Effects of Decentralized Water Regulation on Agriculture in China: A Quasi-Natural Experiment Based on Incentives for Promoting Officials

Abstract

Environmental performance is becoming increasingly essential for promoting local officials in China; thus, their pursuit of promotion may affect agricultural output. This study spatially matched Chinese local official promotion data, regional agricultural output, river-water-quality-monitoring stations, and riverside enterprise discharge data. Based on the difference-in-difference model, the exogenous impact of the natural experiment based on the promotion of officials is quantified as how the promotion behavior of local officials in pursuit of environmental achievements affects agricultural output. This was examined under the decentralization system of China’s environmental governance. The results show that local officials improve agricultural production by controlling environmental pollution through promotion incentives. However, since the central government can observe the regulatory effect of upstream officials through the readings of water monitoring stations, upstream officials strictly enforce the central environmental regulations due to promotion motivation, while downstream officials do not strictly enforce their counterparts. This can result in differentiated impacts on agriculture in upstream and downstream regions. We also carried out a parallel test, placebo test, and measurement error test for the quasi-natural experiment, and the conclusions derived from the analysis remained robust. Our study has important implications for designing compatible environmental governance contracts and incentive policies for promoting agricultural production.

1. Introduction

Agriculture depends on water resources, and the security of these resources affects sustainable production in the context of agriculture. Owing to the increasing pollution of water resources, industrial sewage irrigation has become a key issue in agriculture security [1,2,3,4]. According to the United Nations Food and Agriculture Organization (FAO), the agricultural sector is the largest water user, consuming 70% of the world’s developed freshwater supply. This percentage is considerably higher in developing countries because irrigation accounts for 95% of the total water consumption [5]. With industrial development, and factories discharging pollution into rivers, water pollution is becoming an increasingly severe problem for sustainable agricultural development.

The availability of water resources plays an essential role in agriculture. However, the industrial wastewater pollution of irrigation water can lead to reductions in crop yields, deterioration of soil quality, and other impacts on the surrounding environment. There are also considerable hazards from heavy metals. Industrial production produces substantial amounts of sewage during daily production activities. If industrial wastewater is discharged directly into rivers without treatment, it can cause severe harm to the surrounding ecological and living environments for aquatic organisms [6]. Nitrogen and salt in industrial wastewater can also negatively impact the environment. Although crop growth absorbs a substantial amount of nitrogen, excessive nitrogen in farmland irrigation water leads to a nutritional imbalance in crops and increases vulnerability to disease threats. Excessive salt in sewage causes the dehydration and drying of crops [7,8,9]. Organic wastewater derived from sewage contains a considerable amount of organic matter, which decomposes easily. The use of sewage with organic matter is in contravention of the standard for farmland irrigation and will reduce the quality of food, harming the country’s food safety.

Agriculture is the most fundamental material production department in a national economy. Agricultural development can directly affect the national economy’s development trends. According to data released by the National Bureau of Statistics of China, the gross agricultural production in 2020 was estimated to reach CNY 16.69 billion, accounting for 16.47% of the Gross Domestic Production (GDP). Meanwhile, according to the 2020 monitoring survey report on migrant workers undertaken by the National Bureau of Statistics of China, the working population of rural residents is 362.81 million, of which 208.22 million people work in agriculture year-round, accounting for 21% of the total labor force. Agriculture is directly related to China’s social stability. China has more than one billion people, most of whom are from the domestic agricultural industry. If the agriculture industry cannot continue to provide adequate food security, there will be social panic, production will not develop, and national stability will be affected.

In the 1990s, to achieve high GDP growth in China, the government did not set strict targets for emissions’ reduction and water-quality improvement. China is facing growing environmental challenges, including the deterioration of river water quality. Since the beginning of 2000, China has implemented strict water-pollution control measures. Water quality reported from water-quality-monitoring stations has been included in the assessment indicators for local officials [6]. However, against the background of a central–local decentralization system, officials are subject to personal needs, such as promotion incentives. They will not eliminate water pollution as their only management goal.

