Discipline without Slack: Anti-Corruption in China's Health Sector
Abstract
Why doesn't curbing corruption always improve service outcomes? I argue that gaps in formal institutions create room for corrupt exchanges that provide integral organizational support. As organizations become dependent on these exchanges, curbing corruption can also remove operating capacity. Drawing on over twelve months of fieldwork in China's health sector, I identify two mechanisms through which privately beneficial corruption becomes publicly useful: (1) efforts to conceal payments and advance professional careers channel resources into publicly beneficial investments in hospital equipment, research, and staff training; and (2) trust among bribe givers and recipients, built through repeated exchange, facilitates information sharing and flexible credit.
To test this argument empirically in China's public hospitals, I use AI-assisted extraction of health yearbooks and firm disclosures and construct a novel proxy for reliance on corruption-generated resources with the pre-reform drug share of outpatient expenditure. Leveraging the staggered rollout of China's National Volume-Based Procurement reform, which centralized drug purchasing to constrain hospital discretion and curb corruption, I compare early- and later-exposure provinces in a difference-in-differences analysis covering all 31 provinces over 2010–2023. Contrary to standard accountability expectations, early exposure to reform was associated with higher rather than lower outpatient bills: an additional RMB 46.23 per visit, equivalent to 18% of the 2017 national average bill. Consistent with the argument that corruption fills gaps in formal support, this counterintuitive increase was especially pronounced in provinces with greater prior dependence on drug-linked informal resources. For every ten-percentage-point higher pre-reform drug share, the estimated increase was RMB 76.25 larger, equivalent to 30% of the 2017 national average outpatient bill. This pattern persists across alternative measures and specifications. Complementary analyses show sharp procurement-price reductions for covered drugs and declining selling-expense shares among firms, consistent with a real decline in corruption despite rising bills. The findings suggest that anti-corruption can expose formal deficiencies previously concealed by informal support. Lasting service improvements require not only curbing extraction, but also replacing lost funding and coordination.