Intercity personnel exchange is more effective than policy transplantation at reducing water pollution
Main
Large spatial disparity exists in water pollution governance, and underdeveloped areas suffer more from such environmental inequality1,2,3. Globally, Western Europe performs the best in wastewater collection and treatment, with 88% wastewater being collected and 86% being treated, while South Asia and sub-Saharan Africa have the lowest collection and treatment rates (31% and 16%, respectively)4. Within a country, such spatial disparity also prevails. For instance, in China, southern cities have more wastewater treatment plants and attract more green investments, while less-developed northern cities often have lower levels of wastewater treatment capacity5,6. Disparity in water pollution governance leads to disparity in water pollution exposure as well as in health and economic well-being, with long-lasting effects that persist over decades7,8,9. Unequal vulnerability to environmental risks (for example, unsafe drinking water and water pollution-related disasters) across cities and populations thus calls for more effective policy tools to alleviate inequality in water pollution governance10,11.
A popular solution is that underdeveloped cities can transplant policies from developed cities with successful experiences12,13. Policy transplantation is a process where policies from one city are borrowed by and replicated in another city14,15 (Supplementary Note 1). It enables cities to develop new policies at lower costs16,17. Furthermore, borrowing widely recognized policies increases the legitimacy of policy-making18. Policy transplantation has thus become a standard feature of contemporary policy-making on water management and environmental governance in European countries, the United States and China12,19,20 (Supplementary Note 1). However, policy transplantation often does not work well. First of all, one-size-fits-all policies do not exist, and local contexts must be taken into consideration. Successful policy transplantation depends on many local factors such as local socioeconomic structures. Policy imitations that fail to acknowledge these complexities may not generate expected effects21,22,23. Environmental governance is thus much more than policies. Certain experiences on how to tailor policies according to new local contexts and how to implement policies effectively are important for high-quality environmental governance13. Such governance experiences are often tacit knowledge that cannot be obtained by merely copying policies from successful cities.
In this study, we stress that intercity personnel exchange may work better than policy imitation, and it can bring tacit knowledge from developed cities to less-developed ones. This is in line with previous studies, which have documented that tacit knowledge does not spread evenly across space; rather, it diffuses primarily following the movements of human actors24,25,26,27,28. We use one type of intercity personnel exchange as an example—intercity exchange of city leaders. City leaders that moved from elsewhere to a city are new to the city. They possess experiences of pollution governance gained from elsewhere19,21. They also help formulate intercity cooperation between the current city where they work and cities where they used to work, which can also facilitate the diffusion of knowledge and experiences on pollution reduction29,30. Hence, intercity exchange of city leaders may help alleviate inequality in water governance and contribute to the reduction of water pollution, especially in underdeveloped cities. However, little has been done to examine whether intercity personnel exchange could be a possible policy instrument and its effect on pollution governance and reduction.
Intercity personnel exchange in the public sector, though understudied, is a widely observed institutional arrangement in many countries such as China, Russia, Kenya and Italy31,32,33. It also has a long history, which can be traced back to the period of 1854–1930 in the British Empire and the Qin Dynasty of China (220–207 bc)34,35 (Supplementary Note 2). Among those countries, China provides a perfect context to study the impact of intercity personnel exchange on environmental governance and water pollution. First, unlike Western countries where local politicians are elected by local residents, city leaders in China are appointed by their political superiors. The procedure is complicated. Many factors, including city leaders’ capabilities, available positions and political purposes, are taken into consideration in the process of appointment. Although some factors such as a leader’s capabilities are associated with the possibility of him/her being exchanged, the exact timetable and destination of intercity personnel exchanges are hard to predict ex ante and are beyond an individual’s control29,30. In short, intercity exchanges of city leaders in China are plausibly exogenous. Second, compared with countries such as Russia and Kenya, we observe more frequent personnel exchanges in China, which provides more spatial-temporal variations for valid modeling analysis31 (Fig. 1). Third, China has faced the challenges of severe water pollution and huge intercity disparities in environmental governance. Hence, China’s central government increasingly stresses the idea of sustainable development and has incorporated environmental performance as one of the key criteria for the evaluation of local leaders. The latter have become more motivated to reduce pollution (Supplementary Note 2).

The blue lines represent the personnel exchange of city leaders between two cities, and the width of line is proportional to the number of exchanged city leaders. The base map is from the Standard Map Service System, Ministry of Natural Resources of China (http://bzdt.ch.mnr.gov.cn/index.html).
Using China as an example, this study provides the first nationwide estimates of the effects of intercity personnel exchange on water pollution and on inequality in environmental governance. While previous studies relied on macro-level data and often focused on some regions/industries2,6,36, this paper is conducted for one country using more disaggregated data. Specifically, we employ a number of fine-grained datasets on local leaders’ curriculum vitae, firm behaviors, patents and city policy texts and so on. We first find that intercity exchange of city leaders has resulted in a 4.78–15.26% reduction in different types of water pollutants, which accounts for 39.45–57.98% of China’s total water pollution reduction in this period. Second, we show that intercity exchange of city leaders leads to more water pollution reduction than intercity policy transplantation. Furthermore, in the former case, firms are more likely to improve their production process than to purchase end-of-pipe facilities. Third, we confirm that exchanged city leaders can not only bring experiences of environmental governance, but also help formulate intercity cooperation for green investments and innovations. After personnel exchange, city leaders are likely to initiate new place-based policies targeting at green industries. They also have stronger career incentives to do so.