Under a decentralized environmental governance system, when the central government needs to mobilize local governments to implement specific pollution-control policies, it often implements a goal-oriented incentive mechanism. This links the promotion of officials to specific performance targets. However, if the central government does not appropriately supervise the implementation of these policies, local officials tend to work hard in areas that are easily monitored and do little else. In the context of this study, the central government originally hoped to improve water quality by mobilizing officials’ promotion incentives. However, the central government can only supervise local governments by examining water-quality data from river monitoring stations. The monitoring data only reflect emissions from upstream companies, and local government officials have a strong incentive to regulate upstream companies; however, there are insufficient incentives to encourage the regulation of companies located downstream of river monitoring stations.

Therefore, this study aimed to quantify how promotion incentivized behaviors by local officials affect agricultural output in China under a decentralized environmental governance system.

2. Literature Review and Hypothesis Development

Since China’s reform and opening up in 1978, the long-term goal of economic growth has been used as an incentive for official promotion, which has led to the deterioration of water quality in China [10]. Water quality and environmental challenges in China are still severe, and pollution-control measures need to be paid close attention [11].

After China joined the WTO in 2001, although economic growth is still the most important goal of the central government, China has paid great attention to environmental protection and has begun to solve water-pollution problems [6]. As a result, the central government regards the effect of water-pollution control as an incentive goal for the promotion of local officials [12].

China’s environmental regulatory system is different from that of most developed countries. Its main feature is the so-called environmental decentralization, in which the central government promulgates a set of environmental regulations and goals while local governments implement environmental regulations and achieve specific environmental goals [3], that provides us with a unique opportunity to study the environmental-protection motivation of local governments.

Many works in the empirical literature exist on how environmental regulation affects environmental quality [13,14,15], residents’ health [16,17,18], and labor productivity [19,20,21]. Most of these focused almost exclusively on developed countries. In sharp contrast, there have been few systematic studies on how agriculture is affected by local officials who pursue environmental performance for promotion. This study aimed to find empirical evidence of environmental regulatory loopholes in water-quality monitoring in China and to quantify the impact of such loopholes on sustainable agricultural development. We considered the following hypotheses in this context:

Hypothesis 1.

Agricultural growth positively correlates with official promotion.

Hypothesis 2.

Officials’ pursuing promotions based on environmental performance affects agriculture.

In fact, the length of polluted rivers in China doubled in 2018, and the river environmental degradation trend has yet to be fundamentally reversed [22]. Many studies of water pollution regulation have concluded that monitoring and enforcing the regulation is critical to improving water quality [23] and that limited regulatory resources cause poor environmental enforcement [24]. If policymakers ignore spatially differentiated levels of regulatory stringency and shift water-polluting activities to areas that are more susceptible to polluting discharges, although such an incentive system can effectively achieve environmental governance goals, it will also have governance deviations, resulting in poor river-pollution-governance effects. In this context, consider the following hypothesis:

Hypothesis 3.

Supervised environmental protection is positively related to agricultural growth.

3. Materials and Methods

3.1. Descriptive Statistical Results

Table 1. Descriptive statistics of variables.

Variable		Obs.	Mean	SD	Min	Max
Explanatory variables	Agricultural product (Yuan/CNY)	282,042	139,990.3	116,166.10	0	1,132,368
	Grain product (ton)	1,041,647	253,205	261,762.70	0	3,640,712
	Oil crops product (ton)	1,009,743	14,238.86	23,413.09	0	896,301
	Cotton product (ton)	475,571	4479.587	12,600.20	0	390,177
Independent Variable	Promotion	1,222,322	0.25	0.43	0	1
Moderator Variables	Upstream	1,222,322	0.39	0.49	0	1
	Industrial wastewater discharge (thousand ton)	931,350	231.37	2203.19	0	777,000
	Ammonia nitrogen emissions (ton)	901,049	4.09	246.14	0	200,000
	Chemical oxygen demand (ton)	1,041,423	204.74	8966.35	0	8,210,641
Control Variables	Total power of agricultural machinery (thousand kilowatts)	979,140	40.18	37.58	0	336
	Total sown area of crops (thousand hectare)	650,138	69.31	58.02	0	925.10

