Co-Optating with Political Selection

Co-optation is a mainstay of authoritarian survival strategy, aimed at buying off potential opposition with either rents or policy concessions (Bueno de Mesquita et al. Reference Bueno de Mesquita, Smith, Siverson and Morrow2003; Gandhi and Przeworski Reference Gandhi and Przeworski2006; Reference Gandhi and Przeworski2007; Haber Reference Haber, Weingast and Wittman2006). Most co-optation institutions identified in existing studies share a common feature: in exchange for political support, the regime makes a payout, often in the form of material benefits (e.g., Magaloni Reference Magaloni2006) or political offices (e.g., Malesky and Schuler Reference Malesky and Schuler2010). As a matter of fact, political selection is frequently used as a vehicle for co-optation (Ang Reference Ang2016; Blaydes Reference Blaydes2010; Lust-Okar Reference Lust-Okar2005; Truex Reference Truex2014), as political offices are a credible way of distributing rents (Robinson and Verdier Reference Robinson and Verdier2013). In short, this type of co-optation is based on an immediate transfer of socioeconomic gains to the recipients.

This is where my argument departs from the existing literature. I argue that, besides an immediate transfer of material benefits, a regime can also co-opt with upward mobility, which has a similar effect as actual socioeconomic improvement in dampening demand for redistribution. When an individual believes that she may receive benefits from the regime, even on a later date, her support for (opposition to) the status quo is likely to increase (decrease). Thus, if a regime can find a way to forge such beliefs in its population, it can co-opt it, not with an immediate payout, but an opportunity to earn a future payout.

This co-optation arrangement has been partially identified by studies on single-party regimes (Lazarev Reference Lazarev2007; Svolik Reference Svolik2012, chap. 6), which show that a ruling party can keep low-rank members loyal with a chance of “deferred promotion into rent-paying positions” (Lazarev Reference Lazarev2007). These studies highlight the importance of party hierarchy in this co-optation setup: (1) it structures inequality in the regime to heavily favor political elites (e.g., senior party members) and (2) it maintains a personnel system with regular retirements and promotions in order to create opportunities for non-elites. Together, these two features ensure that there is a nontrivial amount of aggregate upward mobility in the regime.

My argument builds on this theory of party-based co-optation. I argue that, a party hierarchy is necessary but not sufficient to co-opt in large numbers. Another crucial component is the mechanism of political selection, which decides who gets promoted into rent-paying position. Put differently, the selection mechanism determines how upward mobility is distributed in the population, especially between those with patronage ties and those without.Footnote ³ In particular, I identify meritocracy as a selection mechanism that is well suited for co-optation, as it distributes upward mobility more widely and more uniformly.

Distributing Upward Mobility via Meritocracy

In the context of political selection, meritocracy is understood in contrast with patronage: the former adopts objective criteria for recruitment and promotion, whereas the latter relies on incumbent elites’ discretion (Mueller Reference Mueller2009). Although patronage-based selection is very common in authoritarian regimes, its capacity to co-opt is in fact limited. First, only those with ties to incumbent elites have access to the selection process, so it reaches a small portion of the population. Second, even for those with patronage ties, their prospect is far from certain due to unpredictable factional rivalry among elites. The exclusive and contingent nature of patronage means that it distributes the upward mobility generated by political selection in a highly skewed manner with little uniformity or certainty. As such, it can only co-opt the few who have succeeded in selection and are receiving benefits from the regime.

In contrast, meritocracy has the potential to buy off a much larger segment of the population. Shifting from patronage to merit as the main selection criterion, it opens up the process to many who otherwise would not have the access to participate. Of course, who ultimately gets selected depends on how merit is defined and how it is distributed within the population. Generally, merit is considered a valence issue that refers to individual qualities such as competence (Besley Reference Besley2005). Assuming that merit is a function of innate talent, effort, and resource endowment, we can reason that meritocracy helps level the playing field in favor of those without patronage ties, even if it does not completely equalize it. As long as merit is not completely determined by patronage ties, a shift from patronage to meritocracy represents a net improvement of openness in selection, thus distributing upward mobility more widely in the population.Footnote ⁴ Moreover, with standardized procedure and methods of evaluation, meritocratic selection is governed by rules established ex ante, thus eliminating the influence of patronage and connections. This ensures that upward mobility is distributed more uniformly.

