Meritocracy as Authoritarian Co-Optation: Political Selection and Upward Mobility in China

Why does an authoritarian regime adopt meritocracy in its political selection? I argue that meritocracy can be used to co-opt large numbers of ordinary citizens by providing them with an opportunity of socioeconomic advancement instead of income redistribution, as long as the selection process is viewed as inclusive and rule-based. Focusing on the civil service examination in contemporary China, I examine how this meritocratic selection has shaped the relationship between college graduates and the Chinese regime. Exploiting a spatial-cohort variation in applicant eligibility, I find that the exam boosts college graduates’ perceived upward mobility, which in turn weakens their demand for redistribution even in the face of growing inequality. These findings point to an alternative mode of authoritarian co-optation and highlight the role of upward mobility in regime stability.

Note: This figure reports the aggregated yearly numbers of NCSE positions and applicants in the provincial exams from 22 provinces, as well as the aggregated numbers of new college graduates from these provinces.Data used in this graph are compiled by author from media reports; some provinces are missing due to lack of information.Note: Plots in this figure are based on province-level data between 1995 and 2013; they illustrate the size of NCSE recruitment relative to the size of the labor market.The number of all eligible individuals in a province in a give year is calculated by summing the 13 cohorts of college students who graduated between that year and 12 years prior, such that everyone was between age 22 and 35.Data on NCSE recruitment are compiled by author from media reports.2) report OLS coefficients using cross-sectional data from 1999, i.e., the first year NCSE was introduced at the province level, with the outcome being the year of NCSE introduction.Columns ( 3) and ( 4) report coefficients from a fixed effect model using panel data between 1995 and 2009, with the outcome being a dichotomous variable indicating whether a province introduced NCSE in a given year and robust standard errors clustered at the province level.Columns ( 5) to ( 8) report Cox hazard coefficients, first using a cross-sectional model at the 1999 level and then using a panel model.Coefficients in odd-numbered columns are from regressions with one predictor, receptively; coefficients from even-numbered columns are from a horse race regression with all predictors.Robust standard errors are shown in parentheses.Note: This table presents logistic regression estimates of respondents' career preference on years of NCSE in their respective home provinces; standard errors clustered at home city level are reported in parentheses.All regressions include fixed effects of academic major, class year, and college type, which, along with coefficient estimates on province characteristics, can be found in the online replication depository.* p < 0.10, * * p < 0.05, * * * p < 0.01

Figure
Figure A4: Provincial Exams: Trends in Individual Provinces

Figure
Figure A5: Provincial Exams: Size of Recruitment

Table A1 :
Official Policies Governing NCSE

Table A4
Note:This table reports regression results on whether province-level characteristics have any predictive power on the timing of NCSE introduction in each province.Columns (1) and (

Table A7 :
Robustness of Baseline Estimation with Alternative Model SpecificationsThis table reports coefficient estimates on eligibility at introduction; 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.Panel A reports coefficient estimates from the baseline model with p-values for two standard errors, one at the province×cohort level and the other at the province level with wild bootstrap implemented.Panel B controls for GDP, GDP per Note: capita, fiscal revenue, fiscal expenditure, population, and employment at the province×cohort level.Panel C controls for individual covariates including gender, ethnicity, hukou, parent(s) in CCP and civil servant.Panel D includes an interaction between a province's 1999 GDP per capita and cohort fixed effects.*p < 0.10, * * p < 0.05, * * * p < 0.01

Table A8 :
Robustness of Baseline Estimation with Alternative Samples Note: Panel A replicates the baseline model without respondents who obtained Hukou in current locality after NCSE introduction in that province; Panel B replicates the baseline model without respondents with a postgraduate degree.All regressions include province, cohort, and survey year effects, which can be found in the online repository repository.Robust standard errors in parentheses are clustered at the province×cohort level.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A9 :
Effects of Mobility Prospect on Political Attitude This table presents the estimated effects of mobility prospect on individual attitude.Panel A reports OLS estimates; full results are reported in Table A11.Panel B reports reduced form estimates, and Panel C presents IV Note: estimates.All regressions include province, cohort and survey year fixed effects, which can be found in the online replication repository.Robust standard errors clustered at the province×cohort level are reported in parentheses.*p < 0.10, * * p < 0.05, * * * p < 0.01

Table A10 :
OLS Estimate of Mobility Experience on Political Attitude This table represents the OLS estimates of mobility experience on political attitude.All regressions include province, cohort and survey year fixed effects, which can be found in the online replication repository.Robust standard errors clustered at the province×cohort level are reported in parentheses. *

Table A11 :
OLS Estimate of Mobility Prospect on Political Attitude This table represents the OLS estimates of mobility prospect on political attitude.All regressions include province, cohort and survey year fixed effects, which can be found in the online replication repository.Robust standard errors clustered at the province×cohort level are reported in parentheses. *

Table A15 :
Heterogeneous Effects of NCSE Institutionalization This table presents logistic regression estimates of respondents' career preference on years of NCSE in their respective home provinces; standard errors clustered at home city level are reported in parentheses.All regressions include fixed effects of academic major, class year, and college type, which, along with coefficient estimates on province characteristics, can be found in the online replication depository.

Table A16 :
Heterogeneous Effects of NCSE Institutionalization: Alternative Subgroups