Abbreviations: Obs., observations; SD, standard deviation; Min, minimum; Max, maximum.

shows the results from the descriptive statistics of data from 2000 to 2015. The main variables explained agricultural variables, including agricultural products, grain products, oil crops, and cotton products. In 2015, grain output accounted for 80.7% of agriculture output, and oil crops and cotton products accounted for 21.9% and 3.8% of the rest. The explanatory variable was the promotion of local officials. This study divides the experimental group and the control group into quasi-natural experiments based on interactive variables, including upstream and downstream variables and three enterprise pollution emission indicators, namely two control variables that control other factors that affect agriculture. The promotion of officials’ variable and upper small travel variable are dummy variables of 0–1, while the other variables are continuous. The average promotion of officials was 0.39, indicating that officials in 39% of posts were promoted. The standard deviation of the promotion of officials’ variable is higher than the mean. This indicates that the samples are sufficiently diversified to avoid homogeneity of the samples selected and overcome the problem of regression bias.

3.2. Data Source and Variables

To quantify the impact of behavior around promotions of officials in pursuit of environmental achievements in agriculture, this study needed to match and calculate the following data. This includes China’s county-level agricultural data, including China’s total agricultural product, total grain production, total oil production, and total cotton production. Data on the promotion of local officials in China should also be examined. We collected demographic information with career entries of 5032 officials and the past appointments that he/she has served in 2606 counties from Chinese local government websites and Baidu Baike. We then generated the promotion variable for officials whose administrative ranks change to a higher level. Data from rivers upstream and downstream were calculated by using ArcGIS 14.1 software. Based on the electronic map of rivers in China, we divided the upstream and downstream river channels according to the boundaries created by the river-water-quality-monitoring stations. The side of the river with the high terrain was considered to be upstream, and the side with low terrain was considered to be downstream for the purposes of this study. Data derived from water-quality-detection points in the main river basins in China are shown in Figure 1. These data were obtained from the Ministry of Ecology and Environment, PRC. The electronic map of rivers in China, using data from OpenStreetMap and Digital Elevation Model (DEM), is important original data for the research and analysis of terrain, river basins, and ground object identification in China. The DEM is a remote-sensing topographic map with a precision of 12.5 m that is mapped by the ALOS satellite. Pollutant-discharge data from enterprises were collected from the National Bureau of Statistics’ Industrial Enterprises Pollution Discharge Database. These are currently the most detailed enterprise-level pollution-discharge data in China. Using Python to call the Google Map API to query the latitude and longitude of the enterprise address in batches, ArcGIS 14.1 was used to match the latitude and longitude of the enterprise with the river digital map. This was undertaken to determine the presence of sewage enterprises within 10 km of the upstream and downstream riverbanks. Other data, including the total power of agricultural machinery and total sown area of crops, were also collected for inclusion in the analysis. Due to data availability, all the data were matched according to spatial location to form panel data from 2000 to 2015.

The variables examined in this study include the total agricultural production, total grain production, total oil production, and total cotton production. These variables were selected to reflect the four variables of agricultural development. The explanatory variable is a dummy variable for the promotion of local officials in China, with 1 indicating promotion of officials and 0 indicating no promotion of officials occurring. There were two types of interactive variables. The first type comprises upstream and downstream dummy variables, where 1 represents the upstream, and 0 represents the downstream. The second type is the pollutant discharge data of the enterprises. Industrial wastewater discharge, wastewater nitrogen and oxygen discharge, and chemical aerobic discharge were examined according to data availability. Figure 2 shows the relationship between variables.

