The Selection and the Making of Civil Servants: Evidence from China’s College Graduate Civil Services Program

A random half of the job posts will implement the selection algorithm at the status quo, which places heavy weight on candidates’ political obedience. The other random half of the job posts will implement a new algorithm, which maintains the same factors of considerations but places heavier weight on competence.

2022-04-01

2022-09-01

We use both surveys and social security records to examine characteristics of interest along the
selection margins, as well as the outcomes of interests in the treatment margin.

- We elicit a wide range of political and economic preferences, attitudes, and beliefs, including various measures on political obedience. For example, alignment with the Chinese Communist Party’s official policy agenda, support of various public policies, performance evaluation and trust of various levels of the government. We elicit Big 5 personality characteristics, as well as fundamental preferences: risk preference, time preference, altruism, and reciprocity using Falk et al. (2016) module. We use lab-in-the-field behavioral games to elicit respondents’ preference for redistribution, following Fisman et al. (2007). The design allows us to separately identify respondents’ tradeoff between equity and efficiency. In addition, we incorporate the “Public Service Motivation Survey” to measure respondents’ broad alignment with the public service career track.
- We elicit various measures on competence. For example, we collect grades in college. We measure respondents’ knowledge on various socioeconomic events and facts, both regarding China as a whole and the local regions specifically.
- We design a visual-based experimental module to elicit how respondents process politically and economically framed information (e.g. local poverty condition), how they interpret new signals and update their beliefs. The potential existence of bias in information acquisition and belief updating allows us to examine patterns such as motivated beliefs.
- We collect administrative information on job performance during the CGCS program. This includes performance evaluations by peers, supervisors, and villagers, as well as objective and aggerate performance measures at the village level.
- We use social security data to trace career trajectories and income of all individuals. This allows to investigate questions such as which civil servants get promoted, who decide to stay in the public sector, and whether those civil servants who return to private sector converge to their counterparts who start off in the private sector.

We partner with the Guangdong local government to randomly alter the selection algorithm during the 2020 recruitment cycle. A random half of the job posts will implement the selection algorithm at the status quo, which places heavy weight on candidates’ political obedience. The other random half of the job posts will implement a new algorithm, which maintains the same factors of considerations but places heavier weight on competence.

Randomization done in office by a computer.

Job posting.

No

0

According to our simulation based on the application data from the 2017-18 recruitment cycle, the new algorithm would change selection outcomes among 1/3 of the positions. Hence, we will have 600 recruited applicants, and 600 applicants who would have been recruited under the algorithms not implemented in the corresponding posts.

600 applicants in the treatment group, and 600 applicants in the control group.