Results
Effect of intercity personnel exchange on water pollution
Spatial inequality of water pollution governance exists not only between countries but also within one country (Supplementary Fig. 1), which may threaten economic development and public health, especially in underdeveloped cities with low levels of water pollution governance. It is thus urgent to find feasible policy instruments to alleviate the inequality of water pollution governance and to reduce water pollution in underdeveloped cities. We hereby argue that intercity personnel exchange can provide a possibility, and use intercity exchange of city leaders as an example to test the effect.
Figure 2a shows the effect of intercity exchange of city leaders on water pollution by estimating equation (1) using the difference-in-differences (DID) model (Supplementary Table 2). After intercity exchange of city leaders, chemical oxygen demand (COD), ammonia nitrogen and wastewater all decrease, though to different extents (that is, by 8.66%, 15.26% and 4.78%, respectively, at the firm level) (Supplementary Note 5).

a, The impacts of personnel exchange on firm-level COD, ammonia nitrogen and wastewater. The data are presented as the estimated coefficients (dots) ±95% confidence intervals (error bars). The P values for the coefficients of personnel exchange are all 0.000. The sample sizes are 566,765 (COD), 382,730 (ammonia nitrogen) and 583,250 (wastewater). b–d, The contribution of intercity exchange of city leaders to water pollution reduction compared with the national total water pollution reduction during 2006–2013 for COD (b), ammonia nitrogen (c) and wastewater (d). The solid red lines represent the overall change in national industrial water pollution (including the effects of personnel exchanges), and dashed red lines represent the counterfactuals (without the effects of personnel exchanges). The negative values denote water pollution reduction, while positive values represent increases in water pollution. The red shades display the water pollution reduction induced by intercity exchange of city leaders during 2006–2013, which is calculated based on the estimated overall effects in baseline regressions (Supplementary Table 3). The overall change in national industrial water pollution is calculated using data from the Ministry of Ecology and Environment of China (Supplementary Note 6).
We calculate the water pollution reduction induced by the intercity exchange of city leaders during 2006–2013 based on the estimated effects, and then compare that with the overall change in national industrial water pollution (Supplementary Note 6). During the study period, intercity exchange of city leaders has resulted in a 0.93 million ton (MT) reduction in COD, a 0.12 MT reduction in ammonia nitrogen and a 1.93 gigaton (GT) reduction in wastewater (Supplementary Table 3). Figure 2b–d further displays that city leader exchanges contribute to 39.45% (COD), 42.57% (ammonia nitrogen) and 57.98% (wastewater) of the national total water pollution reduction.
We then ran some robustness checks (Supplementary Note 7, Supplementary Tables 4–12 and Supplementary Figs. 3 and 4). First, we testified the validity of the quasi-random feature of intercity personnel exchanges of city leaders. The results suggest that, for each city leader, the exact timetable and destination to be rotated can be hardly predicted ex ante (Supplementary Table 4). Second, we leveraged on the event study approach to check the parallel trend assumption, the primary assumption of the DID estimation. As shown in Supplementary Fig. 3, the effects of personnel exchanges are significant only 2–3 years after a city receives an exchanged city leader, but insignificant in the pre-exchange period. Third, we controlled for more confounding factors by including industry-year and province–industry fixed effects (Supplementary Table 5). Fourth, we utilized several alternative DID modeling strategies to address the concern of heterogeneous treatment effects and multiple switches of treatment status in canonical staggered DID estimators (Supplementary Tables 6–9). Fifth, we ran placebo tests with Monte Carlo simulations for 500 times, to check whether our estimates are biased by other confounding factors, especially those unobservable and hard-to-quantify ones (Supplementary Fig. 4). Sixth, we clustered standard errors at the city level to take into account within-city autocorrelation (Supplementary Table 10). Seventh, we employed alternative indicators, study periods and samples (Supplementary Table 11). We constructed water pollution intensity indicators as the dependent variables and redefined intercity exchange of city leaders in a stricter way. We also sought to partial out the effects of the 2008 financial crisis and major political events by excluding samples in the corresponding years. Provincial capitals were also excluded in another test since they are not comparable to other cities. Finally, we further tested the effectiveness of city leader exchanges and considered the potential pollution substitution between water and air pollution. Supplementary Table 12 documents a negative effect on firm-level air pollution and verifies the effectiveness of intercity personnel exchanges.