These characteristics of meritocratic political selection vis-à-vis upward mobility render it suitable for grassroots co-optation. From the perspective of the regime, it offers several advantages. First, it does not alter the overall redistribution scheme, thus not risking elite discontent. It is also cost effective: with only a small number of payouts going to successful candidates, it garners support from many more who hope to succeed someday.Footnote ⁵

Another advantage of meritocratic political selection is more subtle. While it attracts many to participate, individuals do so for their private gains. By incentivizing citizens to improve their own conditions instead of making collective demands, meritocracy fragments potential opposition to the regime and undermines its collective action capacity. This is especially efficacious when targeted at the most educated group in society, who tend to exhibit greater “destabilizing behavior” when “unemployed, alienated, or otherwise dissatisfied” (Huntington Reference Huntington1968, 68). By siphoning this group off and pacifying them with an opportunity to become political insiders, meritocratic political selection neutralizes a potential threat to regime stability. This is similar to what Lust-Okar (Reference Lust-Okar2005, 36) calls the “divided structure of contestation” in the Arab world, where opposition groups that are allowed to participate in elections become less likely to mobilize against the regime.

Committing to Meritocracy: Institutionalization

For meritocracy to fully function as a co-optation instrument, it is crucial that potential participants have faith in it as a genuine ladder of upward mobility. This is a challenge for the regime: its commitment to meritocracy is not automatically viewed as credible, because incumbent elites may undermine it to reap private gains from patronage ties. Like any other commitment problem facing an authoritarian regime, this can only be solved when the regime is constrained by a credible threat of violence. It is possible in a long-run equilibrium, where reneging on meritocracy would trigger mass revolt that threatens regime survival.Footnote ⁶ Short of that, in the early days following its adoption of meritocracy, the regime can credibly signal its commitment to meritocracy through institutionalization—that is, by laying down rules that bind its own hands and constrain incumbent elites.

In concrete terms, institutionalization of meritocracy involves steps such as codifying the selection process into law, adopting supporting measures to strengthen meritocratic practices, and setting up a system to detect and punish wrongdoing. Although these institutional investments cannot fundamentally prevent the regime from walking away from meritocracy, they create a strong norm and penalize violations. In other words, institutionalization significantly increases the regime’s transaction cost to renege on its commitment, thus bolstering credibility of meritocracy.

The institution of meritocracy also becomes more credible over time by enhancing individuals’ retrospective evaluation of upward mobility. This is certainly the case for those who have participated and succeeded in meritocratic political selection; but more broadly, it also applies to a larger group who have benefited from simply having an additional opportunity, which has added to their overall life chances. As individuals update their perceived upward mobility based on retrospective evaluation, they develop greater confidence that meritocracy is functioning well. Therefore, although meritocracy co-opts by shaping individual expectation of the future, that is, prospective evaluation of upward mobility, it also operates through retrospective evaluation of upward mobility, which reinforces the credibility of meritocracy as a channel of upward mobility.

In summary, I argue that, by extending the opportunity of entering the political elite class to ordinary citizens, meritocracy can serve as an effective instrument of grassroots co-optation. Especially when institutionalized, its inclusive and rule-based nature fosters a widespread, popular belief of upward mobility among ordinary citizens, thus mitigating their discontent with status quo. Next, I apply this argument to the NCSE in China and investigate its effects.

Civil Service: The Political Elite Class

Under the one-party rule of the CCP, party cadres and government officials constitute the political elite class in China. Following a series of personnel reforms in the 1980s and 1990s, the cadre system of the Maoist era was modernized (Manion Reference Manion1985; Reference Manion1993); the official corps was re-branded as the “civil service” with an emphasis on professionalization (Burns Reference Burns1989; Reference Burns1994). The term “civil servant” was adopted in 1993 to refer to all full-time government employees working in administrative capacity with an official rank. Despite the new label, the Chinese civil service retains the essence of a Leninist party-state: it encompasses all levers of political power, ranging from party organs and government agencies to the legislative and judicial branches and key social organizations. Moreover, it covers all ranks from entry-level bureaucrats to high-level political leaders at all five levels of administration.

Although China has a sizable public sector, the civil service is kept lean and selective at around 7 million, roughly 0.5% of the population.Footnote ⁷ More importantly, like the old cadre system, the civil service is a closed hierarchy that recruits newcomers only at the entry level and promotes strictly from within.Footnote ⁸ These characteristics make the civil service highly exclusive and impermeable, thus rendering its entry-level selection an event of great significance.

Besides its exclusivity, the civil service confers to its members’ privileges and advantages that are unavailable to other groups in the Chinese society. First and foremost, civil servants benefit from the regime’s highly differential allocation of redistribution resources (Walder Reference Walder1992) and enjoy unusually generous benefits in areas including healthcare, pension, and even housing. Second, in Chinese society, working in government is considered a significant status attainment based on workplace identification (Bian Reference Bian2002; Lin and Bian Reference Lin and Bian1991). Third, as the Chinese state embeds itself in the nexus of politics and economics, civil servants can accrue valuable social capital to benefit themselves and their families, often in the form of rent-seeking and gray income (Ang Reference Ang2020, chap. 4). It should be noted that there is considerable variation across different ranks, departments, and regions; yet, when compared to others in their immediate surroundings, civil servants clearly belong to a privileged elite class. For ordinary citizens, entering the civil service represents a move up the socioeconomic ladder. Even in the age of market economy, many Chinese still regard entering the political hierarchy as a definitive—if not the preferred—socioeconomic advancement. It is thus important who is eligible to participate in political selection.