3.3. Regression Design

Although regional characteristics can be controlled in the OLS regression model, agricultural planting fertilization and soil fertility are data that cannot be controlled or obtained for this study. The OLS model was used to estimate the impact of sewage discharge into the rivers from riverside agricultural planting. The error in the parameter estimation caused by missing variables is called the endogeneity error. Identifying causality can overcome this endogeneity problem. Therefore, the difference-in-difference (DID) method was used to estimate the influence of the agricultural planting status along rivers.

We divided the samples into experimental and control groups according to whether local officials were promoted or not. The promotion of local officials was taken as the exogenous impact of the natural experiment, and the impact of the promotion of officials on agricultural output was examined by using a DID model. The mechanism of the impact of the promotion of officials on agriculture was analyzed by using the pollutant discharge of enterprises as the intermediary variable. The differentiated impact of official promotion on agriculture was analyzed by using the river-water-quality-monitoring station as the boundary. Industrial wastewater pollution restricts agricultural output in China, but the decentralization mechanism of environmental governance has not effectively addressed this problem.

An approximately random sample grouping was initially conducted to use DID to conduct a quasi-natural experiment. In this study, officials who had been promoted were the experimental group, while unpromoted officials were the control group. $Promotio{n}_{it}$ is the promotion variable of officials at time t of region i, when $Promotio{n}_{it}=1$, indicating that the samples of the experimental group received the exogenous impact. Meanwhile, the control group was not exposed to exogenous shocks. Therefore, the differential measurement model can be expressed as follows:

$$Ag{r}_{it}={\alpha }_0+{\alpha }_{1}Promotio{n}_{it}+{\alpha }_2\ast X_{it}+{\delta }_i+{\delta }_t+{ϵ}_{it}$$

(1)

where $Ag{r}_{it}$ denotes the agricultural variable, $Promotio{n}_{it}$ denotes the promotion variable of officials, and $X_{it}$ denotes other variables affecting agriculture; ${\alpha }_0$, ${\alpha }_1$, and ${\alpha }_2$ are the coefficients to be estimated in the regression analysis. The coefficient of the local officials’ promotion variable ${\alpha }_1$ is the coefficient predominantly examined in this study. This reflects the difference between the experimental group and the control group. It captures the difference before and after the promotion of officials, and the difference between promoted and non-promoted officials. Furthermore, ${\delta }_i$ and ${\delta }_t$ represent the regional and time fixed effects, respectively; and ${ϵ}_{it}$ represents the random disturbance term.

This further quantifies the environmental governance effect achieved by the promotion incentive of local officials and the moderating effect of enterprise pollution emissions on the causal relationship between officials’ promotions and agriculture. In this study, the following regulatory DID model was established:

$$Ag{r}_{it}={\beta }_0+{\beta }_{1}Promotio{n}_{it}+{\beta }_{2}Interac{t}_{it}+{\beta }_{3}Promotio{n}_{it}\ast Interac{t}_{it}+{\beta }_4\ast X_{it}+{\delta }_i+{\delta }_t+{ϵ}_{it}$$

(2)

where $Interac{t}_{it}$ is an interactive variable; ${\beta }_0$, ${\beta }_1$, ${\beta }_2$, ${\beta }_3$, and ${\beta }_4$ are the coefficients estimated by the model; ${\beta }_1$ and ${\beta }_2$ are captured as the main effects of local officials’ promotion and interaction variables, respectively; and ${\beta }_3$ captures the interaction effects of the interaction terms. By calculating the partial derivative of Model (2), the total effect of the promotion of officials can be obtained as ${\beta }_1+{\beta }_3$. The total effect of the promotion of officials depends on the interactive variables. All the regression equations in this study were estimated by using Stata 14 software. To reduce heteroscedasticity and obtain robust regression results, logarithms were calculated for all the continuous variables in the regression process. The design strategy for Model (2) is presented in Figure 3.