We also conducted further analyses to examine the heterogeneous effects of intercity personnel exchange. First, we differentiated intraprovincial and interprovincial exchanges of city leaders, and re-estimated equation (1). Supplementary Table 13 shows that the effect of interprovincial exchanges is much greater than that of intraprovincial exchanges. While intraprovincial exchanges reduce firm-level water pollution by 3.90–14.27%, interprovincial exchanges have led to a 24.87–37.37% decrease in the three types of water pollution. We further show that such a difference is due more to the crossing of provincial borders than to physical distance (Supplementary Note 8 and Supplementary Tables 14 and 15). Second, we tested whether a city leader being exchanged to his/her hometown moderates the relationship between intercity personnel exchange and water pollution reduction. The effect is insignificant (Supplementary Note 8). On the one hand, such leaders are familiar with local contexts and thus know how to reduce pollution more efficiently. On the other hand, they have political and cultural ties with local networks in their hometowns, which may lead to government–business collusion and corruption37. Finally, although intercity exchange of city leaders contributes to water pollution reduction, we are not suggesting city leaders should be exchanged too frequently as it takes time for exchanged city leaders to generate real impacts in new cities. As presented in Supplementary Table 17, most exchanged city leaders have been only exchanged once during our study period, while less than 5% of them have been exchanged more than twice. Given this, frequent intercity exchange does not have a substantial impact on our results (Supplementary Table 18).
Intercity personnel exchange outperforms policy transplantation
Underdeveloped cities often imitate and transplant successful environmental policies from developed cities. We argue that intercity exchange of city leaders could be an alternative policy tool to alleviate inequality in environmental governance, and compare the effect of personnel exchange and policy transplantation. We collected official documents from city-level governments regarding water pollution governance during 2003–2013, and used the text similarity approach to quantify intercity policy transplantation (Methods).
Figure 3 shows that the combination of personnel exchange and policy transplantation has the largest effect, contributing to 10.51%, 21.55% and 7.17% reductions in COD, ammonia nitrogen and wastewater, respectively. The effect of personnel exchange only is smaller (8.47%, 13.64% and 4.56% reductions in COD, ammonia nitrogen and wastewater, respectively). However, the coefficients of ‘transplantation only’ are positive in the case of COD and wastewater and not statistically significant in the case of ammonia nitrogen. If a city merely imitates environmental policies from elsewhere, but does not have specific knowledge and experiences on the implementation of those policies, such simple policy transplantation may not be effective as expected. As noted above, there is no one-size-fits-all policy. Hence, policies copied from other cities may be not suitable for a new city, and may even have some negative effects16,21,22,23. We further show that exchanged city leaders motivate firms to improve their production process, rather than purchasing more end-of-pipe facilities (Supplementary Note 9, Supplementary Fig. 5 and Supplementary Tables 20–30).

a–c, The effects of intercity personnel exchange and policy transplantation on firm-level COD (a), ammonia nitrogen (b) and wastewater (c) estimated from equation (2) (Supplementary Table 19). For COD regression, the P values for the coefficients are 0.000 (exchange and policy transplantation), 0.000 (exchange only) and 0.002 (transplantation only) (N = 566,678). For ammonia nitrogen regression, the P values for the coefficients are 0.000 (exchange and policy transplantation), 0.000 (exchange only) and 0.871 (transplantation only) (N = 382,730). For wastewater regression, the P values for the coefficients are 0.000 (exchange and policy transplantation), 0.000 (exchange only) and 0.044 (transplantation only) (N = 583,250). All dependent variables are logged. The data are presented as the estimated coefficients (dots) ±95% confidence intervals (error bars).
Mechanisms
Understanding the mechanisms of how intercity exchange of city leaders contributes to water pollution reduction could help us to understand why it can outperform intercity policy transplantation and provide important lessons for future policy-making. Here, we examine three mechanisms.
First, exchanged city leaders carry tacit knowledge such as governance experiences with them while moving. They could obtain knowledge and experiences on environmental governance from their previous positions in other cities and carry those knowledge and experiences to their new positions. However, it is hard to quantify governance experiences directly. Here, we adopt two indirect measures: (1) if a city leader used to work in environment-related departments, he/she may obtain relevant knowledge and experiences. We differentiate between exchanged city leaders with and without working experiences in environment-related departments, which denote departments of environmental protection and regulation, natural resources and water resources at various levels (Supplementary Note 10). Figure 4a–c reveals that exchanged city leaders with working experiences in environment-related departments contribute more to water pollution reduction than those without such experiences (Supplementary Tables 31 and 32). (2) It is fair to assume that city leaders moving from first-mover cities with good environment governance to late-comer cities with poor environmental governance tend to have more advanced governance experiences that can be used to reduce water pollution in late-comer cities (Supplementary Note 10). We therefore extend equation (1) by differentiating two types of intercity personnel exchanges: ‘exchanges from first-movers to late-comers’ and ‘other exchanges’. Since cities with poor environmental governance are likely to have more severe water pollution and subsequently more room for improvement, we controlled for a city’s initial pollution intensity. We find that city leaders moving from first-mover cities to later-comer ones contribute more to water pollution reduction (Supplementary Tables 33–37). Specifically, as shown in Fig. 4d–f, intercity exchange of city leaders from first-mover cities to late-comer ones is associated with 15.09%, 35.11% and 5.26% reductions in COD, ammonia nitrogen and wastewater, respectively (Supplementary Table 33). In the case of other exchanges, the effects are relatively lower (7.93% in COD, 12.55% in ammonia nitrogen and 4.59% in wastewater). Intercity exchange of city leaders can thus help alleviate inequality in water pollution and environmental governance.