Institutionalization and Popularity

Of course, official propaganda alone cannot convince the public that NCSE provides a meaningful opportunity to ordinary citizens, unless the selection process lives up to its claim of meritocracy. Since its introduction, NCSE has not been immune to corruption and other undue influences; patronage and nepotism continue to influence selection in some cases. To root out malfeasance and improve the credibility of NCSE, the CCP has taken steps to institutionalize it.

Part of the institutional investment focuses on improving NCSE procedure and transparency with clear rules and regulations, especially at the interview stage, in order to reduce space for manipulation.Footnote ¹³ Other efforts seek to establish a robust legal and disciplinary framework so that any wrongdoing in NCSE is promptly investigated and punished. Following the Civil Service Law, the central government has issued a long list of auxiliary documents to govern various aspects of NCSE (see Table A1 in the Supplementary Material). While these efforts cannot completely eliminate patronage, they have reduced such influence and increased the weight of merit in the selection process (Liu Reference Liu2019). They also make NCSE more rule-based and offer some accountability to potential applicants.

The growing credibility of NCSE is reflected in the public enthusiasm it has received. Even faced with competition from a robust private sector, NCSE attracts many college-educated youths every year. Figure 1 illustrates trends in the national exam between 2003 and 2018: the yearly number of applicants has consistently exceeded one million since 2009, with an applicant–position ratio of over 50:1. Provincial exams are slightly less competitive, with the applicant–position ratio averaging around 30:1. This does not mean provincial exams are less popular: the ratio of applicants to new college graduates each year is high (median = 0.587 and mean = 0.690), even exceeding 1 in some provinces (Figures A3 and A4 in the Supplementary Material). It shows that a large proportion of college graduates participate in NCSE, sometimes even after they have graduated. It also indicates that the popularity of NCSE is not concentrated among graduates of top universities; many from lower-ranked school also participate.

Figure 1. Rising Popularity of NCSE—National Exam

Note: Statistics used in this figure are compiled by the author from media reports.

As NCSE has become more institutionalized over time, it has built its credibility as a relatively open and rule-based system of political selection, especially when compared with the old job assignment system. In the rest of the article, I examine how it functions as an institution of grassroots co-optation. Specifically, I investigate its effect on individual perception of upward mobility and political attitude.

Empirical Strategy

To understand the effects of NCSE, I focus on the provincial exams administered by respective provincial governments. Although the national exam is considered more prestigious by some, provincial exams are responsible for the lion’s share of openings each year and cover a greater geographical span.Footnote ¹⁴ For most people, it is more practical or even preferable to pursue a government career in their locality of residence; provincial exams are thus a more realistic choice.

More importantly, focusing on the provincial exams allows me to take advantage of the staggered introduction of NCSE across localities. Following the central government’s call to adopt the principle of “all entries via exam,” provinces began to implement NCSE in full scale between 1999 and 2009. Full-scale implementation of NCSE in each province means that there was a sharp increase in government job opportunities for the local population in the year when it was first introduced. To determine the date of NCSE introduction at the province level, I use an online database of Chinese laws and regulations and identify, for each province, the year when the phrase “all entries via exam” (fanjin bikao) first appeared in a provincial government official document on the issue of civil service recruitment. Figure 2 shows a map of province-level NCSE introduction by year.

Figure 2. Year of NCSE Introduction at the Province Level

Coupled with NCSE eligibility requirement on age (i.e., below 35 years of age), the staggered introduction across provinces means that college-educated citizens who entered the workforce in the same cohort may or may not have been able to take advantage of this new selection mechanism, depending on their locality of residence. Exploiting this exogenous spatial-cohort variation, I employ a generalized DID design to examine whether, among college-educated citizens, there is a significant difference in perceived upward mobility between those who became eligible for NCSE and those who did not due to age cutoff. Moreover, I apply this generalized DID design to investigate whether NCSE has any effect on individual political attitude, as a result of enhanced upward mobility.