4. Results

4.1. The Effect of Promotion of Officials on Agriculture

In this study, we considered the promotion of local officials as an exogenous impact and quantified its impact on agriculture. A regression analysis was performed based on Model (1). The variables explained were agricultural added value, total grain output, total oil output, and cotton output. The core explanatory variables were fictitious variables of official promotion. In addition to controlling for regional and time-fixed effects, the regression process also controlled for and affected the total power of agricultural machinery and the total sown area of crops. The regression results are presented in

Table 2. Effect of promotion of local officials on agriculture.

Variables	(1) Agricultural Production (CNY Ten Thousand)	(2) Total Grain Production (Ton)	(3) Oil Production (Ton)	(4) Cotton Production (Ton)
Promotion	0.0071 ***	0.0155 ***	0.0275 ***	0.0173 ***
	(0.0013)	(0.0008)	(0.0018)	(0.0037)
Total power of agricultural machinery (thousand kilowatts)	0.0038 *	0.1345 ***	0.1655 ***	0.0126 *
	(0.0037)	(0.0019)	(0.0045)	(0.0100)
Total sown area of crops (thousand hectare)	0.4396 ***	0.7128 ***	0.8378 ***	0.5345 ***
	(0.0064)	(0.0030)	(0.0071)	(0.0181)
Observations	78,593	342,836	330,958	177,545
R-squared	0.9815	0.9664	0.9439	0.9500

Standard errors in parentheses; *** p < 0.01 and * p < 0.1.

. Each column represents a separate regression model. All the regression results are statistically significant, and the coefficient symbols are in line with expectations. The coefficient of local officials’ promotion of the core explanatory variable is significantly positive. This indicates that the promotion of local officials improves agricultural growth. Taking the first column in Table 2 as an example, the coefficient of the local officials’ promotion variable is 0.0071. This means that the promotion of local officials increases the agricultural added value by 0.71% compared with the agricultural added value of the region where officials are not promoted and if other conditions remain unchanged.

4.2. Parallel Test of Promotion of Officials Effect

The difference-in-difference model did not require the experimental and control groups to be identical. There may be some differences between the two groups, but the difference-in-difference model requires that the differences do not change over time. The processing and control groups must have the same development trend before the promotion of local officials. Therefore, the target variables for the processing and control groups can only be used if the parallel trend hypothesis is satisfied before the promotion of local officials. This is called the parallel trend hypothesis of the DID model. In contrast, if there are certain differences between the treatment and control groups in advance, the results produced by the DID model can no longer represent the net effect of the promotion of local officials. It is then highly likely that there are other factors affecting the changes in the variables explained.

This study used the first four periods of local officials’ promotion, the current period of local officials’ promotion, and the last four periods of local officials’ promotion as explanatory variables for regression. At the same time, the control variable, time fixed effect, and region fixed effect were added. The coefficients of the explanatory variables and the corresponding 95% confidence intervals were then regressed. This is shown graphically in Figure 4. In the first four periods of local officials’ promotion, the confidence interval of the coefficient includes 0. This indicates that the coefficient is not significantly different from 0, and the coefficient fluctuates around 0. This shows that there is no significant difference between the control and experimental groups. However, after the promotion of officials, the coefficients are all significant and greater than zero. This indicates that the promotion effect of local officials is positive, which leads to the growth of agriculture. Therefore, the DID method was designed in this study.

4.3. Placebo Test of the Effect of Promotion of Officials

The causality test shows that the promotion of local officials has an exogenous impact that may be disturbed by the missing variables. There may be circumstances that lead to a bias in the difference-in-difference regression design. First, certain factors vary over time and place, which are difficult to observe and control. Second, there may be a gap in agricultural production in the watershed due to the implementation of agricultural policies by different regional governments. Third, other sudden problems and farmers’ planting customs may impact agricultural production. Hence, the placebo test of exogenous impact by Chetty et al. (2009) and La Ferrara et al. (2012) was referred to in this study [6,25]. Through random promotion of officials and promotion time as the false variable of promotion of officials, the false variable of promotion of officials was used for differential regression analysis. Based on randomly selected samples, the differential regression analysis was conducted 500 times in this study. Figure 5 shows the distribution of the coefficients obtained from the 500 regressions. The coefficient estimated based on random samples is distributed around 0. The coefficient distribution is not significantly different from 0. The real regression coefficient, namely the coefficient of the local promotion of officials’ variable in Table 2, is outside the distribution. This means that there is no causal relationship between the impact of fake official promotion and agricultural production, except for the promotion of real local officials. This indicates that the causal analysis of the DID model established in this study was not affected by the missing variables.