a–c, The effects of exchanged city leaders with and without working experiences in environment-related departments on firm-level COD (a), ammonia nitrogen (b) and wastewater (c) (Supplementary Note 10 and Supplementary Table 31). By environment-related departments, we refer to departments of environmental protection and regulation, natural resources and water resources at various levels. The P values for the coefficients of ‘with experiences in environment-related departments’ and ‘without experiences’ are all 0.000. d–f, The effects of two types of intercity personnel exchanges on firm-level COD (d), ammonia nitrogen (e) and wastewater (f). We ranked all cities according to their water pollution reduction rates. The top 33% cities with the highest levels of water pollution reduction rate are defined as first-mover cities since they have the best performance in pollution control and environmental governance, while the bottom 33% cities are defined as late-comer ones. We then decomposed intercity exchange of city leaders (exchange) into two types: ‘exchanges from first-movers to late-comers’ and ‘other exchanges’, and include both in equation (1), while controlling for a city’s initial pollution intensity (Supplementary Note 10 and Supplementary Table 33). For the COD and ammonia nitrogen regressions, the P values for the two coefficients are all 0.000. For wastewater regression, the exact P values are 0.047 and 0.000, respectively. The data are presented as the estimated coefficients (hollow circles) ±95% confidence intervals (error bars).
Second, an exchanged city leader helps formulate intercity cooperation, especially between the city where he/she used to work and the city where he/she is currently working30,36. Here, we examine three types of intercity cooperation: intercity innovation collaborations and new investments into green industries and high-tech firms. Investments in green industries can push forward green transitions and reduce pollution directly. Innovation collaborations and investments in high-tech firms contribute to the development of technologies and innovation, which subsequently increase production efficiency and reduce pollution intensity in all industries. We used the Chinese Intellectual Patent Office (CIPO) dataset that contains information on patents to calculate intercity innovation collaboration and the National Enterprise Credit Information Publicity System (NECIPS) dataset that contains information of all registered firms to calculate intercity investments. We adopted an extended DID model to investigate the effects on intercity cooperation (Methods).
Figure 5a–c shows that a city leader moving from city c1 to c2 leads to an increase in investments in green industries from c1 to c2 by 9.31%, and an increase in investments in high-tech firms from c1 to c2 by 5.10%. It also results in a 4.03% increase in patent collaborations between c1 and c2. We further explore whether the increase in investments and patent collaboration is driven by the creation of new connections (extensive margin) or the extension of existing connections (intensive margin) (Supplementary Tables 38 and 39). The coefficients of ‘exchange’ are significant and positive in extensive margin models, but are not significant or only slightly significant in intensive margin models. This suggests that intercity exchange of city leaders stimulates intercity investments and patent collaborations, and such effects are predominantly driven by the extensive margin. A city leader moving from city c1 to c2 leads to the increase in the extensive margin of investments in green industries from c1 to c2 by 5.42%, in the extensive margin of investments in high-tech firms from c1 to c2 by 3.97% and in the extensive margin of patent collaborations between c1 and c2 by 2.74% (Supplementary Tables 38 and 39). Such effects are sizable, accounting for 27.53%, 19.74% and 13.75% of the current investment flows in green industries and high-tech firms, and patent collaborations, respectively (for example, 27.53% = 0.0542/0.1969 × 100%).

a–c, The effects of intercity exchange of city leaders on intercity investments in green industries (a), intercity investments in high-tech firms (b) and intercity patent collaborations (c). Each dot and error bar represents a separate regression by estimating equations (3)–(5) (Supplementary Tables 38 and 39). The blue dots and bars represent the estimated coefficients and the 95% confidence intervals for the overall effects on the logarithms of intercity cooperation indicators (that is, intercity investments in green industries, intercity investments in high-tech firms and intercity patent collaborations). The orange dots and bars represent the estimated effects of personnel exchanges on the probabilities of creating new connections (extensive margin), while red dots represent the effects of personnel exchanges on existing connections (intensive margin). In a, the P values for the coefficients of personnel exchange are 0.002 (overall effect), 0.000 (extensive margin) and 0.056 (intensive margin), respectively, while in b, the P values are 0.017, 0.041 and 0.291, and in c, the P values are 0.003, 0.011 and 0.948, respectively. d, The effects of personnel exchange on firm taxes. We regress the log form of tax on personnel exchange (Supplementary Table 40). The blue dots and error bars represent the estimated coefficients and the 95% confidence intervals for the overall effects. The orange dots and bars represent the estimated effects of personnel exchanges on tax in the case of green industries, while red dots represent the effects of personnel exchanges in the case of other industries. The P values for those three coefficients are all 0.000. The data are presented as the estimated coefficients (dots) ±95% confidence intervals (error bars).
