Answering Politically Sensitive Questions in China

Although political scientists have learned a great deal about answering politically sensitive questions in China, inquiry has been primarily limited to trust in the central government. We advance the use of list experiments to examine trust in local government and support for the removal of the constitutional term limit for the state presidency. In China, local governments not only provide public goods and services but also implement national policies (Zhong Reference Zhong2014). Nevertheless, the Chinese people view the two levels differently, with the public expressing greater anger and blame toward local government (Dickson Reference Dickson2016; Li Reference Li2016; Zhong Reference Zhong2014). As a result, protests in China occur primarily at the local level and local governments carry out most of the repressions (Cai Reference Cai2008; Tang Reference Tang2016). Whereas political trust inquiries are about national and local governments generally, the term limit inquiry gauges support for a dramatic shift away from collective leadership and regularized leadership succession (Shirk Reference Shirk2018), allowing a president to stay in power indefinitely. Because local government is an integral part of China’s political system and the term limit removal represents a deepening of authoritarian rule, we expect participants to overreport trust in local government and support for the removal of the presidential term limit in addition to overreporting trust in the central government (H1).

Studies of political trust in authoritarian countries typically assume that fear of government retribution is the motivation for overreporting (Blair, Coppock, and Moor Reference Blair, Coppock and Moor2020). Despite the important role that fear plays, overreporting is also likely due to social norms and cultural traditions in China that require deference toward authority and social hierarchy (Shi Reference Shi2001), a type of social desirability bias. Indeed, an individual could be motivated by both fear and social desirability. Therefore, we introduce a second indirect measure, self-monitoring, to examine individual differences in susceptibility to social desirability bias (Snyder Reference Snyder1974). Self-monitoring, the degree to which individuals modify their behavior to make positive impressions on others, has been used to examine social desirability bias on sensitive topics including political trust in the United States (Intawan and Nicholson Reference Intawan and Nicholson2018). If there is an element of social desirability to these questions, we expect high self-monitors to be more likely to overreport political trust and regime support (H2).

Last, we look at social and financial well-being as correlates of overreporting. In keeping with previous research (Jiang and Yang Reference Jiang and Yang2016; Robinson and Tannenberg Reference Robinson and Tannenberg2019), we posit that well-to-do people, such as those with higher income, life satisfaction, and satisfaction with the current national situation, are more likely to overreport trust in government and support for removing the presidential term limit because they have benefited more from the existing political system and have more to lose from sanctions (H3). In addition to being associated with overreporting, previous research has found that social and financial well-being are important correlates of political trust and regime support (Guriev and Treisman Reference Guriev and Treisman2020; Tang Reference Tang2016; Zhong Reference Zhong2014).

Study Design and Measures

We surveyed a nonprobability internet sample aged 18 or above in China through Qualtrics from April to June of 2018 (Nicholson and Huang Reference Huang, Intawan and Nicholson2022), which happened soon after removal of the presidential term limit in March. Altogether, 1,602 individuals participated and, as Appendix A shows, their demographic breakdown is somewhat comparable to that of the internet-active adult subsample of the 2018 China Family Panel Studies (CFPS), a nationally representative survey. We used quotas to make the sample resemble the Chinese internet population, but due to the characteristics of internet panels (including Qualtrics) our sample was nevertheless younger and more highly educated, which is typical of online surveys in China (Li, Shi, and Zhu Reference Li, Shi and Zhu2018).

We use both direct and indirect questions to measure trust in government and support for the presidential term limit removal. Comparing direct questions with indirect questions (using list experiments) helps reveal whether, and to what extent, topics are sensitive. To measure trust in the national and local governments directly, all participants answered the following questions: “To what extent do you trust the central government/your municipal (county) government to do what is right?” The corresponding question to directly measure support for the term limit removal was “Do you support the removal of the term limit for the national leader from the constitution?”