Data

In the main analysis, I use data from four consecutive waves of the Chinese General Social Survey (CGSS), conducted between 2010 and 2013. Given that the last provinces to fully implement NCSE did so in 2009, incorporating data dating from 2010 ensures that there are respondents in both treatment and control groups in each province. One advantage of using repeated cross-sectional data is the large sample size from pooling together multiple waves. Because this study focuses on college-educated citizens, it is important to have a decent sized sample of college-educated respondents.Footnote ¹⁵ Combined, the four waves of CGSS data provide a sample of over 4,500 college-educated respondents between the ages of 18 and 60.

For the treatment variable, I construct a dichotomous variable, eligibility at introduction, by matching each respondent’s province of residence to the NCSE introduction timetable and determining her age when NCSE was introduced in her province. Those younger than 35 years of age had at least one chance to take the exam and are hence coded as 1; those at the age of 35 or above, on the other hand, were excluded from the selection process altogether and are hence coded as 0. In essence, the treatment variable indicates whether a respondent was ever given an opportunity to compete to enter the civil service, regardless of whether she had taken advantage of it and whether she had succeeded. To the extent that almost all respondents were aware of NCSE introduction in their home provinces, there should be little to no treatment noncompliance in the analysis.

The outcomes of interest in this study are twofold, including respondents’ perceived upward mobility and their political attitude. To measure the former, I use three survey questions on respondents’ self-reported socioeconomic status at different points in their lives: when they took the survey (current), 10 years before the survey (past), and 10 years after (future), respectively; I then take the differences to construct two variables: mobility experience and mobility prospect.

Mobility experience measures the difference between a respondent’s current and past socioeconomic status—that is, a retrospective evaluation of upward mobility. Similarly, mobility prospect measures the difference between her future and current socioeconomic status—that is, a prospective evaluation of upward mobility. Because NCSE had been introduced prior to the survey, it is expected to have a direct effect on mobility experience for those eligible for the exam. It should be noted that this effect is not only caused by some eligible respondents becoming civil servants via NCSE. Rather, the introduction of NCSE is expected to have enhanced how eligible respondents perceived the amount of upward mobility available to them, whether they became civil servants or not.Footnote ¹⁶ By comparison, mobility prospect is less directly affected by a respondent’s past eligibility in NCSE; still, those who benefited from having an additional career choice in the past may be better positioned to maximize their potential, thus leading to greater upward mobility in the future.

As discussed in the theory section, the argument of co-optation with meritocracy is anchored in prospective evaluation of upward mobility, while retrospective evaluation also plays a key role in reinforcing the credibility of meritocratic selection. It is thus appropriate to examine both mobility experience and mobility prospect. Moreover, because NCSE introduction took place prior to the CGSS surveys, mobility experience captures—or recovers—how the exam has shaped respondents’ perception of prospective upward mobility at that time. Therefore, in the empirical analysis, mobility experience is the outcome of most interest.

The second set of outcome variables measures respondents’ political attitude, using a battery of survey questions. Among them, I am most interested in redistributive preference, which is measured by a survey question asking whether respondents agree with the statement, “The government should tax the rich more to help the poor.” Other variables include respondents’ perceived political influence, their right to political participation, and their political trust in different levels of government. There is no strong theoretical support that upward mobility has a direct impact on aspects of political attitude other than redistributive preference; however, it is plausible that NCSE could induce some changes in respondents’ attitude toward the regime. It should be noted that these survey questions were not asked in every wave of CGSS, so there are fewer observations, but still sufficient for the purpose of this analysis.Footnote ¹⁷

Identification Models

To estimate the effect of NCSE introduction on individual perception of upward mobility, I apply a generalized DID framework and estimate the following model:

(1) $$ \begin{array}{rl}mobilit{y}_{icps}=\beta \hskip2.84526pt eligibl{e}_{icp}+\Sigma {\gamma}_c+\Sigma {\delta}_p+\Sigma {\theta}_s+{\epsilon}_{icp},& \end{array} $$

where $ mobilit{y}_{icps} $ represents the perceived upward mobility, either in retrospect or prospect, of individual i of cohort c in province p surveyed in year s; $ eligibl{e}_{icp} $ indicates whether individual i of cohort c in province p was eligible for NCSE when it was first introduced; $ {\gamma}_c $ , $ {\delta}_p $ , and $ {\theta}_s $ capture the province, cohort, and survey year fixed effects; and $ {\epsilon}_{icp} $ represents any idiosyncratic differences that are correlated across individuals within a province×cohort cell.

By using a DID framework, the baseline model allows me to address a variety of concerns in identifying the effects of NCSE introduction. First, province-level differences in economic development, overall upward mobility, income level, and political culture may be correlated with individual-level outcomes of interest. By controlling for province fixed effects, this model exploits cross-cohort variation within each province as a result of staggered NCSE introduction. Similarly, cohort-level differences could lead to differences in individual perception and attitude that are independent of NCSE; by including cohort fixed effects, this model is able to difference out cross-cohort changes that occur even in the absence of NCSE introduction.

