esarey, and Zhang (Reference Stockmann, Esarey and Zhang2018) found no evidence that priming participants with the central government produced a fearful response in evaluations of a non-political advertisement. Across a variety of studies using different methodologies, a good deal of evidence suggests that the Chinese people are willing to answer politically sensitive questions in a truthful manner.

Yet there is also evidence that fear and social desirability might play a role in survey responses about the Chinese government. Two list experiments, respectively by Li, Shi, and Zhu (Reference Li, Shi and Zhu2018) and Robinson and Tannenberg (Reference Robinson and Tannenberg2019), found that around twenty-five percentage points fewer respondents trust the regime in indirect questioning than in direct questioning. Tang’s (Reference Tang2016) list experiment, however, found misreporting affected about only 4% of respondents. While not directly measuring political trust, Jiang and Yang (Reference Jiang and Yang2016) found that, after a high-profile political purge, responses to politically sensitive and insensitive survey items diverged, which they interpret as evidence of preference falsification. Li (Reference Li2016) found that people who trust the central government more than local governments have significantly stronger latent doubts about the former. Even though these studies nevertheless find majority trust in government, questions about the amount of regime support and political trust in China persist.

Implicit Trust in Government

Since the literature on political trust in China has been preoccupied with questions of intentional misreporting it has largely neglected another limitation: how people can be insufficiently introspective or self-aware to self-report attitudes. An extensive literature has helped establish dual process theories of the mind (e.g., Banaji and Greenwald Reference Banaji and Greenwald2013; Kahneman Reference Kahneman2011; Pérez and Riddle Reference Pérez, Riddle and David2020). The intuitive mind makes fast, often pre-conscious, involuntary judgments whereas the reflective mind makes slow, conscious, controlled judgments. Scholars have labeled the automatic attitudes that arise from the intuitive mind as implicit and the purposeful attitudes that arise from the reflective mind as explicit.

Implicit and explicit attitudes are essentially different types of thinking. Implicit cognition involves lower-order, rudimentary associations rooted in past experiences (developmental events) and culture whereas explicit cognition reflects more recent events (Greenwald and Banaji Reference Greenwald and Banaji1995). Although it is possible for implicit attitudes to change, they are much less flexible than explicit attitudes. Implicit attitudes are akin to mental habits whereas explicit attitudes are much more adaptable to new information. Since implicit attitudes are automatically activated by familiar objects, they “facilitate one’s explicit thinking by quickly providing working knowledge about previous decisions, judgments, or evaluations that resemble the current context demanding one’s more effortful thoughts” (Perez Reference Pérez2016, 27).

Since all attitudes are unobserved, both implicit and explicit attitudes face notable measurement challenges (see Pérez Reference Pérez2013). Although explicit attitudes are measured directly with self-reports, their accuracy can be problematic when people are unwilling to honestly answer some types of questions (due to sensitivity bias), or they may lack self-awareness. Indeed, the use of list experiments to study trust in government in China assumes that participants will not provide honest responses because they fear government retribution (Li, Shi, and Zhu Reference Li, Shi and Zhu2018; Robinson and Tannenberg Reference Robinson and Tannenberg2019; Tang Reference Tang2016). However, the cognitive process underlying the list experiment is a reasoned, conscious decision that is markedly different from the automatic, gut-level response assumed to underly implicit attitudes.

A formidable challenge to the study of implicit attitudes has been how to capture them since they are automatic and may reside outside of conscious awareness. To address this challenge, psychologists have developed tasks that measure attitudes indirectly such as the Implicit Association Test (IAT) (Greenwald, McGhee, and Schwartz Reference Greenwald, McGhee and Jason1998). The IAT requires participants to rapidly match attributes (e.g., good and bad) to contrasting category items (e.g., white and black), akin to playing a fast-paced video game. The faster the match between a category and an attribute, the stronger the association. In our study, respondents faster at matching government with trust than distrust would hold an implicit trust in government. Since the IAT requires an immediate response, respondents are largely unable to censor themselves thus precluding social desirability effects (Rudman Reference Rudman2004). Although the IAT was initially developed to measure racial bias, it has been used on a wide variety of topics in the study of politics including gender and voting behavior (Mo Reference Mo2015), the nature of party identification (Hye-Yon Lee et al. Reference Lee, Amber, Hawkins and Theodoridisn.d.; Theodoridis Reference Theodoridis2017), religion and voting behavior (Albertson Reference Albertson2011), and perceptions of courts (Hansford, Intawan, and Nicholson Reference Hansford, Intawan and Nicholson2018), to name only a few. In research closest to our own, studies have found that implicit trust in democracies is routinely higher than explicit, self-reported trust (Intawan and Nicholson Reference Intawan and Nicholson2018; Murtin et al. Reference Murtin, Fleischer, Siegerink, Aassve, Algan, Boarini, González, Lonti, Grimalda, Vallve, Kim, Lee, Putterman and Smith2018).

