Effects of anti-corruption campaigns on cesarean section rate: evidence from China

The study addresses the problem of rising and often unnecessary cesarean section (CS) rates, which exceed WHO recommendations and pose risks to mothers and infants. In China, financial incentives and entrenched corrupt practices in public hospitals are believed to encourage higher CS use due to greater revenue and convenience relative to vaginal deliveries. Against this backdrop, China’s large-scale anti-corruption campaign launched in late 2012 under President Xi Jinping may have altered social norms and increased the costs of corruption across sectors. The research question is whether anti-corruption campaigns reduced unnecessary CS by shifting the social environment that shapes physician behavior. The authors focus on Shanxi Province—known for severe corruption and a sweeping anti-corruption effort—to examine how campaign intensity affected local corruption and CS rates, with implications for reducing corrupt practices in healthcare.

The literature review covers: (1) corruption and anti-corruption efforts, noting pervasive health-sector corruption, its adverse health and system effects, and mixed or unintended outcomes of prior anti-corruption initiatives; the Xi-era campaign is unprecedented and shown to influence economic and social outcomes and change perceptions and norms around corruption. (2) CS prevalence and physician incentives: physicians can shape patient demand for CS due to financial gains, convenience, litigation fears, and organizational norms; in China, CS is a major revenue source and quicker than vaginal delivery, contributing to high national CS rates. (3) Research gap and hypothesis: few interventions target physician behavior by shifting social norms. The authors hypothesize that anti-corruption campaigns reduce unnecessary CS by influencing the local social environment of corruption. The study aligns with WHO’s Anti-Corruption, Transparency, and Accountability initiatives by exploring system-level approaches.

Data come from three sources: (1) China Corruption Investigation (CCI) Dataset (Tencent) listing officials investigated for corruption (2011–), used to construct city-year anti-corruption intensity (ACI) in Shanxi; (2) China Stock Market and Accounting Research (CSMAR) database for publicly listed firms’ business entertainment spending and revenue, aggregated to city-year ratios to proxy local corruption (ETC); (3) patient discharge records from 154 public hospitals in Shanxi (2013–2017), restricted to obstetrics/gynecology deliveries, with demographics, diagnoses (ICD-10), procedures (ICD-9-CM3), and text diagnoses. After merging on city-year and CCI availability (2013–2015), the analysis sample includes 298,311 delivery episodes from 137 public hospitals.

Outcome: CS is a binary indicator defined by primary ICD-10 diagnosis starting with O82. Exposure (endogenous): city-year corruption level ETC (listed firms’ entertainment spending/revenue ratio). Instrument: lagged anti-corruption intensity ACI_{t-1} (city-year number of officials investigated in the prior year). A cumulative ACI measure is also considered. Controls: patient age, marital status, occupation, number of births, insurance status, comorbidities; city-level population density, GDP per capita, consumption per capita; city and year fixed effects.

Econometric approach: Two-stage least squares (2SLS). First stage: ETC_ct = α + α1·ACI_{c,t-1} + X'β + city FE + year FE + ε. Second stage: outcome (CS indicators) on predicted ETC with same controls and fixed effects.

Instrument validity: Relevance demonstrated by strong first-stage associations showing higher ACI_{t-1} predicts lower ETC. Exclusion restriction discussed: potential confounding via hospital financial constraints is examined using Shanxi Public Health Yearbook data (physician counts, total revenue, government transfers); no significant relationship with ACI was found. Robustness: Using ΔETC = ETC_t − ETC_{t-1} as the second-stage regressor yields similar results. Mechanism checks: AMI subsample shows reduced stent use (surgery rate and number of stents) with higher anti-corruption intensity. Physician-level analysis: data reorganized to attribute cases to head of department, attending, or resident physicians to examine heterogeneous behavioral responses.

Descriptive: Among 2013–2015, overall CS rate fell from 26.05% (2013) to 21.1% (2015). CS rates were much higher for outpatient admissions than emergency admissions. Anti-corruption intensity rose sharply (average officials investigated: 9.6 in 2013 to 41.8 in 2015).

First stage (Table 2): Greater anti-corruption intensity reduces measured local corruption (ETC). A one standard deviation increase in ACI_{t-1} is associated with a 1.02% decrease in ETC; using cumulative ACI, a one SD increase reduces ETC by 0.5%. Both effects are statistically significant at the 1% level.

Main IV results (Table 3): Reductions in local corruption (ETC) driven by anti-corruption campaigns are associated with significant declines in CS rates. Overall, a one standard deviation decrease in ETC corresponds to about a 3.3% reduction in CS rate; effects are concentrated among outpatient admissions (≈3.5% reduction), with no significant effect for emergency admissions—consistent with reductions in discretionary/unnecessary CS.

Heterogeneity (Table 4): Effects are larger for insured patients (significant reductions, nearly double the main estimate; coefficient on ETC: −4.543***) and not significant for uninsured (out-of-pocket) patients. By parity, effects are smaller for first births (−2.425***) and much larger for second or later births (−12.745***), suggesting greater influence where patients are more likely to follow physician recommendations.

Physician-level (Table 5): No significant effects for heads of department. Significant and larger reductions for attending (overall −4.681***; outpatient −7.819***) and resident physicians (overall −4.558***; outpatient −7.623***), indicating that junior physicians’ behaviors drove the decline in unnecessary CS, particularly in outpatient settings.

Robustness: Using ΔETC yields similar magnitudes and signs. In an AMI subsample, higher anti-corruption intensity reduced both the stent surgery rate and the number of stents per surgery, supporting a broader reduction in high-margin, potentially overused procedures.

Findings support the hypothesis that anti-corruption campaigns reduced unnecessary CS by altering the social environment: decreasing the pressure to conform to corrupt norms, increasing the perceived costs and risks of corrupt actions via reporting and enforcement, and shifting social norms around corruption. Concentration of effects in outpatient admissions and among insured and higher-parity patients suggests reductions in discretionary use of CS where physician persuasion and financial incentives matter more. Larger effects among attending and resident physicians indicate changes in frontline clinical decision-making. The results imply that system-level integrity reforms can indirectly improve clinical appropriateness in sectors not directly targeted by enforcement, offering policy-relevant strategies to reduce unnecessary interventions in healthcare.

The study provides novel evidence that China’s anti-corruption campaigns substantially reduced unnecessary cesarean sections in Shanxi by lowering local corruption levels and influencing physician behavior, particularly among attending and resident physicians and in outpatient admissions. It contributes by linking broad anti-corruption efforts to health-sector outcomes, illuminating mechanisms through social norms and incentive changes, and highlighting organizational-level interventions as effective levers to reduce unnecessary CS. Future research should examine long-term effects, broader geographic contexts, and the evolution of social norms among healthcare providers, as well as potential spillovers to other clinical domains.

The analysis focuses on short-term effects (2013–2015) due to data availability, limiting insight into long-term impacts. External validity may be constrained: Shanxi’s unique economic-political context and the highly salient “Great Shanxi Political Earthquake” may amplify effects relative to other regions. Differences in healthcare systems, culture, and corruption patterns elsewhere may limit generalizability. While instrument validity checks and robustness analyses were conducted, unobserved factors correlated with anti-corruption intensity cannot be fully ruled out.