Third, exchanged city leaders tend to initiate new favorable policies for green industries in new cities. Though both exchanged city leaders and locally promoted city leaders can initiate new industrial policies, exchanged leaders are new comers and more likely to attack old institutions and initiate new ones, especially for green industries38. Here, we focus on one type of favorable industrial policy, tax credit, which is a widely used policy tool. Our dataset provides information on firm-level tax during the study period. We show that exchanged city leaders tend to initiate more tax cuts and, more importantly, such effects are greater in green industries (Fig. 5d and Supplementary Tables 40 and 41). Overall, firms in green industries have a 5.67% reduction in tax, which is 30.05% higher than those in other industries (that is, 30.05% = (0.0567 – 0.0436)/0.0436 × 100%). This indicates that exchanged city leaders would channel more fiscal resources toward green industries.
Motivations of exchanged city leaders
We then examined exchanged city leaders’ motivations to reduce water pollution and see if they are more motivated than locally promoted leaders. Here, we use two types of indicators to proxy the motivation of city leaders.
The first type is associated with the ages of city leaders. In China, city leaders above 60 years are forced to retire39. Old city leaders, especially those close to 60 years, are less motivated and less likely to initiate changes, while young city leaders are more motivated to have good performance with the hope of standing out and being promoted40. We thus compare the ages of exchanged and local city leaders. Figure 6a plots the cumulative distribution functions of the ages of exchanged and local city leaders in the first year of their terms. It provides some preliminary graphical evidence that exchanged city leaders are younger and more likely to be motivated to make some progress when they start new terms. We then formally testified this by regressing city leaders’ ages in the first year of their terms on ‘exchange’ (Fig. 6b), which is similar to between-group t-tests but taking into account heteroscedasticity (model 1). We can see that exchanged city leaders are younger. This finding still holds after controlling for year fixed effects (model 2) and year and city fixed effects (model 3).

a, Cumulative distribution function of the ages of exchanged and local city leaders in the first year of their terms. b, We regressed city leaders’ ages in the first year of their terms on ‘exchange’ in model 1, and control year fixed effects (FEs) in model 2 and year and city FEs in model 3 (Supplementary Table 42). The exact P values are 0.026 (model 1), 0.006 (model 2) and 0.001 (model 3). The data are presented as the estimated coefficients (dots) ±95% confidence intervals (error bars). c, We divide cities into three groups according to the difficulty of their assigned pollution reduction targets for COD and ammonia nitrogen, and then define three dummies, ‘high/medium/low target’. We interact those three dummies with ‘exchange’ and include them in equation (1) (Supplementary Note 11 and Supplementary Table 43). China’s central government began to set reduction targets for ammonia nitrogen in 2011 and does not set targets for wastewater. We therefore exclude the samples before 2011 in the analysis of ammonia nitrogen and do not take into account wastewater in this analysis. For COD regression, the exact P values are 0.000 (high target), 0.000 (medium target) and 0.030 (low target). For ammonia nitrogen regression, the exact P values are 0.000 (high target), 0.001 (medium target) and 0.014 (low target). The data are presented as the estimated coefficients (dots) ±95% confidence intervals (error bars).
Second, exchanged city leaders’ strong motivations can be also seen from their responses to pollution control targets set by the central government. Pollution control has become one of the key factors for the central and provincial governments to evaluate city leaders’ performance, and city leaders who fail to achieve preassigned pollution reduction targets will have less chance to be promoted and may even be removed from their current positions41,42. The central government has set up some place-specific pollution reduction targets for major pollutants. The targets for COD and ammonia nitrogen have been issued since 2006 and 2011, respectively. No target has been set for wastewater. We divided cities into three groups according to the difficulty of their assigned pollution reduction targets for COD and ammonia nitrogen, and then define three dummies, ‘high/medium/low target’, to capture this difference. We interact those three dummies with ‘exchange’ and include them in equation (1) (Supplementary Note 11 and Supplementary Table 43). Exchanged leaders in cities with more difficult targets are likely to face larger pressure from the central government, and they tend to perform better in water pollution reduction (Fig. 6c and Supplementary Tables 43 and 44). Compared with local city leaders, exchanged city leaders are more motivated to respond to targets set by the central government and to have better performance under greater pressure from the top. This finding is less evident in the case of ammonia nitrogen as the targets were not announced until 2011, which is not long enough to see its effect in our study period (2006–2013).
Simulations with more intercity exchanges of city leaders
In our study period, less than 50% cities have exchanged leaders (Supplementary Table 1). We thus see great potential in water pollution reduction if exchanged city leaders, as a policy tool, could be introduced to more cities. We estimate the potential water pollution reduction in three scenarios (Supplementary Table 46):
-
(1)
Scenario 1: if all cities without exchanged leaders are assigned with city leaders exchanged from cities in the same province.
-
(2)
Scenario 2: if all cities without exchanged leaders are assigned with city leaders exchanged from cities outside the province.
-
(3)
Scenario 3: if all cities with leaders exchanged from the same province are reassigned with exchanged leaders from other provinces.
We first show the estimated effects of intercity exchange of city leaders on water pollution reduction in our baseline models (Fig. 2b–d and Supplementary Note 6). We then calculate the potential reductions in three scenarios during the study period based on the estimated effects of intraprovincial and interprovincial exchanges (Methods and Supplementary Table 46). Finally, we compare estimated effects in baseline models and three scenarios to the real value of national total water pollution.