Since people hold both implicit and explicit attitudes towards the same (familiar) object it is possible that they may be of two minds. Indeed, the impetus behind the IAT was to examine attitudes on sensitive topics in which people either censor themselves due to social desirability bias or because of a lack of self-awareness. The most prominent implicit attitudes research has focused on racial attitudes, a sensitive topic in the United States. In these studies, there is typically a significant divergence between self-reported or explicit attitudes (people profess racial equality) and implicit attitudes (people also prefer white faces to Black faces) (Banaji and Greenwald Reference Banaji and Greenwald2013). Yet implicit and explicit attitudes need not diverge. For instance, scholars have found that implicit and explicit party identity are strongly related (Theodoridis Reference Theodoridis2017). In short, the association between implicit and explicit attitudes depends on the concepts under investigation (Kurdi, Ratliff, and Cunningham Reference Kurdi, Ratliff and Cunningham2021).

Whether people are of one or two minds about a topic has important consequences for opinion and behavior. Since implicit attitudes are spontaneously expressed, they represent initial judgments. These snap judgments or intuitions may structure opinion and behavior (Kahneman Reference Kahneman2011; Perez Reference Pérez2016). Explicit attitudes represent deliberative and controlled thinking, the types of attitudes that are reported in surveys. If implicit and explicit attitudes on a topic converge, they are mutually reinforcing, and the person is of one mind. On the other hand, if implicit and explicit attitudes meaningfully diverge, meaning that the person is of two minds about the topic, responses are conditional on the degree to which people engage in thoughtful, deliberative reasoning. If a person responds effortlessly, with little to no thought, the default response will likely follow implicit attitudes whereas an effortful, thoughtful response is likely to follow explicit attitudes.

In the case of political trust, if a person is of one mind about government, we should expect her to be trusting of the regime. Yet if the person is of two minds about government (e.g., they are implicitly trustful but explicitly distrustful), implicit trust could be undermined by explicit distrust if the person was to give an effortful, considered response. The empirical implication is that the implicit measure is uncontaminated by explicit cognition since it is, by design, a pre-conscious measure. On the other hand, as mentioned, self-reported measures are likely to be informed by implicit attitudes, although ultimately, they reside within the domain of conscious, more thoughtful answers. As discussed in the next section, we offer rival hypotheses on whether implicit trust will converge or diverge from explicit trust.

Given that the IAT measures the association between categories and attributes, one might question whether it is measuring familiarity rather than an attitude. In other words, faster associations between some categories and attributes are taken to mean a stronger preference for one category over another but it might also be indicative of greater category familiarity. Since this is a question of construct validity, early research on the IAT examined the issue carefully. In a variety of tests that accounted for familiarity (either controlling for it or using unfamiliar stimuli), these studies largely ruled out familiarity (see Dasgupta, Greenwald, and Banaji Reference Dasgupta, Greenwald and Banaji2003). Furthermore, some studies have even reported participants showing greater favorability toward less familiar stimuli from an ingroup than more familiar stimuli from an outgroup (Rudman et al. Reference Rudman, Greenwald, Mellott and Jordan1999).

The Correlates of Implicit Trust

Aside from a recent study about university students’ implicit preferences for the central government versus university leaders (Zhou, Tang and Lei Reference Zhou, Tang and Lei2020), we are unaware of any research on implicit political attitudes among the general public in China. As a first step to measure implicit trust in government, our focus is the general government system, and we leave the potential difference between China’s central government and local governments to future research. This seems appropriate given China’s unitary and authoritarian political system, particularly since the sweeping centralization and even personalization of power in the past decade (Economy Reference Economy2018; Shirk Reference Shirk2018). For example, local officials are now judged by how well they uphold the central leadership and General Secretary Xi Jinping’s authority (the “Two Upholds” or liangge weihu) and are forbidden to have “improper discussions of central policies” (wangyi Zhongyang). For these reasons, we focus on political trust in the whole system but recommend that future research examine constituent parts.

