Introduction

Cesarean section (CS) is a vital and life-saving medical procedure for both mothers and infants. However, in many countries, the CS rate exceeds the recommended population rate of 10–15% set by the World Health Organization (WHO) (World Health Organization, 2015). For example, in 2018, the CS rate reached a high of 47.8% in urban areas of China (Long et al., 2022). A growing body of literature emphasizes that unnecessary CS procedures can lead to a spectrum of short-term and long-term adverse effects for mothers and infants (Althabe et al., 2006; Villar et al., 2006; Betran et al., 2015; Costa-Ramon et al., 2017; Sandall et al., 2018). Many countries have undertaken efforts to reduce the unnecessary use of CS (Bertran et al., 2018).

One potential factor contributing to the increasing CS rate is physician incentives. Physicians often steer patient demand to align with their own self-interest, which may encompass financial incentives, personal convenience, and concerns related to malpractice (Gruber and Owings, 1996; Brown III, 1996; Costa-Ramon et al., 2017; Halla et al., 2016; Dubay et al., 1999; Currie and MacLeod, 2008; Pilvar and Yousefi, 2021). In many countries, CS procedures can generate higher revenue than vaginal births, providing financial incentives for physicians to encourage women to opt for CS even when it may not be the most appropriate choice for their situation. Such behaviors exhibited by physicians can be interpreted as a manifestation of corruption, defined as the “misuse of entrusted power for personal gain” (Pozsgai-Alvarez, 2020; Fu et al., 2023).

In China, the prevalence of corrupt behaviors among physicians, driven by financial incentives, posed a significant challenge to the public health sector. As China underwent market-oriented economic reforms in the 1980s, the government substantially reduced financial support for public hospitals, forcing these institutions to heavily rely on self-financing (Fu et al., 2017). The inflexible pricing structure that undervalued doctor services compelled physicians to seek profits through alternative means, including over-prescribing expensive drugs and consumables, performing unnecessary surgeries, and accepting bribes from patients (Zhang et al., 2014; Shi et al., 2018; Li et al., 2022; Fu et al., 2023). Consequently, corrupt behaviors became ingrained as a social norm within the public hospital environment (Fu et al., 2023). CSs have long been considered the primary source of revenue for obstetrics Department in China (Bogg et al., 2010). The cost of a vaginal delivery typically ranges between 500 RMB to 1500 RMB, while CS prices can vary from 3000 RMB to 12,000 RMB depending on the hospital’s level (Bogg et al., 2010; Liu et al., 2018; Meng et al., 2019). In addition, the average labor of vaginal delivery lasts 12–24 h, while the operation of CS only takes 30–60 min. Given the amount of resources involved and the substantial profit margins associated with CSs compared to vaginal deliveries, it is an open secret that physicians prefer CSs over vaginal deliveries in Chinese public hospitals.

In recent years, the Chinese government has made significant efforts to combat corruption, most notably through the anti-corruption campaign initiated by President Xi Jinping in late 2012. Enforced by the Central Commission for Discipline Inspection (CCDI), this campaign targeted all sectors in China and has successfully identified and convicted over 1.5 million government officials. Given the scale and importance of the anti-corruption campaign, recent studies have examined its effects on a range of outcomes, including economic growth, land prices, car purchasing, entrepreneurship, poverty incidence, and coal mine mortality (Nie et al., 2016; Chen and Zhong, 2020; Kong and Qin, 2021; Xu et al., 2021; Han et al., 2022), as well as corporate activities such as firm innovation, sales, productivity, profitability, financial report quality, corporate fraud, asset allocation, and security market crash risk (Nie et al., 2016; Xu and Yano, 2017; Chen et al., 2018; Zhang, 2018; Jin et al., 2019; Hope et al., 2020; Kong et al., 2020; Cai et al., 2022). Importantly, in addition to these findings, research has demonstrated that the anti-corruption campaigns implemented under the Xi administration have reshaped societal perceptions of corruption in China (Wang and Dickson, 2021; Fang, 2023; Pan et al., 2023).

To comprehend medical corruption in China, previous studies have identified three primary driving forces: pressure to abuse, opportunity to abuse, and rationalization (Vian, 2008; Gaitonde et al., 2016; Fu et al., 2017, 2023). Within the context of public hospitals in China, physicians may engage in corrupt practices influenced by social normative pressure and financial pressure. Moreover, in environments where there is low punishment for corruption, physicians may perceive their actions as justified when directing patients toward procedures with higher profit margins. Although the anti-corruption campaigns under the Xi administration may not explicitly target the public health sector, they could potentially diminish physicians’ corrupt behaviors by alleviating social normative pressure and raising the opportunity costs associated with corruption.