We use list experiments as our primary indirect method. In a typical list experiment, participants are randomly assigned to two conditions, a control group including N nonsensitive items and a treatment group including a sensitive item and the N nonsensitive items. Participants are asked to report how many of the items in the list apply to them without identifying which items apply. The difference between the average number reported by each group reveals the prevalence rate for the sensitive item. In the political trust list experiments, we assigned participants to one of three groups: control, central government, and local government. The control group was asked how many of the following items they trust: (a) neighbors, (b) schoolmates/colleagues, (c) drug users, and (d) the police.Footnote ¹ For the central and local treatments, “the central government” and “county/municipal government” were respectively added as the sensitive item. For the term limit list experiment, the control group was asked how many of the following they supported/approved of: (a) bike sharing, (b) smoking, (c) restricting the celebration of western holidays, and (d) sending children to study abroad. For the treatment group, the sensitive item of “removing the term limit for the national leader” was added. We did not find evidence of ceiling or floor effects.Footnote ² For both political trust and support for the term limit removal, the order of direct questioning and list experiments was randomized.

To explore whether social desirability is an additional source of misreporting, we use an index of three items to construct the self-monitoring scale. Along with standard demographics and other attitudinal controls, the survey included items on social and financial well-being such as income, life satisfaction, and government performance (proxied by China’s current situation). Refer to Appendix B for question wording, Appendix C for the balance of covariates, and Appendix D for tests of question order effects.

Results for Political Trust

We begin with the results for trust in the national and local governments. Table 1 shows the proportions of participants trusting the central and local governments, as measured by list experiments and direct questioning, and the differences between the two measures. As our sample was not nationally representative, we weight it by the internet active sample of 2018 CFPS to facilitate comparisons to previous research (refer to Appendix D for the unweighted results). The list experiment shows 76.7% trust the central government, compared with 90.4% in direct questioning. The difference between these values suggests that 13.7 percentage points more people reported trusting the central government in direct questioning than in the list experiment. Nevertheless, even after accounting for overreporting in direct questioning, majority trust in the central government persists.

Table 1.

Trust in Different Levels of Government by List Experiment and Direct Questioning with Poststratification Weighting

Note: Point estimates of trust in government with 95% confidence intervals in parentheses. The list experiment results are difference-in-means estimates. The sample is weighted by the 2018 CFPS internet active adult sample data on gender, age (below 30 vs. above 30), education (college vs. no college), and CCP membership. ^***p < 0.001.

To better situate our findings, we compare these results with Tang’s (Reference Tang2016) seminal study, which offers the only list experiment on political trust in China with a nationally representative sample. Our direct questioning result on trust in the central government is similar to the direct questioning result in Tang’s study (92%) as well as those from many other major, nationally representative surveys. But the level of trust in our list experiment is somewhat lower than that in Tang’s list experiment (76.7% versus 88%), and the level of overreporting is higher (13.7 versus 4 percentage points). Although the dissimilarity may be partly due to the differences in survey modes (online versus face to face) and timing (2012–13 versus 2018), they also suggest that despite that reported trust in the central government has been fairly stable, the level of actual trust may have decreased somewhat during this time (but still a majority).Footnote ³ In addition, our findings suggest that the sensitive nature of the question has increased in recent years, although perhaps not as much as the unweighted results of Li, Shi, and Zhu (Reference Li, Shi and Zhu2018) and Robinson and Tannenberg (Reference Robinson and Tannenberg2019) suggest. Table 1 also shows that the level of trust in local government was 67%, compared with 72.3% in direct questioning. In other words, 5.3 percentage points more people reported trusting local governments in direct questioning than in the list experiment.

In summary, comparing trust in the central and local governments provides new insight into the nature of political trust in China. First, participants trust the central government more than they do local governments whether measured by direct questioning or list experiments (the 95% confidence interval for the level of trust in local government in the list experiment crosses 50% even though the point estimate suggests majority trust). Second, the degree of overreporting trust in direct questioning is higher with the central government than with the local government. This finding suggests that the central government is more trusted, feared, and as we find below, apparently more revered.

Who Overreports Political Trust?

We also explored whether individual characteristics are associated with misreporting political trust by using Blair and Imai’s (Reference Blair and Imai2012) multivariate regression method, a procedure that allows for the estimation of relationships between preferences over the sensitive item and personal characteristics. Figure 1 depicts the differences between estimates from direct questioning and list experiments by demographic and attitudinal covariates (refer to Appendix E for numerical results). To begin with our inquiry into social desirability bias, high self-monitors overreport trust in the central government significantly more than do low self-monitors, whereas the level of self-monitoring does not affect overreporting trust in local government. It appears that expressing distrust in the central government might violate a social norm in China. Although fear is likely to be the dominant factor in explaining overreports in trust in the national government, self-monitoring has the second largest effect among the individual covariates, suggesting that social desirability may play a prominent role in overreporting.