Is the Chinese public implicitly trustful or distrustful of government? We offer rival expectations, one anticipating negative implicit trust and the other positive. The basis of negative trust has already been discussed since it shares the same root causes thought to motivate people to overreport trust; namely living under an authoritarian regime that controls, sometimes in cruel fashion, daily political, social, and economic life. Simply put, government repression, even if not directly visited upon the individual, orients the citizen in opposition to government, potentially engendering a deep-seated, visceral distrust.

The expectation that we will find positive implicit political trust is rooted in research on political socialization. Trust in government is an attitude acquired early in life (Easton and Dennis Reference Easton and Dennis1969). In the United States, Jennings and Niemi (Reference Jennings and Niemi1968) found high school students to be more trusting in government than their parents, attributing the greater trust to an emphasis on civics education that recedes as political awareness grows across the life span. Yet, as Intawan and Nicholson (Reference Intawan and Nicholson2018) show, Americans largely retain an automatic, implicitly held trust in government. Such a process of fostering positive implicit trust in government is likely to be even more pronounced in China. The Chinese educational system provides highly positive and idealized portrayals of the political system, starting with civics lessons in primary and secondary school and continuing with political ideology education in universities (Cantoni et al. Reference Cantoni, Chen, Yang, Yuchtman and Zhang2017; Kennedy Reference Kennedy2009). Outside the classroom, long-term exposure to the regime’s “thought work” through state-controlled media and socialization in state-sanctioned political participation also make likely that a trustful orientation toward government inhabits the minds of the Chinese people (Repnikova and Fang Reference Repnikova and Fang2018; Shambaugh Reference Shambaugh2007).

As a window into the origins of implicit trust, and whether they differ from explicit trust, we examine the correlates of each. The research on (explicit) trust in government suggests factors related to political socialization (e.g., Almond and Verba Reference Almond and Verba1963; Cook and Gronke Reference Cook and Gronke2005; Easton and Dennis Reference Easton and Dennis1969) may also be correlates of implicit trust. Given their deeper and sustained exposure to these political lessons, we expect those with higher education to have greater implicit trust. Similarly, we expect to find this relationship among urban dwellers since, in communist systems, the state has a deeper penetration in urban areas than in rural areas and individuals living in cities are more heavily exposed to government messages (Jowitt Reference Jowitt1992; Pop-Eleches and Tucker Reference Pop-Eleches and Tucker2020). This may especially be the case in China after the dismantling of agricultural collectivization since the late 1970s. In line with previous research, we expect Confucian values (Shi Reference Shi2001; Tang Reference Tang2016) and perceptions of government performance (Guriev and Treisman Reference Guriev and Treisman2020; Shi Reference Shi2001; Tang Reference Tang2016; Wu, Yang, and Chen Reference Wu, Yang and Chen2017) to also play important roles in fostering explicit trust; we do not have expectations about their role in fostering implicit trust.

Finally, since social desirability may affect the willingness to report political trust, both implicit and explicit, we examine self-monitoring, the degree to which individuals alter their behavior to make a positive impression on others (Snyder Reference Snyder1974). Self-monitoring has been used to account for social desirability bias in research on racial attitudes (Berinsky Reference Berinsky2004) and party identification (Klar and Krupnikov Reference Klar and Krupnikov2016). Intawan and Nicholson (Reference Intawan and Nicholson2018), however, did not find differences in explicit (self-reported) trust between high and low self-monitors suggesting that low trust in government in the United States is not influenced by social desirability. Since the motivation for underreporting political trust in the United States and overreporting trust in China are believed to stem from different sources, the desire to not appear naïve versus fear, respectively, we include self-monitoring in our study. Nevertheless, we only expect self-monitoring to be relevant to explicit trust (high self-monitors are more likely to overreport trust) since implicit attitudes are not easily self-censored.

Study Design and Measures

For our study, a nonprobability internet sample of respondents in China aged 18 or above were recruited by Qualtrics and surveyed from April to June 2018. As online appendix A shows, their demographic breakdown is somewhat comparable to the internet-active adult subsample of the 2018 China Family Panel Studies (CFPS), a nationally representative survey conducted in the same year.Footnote ¹ However, there were significantly more respondents with four-year college education or above in our sample, which is typical in online surveys in China.