In this study, we aim to examine the effects of anti-corruption campaigns on CS rates and explore the underlying mechanisms. Utilizing data from the China Corruption Investigation Dataset and publicly listed firms between 2013 and 2016, we first estimate the effects of anti-corruption campaigns on local corruption levels, indicated by the ratio of firms’ business entertainment spending to revenue. These findings offer insights into how anti-corruption campaigns influence the social environment surrounding corruption. Subsequently, we utilize discharge records from 137 public hospitals and employ an instrumental variable (IV) approach to analyze how anti-corruption initiatives affect physician decision-making regarding CS procedures through alterations in the social landscape of corruption. To delve deeper into the mechanisms at play, we restructure the data at the physician level and meticulously examine the impacts on physician behaviors.

Our findings contribute to the literature in the following aspects. Firstly, we provide novel empirical evidence regarding the impact of anti-corruption campaigns on the public health sector, illustrating their effectiveness in curbing unnecessary CSs by influencing physician behaviors. Secondly, our results offer insights into the mechanisms through which anti-corruption campaigns exert their influence. Thirdly, our findings offer valuable insights into actionable strategies for combating corruption within the healthcare sector and implementing interventions aimed at reducing the prevalence of CS among healthy women and newborns.

The remainder of the paper is organized as follows. The next section reviews the related work and proposes the research questions. The section “Data and methods” introduces the details of data and empirical approaches. In the section “Results”, the results are presented. The last section presents a discussion of the findings and implications for public health policies.

Literature review and research questions

Corruption and anti-corruption efforts

Corruption represents a pervasive challenge within the healthcare systems of many developing countries (Transparency International, 2019). Its impact extends beyond financial losses, limiting access to health services and undermining the various dimensions crucial for effective health system performance (Garcia, 2019). The direct financial toll of medical corruption alone is estimated to constitute at least 6% of global healthcare spending (Fu et al., 2023). Furthermore, corruption exerts detrimental effects on population health, contributing to increased infant and child mortality, exacerbating antibiotic resistance, and impeding progress in controlling both communicable and chronic diseases (Rönnerstrand and Lapuente, 2017; Dincer and Teoman, 2019; Ferrari and Salustri, 2020).

Given the adverse effects of corruption, governments around the world have exerted various efforts in combating corruption (Mungiu-Pippidi, 2017). However, the successful story is few. Instead, many are shown to have unintended consequences (Doig et al., 2007; Mutebi, 2008; Batory, 2012; Lichand et al., 2016; Johnston, 2017; Nishijima et al., 2022). For example, Lichand et al. (2016) and Nishijima et al. (2022) show that the anti-corruption audits in Brazil have caused worsened health indicators.

Among all the anti-corruption efforts, the campaign initiated under the Xi administration stands out as one of the most unprecedented in terms of magnitude and duration (Fang, 2023). Previous studies examining this anti-corruption campaign have found significant influences on economic growth, firm activities, poverty alleviation, and coal mine mortality (Nie et al., 2016; Lin et al., 2016; Xu and Yano, 2017; Chen et al., 2018; Zhang, 2018; Jin et al., 2019; Hope et al., 2020; Kong et al., 2020; Chen and Zhong, 2020; Kong and Qin, 2021; Xu et al., 2021; Han et al., 2022; Cai et al., 2022). Crucially, beyond these outcomes, studies have shown that the anti-corruption campaign under the Xi administration has altered people’s perceptions, leading to a fundamental shift in social norms regarding corruption (Wang and Dickson, 2021; Fang, 2023; Pan et al., 2023).