Extended Data Fig. 1 shows that in all three scenarios, there will be additional water pollution reduction. While the simulated reduction in scenario 1 is slightly greater than the estimated pollution reduction in our baseline models, the simulated reductions in scenarios 2 and 3, in which we take into account interprovincial exchanges, are much greater. Specifically, in scenario 1, there will be a 0.94 MT additional pollution reduction in COD, 0.12 MT in ammonia nitrogen and 1.84 GT in wastewater, which are around 2.76% (COD), 5.01% (ammonia nitrogen) and 0.99% (wastewater) of the national total water pollution during 2006–2013. In scenario 2, there will be an additional reduction of 3.32 MT in COD, 0.32 MT in ammonia nitrogen and 1.17 GT in wastewater, about 9.77% (COD), 13.13% (ammonia nitrogen) and 6.28% (wastewater) of the national total. In scenario 3, there will be an additional reduction of 2.26 MT in COD, 0.21 MT in ammonia nitrogen and 8.30 GT in wastewater, respectively. The potential reduction is about 6.64% (COD), 8.55% (ammonia nitrogen) and 4.45% (wastewater) of the national total. In short, intercity exchange of city leaders could be a useful policy tool to generate water pollution reduction.
Discussion
To reduce water pollution and pollution-induced risks, lagging cities can learn lessons and transplant successful policies from first-mover cities. However, many failures in policy transplantation have occurred in recent years, as lagging cities just simply copied policies from elsewhere without taking into account complicated local socioeconomic contexts. We hereby argue that intercity personnel exchange can be a feasible and more promising policy instrument in alleviating inequality in water governance and water pollution. In this research, we focus on one special group of people as an example—city leaders exchanging between cities—who can do more than merely imitating policies from elsewhere. So far, little is known about the effect of intercity personnel exchange on environmental governance and pollution reduction. However, such studies, especially those conducted for a large geographical territory with disaggregated data, are necessary and can provide important policy implications.
Using large-scale disaggregated data on city leaders, policy texts, firm-level water pollution, investments and patents, we provide the first nationwide estimates of the effect of intercity exchange of city leaders on water pollution, and compare that with the effect of intercity policy transplantation. Our results show that intercity exchange of city leaders could lead to an 8.66% reduction in COD, a 15.26% reduction in ammonia nitrogen and a 4.78% reduction in wastewater at the firm level. This effect is sizable, as it contributes to 39.45% (COD), 42.57% (ammonia nitrogen) and 57.98% (wastewater) of the national total water pollution reduction during our study period. Furthermore, we compare the effect of personnel exchange and that of policy transplantation. We find strong evidence that intercity exchange of city leaders as a policy tool not only outperforms policy imitation in the magnitude of water pollution reduction, but also provides a better way to reduce water pollution, that is, pushing firms to improve their production process rather than purchasing more end-of-pipe facilities. To explain these patterns, we further examine three mechanisms. First, exchanged city leaders often carry governance experiences to new cities. Second, they help formulate intercity cooperation. Finally, they often initiate new favorable policies, especially for green industries in new cities. Exchanged city leaders are often more motivated to perform better in pollution reduction. Our simulation points out the huge potential in additional water pollution reduction if intercity exchange of city leaders, especially interprovincial exchange, could be introduced to more cities.
The first contribution of this research is to propose an alternative policy instrument for improving environmental governance and reducing water pollution, especially in lagging cities. Intercity exchange of city leaders not only helps the diffusion of advanced experiences on environmental governance, especially when leaders move from first-mover cities to later-comer ones, but also fosters the formulation of intercity cooperation. Both aspects are important for environmental governance and pollution reduction, but none can be realized by merely replicating policies from successful first-mover cities. Policy transplantation could even have negative impacts on pollution control, since policies copied from other cities may be not suitable for a new city. We also show that policy transplantation could be beneficial if being used together with intercity exchange of city leaders. Exchanged city leaders with advanced experiences know how to adjust transplanted policies accordingly and how to implement those policies in ways that are efficient and suitable for their cities.
This research also contributes to the literature on the mobility of different groups of people. While existing studies pay much attention to the mobility of entrepreneurs, scientists and scholars24,25,26, less has been done to explore the mobility of political leaders. On the one hand, while moving across space, political leaders, like entrepreneurs and scholars, can also spread knowledge to new cities and promote intercity cooperation. On the other hand, the movement of political leaders has some distinct effects. First, unlike entrepreneurs who spread business knowledge or scientists who carry the knowledge of specific technologies, political leaders help diffuse knowledge and experiences on governance21,43. Second, political leaders could also facilitate intercity investments and innovation collaborations like entrepreneurs and scientists, but they do not directly participate in intercity investment and innovation activities. Rather, they act more like facilitators, and utilize their formal or informal connections in cities where they used to be and cities where they are currently working to support the formulation of intercity cooperation27,29,30. Third, political leaders can push forward the issuance of new place-based policies38, which cannot be done by other groups of migrants, such as entrepreneurs and scientists. Finally, while entrepreneurs and scholars are more likely to move from underdeveloped cities to developed ones44, which increases spatial inequality, we point out that intercity exchange of city leaders has the potential to mitigate spatial inequality in water pollution and water governance.