Before taking the IAT near the end of the survey, we asked about explicit trust in government, system justification, support of the social credit system, acceptance of government surveillance, trust in government during crisis events, other attitudinal and belief questions, and demographic characteristics (refer to online appendix B for question wording). To measure explicit trust, we used the American National Election Study (ANES) trust in government index and adapted it to the Chinese context. Although this measure is not without critics (e.g., Cook and Gronke Reference Cook and Gronke2005; Gershtenson and Plane Reference Gershtenson, Plane, Aldrich and McGraw2011), the ANES trust items are among the most widely used and provide a valuable basis for comparison. As is customary, we recoded it to range from 0 to 100 with higher numbers indicating greater trust.

We use a variant of the IAT called the Single-Target IAT (ST-IAT) (Karpinski and Steinman Reference Karpinski and Steinman2006) to measure implicit trust in government. Whereas the standard IAT features two contrasting categories (Greenwald, McGhee, and Schwartz Reference Greenwald, McGhee and Jason1998), the ST-IAT only includes one category. Although it is valuable to compare the Chinese government to other entities (Zhou, Tang, and Lei Reference Zhou, Tang and Lei2020) we chose not to so as to ensure that the implicit measure is directly comparable to the explicit trust measure. In addition, the ST-IAT has been found to have sufficient reliability and stability (Bluemke and Friese Reference Bluemke and Friese2008, 988; also see Karpinski and Steinman Reference Karpinski and Steinman2006) and has been used to study implicit trust in the United States (see Intawan and Nicholson Reference Intawan and Nicholson2018).

The ST-IAT task consists of matching trust and distrust to the target category government (政府). Since there are not proper synonyms for government,Footnote ² we altered fonts to introduce variability in much the same way that some IATs use pictures instead of text (see Intawan and Nicholson Reference Intawan and Nicholson2018). The “trust” words in our ST-IAT task include trust (信任), trustworthy (值得信赖), dependable (可靠), reliable (靠谱), and honest (诚实) while the “distrust” words are distrust (不信任), untrustworthy (不值得信赖), undependable (不可靠), unreliable (不靠谱), and dishonest (不诚实). We randomize the blocks so that government is initially paired with trust or distrust. If the respondent is initially assigned to the trust block, the instructions require them to match government with trust. In the second block, respondents are instructed to match government with distrust. The initial block (trust) corresponds to a key response on the left side of the keyboard and the second block (distrust) corresponds to a key response on the right side. Respondents faster at matching government with trust relative to distrust possess implicit trust whereas the opposite hold implicit distrust. Figure 1 depicts two separate illustrative screens in the IAT (the actual experiment uses Chinese characters), one matching government with trust and the other matching government with distrust.

Figure 1 Illustrative screens in the Implicit Association Test for trust in government

Note: The left screen matches government with trust words and the right screen matches government with distrust words.

Following convention (Greenwald, Nosek, and Banaji Reference Greenwald, Nosek and Banaji2003), we removed respondents where more than 10% of their trials were less than 300 milliseconds and then converted the remaining 1,543 respondents’ trust ST-IAT data into D-scores. The D-score is the mean difference in response times between the blocks where trust is paired with government from the blocks in which distrust is paired with government, divided by the standard deviation in the test blocks; it ranges from -2 to +2 wherein a positive score indicates implicit trust, and a negative score indicates implicit distrust. The D-score thus reveals both the direction and strength of an implicit attitude wherein 2 represents maximum implicit trust and -2 represents maximum implicit distrust. The algorithm also incorporates errors (incorrect matches) and accounts for the variability of response latencies across trials (see Greenwald, Nosek, and Banaji Reference Greenwald, Nosek and Banaji2003).

Understanding Implicit Trust

We begin by looking at implicit trust and its relationship to explicit trust. Figure 3 depicts implicit trust in the Chinese government. About 83.7% of the respondents have positive D-scores, and the mean is .29, t(1542)=35.5, p < .001 with a standard deviation of .32. Since our sample is not nationally representative, we adjusted the weights on demographic variables to obtain national estimates. Li, Shi, and Zhu (Reference Li, Shi and Zhu2018) show that samples from crowd-sourcing internet platforms, such as ours, are particularly useful for predicting the attitudes of Chinese internet users. Therefore, we use the internet active adult sample of the nationally representative 2018 CFPS data to reweight gender, age group, education, CCP membership, and urban residency. This exercise yields a national estimate very similar to our sample estimate: 85.8% of the public possesses a positive D-score with a mean of .32.