Prevalence of CS and physician incentives

CS stands out as the most prevalent major surgical intervention globally (Boerma et al., 2018). While some attribute the escalating CS rate to maternal requests, recent studies suggest that, in the absence of current or previous complications, most women worldwide do not express a preference for CS (Betran et al., 2018). Instead, pregnant women often defer to health providers as the pivotal influence on their delivery mode decision (Ji et al., 2015; Kingdon et al., 2018). Physicians, however, often shape patient demand to align with their self-interest (McGuire, 2000). For example, Gruber and Owings (1996) demonstrated that American gynecologists, faced with income decline due to a local fertility rate drop, compensated by favoring CS over vaginal delivery. This behavioral pattern raises concerns regarding the potential correlation between physicians’ incentives and the increasing CS rates, shedding light on broader issues of corruption within healthcare systems. Drawing from prior research, concerns about litigation, organizational and peer group norms, financial benefits, and convenience are the main drivers of physician’s use of CS (Brown III, 1996; Costa-Ramon et al., 2017; Halla et al., 2016; Dubay et al., 1999; Currie and MacLeod, 2008). Studies have shown that interventions such as educational packages, mandatory second-opinion policies, audits, and feedback could reduce CS but also bring psychological pressure to physicians, causing unintended consequences (Althabe et al., 2004; Chaillet et al., 2015; Chen et al., 2018). Instead, interventions targeted at the organizational level are proven to be more effective (Betran et al., 2018; Melo and Menezes‐Filho, 2023; Pilvar and Yousefi, 2021).

In China, CSs are frequently considered the primary revenue source for obstetrics departments, comprising ~70–85% of the department’s total revenue (Bogg et al., 2010). The cost of a vaginal delivery typically ranges between 500 RMB and 1500 RMB, while CS prices can vary from 3000 RMB to 12,000 RMB depending on the hospital’s level (Bogg et al., 2010; Liu et al., 2018; Meng et al., 2019). In addition, the average labor of vaginal delivery lasts 12–24 h, while the operation of CS only takes 30–60 min. Given the amount of resources involved, the substantial profit margins associated with CSs compared to vaginal deliveries serve as an incentive for physicians to perform more CS procedures. Consequently, the CS rates in China were above 40% in 2008, one of the highest around the world (Long et al., 2022).

Research gap and hypothesis

Drawing on existing research, it is clear that shaping physician decisions regarding the mode of delivery is crucial in reducing unnecessary CS. While prior literature has delved into interventions aimed at decreasing unnecessary CS, few have specifically targeted interventions to influence physician behaviors by instigating shifts in social norms. It is not unexpected that research in this realm is limited, given the formidable challenge of enhancing public integrity. Many previous anti-corruption initiatives have been demonstrated to fall short in addressing such a task.

The anti-corruption campaigns initiated under the Xi administration present a distinctive opportunity to explore the potential impact of changes in the social environment on reducing unnecessary cesarean sections (CS). Previous research has demonstrated that these campaigns not only successfully transformed public perceptions of corruption but also exerted influence on local social environments. Given that the campaigns are not specifically directed at the health sector, this study seeks to investigate whether shifts in the social environment related to corruption can exert an influence on physician behaviors, thereby leading to a reduction in CS rates. Expanding on this analysis, the study proposes the following research hypothesis:

Hypothesis (H): Anti-corruption campaigns could reduce unnecessary CS rates by influencing the social environment of corruption.

Indeed, combating corruption in the health sector has garnered increased attention in recent years. Instances of corrupt practices, including embezzlement, pilfering of drugs and supplies, and overbilling, have regrettably been prevalent in the public health sector. Consequently, there is a pressing need to unravel effective measures to curtail corruption in the health sector. Understanding these measures is imperative for policymakers seeking precise strategies to combat corruption effectively. The findings of this study lend substantial support to the Anti-Corruption, Transparency, and Accountability (ACTA) initiatives advocated by the World Health Organization (WHO). By endorsing these initiatives, the study underscores their potential to serve as robust frameworks in the ongoing global efforts to promote integrity and transparency in healthcare systems.

Data and methods

Data sources

The primary data sources for this study encompass three datasets: the China Corruption Investigation (CCI) Dataset, the China Stock Market and Accounting Research (CSMAR) Database, and hospital discharge records. The CCI Dataset was compiled from Tencent, China’s largest internet company. Tencent established an online repository containing information on all corruption investigations conducted across China since 2011 (Wang and Dickson, 2021). Covering officials from the central to local levels, this database includes details such as each official’s name, position, locality, rank, and the reason for the investigation. The dataset’s accuracy has been corroborated by previous studies on corruption investigations in China (Wang and Dickson, 2021; Chen et al., 2023; Chu et al., 2024).