As a robustness check, I estimate several alternative model specifications, including (1) regressions controlling for province×cohort-level covariates, (2) regressions controlling for individual-level covariates, and (3) regressions controlling for province-specific cohort fixed effects. Moreover, to address potential concerns with the spatial-cohort variation, I replicate the baseline estimation by excluding certain respondents from the sample. First, to make sure each respondent’s province of residence does not suffer from post-treatment bias, I drop those who obtained their Hukou in current county after NCSE introduction in that province. Similarly, to ensure that the age cutoff at 35 applies to all respondents, I drop those with postgraduate degrees.

In addition to the direct effect of NCSE on individuals’ perceived upward mobility, I am also interested in exploring whether the resultant change in perceived upward mobility has any effect on their political attitude. To this end, I adopt an instrumental variable (IV) approach, where the treatment of upward mobility is instrumented by eligibility to participate in NCSE. To elaborate, in the first stage, I use eligibility at introduction to predict mobility experience and mobility prospect based on Equation 1; in the second stage, I estimate the following structural equation:

(2) $$ \begin{array}{rl}attitud{e}_{icps}=\theta \hskip2.84526pt \hat{mobilit{y}_{icps}}+\Sigma {\gamma}_c+\Sigma {\delta}_p+\Sigma {\theta}_s+{\zeta}_{icp},& \end{array} $$

where $ attitud{e}_{icps} $ is one of the six outcomes, including redistributive preference, and $ mobilit{y}_{icps} $ is the predicted value of mobility experience or mobility prospect from the first stage. I include full sets of cohort and province fixed effects, $ {\gamma}_c $ and $ {\delta}_p $ , as well as survey year fixed effects, $ {\theta}_s $ ; $ {\zeta}_{icp} $ represents standard errors clustered at the province×cohort level.

Exploring Causal Mechanisms

The analysis plan laid out so far is designed to estimate the causal effects of NCSE on individual perception and attitude. It does not shed light on how NCSE accomplishes it, due to lack of relevant information in the CGSS data. To explore the underlying causal mechanisms, I resort to a different dataset.

Qualitative evidence, such as CCP’s publicity campaigns discussed earlier, points to one possible causal mechanism: NCSE has led more college-educated citizens to consider working in government. Prior to NCSE, a career in government was largely out of reach for most people who did not have the political connections to access or navigate the old job assignment system. By introducing an open, rule-based selection system, NCSE provides them with a chance at government employment based on their ability and effort. As more college graduates come to regard civil service as a viable career path, regardless of family background or social networks, their perceived upward mobility in the regime increases on average.

Empirically, this shift can be measured by individual career preference in response to NCSE institutionalization. As NCSE becomes more institutionalized, the influence of political connections further diminishes; correspondingly, more college graduates would consider a career in government, especially those without political connections. Granted, institutionalization is a gradual process. It takes time for provincial governments to adopt additional measures; it also takes time for local residents to build up confidence in NCSE.

Thus, I explore whether college-educated youths adjust their career preference in response to NCSE institutionalization, as proxied by the years of NCSE implementation. I draw on data from the 2009 wave of the Beijing College Students Panel Survey (BCPS), which interviewed 4,752 college students enrolled in 15 universities in Beijing.Footnote ¹⁸ Because all respondents would automatically become eligible for NCSE upon graduation, I examine whether those from provinces where NCSE has been implemented for longer exhibit a stronger preference for a career in government.

For the predictor, I follow the same identification strategy as before. I match each respondent’s home province to the timetable of province-level NCSE introduction and create a variable, years of NCSE, that measures the number of years lapsed between NCSE introduction in her home province and the year 2009 when the survey was conducted. The outcome of interest is respondents’ career preference. Using two questions in the survey, I construct two dichotomous variables, government organs and civil service, to indicate respondents who chose “CCP and government organs” as their most preferred workplace and those who chose “civil service” as their most desired profession, respectively.