Methods
Firm-level water pollution and economic data
The firm-level water pollution data we use are from the Pollution Emissions of Industrial Firms (PEIF) database, compiled by the Ministry of Ecology and Environment. This database has three features. First, it covers 85% industrial pollution in China, which is of high representativeness. Second, nearly all main pollutants are recorded in this database, such as sulfur dioxide (SO2), COD, ammonia nitrogen (NH3) and nitrogen oxides (NOx). Third, the database provides detailed information on the amount of pollution generated, pollution emission and pollution removal, as well as information on pollution abatement facilities, which allows us to study firms’ pollution reduction. This study focuses on water pollution. Air pollution is analyzed as a robustness check for the effectiveness of personnel exchange.
Our second firm-level dataset is the Annual Survey of Industrial Firms dataset, compiled by the National Bureau of Statistics of China. It contains all state-owned firms and non-state-owned firms with annual sales above 5 million RMB, covering 85% industrial outputs of China. It provides firm-level information, including basic firm information (for example, each firm’s name, identification number, registration type, start year and the industry it belongs to) and some other economic information (for example, each firm’s employment, output, tax and profit).
We followed Brandt et al.45 and Qi, Zhang and Chen46 to clean our data. First, we removed all incorrect values that violate the accounting rules and those that are wrongly recorded, such as negative asset, negative output and liquid/fixed asset that is greater than total asset. Second, for each pollutant, we restricted samples to firms whose pollution emission is above 0. Finally, based on firms’ names and their unique registration codes, we merge the PEIF and Annual Survey of Industrial Firms databases. Our final dataset contains 585,953 observations of 90,104 industrial firms during 2006–2013. See Supplementary Note 3 for further details on our firm-level data.
City leader data
By hand-collecting leaders’ biographical information from government websites and other online official records, we constructed a database of city leaders’ features and political careers for two kinds of city leaders in China—mayors and party secretaries of cities. For each leader, the dataset records personal traits such as his/her name, gender, birthplace, year of birth and educational background, and political careers such as the location and job title of his/her current (and last) position, and the years when he/she starts and ends each term.
The key independent variable of interest is the intercity exchange of city leaders. By tracing a city leader’s political career, we can know whether a city leader is moving. Specifically, we compared the cities of each city leader’s current and last position, and define him/her as an exchanged city leader if these two cities are not the same. We further differentiated two types of personnel exchanges: intraprovincial and interprovincial exchanges. We identified each exchanged city leader’s origin and destination cities, and see if those two cities are in the same province or not. See Supplementary Note 3 for further details.
Investment and patent data
Investment data are from the NECIPS dataset, which is compiled by the State Administration for Market Regulation of China. The data include information about firm locations, the year of firm establishment/exit and shareholders of each firm in China. With the information on shareholders and locations, we can trace interfirm investments and then calculate intercity investment flows. Furthermore, this study stresses two types of investments that are beneficial for green development and pollution reduction—investments in green industries and high-tech firms. First, to identify green industries, we used the PEIF database and calculated the average emission intensity of each three-digit industry in 2006, the starting year of our study period, and define green industries as those with pollution intensities below the median. Second, the information on whether a firm is a high-tech enterprise is from the list of high-tech firms identified by the Torch High Technology Industry Development Center, Ministry of Science and Technology.
Our patent data are from the CIPO dataset. The data include information on patent ID, applicants, application year, application type, International Patent Classification code and addresses. To study patent collaboration, we focused on patents with at least two applicants, and at least one of which is a firm. Using the locations and application years of patents, we can calculate intercity patent collaborations in various years.
City-level socioeconomic data
City-level socioeconomic indicators, including gross domestic product (GDP) per capita, population density and fiscal expenditure per capita, are calculated using data from China’s City Statistical Yearbooks.
Baseline model
To examine the effect of intercity exchange of city leaders on firm-level water pollution, we design the following generalized DID regression and estimate it at the city-firm-year-leader level
where c, i, m and t denote city, firm, leader position (city mayor or party secretary) and year, respectively. Yc,i,m,t represents a series of firm-level outcomes, including firm-level emissions of water pollutants (COD, ammonia nitrogen and wastewater) and firm-level taxes. All outcome variables are logged. The dummy variable exchangec,m,t takes the value of 1 if the current city leader is exchanged from another city and 0 otherwise. Zc,i,m,t is a set of city-, leader- and firm-level control variables, including GDP per capita, population density, fiscal expenditure per capita, each leader’s age and tenure, and each firm’s size, age and output. We also control for firm fixed effects λi, leader fixed effects ξl and province–year fixed effects ζp,t in the model. Both city mayors and party secretaries are included.
During our study period, 730 out of 1,243 city leaders had been exchanged across cities, accounting for 58.73% of the total. Hence, the effect of intercity personnel exchange is not driven by a small number of exchanged city leaders. Our event, that is, intercity exchange of city leaders in China, has provided enough spatial-temporal variations for statistical analysis. Furthermore, around 44% of the total firm observations have been impacted by intercity exchange of city leaders.