For our analysis, we concentrate solely on Shanxi Province due to data availability. Shanxi, situated in the north-central region of China, is an underdeveloped inland province covering ~156,300 square kilometers. In 2017, its GDP per capita stood at 41,242 RMB, which is ~69% of the national average (59,660 RMB). Renowned for its abundant coal resources, Shanxi boasts an estimated one-third of the People’s Republic of China’s total coal reserves. The socio-economic landscape of Shanxi provides fertile ground for collusion between local politicians and business elites, rendering it one of China’s most corruption-ridden provinces. The anti-corruption campaigns in Shanxi proved highly effective, resulting in a sweeping overhaul of the political establishment with unprecedented ferocity, an event referred to as the “Great Shanxi Political Earthquake.” In the CCI Dataset, there were, in total, 729 cases between 2013 and 2015 for officials located in Shanxi. The intensity of anti-corruption campaigns is defined as the city-year level of the total number of officials under investigation.

We retrieve firm information from the CSMAR database. The dataset contains detailed information on publicly listed firms, including categories of business spending and revenue. We first restrict the sample to firms located in Shanxi Province based on the registration address and then calculate the ratio between a firm’s entertainment spending and revenue. For our analysis, we aggregate the ratio to the city-year average to define a city’s level of corruption.

Information on physician behavior, diagnosis, and patient characteristics came from the discharge record data provided by the Health Commission of Shanxi Province. They contain patient records from 2013 to 2017 for 154 public hospitals. For our main analysis, we restrict the sample to Obstetrics and Gynecology patients with delivery records. The sample consists of 685,842 inpatient episodes. The dataset contains information on hospital identifier, expenditure during the hospitalization, inpatients’ demographics (e.g. age, gender, marital status), and clinical information (procedures performed identified by International Classification of Diseases, 9th Revision, Clinical Modification, Volume 3, ICD-9-CM3, and diagnoses assigned identified by International Classification of Diseases, 10th Revision, ICD-10). The advantage of this dataset is that it also contains descriptive texts on up to 9 diagnoses for each patient, which provides essential information on identifying CS. Our final sample requires the merging of the three datasets. We merge the three datasets based on city and year. Since the CCI dataset is only available from 2013 to 2015, the sample size is 298,311 after merge with 137 public hospitals.

Econometric analysis

We employ a two-stage least-square (2SLS) estimation to examine the effects of anti-corruption campaigns on CS rates through influences in social environments. Our second-stage regression is as follows:

$${Y}_{{ict}}={\beta }_{0}+{\beta }_{1}{ET}{C}_{{ct}}+{X}^{{\prime} }\theta +{\gamma }_{c}+{\delta }_{t}+{\epsilon }_{{ict}}$$
(1)

where \({Y}_{{ict}}\) represents our outcomes of interest for patient i at city c in year t. The outcome of primary interest is CS. We define whether a patient experienced CS as a dummy variable, with 1 being the first three digits of the ICD-10 code of the patient’s primary diagnosis as “O82” and 0 otherwise. The level of corruption in a current social environment is measured by \({{ET}{C}}_{{ct}}\), the city-year average of firms’ business entertainment spending to revenue ratio. \(X\) is a vector of control variables, including the patients’ age, marital status, occupation, number of births, insurance status, and number of comorbidities. For the city level, we control for population density, gross domestic product (GDP) per capita, household consumption per capita. We also control city and year-fixed effects, respectively.

If we directly apply Eq. (1), we would anticipate a positive association between CS rates and city-level corruption, as unnecessary CS is also indicative of corrupt behavior (Supplementary Table S1). Naturally, there exists an endogenous relationship between the current corruption level and anti-corruption intensity. In other words, a city with a higher corruption level is likely to undergo a higher level of anti-corruption efforts. To mitigate the issue of reverse causality, we hypothesized that a city’s current level of corruption is an outcome of the anti-corruption campaign efforts from the previous year. Thus, we use \({{AC}{I}}_{c,t-1}\) as an instrument for \({{ET}{C}}_{{ct}}\). Our first-stage regression is as follows:

$${{ET}{C}}_{{ct}}={\alpha }_{0}+{\alpha }_{1}{{AC}{I}}_{c,t-1}+{X}^{{\prime} }\theta +{\gamma }_{c}+{\delta }_{t}+{\epsilon }_{{ict}}$$
(2)

where \({{ET}{C}}_{{ct}}\) is the city-level corruption and \({{AC}{I}}_{c,t-1}\) is the city-year total number of officials being investigated in year t−1.