Figure 1.

Individual Characteristics and Overreporting Trust in Government

In terms of financial and social well-being, Figure 1 also shows that people with high income overreport trust in both the central and local governments, whereas those with low income do not, and the differences are statistically significant. Additionally, people with high life satisfaction overreport trust in both the central and local governments but those with low life satisfaction do not overreport trust in local government. Overall, people with more comfortable life situations or perspectives are more likely to overreport political trust. Last, Figure 1 suggests differences in life satisfaction, political interest, and Confucian values matter more in evaluations of the local government than in those of the central government. This is consistent with the above finding that expressing trust in the central government is likely a social norm in China.

Results for the Term Limit Removal

The direct question about the presidential term limit removal allows five choices ranging from “support” to “oppose.” Because the survey was conducted immediately after the removal, we provided a neutral response in the middle of the scale, “neither support nor oppose,” to reduce its sensitivity.Footnote ⁴ To make the direct questioning analysis comparable with that for the list experiment (which does not have a neutral option), we dropped participants who chose the neutral response in direct questioning in our first analysis.Footnote ⁵ However, we also conduct a separate analysis looking at neutral responders.

Table 2 shows the proportions of nonneutral respondents supporting the term limit removal measured by the list experiment and direct questioning and the difference between the two measures. Participants overreported support for the removal of the term limit in direct questioning: Although 59.6% explicitly support it, the point estimate in the list experiment percentage is lower than 50%. However, as the 95% confidence interval crosses 50%, we cannot conclude that only a minority supported the term limit removal. Nevertheless, it is clear that we do not find evidence of majority support for the removal, in contrast to our results on trust in the central government. Furthermore, the differences between the list measure and direct questioning on the term limit removal was 22.6 percentage points, higher than the differences in overreporting political trust. This comparison suggests that the term limit removal was especially sensitive.

Table 2.

Support for the Term Limit Removal by List Experiment and Direct Questioning with Poststratification Weighting

Note: Point estimates of support for the term limit removal with 95% confidence intervals in parentheses. The list experiment results are difference-in-means estimates. The sample is weighted by the 2018 CFPS internet active adult sample data. ^**p < 0.01, ^***p < 0.001.

Next, we examine the list experiment results for the neutral responders in direct questioning. The last row of Table 2 shows that the proportion of neutral responders in direct questioning who supported the term limit removal in the list experiment was also lower than 50% (44.2%). Thus, most neutral responses appeared to be tacit opposers by the list measure. Taken together, our results suggest that there was a substantial reservation about the term limit removal. Previous research suggests that the Chinese people are relatively uncritical of the national government, reserving most of their blame for local government (Chen Reference Chen2017; Dickson Reference Dickson2016; Li Reference Li2016; Zhong Reference Zhong2014). The lack of evidence for majority support for the term limit removal indicates the Chinese public is not unquestioning or naïve; they are capable of expressing reservations about government, at least indirectly. This finding further suggests the relatively high level of trust in the national government is largely genuine.

Who Supports the Removal of the Presidential Term Limit?

We also explored subgroup differences between the list and direct measures for the presidential term limit removal again using Blair and Imai’s (Reference Blair and Imai2012) multivariate regression method. Figure 2 shows people low in self-monitoring overreport support, but the difference between them and high self-monitors is not significant, suggesting that social desirability is not a factor, different from our expectation. One likely reason is that the political sensitivity of the term limit removal question is notably higher than that of trust in government, and therefore the effect of fear dominates. With respect to financial and social well-being, people with high life satisfaction, high satisfaction with China’s current situation, and high income (i.e., those leading a more comfortable life) are more likely to overreport support for the term limit removal in direct questioning than those with opposite characteristics.Footnote ⁶

Figure 2.

Individual Characteristics and Overreporting Support for the Term Limit Removal