Figure 2 Density of implicit trust in the Chinese government (ST-IAT D-Scores)

Figure 3 Density of explicit trust in the Chinese government

Recall that explicit trust ranges from 0 to 100 with higher numbers indicating greater trust. As shown in figure 3, about 71% of respondents exhibit explicit trust higher than 50, with a mean of 60.2 (SD=24.9; refer to online appendix C for summary statistics). Re-weighting the demographic variables by 2018 CFPS produces similar estimates: about 70.6% of the Chinese people exhibit positive explicit trust in government and the mean is 60.4.Footnote ³ This level of trust is slightly lower than found in many surveys, but we cannot rule out question wording differences. Regardless of weighting, the percentage of respondents with positive implicit trust appears to be higher than the percentage with explicit trust.

Figure 4 depicts the relationship between our measures of implicit and explicit trust. The correlation between implicit and explicit trust is nearly zero (r = -.0015) indicating that they are unrelated.Footnote ⁴ How might we explain a weak correlation between two largely positive measures? As seen in figure 4, implicit trust is consistently positive across both high and low values of explicit trust, so the absence of a correlation is largely due to respondents who are implicitly trustful but explicitly distrustful. Using an IAT that pitted trust in university officials against trust in government, Zhou, Tang, and Lei (Reference Zhou, Tang and Lei2020) reported a correlation of approximately .10 between implicit and explicit measures also suggesting little relationship. The disassociation of implicit and explicit trust was also found in the United States (Intawan and Nicholson Reference Intawan and Nicholson2018), but other research has reported stronger, albeit modest, correlations in the United States and other countries (see Murtin et al. Reference Murtin, Fleischer, Siegerink, Aassve, Algan, Boarini, González, Lonti, Grimalda, Vallve, Kim, Lee, Putterman and Smith2018). As established later, implicit and explicit trust are different measures with distinct correlates.

Figure 4 The relationship between implicit and explicit trust in China

To explore the origins of implicit trust, we considered a variety of correlates that previous studies have found to be associated with explicit trust. We start by examining demographic characteristics, including gender, age, education, household residence registration (urban versus rural), family income status, and CCP membership. We then consider two major sources of trust in the Chinese government identified in previous studies: Confucian values and government performance. We use an index of items borrowed from the Asian Barometer Survey for Confucian values and evaluations about China’s overall situation as a measure of government performance.

Given the vexing problem of sensitivity bias concerns in the study of trust in authoritarian settings, we also examined self-monitoring, the degree to which individuals alter their behavior to make positive impressions on others (Snyder Reference Snyder1974). If there is an element of social desirability in self-reported measures of trust, we expect those higher in self-monitoring to be more likely to misreport trust in government. As mentioned, although we include self-monitoring in models of explicit and implicit trust, we do not expect it to affect implicit trust since implicit measures are pre-conscious and largely immune to social desirability problems.

Figure 5 depicts regression results for the correlates of implicit and explicit trust (refer to table D.1 in the online appendix for the numerical results and table D.2 for the correlation matrix between the independent variables). We ran two models for each type of trust, one with demographics only (coefficients represented by triangles), and the other with demographic and attitudinal and belief variables that previous studies have suggested matter for explicit trust (coefficients represented by circles). The results of the two models are consistent. To aid interpretation, we rescaled all variables to range from 0 (low) to 1 (high). This means each estimate represents the percentage change in the dependent variable when the independent variable is changed from the minimum to the maximum level while all other variables are held constant.

Figure 5 Sources of implicit and explicit trust in government

Notes: OLS estimates with 95% confidence intervals. All variables are rescaled to range from 0 to 1. Triangles are coefficients from models with demographic variables only and circles are coefficients from the full models.

Beginning with demographics, figure 5 shows that education has a positive effect on implicit trust, regardless of model specification. This result suggests that education instills implicit trust via political and civic education beyond primary and junior high school. Based on the full model, increasing education from the minimum (primary school or below) to the maximum (graduate school) level increases implicit trust by 9.1 percentage points (from 54.5% to 63.6%). Equivalently, an additional level of educational attainment increases implicit trust by 1.8 percentage points.Footnote ⁵ Although it has a more modest effect compared to education, residency in urban areas has a positive effect on implicit trust in government. The average marginal effect of urban residency relative to rural residency is 2.4 percentage points (from 59.15% to 61.54%). The positive effects of education and urban residency on implicit trust suggests a critical role for political education and socialization in the fostering of implicit political trust.