Instrument validity

We interpret the IV estimates as local average treatment effects, providing the association of being exposed to different anti-corruption campaign intensities. One requirement for a valid instrument is that it is correlated with the endogenous explanatory variable. We show in the first stage that \({{AC}{I}}_{c,t-1}\) is strongly associated with a reduction in city-level corruption at time t.

The exclusion restriction necessitates that the intensity of the anti-corruption campaign does not correlate with unobserved factors influencing CS rates. Previous studies have pointed out that financial constraints faced by hospitals may influence physicians’ decisions on CS (Gruber and Owings, 1996). To explore this hypothesis, we gather data on hospital operations, including the number of employed physicians, total annual revenue, and total annual transfers from the government, from the Shanxi Statistical Yearbook in Public Health. We replicate our first-stage regression with these variables as the outcomes. The results are presented in Supplementary Table S2, where we observe no significant relationship between anti-corruption campaigns and hospital financial burdens.

Results

Descriptive statistics

Table 1 presents summary statistics for the entire sample and by different years. The average age of the patients is around 28 years old, with more than 90% being married. About 20% of the patients are formally employed, and over 60% of them are insured by health insurance. The average number of comorbidities is around 2, and the number of births is slightly above 1, indicating a relatively young and healthy sample with either their first or second deliveries. Since the inception of the anti-corruption campaigns in 2013, the overall CS rate has consistently decreased, from 26.05% in 2013 to 21.1% in 2015. Compared to the CS rates among patients admitted through emergency care, the CS rates for patients admitted through outpatient schedules are much higher. In 2013, the CS rate for patients admitted through outpatient schedules was 30.6%, significantly exceeding the WHO recommended rate. The intensity of the anti-corruption campaigns increased rapidly over time, with an average of 9.6 officials being investigated in 2013, rising to 41.8 officials in 2015.

Table 1 Summary statistics.
Full size table

Figure 1 compares the time series patterns of the intensity of anti-corruption campaigns across the sample cities. The campaign intensity varied substantially over time and across cities. For example, Taiyuan, the capital city of Shanxi Province, witnessed the largest number of officials being investigated in 2014, while Xinzhou had only fewer than 10 investigated officials. The substantial variation in intensity over time and across cities helps us identify the relationship between anti-corruption campaigns and local corruption levels.

Fig. 1
figure 1

Variation in anti-corruption Intensity across city and over time. This figure presents the city-year average number of officials being investigated from 2013 to 2015.

Full size image

First-stage results

Table 2 presents the relationship between anti-corruption campaigns and local corruption levels. In Column 1, the total number of officials being investigated in the year before used as an instrumental variable (IV), shows that a one standard deviation increase decreases the local corruption level by 1.02%. Additionally, we construct a cumulative variable that calculates the total number of officials being investigated in a city from the year 2013 until the sample year, as shown in Column 2. A one standard deviation increase in the cumulative intensity leads to a 0.5% reduction in local corruption levels. Both estimations are statistically significant at the 1% level. We adopt the first specification for our second-stage results.

Table 2 Effects of anti-corruption campaigns on corruption levels (IV first stage).
Full size table

Main results

Table 3 presents the instrumental variable (IV) estimates, shedding light on the relationship between anti-corruption campaigns and CS rates. In Column 1, we find a positive correlation, indicating that anti-corruption campaigns are linked to a significant reduction in CS rates. Specifically, a one standard deviation reduction in the local corruption level, as measured by ETC, through anti-corruption campaigns is associated with a 3.3% decrease in the CS rate.

Table 3 Effects of anti-corruption campaigns on CS rates (IV second stage).
Full size table

To delve deeper into the impact across different admission scenarios, Column 2 examines patients admitted through emergency care, while Column 3 focuses on those admitted through outpatient services. Interestingly, no significant effects of anti-corruption campaigns on CS rates are observed among patients admitted through emergency care. In contrast, for patients admitted through outpatient services, a one standard deviation reduction in local corruption is associated with a substantial 3.5% decrease in CS rates. One potential explanation is that obstetric patients admitted through emergency care are more likely to experience complications or urgent conditions necessitating CS, whereas those admitted via outpatient services are less likely to do so. These findings suggest that the observed reductions in CS rates are predominantly driven by a decline in unnecessary CS operations, particularly among patients admitted through outpatient services.