To estimate the effect of NCSE institutionalization on college-educated youths’ career preference, I estimate the following model using a logit regression:

(3) $$ \begin{array}{rl}{Y}_{ip}=\alpha +\beta \hskip0.3em NCS{E}_p+\Sigma \delta {X}_i+\Sigma \gamma {Z}_p+{\epsilon}_i,& \end{array} $$

where $ {Y}_{ip} $ represents the career preference of respondent i from province p and $ NCS{E}_p $ measures the number of years NCSE had been implemented in province p. The model controls for both individual- and province-level covariates that could influence a respondent’s career preference: $ {X}_i $ represents a set of individual-level characteristics, including gender, ethnicity, parent(s) in CCP, parent(s) in government, hometown in big city (i.e., provincial capitals and municipalities), score on college entrance exam (gaokao), CCP membership, extra-curricular activities, academic major, class, and university type; $ {Z}_p $ represents a set of province-level characteristics, including GDP, GDP per capita, fiscal revenue, fiscal expenditure, FDI inflow, population, employment rate, and the number of jinshi in the imperial civil service exam in Ming dynasty.Footnote ¹⁹ For each outcome, I first estimate the model without any covariates; I then estimate two specifications with province-level characteristics, first in the current year and then with a 1-year lag.

Effects on Perceived Upward Mobility

I begin the analysis by estimating the baseline DID model. Before pooling all respondents in two groups based on eligibility at introduction, I first examine whether their perceived upward mobility varies with the “distance” between their age and the NCSE eligibility cutoff. This allows me to discern whether perception varies with age even in the absence of NCSE and whether there is a sharp increase following NCSE introduction. To do so, I split survey correspondents into 11 groups based on their age at the time of NCSE introduction. The group in the middle as the reference category consists of those who just missed the age cutoff (i.e., 35 years old) and were hence ineligible for NCSE; on the older side are five age groups ineligible for NCSE; on the younger side are five other age groups eligible to take the exam at least once. I restrict the sample to 10 age groups on either side of the reference category.

Figure 3 reports the coefficient estimates for each group. In plot (a), there is a clear jump in positive mobility experience. Compared with respondents of age 35, those of age 34 experienced greater upward movement in the past 10 years. This sharp increase also persists for the other younger age groups. In contrast, the older age groups do not differ significantly from the reference group. The clear jump in perceived mobility experience at the NCSE age cutoff shows that it is not a general upward trend driven by age. A similar pattern is also observed in plot (b) for positive mobility prospect, albeit somewhat attenuated. This is expected, as NCSE introduction is a past event that most directly affects respondents’ retrospective evaluation of mobility, hence the sharp increase in plot (a); on the other hand, its effect on respondent’s prospective evaluation of mobility is less direct and may be mediated by many other factors.

Figure 3. Perceived Upward Mobility by Age Groups

Note: Each plot reports the coefficient estimates with 95% confidence intervals. The outcome in plot (a) is a dummy variable indicating positive mobility experience; the outcome in plot (b) is a dummy variable indicating positive mobility prospect. The estimates are based on a linear probability regression of each outcome on a set of dummy variables indicating respective age groups, as well as province, cohort, and survey year fixed effects; standard errors are clustered at the province×cohort level. See Table A4 in the Supplementary Material for details.

Next, I estimate the standard DID model based on Equation 1. The regression results are reported in Table 1. For each outcome of interest, I estimate the model first using the full sample (Panel A), then using subsamples with fewer age groups around the cutoff at the time of NCSE introduction (Panels B and C). Restricting the sample to narrower bandwidths reduces imbalance of respondent characteristics between the control and treatment groups, allowing me to better isolate the effect of NCSE.

Table 1. Effects of NCSE on Perceived Upward Mobility

Note: Robust standard errors clustered at the province×cohort level are reported in parentheses. All regressions include province, cohort, and survey year fixed effects, which can be found in the online replication repository. * $ p<0.10 $ , ** $ p<0.05 $ , *** $ p<0.01 $ .

Column 1 of Table 1 presents the estimated effect of NCSE introduction on mobility experience. The positive coefficient estimates of eligibility at introduction across all three panels demonstrate clearly that individuals who became eligible for NCSE following its introduction report significantly greater upward mobility in the past decade than those who were locked out of this opportunity. Additionally, columns 2 and 3 report the estimated effect of NCSE introduction on mobility prospect. Because this outcome is about future prospect, which can also be affected by whether respondents were still eligible for NCSE at the time of the survey, I include an additional predictor, current eligibility, in an alternative model specification in column 3. Again, the positive coefficient estimates of eligibility at introduction across all three panels indicate that those who had the choice to take advantage of NCSE are more optimistic about their future mobility as well. In addition, the positive coefficient estimates for current eligibility also suggest that the presence of NCSE increases a respondent’s optimism about her future upward mobility.Footnote ²⁰ Combined, these results show that NCSE introduction has had a significant positive effect on the perceived upward mobility of those who became eligible to participate.