Among other demographic characteristics, the results also indicate that CCP members have a slightly lower level of implicit trust than non-members (1.9 percentage points in the full model). More research is needed to understand this result, but it is consistent with the fact that in contemporary China individuals often join the party for instrumental and career reasons and cannot simply be viewed as party loyalists (Dickson Reference Dickson2016). The process of joining the Party does involve some degree of political socialization, but since it occurs in adulthood (either after a person enters the work force or at the late stage of one’s education), its role in fostering implicit trust is limited. Figure 5 also shows that none of the attitudinal variables including evaluation of China’s current situation and self-monitoring had significant effects on implicit trust at the conventional .05 level.

The correlates of explicit trust are different. In contrast to implicit trust, education has a negative effect on explicit trust: increasing education from the minimum to the maximum level decreases explicit trust by 8.5 percentage points (from 66.5% to 58.0%). This is consistent with Tang’s (Reference Tang2016) finding that higher education makes Chinese citizens more critical of government policies. Education’s role in cultivating “critical citizens” (Norris Reference Norris1999) suggests something of a paradox. Despite early political education instilling implicit pro-government orientations, additional education fosters more deliberative and critical thinking about government (as opposed to consisting merely of rote memorization as is often described in the popular press) which decreases explicit trust. At the same time, additional education appears to reinforce early political education through increased implicit trust. Similarly, while urban residence has a positive effect on implicit trust, it has a negative though small effect on explicit trust (effect size = 2.7 percentage points; from 59.8% to 62.5%). That both education and urban residence have opposite effects on implicit and explicit trust in government underscores how the two types of trust are distinct. These agents of political socialization matter to both types of trust, but in different ways. CCP members do not have higher explicit trust in government (according to the full model), once again suggesting the instrumental rather than political role of party membership.

Contemporaneous judgments and considerations also play a role in explaining explicit trust but not implicit trust. Consistent with previous research, respondents more satisfied with China’s performance are more likely to hold explicit trust, and the effect size is the largest among all covariates. Based on the full model, moving from the lowest to the highest value of performance produces a 55.9 percentage points increase in explicit trust (from 19.1% to 75.0%). Income is also positive, with explicit trust increasing 16.1 percentage points (from 50.5% to 66.6%) from the lowest to the highest levels of income. Confucian values are positively correlated with explicit trust, but the relationship is not significant at the conventional level. Lastly, figure 5 shows that high self-monitors are more likely to hold explicit trust, novel evidence that social desirability bias affects self-reported political trust. Moving from the lowest to highest value of self-monitoring produces a 9.8 percentage points increase in explicit trust (from 56.1% to 65.9%). This result suggests that explicit trust is somewhat inflated by social desirability bias. As expected, however, self-monitoring does not affect implicit trust, which is consistent with research showing that people are largely unable to censor implicit biases.

In sum, implicit trust and explicit trust in government are distinct. Not only are they unrelated to each other but our analysis suggests that implicit trust appears to be rooted in (political) education and socialization but detached from the sources that inform explicit trust such as government performance, social desirability, and family income. Such differences suggest that implicit trust is primarily rooted in early socialization whereas explicit trust is heavily informed by current circumstances. These results are consistent with research on implicit and explicit cognition that show that the former is largely shaped by early experiences and culture and the latter are more adaptable to new information. As such, our results also suggest that the factors that predict implicit trust appear to be consistent with Easton’s (Reference Easton1965) description of diffuse support (a reservoir of goodwill) and the factors that predict explicit trust are more consistent with specific support (evaluations of political authorities). Furthermore, the correlates of political socialization that matter to both, including education and urban (versus rural) residence, affect implicit and explicit trust differently. For instance, whereas higher education appears to heighten implicit trust through reinforcement of early formative lessons, it diminishes explicit trust by promoting more deliberative, thoughtful evaluations.

A final difference between the two models worth mentioning is that the model goodness-of-fit measure for explicit trust is considerably larger than for implicit trust. The R-squared for the former is .381 and .031 for the latter. Although some of the difference is likely attributable to the fact that implicit measures are noisier than explicit measures, another possibility is that there is a great deal more we need to learn about the origins of implicit trust. In the following section, we investigate whether implicit trust affects different types of support for the regime.

Does Implicit Trust Matter?