Heterogeneous effects

To explore potential heterogeneous effects and potential mechanisms, we analyze subgroups of the patients according to insurance status and number of births. In Table 4, columns 1 and 2 show the results when the sample is split with respect to insurance status. Patients with insurance are usually less sensitive to price compared to those without coverage. Consequently, they are more inclined to adhere to physicians’ recommendations for more expensive procedures (Xiang, 2021). The findings in Table 4, columns 1 and 2, validate this hypothesis. The reduction in CS rates is found to be significant only among patients with health insurance, with no significant reduction observed for those who paid completely out-of-pocket. The estimates for the reduction in CS rates among insured patients are larger than the main estimates, almost twice in magnitude.

Table 4 Effects of anti-corruption campaigns on CS rates by insurance status and number of births.
Full size table

In Table 4, columns 3 and 4 report the results when the sample is split according to number of births. Column 3 restricts the sample to patients with their first delivery, and column 4 is for patients with more than 1 delivery. Smaller effects are observed in the reduction of CS rates among first deliveries, while significantly larger effects are found for second or later deliveries. Parents are more discerning in their choice of delivery method for the first birth. Unlike subsequent births, where parents may be more open to accepting medical recommendations for CSs, they are generally less inclined to opt for CS during the first birth unless deemed medically necessary. This cautious approach is often taken to preserve the opportunity for a second child. The point estimate for the second and later births is almost six times that of the main estimate.

The results pertaining to patients with insurance coverage and those undergoing second or later births underscore the importance of altering physicians’ behaviors to decrease unnecessary CSs. The impact of anti-corruption campaigns is most pronounced when patients exhibit lower sensitivity to their delivery choices and are more inclined to adhere to physician recommendations for CSs.

To explore potential heterogeneous effects by the level of physician leadership, we take advantage of the anonymized physician information and reorganize the data at the physician level. We categorize physicians into three levels: head of department, attending physician, and resident physician. The attending physician is often the primary obstetrician overseeing the patient’s care and evaluates various factors such as the mother’s health, the baby’s condition, and any complications during labor to determine if a CS is necessary. Therefore, the decision of CS is usually made by the attending physician in consultation with the patient. The resident physicians are usually under the supervision of the attending physicians. Therefore, we would expect that the anti-corruption campaigns would reduce the CS rates among attending and resident physicians. Table 5 reports the estimates. No significant effects on CS rates are observed when grouped by the head of the department. However, larger effects in the reduction of CS rates are found for both attending physicians and resident physicians, almost twice the magnitude of the main estimate. Furthermore, the reductions in CS rates for attending and resident physicians are primarily associated with patients admitted through outpatient care, aligning with our primary findings. These results suggest that anti-corruption campaigns significantly influenced behaviors among junior physicians, leading to a reduction in unnecessary CS practices.

Table 5 Effects of anti-corruption campaigns on CS rates by levels of physicians.
Full size table

Robustness checks

Readers may express concern about the relationship between corruption in the current year (\({{ET}{C}}_{{ct}}\)) and the intensity of anti-corruption campaigns in the previous year, suggesting that it may not accurately capture changes in local corruption levels. In response to this concern, we introduce the difference between the corruption level in the current year and that of the previous year (i.e. \(\Delta {{ET}{C}}_{{ct}}={{ET}{C}}_{{ct}}-{{ET}{C}}_{c,t-1}\)). We then conduct the instrumental variable (IV) estimation with \(\Delta {{ET}{C}}_{{ct}}\) as our second-stage main regressor. Supplementary Table S3 provides details on the first-stage and second-stage estimates. Remarkably, the results closely align with the main findings in terms of both magnitudes and signs, reinforcing the robustness and reliability of our primary results.

To investigate whether the anti-corruption campaigns altered physician behaviors, we performed additional analyses by concentrating on patients diagnosed with acute myocardial infarction (AMI). We assess whether the anti-corruption campaigns resulted in a decrease in the utilization of stents. Stents are high-value medical consumables with very high prices (an average of 13,000 RMB per stent). Our hypothesis is that among the treatment alternatives for AMI—drug therapy and angioplasty—physicians may be motivated to opt for more stent surgeries due to their higher financial returns, a similar logic comparable to CS surgery versus vaginal delivery. Supplementary Table S4 reports the effects of the anti-corruption campaigns on the rate of stent surgery and number of stents used per surgery. We find that the increases in anti-corruption campaign intensity lead to a reduction in both the stent surgery rate and the number of stents used per surgery. Supplementary Table S5 reports estimated results by the level of physician leadership. We find that physicians at all levels reduced stent surgery rates, with the largest effects being found among the head of the department. Similar to the main results on CS rates, the reductions in stent surgery rates are associated with patients admitted through outpatient care. These results provide further evidence that anti-corruption campaigns significantly influenced physicians’ behaviors, resulting in reductions in corrupt behaviors.