To address concerns associated with the identification strategy, I perform a series of robustness checks. The results are reported in the Supplementary Material. First, I examine potential imbalance between the treatment and control groups. Table A3 in the Supplementary Material presents the mean respondent characteristics for each group, as well as the unconditional and conditional differences in means between the two. Despite some expected unconditional imbalance, the two groups become more similar once the province- and cohort-level characteristics are taken into account, and increasingly so as the age bandwidth narrows. Second, to ensure that the model captures the effect of an exogenous shock instead of a general trend, I conduct a placebo test of the baseline estimation by arbitrarily shifting the NCSE introduction date in each province by 5 years forward and backward, respectively. I find no significant results on either outcome of interest (Table A5 in the Supplementary Material). Third, I address the concern that the timing of NCSE introduction might not be random and certain province-level characteristics could simultaneously affect its timing and impact individual perception of upward mobility. To address this, I use several models to predict the timing of NCSE introduction at the province level but find no significant or consistent results (Table A6 in the Supplementary Material). Furthermore, I estimate several alternative model specifications to account for province×cohort trends, individual-level characteristics, and province-specific cohort fixed effects, respectively; the results are highly consistent (Table A7 in the Supplementary Material). Last but not least, to address potential concerns with the identification of treatment, I replicate the baseline estimation with alternative samples; the results remain highly consistent (Table A8 in the Supplementary Material).

Effects on Political Preference and Attitude

In this section, I examine whether the effect of NCSE on perceived upward mobility translates into attitudinal changes. To do so, I estimate a two-stage least-squares model, where the treatment, mobility experience, is instrumented by eligibility at introduction, and the outcomes are respondents’ political preference and attitude.

Two assumptions must be met for the IV estimates to be consistent—that is, strong instrument and exclusion restriction. First, as shown in Table 1, results from the first stage are robust, regardless of age bandwidth. A strong-IV test also yields F statistics that mostly clear the standard critical value of 10 (Staiger and Stock Reference Staiger and Stock1997). Second, the exclusion restriction requires that the instrument, eligibility at introduction, affects political attitudes only through its effect on upward mobility. To meet this assumption, I restrict the analysis to respondents in the vicinity of the age cutoff, such that the assignment of treatment is random or at least as-if random.

Following common practice, I first perform an ordinary least-squares (OLS) regression and a reduced form estimation, respectively, on the relationship between political attitude and perception of upward mobility. Next, I conduct the IV analysis.

Table 2 presents results using mobility experience as the predictor. We first examine the coefficient estimates for redistributive preference in column 1. In Panel A, the OLS estimate indicates a negative correlation: when an individual has experienced greater upward mobility, her preference for income redistribution tends to diminish. However, this relationship is correlational. To estimate the causal effect of mobility experience, we look at the other two panels.

Table 2. Effects of Upward Mobility Experience on Individual Attitude

Note: This table presents the estimated effect of mobility experience on individual attitude; robust standard errors clustered at the province×cohort level are reported in parentheses. All regressions include province, cohort, and survey year fixed effects, which can be found in the online replication repository. For full results of Panel A, see Table A10 in the Supplementary Material. * $ p<0.10 $ , ** $ p<0.05 $ , *** $ p<0.01 $ .

Panel B presents the reduced form estimates, where the predictor is eligibility at introduction. The coefficient estimates are consistently negative and statistically significant. Panel C presents the IV estimates, where mobility experience is instrumented by eligibility at introduction. Regardless of age bandwidth, upward mobility that results from NCSE introduction has a significant, negative effect on redistributive preference. Taking the mean value, we can infer that every unit increase in mobility experience reduces preference redistribution by 0.605 on a scale of 1–5, which is sizable.

The consistent coefficient estimates of redistributive preference on mobility experience across all three panels indicate that individuals whose perceived upward mobility has improved as a result of NCSE introduction prefer less redistribution than those excluded from the process.

Besides redistributive preference, the rest of Table 2 presents some interesting findings. It is noticeable how individual experience with upward mobility is positively correlated with many other aspects of political attitude, including perception of political influence (Panel A, column 2) and trust in government (Panel A, columns 4–6). These results are in line with the expectation that greater social mobility has a stabilizing effect on a political system. However, these relationships disappear in Panels B and C, suggesting that the resultant upward mobility from NCSE does not directly change these political attitudes. Even though NCSE introduction has broadened upward mobility paths for those eligible to participate, it did not lead them to perceive greater political influence (column 2), nor did it make them trust the government more (columns 4–6).

Overall, the results in Table 2 are illuminating. On the one hand, NCSE eligibility does not seem to enhance citizens’ trust in government. On the other hand, by opening up entry-level political selection to all age-eligible, college-educated citizens, NCSE tempers their demand for redistribution, thereby lessening the pressure on the regime to systematically address inequality, which has grown more acute in recent decades. NCSE thus helps placate an important—and potentially combustible—segment of the Chinese society without engaging in large programmatic payouts.