The second major objective of our study is to examine whether implicit trust matters for understanding diffuse and specific regime support (Easton Reference Easton1965). As mentioned, we expect implicit trust to affect diffuse support. Since implicit trust is deep-seated, its effects are likely to shape broad orientations toward government and society. To examine diffuse support, we chose system justification, the tendency of individuals to endorse the social, economic, and political status quo (Jost Reference Jost2020; Jost, Banaji, and Nosek Reference Jost, Banaji and Nosek2004; Kay and Jost Reference Kay and Jost2003). Items in the system justification scale, for instance, ask respondents to agree or disagree with statements such as “In general, our country’s political system operates as it should” (refer to online appendix B for items in the scale). Research on system justification theory has found that implicit attitudes about social groups play an important role in understanding system justification (Jost Reference Jost2020; Jost, Banaji, and Nosek Reference Jost, Banaji and Nosek2004). In research similar to our own, Intawan and Nicholson (Reference Intawan and Nicholson2018) found that implicit trust was a significant predictor of system justification in the United States. Accordingly, we expect respondents high in implicit trust to express greater system justification whereas we expect low implicit trust to diminish it.

We also expect implicit trust to predict attitudes towards various types of government policy that involve specific support including attitudes toward social control and trust in the government during crisis events. China’s emerging social credit system is a major initiative of social management and control, and an exemplary case to study the relationship between implicit trust and specific support. The social credit system was first proposed in 2014, and currently exists in the form of local pilots and experiments. The government’s stated goal for the program is to boost social trust and fight corruption and fraud by tracking the creditworthiness of individuals, businesses, and government agencies (Chorzempa et al. Reference Chorzempa, Triolo and Sacks2018; Kostka Reference Kostka2019). Many Western observers, however, regard it as a sweeping Orwellian surveillance apparatus for social control and privacy infringement (e.g., Acemoglu and Robinson Reference Acemoglu and Robinson2019). Given the vast reach of the social credit system, we found it an ideal policy for studying implicit trust and expect those with higher implicit trust to be more likely to accept the program.

Since evaluations of government are strongly informed by emotion during times of crises (Albertson and Gadarian Reference Albertson and Gadarian2015; Merolla and Zechmeister Reference Merolla and Zechmeister2009), they also represent an ideal test of whether implicit trust matters. Given the danger involved in times of crises, Kahneman (Reference Kahneman2011, 35) argues that implicit attitudes are likely to dominate since they are central to “fight or flight” responses. Following Intawan and Nicholson (Reference Intawan and Nicholson2018), we examined trust in government during “imagined” crisis events. As they discuss, times of crisis are likely to draw on automatic, gut-level processes given the inherent danger involved. In times of crisis then, citizens are likely to fall back on implicit trust, looking to government for safety and protection. On the other hand, implicit distrust should do the opposite since a lack of confidence in government is likely to make people desire that it not be involved.

Although our hypotheses address substantive questions central to the study of authoritarian politics, they also address a critical methodological consideration about the implicit trust measure, namely validity. Since implicit attitudes often diverge from explicit attitudes, there is little guidance on how to examine construct validity. However, the analyses of the various forms of diffuse and specific support provide a reassuring breadth of tests of the predictive validity of implicit trust. Therefore, these analyses not only help us better understand the effects of implicit trust but also validate an implicit measure that is novel in the study of trust in an authoritarian setting.

Analysis of Diffuse and Specific Support

All governments desire legitimacy, including authoritarian regimes (Zhao Reference Zhao2009). Easton’s (Reference Easton1965) notion of diffuse support is especially important here since it concerns a commitment to the structures and norms of a regime regardless of its current incumbents or performance. To measure broad system support, we look at how likely respondents are to agree with justifications about the social, economic, and political order using the system justification scale (Kay and Jost Reference Kay and Jost2003). Since the eight items that make up the scale were originally created to examine system justification in the United States, we made simple modifications to make them suitable for the Chinese context. The response scale (nine point) for each item ranges from strongly disagree to strongly agree and is coded such that higher scores indicate higher system justification. Combining the eight items, we created an index ranging from 1 to 9. For the sample, the mean for system justification is 6.46 (SD=1.34).

In the following statistical analyses, we use both OLS regression, with all variables including the dependent variables rescaled to range between 0 and 1 (to facilitate interpretation), and ordered probit regression with the dependent variables in their original scales. Although we expect explicit trust to be endogenous to system justification, it is primarily included as a control variable. As shown in regressions reported in online appendix D (table D5), results with or without controlling for explicit trust are consistent. Because the Chinese public’s political attitudes are often influenced by their life satisfaction and political interest (Huang Reference Huang2015a; Zhong Reference Zhong2014), we also include these variables as controls in addition to standard demographic variables. As found in online appendix D (table D6), the results of the OLS and ordered probit regressions are consistent.