Discussion and conclusion

We conducted an analysis examining the impact of anti-corruption campaigns on CS using discharge record data from public hospitals in Shanxi Province. Our primary discovery indicates that anti-corruption campaigns have the potential to significantly decrease CS rates in public hospitals. The influence of anti-corruption campaigns in Shanxi, while not explicitly directed at the healthcare sector, has manifested in observable shifts in physicians’ behaviors.

Our findings regarding the CSs align with results from prior studies confirming the connection between physician incentives and CS rates (Gruber and Owings, 1996; Costa-Ramon et al., 2017; Betran et al., 2018; Pilvar and Yousefi, 2021). Driven by personal motivations such as financial gains and personal convenience, physicians may be inclined to perform CS procedures with little or no additional benefits for patients. Interventions that modify physicians’ incentives have been demonstrated to effectively control CS rates (Gruber and Owings, 1996; Pilvar and Yousefi, 2021). In comparison to existing literature, our study suggests that anti-corruption campaigns may serve as a potentially effective means to influence physician behaviors, complementing traditional health policy reforms such as educational interventions, health system upgrading, hospital policy and payment reforming, and midwives/doulas training (Althabe et al., 2004; Chaillet and Dumont, 2007); Betran et al., 2018; Zhang et al., 2020).

The reduction in CS rate was particularly larger after 2014, the year when the Shanxi political environment experienced the “Great Shanxi Political Earthquake”. As physicians were not the primary target of the investigations between 2013 and 2015, we propose the following potential mechanisms that anti-corruption campaigns may change physician behavior based on conceptual frameworks from earlier studies (Vian, 2008; Gaitonde et al., 2016; Onwujekwe et al., 2019; Fu et al., 2023). First, the anti-corruption campaigns reduced physicians’ pressure to abuse power. Some physicians may be peer-pressured to conduct CS as preferring CS to vaginal delivery might be prevalent among colleagues. The anti-corruption campaigns relieved these physicians and gave them reasons to doge potential social isolation. Our estimated results at the physician level provide suggestive evidence.

Secondly, the anti-corruption campaigns curtailed opportunities for corruption. Throughout these campaigns, citizens were empowered to directly contact all levels of the CDI by mail, phone, email, or Internet, reporting instances of corruption in public sectors. The existence of whistleblowing mechanisms could potentially impact physicians’ behaviors, instigating concerns about the possibility of being reported. This increased physicians’ opportunity cost of corruption.

Thirdly, the anti-corruption campaigns might have influenced physician decisions by altering the social norms associated with corruption. Prior studies have demonstrated that interventions modifying descriptive norms of corruption can significantly reduce corrupt behaviors (Gächter and Schulz, 2016; Mungiu-Pippidi, 2017; Fehr, Schurtenberger, 2018; Köbis et al., 2022). Performing surgical procedures with higher financial returns could have become a norm. The anti-corruption campaigns may have shifted social norms away from corrupt behaviors in Chinese society, thereby influencing physician conduct.

We conclude by acknowledging certain limitations and identifying areas for future research. Owing to data constraints, our analysis has focused on the short-term effects of anti-corruption campaigns on CS. It is imperative to delve into the long-term effects and assess the evolution of social norms pertaining to corrupt behaviors among physicians. While our findings align with previous studies on the efficacy of anti-corruption campaigns under the Xi administration (Hope et al., 2020; Hao et al., 2020; Xu et al., 2021; Han et al., 2022), we recommend exercising caution when extrapolating our results to external settings. Due to data availability constraints, our analysis exclusively examines the impact of anti-corruption campaigns on CS rates within Shanxi Province. The effects of these campaigns may be particularly pronounced in Shanxi compared to other provinces, owing to the unique social and economic conditions prevalent in the region. The notable “Great Shanxi Political Earthquake” garnered significant social attention and could potentially result in broader societal impacts. As suggested by earlier studies, anti-corruption efforts should take local circumstances and needs into consideration. Due to potential variations in healthcare systems, cultural contexts, and the nature of corruption across different regions, this may impact the generalization of our conclusions.