When gender quotas protect the powerful: Lessons from China’s civil service. Guest post by Jiawei Lyu

This is the 24^th in this year’s series of posts by PhD students on the job market.

Affirmative action policies are typically designed to help underrepresented groups gain access to opportunities. But what happens when they are used to protect the historically dominant? In my job market paper, I study a “reverse” gender quota in China’s civil service - an effort to preserve male representation that ultimately reduced candidate quality and weakened performance.

A surprising case of affirmative action for men

In many parts of the world, women are now entering public service in large numbers. China is no exception. Since the launch of its national merit-based civil-service exam in 2011, women have consistently outperformed men. By the mid-2010s, women made up almost two-thirds of new civil-service recruits in China’s tax bureaus for example.

Some agencies saw this as progress; others saw imbalance. Beginning in 2016, some[JL1]  county tax bureaus - the frontline of China’s revenue administration - introduced a “one-to-one” gender quota, splitting every hiring batch evenly between men and women. Job postings were literally paired: one slot “for male candidates,” one “for female candidates.” Figure 1 shows the initiation year of the quotas.

The goal, officials said, was gender balance. But the effect was to cap women’s entry into a system that remained overwhelmingly male in leadership and culture. Because tax bureaus are central to fiscal capacity, the policy created a natural test of how such hiring policies affect both representation and performance.

Building a new dataset

To answer this, I built a new dataset combining over 90,000 civil service job postings from 2011 to 2023 with county-level tax revenue data from 2011–2022. The postings - scraped and digitized from China’s National Civil Service Examination portal and public announcements- show exactly when and where quotas were adopted, how many positions were open to each gender, and who was ultimately hired.

Because the quotas were adopted gradually and voluntarily by different counties, I compare counties before and after adoption to otherwise similar ones that never adopted, covering 601 counties with non-missing tax revenue data. The analysis also accounts for differences across counties and years, as well as city-level policies that may evolve over time. Standard errors are adjusted for patterns within each county over time. By the end of 2023, 2,410 counties—about 85 percent of all counties—had implemented the quota at least once. This variation makes it possible to show the effects of the policy on women’s representation, candidate quality, and institutional performance.

Who chose to adopt the quota—and why?

Before looking at its effects, I first examined what predicts which counties adopted the quota.

Adoption was not random. Counties that were larger, wealthier, and less urbanized were significantly more likely to introduce the quota. Local history also mattered: counties that had hired a higher share of female recruits in recent years were more prone to adopt the policy soon after.

The pattern suggests both economic and institutional motives. In places where tax departments already had many women in new entry positions, officials appeared eager to “rebalance” the intake. And within counties, the quota was far more likely to appear in tax-collection roles - jobs involving audits, fieldwork, and client contact - than in administrative or clerical posts.

In short, the policy tended to emerge where women’s representation was rising fastest, and where senior managers viewed men as more suited to external or revenue-facing work. These adoption patterns foreshadow the results that follow.

What happened when the quota arrived?

The findings reveal a consistent pattern of unintended consequences.

• Fewer women applied and were hired.

In the first year of adoption, the share of women among new hires fell by about 15 percentage points relative to a baseline average at 67.7%, and the total number of applicants dropped by nearly 30 percent, largely because female candidates opted out once gender-restricted postings appeared.

• Candidate quality declined.

Written-exam cut-off scores fell by about 0.12 standard deviations overall, and by 0.77 standard deviations for male-targeted positions. In other words, men admitted under the quota scored lower than the women they replaced.

• Productivity fell.

Counties that adopted the quota collected less tax revenue after implementation. A one-percentage-point increase in the share of quota positions is associated with a 0.085 percent drop in revenue. Given that quota positions account for about 55 percent of jobs in adopting counties, this translates into an average 4.7 percent decline in total tax revenue—roughly US $4 million per county each year. These losses are large for an institution whose effectiveness directly funds local public services. Figure 2 below illustrates the revenue decline following quota adoption.

Could the decline simply reflect broader economic conditions? Unlikely. County GDP actually rose slightly after quota adoption. Administrative-case data show fewer approvals processed by tax officials but no change in penalties issued to taxpayers, implying that the fall stemmed from bureaucrats’ productivity, not taxpayer compliance.

Other robustness checks - alternative estimators, spatially adjusted errors, and an instrumental-variable approach using even-numbered job postings - all confirm the main pattern: performance declined after the quota.

Why did performance fall?

Two mechanisms may explain the drop in performance.

• Lower personnel ability.

Written-exam performance strongly predicts later revenue outcomes. A one-standard-deviation increase in cut-off scores is associated with about a 1.5 percent increase in tax revenue. By forcing bureaus to hire a fixed number of men, the quota replaced higher-scoring women with lower-scoring men - particularly in tax-collection roles that directly drive revenue.

• Biased institutional gender norms.

The policy also reinforced male-dominated workplace norms. In more corrupt counties- where male networks and informal socializing often shape deals and promotions - the revenue loss after quota adoption was much larger. The corruption level is measured as the number of corruption cases before the quota adoption from 2011 to 2015. Evidence from local drinking cultures and promotion data supports this interpretation: where male social networks are stronger, the productivity penalty of the quota is worse.

Summary and Policy Implications

Supporters of these quotas argued that they would “restore balance” by keeping both genders equally represented. Yet balance on paper can hide inequality in practice. When the starting point is a male-dominated institution, imposing equal hiring by gender effectively protects incumbents rather than creating opportunity.

This study shows that reverse affirmative action can be costly - not only for women, but also for the organizations implementing it. By prioritizing gender balance over merit, counties reduced both representation and productivity.

The lesson extends beyond China. Around the world, debates over “too many women” in schools, public administration, or professional fields are emerging. Some universities have considered “affirmative action for boys”; others argue for gender quotas to preserve balance. These policies may sound equitable but risk undermining efficiency and productivity when applied in settings where men still dominate leadership and culture.

Ultimately, hiring rules that appear neutral can alter both opportunity and performance. Understanding when and why institutions adopt them is key to designing reforms that strengthen, rather than weaken, state capacity. When efforts to ensure fairness end up preserving advantage, both inclusion and effectiveness are diminished.

Jiawei Lyu is a PhD student at the University of Pittsburgh.