The left panel of figure 6 shows the predicted values of system justification when implicit trust in government changed from low to high levels (with both variables rescaled to range from 0 to 1), based on an OLS regression controlling for explicit trust and other covariates. As hypothesized, implicit trust is a positive and significant predictor of system justification. In other words, as implicit trust increases so does agreement with statements justifying China’s political, social, and economic systems. Moving across the range of implicit trust from its minimum to maximum value in the data while holding all else constant increases system justification by 6.3 percentage points (from 64.4% to 70.7%).

Figure 6 Implicit trust in government, system justification, and forms of specific support

Notes: These are predictive margins with 95% confidence intervals based on OLS regressions controlling for demographic variables, explicit trust, and other attitudinal covariates. Variables are rescaled to range from 0 to 1.

Next, we examine whether implicit trust affects specific support in the guise of support for the state’s social credit system, perhaps an ideal example of authoritarian policy as discussed earlier. Since the social credit system is often regarded outside of China as a type of surveillance, we asked a related question about general attitudes toward government surveillance and the collection of personal information. In our sample, 87.2% “support” or “somewhat support” the social credit system on a five-point scale ranging from “oppose” to “support” (mean=4.37, SD=0.84). In contrast, on the question of government surveillance, 57.2% said they were “very concerned” or “somewhat concerned.” This variable uses a four-point scale and we recoded it so that higher numbers indicate less concern about (or more support for) government surveillance, with a mean of 2.39 (SD=0.91).

The middle and right panels of figure 6 show the relationship between implicit trust and opinion towards the social credit system and government surveillance. As hypothesized, implicit trust is a significant and positive predictor of support for the social credit system. Moving from the minimum to maximum value of implicit trust increased support for the social credit system by 10.9 percentage points (from 77.4% to 88.3%). In contrast, implicit trust is not a significant predictor of attitudes toward government surveillance. This result suggests that the Chinese public does not equate the social credit system with government surveillance. While the social credit system bundles prosocial behavior with political compliance (Tirole Reference Tirole2021), the Chinese people, particularly those trusting the government, accept the government’s justification of the program. Consistent with this interpretation, tables D5 and D6 in the online appendix show that people with higher education and higher life satisfaction are more supportive of the social credit system (see also Kostka Reference Kostka2019), indicating that they are more likely to interpret the social credit system through frames of promoting honest dealings in society and the economy rather than privacy encroachment.

Analysis of Crisis Events

Lastly, we examine trust in the Chinese government during crisis events. We borrowed crisis items on a natural disaster and a foreign military attack from Intawan and Nicholson (Reference Intawan and Nicholson2018) and created a new item on a trade war. While the questions are framed as hypotheticals, they are highly relevant to Chinese society. In particular, the survey was conducted amid the U.S.–China trade war, so the trade war question was related to a real and on-going event (even though its wording is about a generic foreign country rather than specifically the United States). Natural disasters such as floods and earthquakes are recurrent phenomena in China. Tensions have also been increasing in China’s foreign relations in recent years, and a military conflict with a foreign power is not unimaginable. The response scale (seven point) for each item ranges from low to high trust such that higher values indicate greater trust to successfully address the crisis. Overall, respondents expressed relatively high political trust for each crisis (trade war M = 5.69, SD=1.25; natural disaster M = 5.97, SD=1.21; foreign attack M = 5.97, SD=1.20).

Figure 7 presents the predicted values of trust in government during crisis events. The results indicate that higher implicit trust predicts trust in government in all three crisis settings. In terms of effect sizes, increasing implicit trust from the minimum to the maximum increased trust of the government in addressing a trade war by 14.9 percentage points (from 69.1% to 84.0%), a natural disaster by 8.2 percentage points (from 77.8% to 86.0%), and a foreign military attack by 19.2 percentage points (from 71.2% to 90.4%). Each result not only suggests how implicit trust is likely to matter in explaining public response to crises, and the upswell of support given to government, but also helps validate the implicit trust measure apart from the more traditional inquiries of system support. These results are also consistent with China’s experience during the 2020 coronavirus outbreak, with a trusting society exhibiting high policy compliance. In early 2020 when the government put most of the country under perhaps the strictest quarantine in the world (Shih Reference Shih, Greer, King, da Fonseca and Peralta-Santos2021), Chinese society generally accepted the rule with order and little unrest, even though at the time the measure had not been adopted by any other country (Stasavage Reference Stasavage2020).

Figure 7 Implicit trust in government during crisis events

Notes: These are predictive margins with 95% confidence intervals based on OLS regressions controlling demographic variables, explicit trust, and other attitudinal covariates. Variables are rescaled to range from 0 to 1.