There is an alternative interpretation of these results: as NCSE absorbs some of the most competent individuals, they free up other attractive employment opportunities for the rest, thus resulting in a general improvement of job market prospect for everyone, which in turn dampens their redistributive preference. In reality, however, the size of NCSE recruitment is too minuscule to cause any significant effect on the job market: at the province level, the number of NCSE openings accounts for an average of 0.019% of all employment or 0.385% of all eligible college graduates. It suggests that this alternative interpretation is unlikely due to its very limited reach. In contrast, there are many times more NCSE applicants, that is, an average of 10.02% of all eligible college graduates, reflecting significant enthusiasm over a career in civil service specifically (see Figure A5 in the Supplementary Material for more information). Based on this information, we can reason that the effect of NCSE on redistributive preference stems primarily from its capacity to move individuals into the political elite class—that is, the capacity to co-opt.

I also perform the same series of estimation with mobility prospect as the predictor (Table A9 in the Supplementary Material). The coefficient estimates of redistributive preference across all three models are highly consistent, indicating that those who perceive greater upward mobility in the future due to NCSE eligibility prefer less redistribution. The findings lend further support to my argument, as mobility prospect is based on prospective evaluation: when an individual has an opportunity to advance socioeconomically that is yet to be realized, she tends to moderate her demand for redistribution.

NCSE Institutionalization and Career Preference

As discussed earlier, NCSE boosts perceived upward mobility by availing civil service as a viable career choice to more college-educated citizens, regardless of their socioeconomic background or political connections. To test this mechanism, I first use BCPS data to plot respondents’ average level of interest in a career in government against the degree of NCSE institutionalization in their home province, as measured by years of NCSE.

As shown in Figure 4, even without taking into account any individual- or province-level covariates, the longer NCSE has been implemented in a province, the greater the interest by college-educated youths in a career in government. The sharp drop at year 10 in plot (a) is caused by one province-level municipality, Tianjin, which was the first to adopt NCSE in 1999. The low level of interest in government career in Tianjin is likely due to it being a large metropolitan city, where college-educated youths have many other attractive career opportunities.Footnote ²¹ In plot (b), I exclude the four province-level municipalities in China (i.e., Beijing, Shanghai, Tianjin, and Chongqing); the upward trend becomes even more robust.

Figure 4. Average Career Preference by NCSE Implementation Years

Note: Each plot represents the average percentage of respondents interested in a career in government, given the number of years NCSE has been implemented in their home province. The error bars represent 95% confidence intervals. See Table A14 in the Supplementary Material for details.

Next, I estimate a logit regression of career preference on years of NCSE; the results are reported in Table 3. Regressions in Panel A use the full sample. For both government organs and civil service, the coefficient estimates on years of NCSE are consistently positive and significant. On average, with each additional year of NCSE implementation in their home province, college students’ preference for a career in government increases by approximately 10%. This upward trend would eventually plateau, but it is telling that, in the initial years after NCSE introduction, it attracted more college-educated youths to considering this career path.

Table 3. Effect of NCSE Institutionalization on Career Preference

Note: This table presents the logistic regression estimates of respondents’ career preference on years of NCSE in their home province; standard errors clustered at the home city level are reported in parentheses. All regressions include fixed effects of academic major, class year, and college type. For full results in Panels B and C, see Table A15 in the Supplementary Material. * $ p<0.10 $ , ** $ p<0.05 $ , *** $ p<0.01. $

In Panels B and C, I estimate the same regression while restricting the sample to respondents with or without political connections, respectively, as measured by whether they have at least one parent who is a state official (jiguan ganbu).Footnote ²² The results show that politically unconnected students are especially responsive to NCSE institutionalization (Panel C). These are students for whom NCSE makes a meaningful difference in terms of their chances of ever becoming part of the political elite class. In contrast, students with political connections do not appear to update their career preference in response to NCSE institutionalization (Panel B). These findings lend further support to the hypothesized causal mechanism that NCSE makes more college graduates see working in government as a viable path of socioeconomic advancement.

Some other results in Table 3 are also worth highlighting. Specifically, in Panel A, both Han ethnic and hometown in big city are negatively correlated with the outcome variables, suggesting that a career in government is more attractive to students with fewer resources or alternative upward mobility paths. On the other hand, unsurprisingly, students with parent(s) in government are more likely to prefer a career in government.

Although this analysis is not strictly causal, it provides some exploratory evidence that the presence of NCSE encourages more college-educated youths—especially those without political connections—to consider civil service as a viable career choice, thus bolstering their overall perception of upward mobility. It also points to the importance of institutionalization: as meritocratic selection becomes more entrenched and better enforced, it helps convince more citizens that the channel is open to them no matter their current socioeconomic status.