Text analysis and policy transplantation
We collected city-level official documents via the database constructed by the Center for Legal Information, Peking University (www.pkulaw.com), which covers all official documents of laws and the majority of regulatory documents in China. We collected the whole text of each policy initiated by city-level governments concerning water governance. We searched for city-level regulations and laws whose titles include one or more of the following keywords: water pollution (shuiwuran in Chinese), water body (shuiti), water environment (shuihuanjing), water quality (shuizhi) or water resource (shuiziyuan). We kept city laws and regulations, but excluded official replies for administrative licensing and revisions for official documents. Official replies for administrative licensing cannot be seen as policies. Revisions for official documents were excluded since their first editions have been taken into account. We also excluded official documents issued by provincial governments as we focus on policy transplantations among cities. Considering that the average of a leader’s term is 2–3 years and the time lag effects for policy imitation, we examined all regulations and laws during 2003–2013 (while the study period for baseline models is 2006–2013). Finally, we end up with 442 city-level official documents.
We employed the cosine similarity approach and calculated pairwise similarity between documents. While studying if city c1 transplants policies from other cities, for policy i initiated by city c1 in year t, we only calculated text similarity between policy i and policy j proposed by another city c2 before year t. For each city, we calculated the maximum value of these text similarities. Eventually, we considered that city c1 transplants policies on water governance from other cities if the maximum value for city c1 is beyond the median value of those maximums. We also ran several sensitivity tests to check for robustness by changing the thresholds (Supplementary Tables 22–30).
We modified the baseline model (that is, equation (1)) by replacing exchange with three dummy variables: personnel exchange (exchange only), policy transplantation (transplantation only) and the combination (exchange × transplantation)
where ({mathrm{exchang{e}}},{mathrm{onl{y}}_{c,m,t}}) takes the value of 1 if leader m of city c in year t is transferred from another city and city c does not transplant policies from elsewhere in year t, and 0 otherwise; ({mathrm{transplantatio{n},{onl{y}}}_{c,t}}) equals 1 if city c imitates policies from elsewhere and city leader m of city c is chosen locally rather than transferred from other cites, and 0 otherwise. In this case, the control group includes cities that have neither exchanged leaders nor conducted policy imitations. Other specifications remain the same as in equation (1). During our study period, 174 out of 284 cities had transplanted other cities’ water-related policies, among which 85 cities had also had exchanged city leaders.
Intercity cooperation
We first testified the effects of intercity exchange of city leaders on intercity cooperation that is beneficial to water pollution reduction by estimating equation (3), and next investigated whether such effects are driven by the formation of new connections (extensive margin; equation (4)) or the extension of existing connections (intensive margin; equation (5)) for the
full model
extensive margin
and intensive margin
where flow can be intercity investment flows or knowledge flows, defined as the number of intercity investment flows from c1 to c2, or the number of patent collaborations between c1 and c2 (c1 ≠ c2). ({mathbb{1}}({mathrm{flow}}_{c1,c2,m,t+1} > 0)) is an indicator function, which is equal to 1 if ({mathrm{flow}}_{c1,c2,m,t+1} > 0), and 0 otherwise. The key variable of interest, ({mathrm{exchange}}_{c1,c2,m,t}), is similar to ({mathrm{exchange}}_{c,m,t}), but it takes into account the origin and destination of each personnel exchange. Specifically, it takes the value of 1 if the current city leader m of city c2 is transferred from c1, and 0 otherwise. ({delta }_{a,t}({theta }_{b,t})) is the origin (destination) city–year fixed effects, and ({eta }_{p1,p2}) is the province-pair fixed effects. The former two types of fixed effects capture all time-variant factors of origin/destination cities (for example, economic factors such as GDP per capita and population density). We consider city pairs that have at least one flow during the study period.
Simulation analysis under different scenarios
We constructed three scenarios to simulate the potential in achieving more water pollution reduction by rearranging intercity exchange of city leaders. For the first scenario, we considered that all cities without exchanged leaders are assigned with city leaders exchanged from cities in the same province. For the second scenario, we considered that all cities without exchanged leaders are assigned with city leaders exchanged from cities outside the province. For the third scenario, we considered that all cities with leaders exchanged from the same province are reassigned with exchanged leaders from other provinces.
We first calculated the simulated water pollution reduction for each year in scenarios 1 and 2
where (hat{beta }) is the estimated coefficient of intraprovincial exchange (scenario 1) or the estimated coefficient of interprovincial exchange (scenario 2) from Supplementary Table 13, (overline{mathrm{pollutio{n}}_{t}}) is the average annual pollution of firms that are located in cities without exchanged city leaders and Nt refers to the number of such firms in year t. We sum up the simulated reduction in all years as
Next, we calculated the simulated water pollution reduction for each year in scenario 3 as
where ({hat{beta }}_{mathrm{interprovincial}}) and ({hat{beta }}_{mathrm{intraprovincial}}) are the estimated coefficients of interprovincial and intraprovincial exchanges from Supplementary Table 13, respectively; (overline{mathrm{pollutio{n}}_{t}}) denotes the average annual pollution of firms that have already been affected by intraprovincial personnel exchanges and Nt represents the number of such firms. We sum up the simulated reduction in all years as
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Responses