4.4. Impact of Environmental Governance Mechanisms

To test how local officials’ pursuit of environmental achievements and promotion affects agriculture, we used pollution discharge as the interactive variable. The Interaction Model (2) set up in this study was analyzed by using regression. Given that businesses along rivers discharge pollutants into the river and water is taken from the river for agricultural irrigation, pollution can affect agricultural cultivation. If the irrigation water contains too much nitrogen, it can cause a nutritional imbalance in crops, resulting in crop stems that are too long; fall easily; and have poor stress resistance, diseases, poor ripening, and other problems. Such crops will inevitably reduce production and are poor quality. The second factor is the harmful effects of organic matter. After the organic matter in sewage enters farmland, the decomposition process consumes a substantial amount of oxygen, leading to a lack of oxygen intake by rice and other crops. Meanwhile, some substances produced during the decomposition process are toxic to rice and directly affect the rice yield. The harmful effects of heavy metals were also discussed. Damage caused by heavy metals to crops results in the leaves quickly curling, leading to the gradual death of the plants.

In this study, enterprise sewage discharge, industrial wastewater ammonia nitrogen discharge, and industrial wastewater chemical aerobic were selected as the exchange variables. A regression analysis was conducted on Model (2). The results are presented in

Table 3. Influence of promotion of officials on agricultural production through pollution mechanism.

	(1)	(2)	(3)	(4)
Variables	Agricultural Added Value (CNY Ten Thousand)	Total Grain Production (ton)	Oil Production (ton)	Cotton Production (ton)
Panel A: Industrial wastewater discharge
Promotion	0.0116 **	0.0494 ***	0.0185 **	0.1059 ***
	(0.0058)	(0.0033)	(0.0077)	(0.0151)
Industrial wastewater discharge (ton)	−0.0029 ***	−0.0012 ***	−0.0062 ***	−0.0067 *
	(0.0003)	(0.0002)	(0.0004)	(0.0009)
Promotion × industrial wastewater discharge	0.0024 ***	0.0014 ***	0.0057 ***	0.0061 ***
	(0.0005)	(0.0003)	(0.0007)	(0.0014)
The total power of farm machinery (thousand kilowatts)	0.0056	0.1412 ***	0.1836 ***	0.0185
	(0.0044)	(0.0022)	(0.0051)	(0.0113)
The total sown area of crops (thousand hectares)	0.4218 ***	0.6836 ***	0.7065 ***	0.6131 ***
	(0.0082)	(0.0034)	(0.0080)	(0.0200)
Observations	46,327	232,221	224,277	124,183
R-squared	0.9809	0.9656	0.9423	0.9500
Panel B: Emissions of ammonia nitrogen from industrial wastewater (ton)
Promotion	0.0130 ***	0.0359 ***	0.0314 ***	0.0962 ***
	(0.0025)	(0.0013)	(0.0033)	(0.0064)
Emissions of ammonia nitrogen from industrial wastewater (ton)	−0.0003 *	−0.0008 ***	−0.0008	−0.0092 ***
	(0.0004)	(0.0002)	(0.0006)	(0.0013)
Promotion × emissions of ammonia nitrogen from industrial wastewater	0.0001 *	0.0006 *	0.0005 *	0.0017 ***
	(0.0006)	(0.0004)	(0.0010)	(0.0020)
The total power of farm machinery (thousand watts)	0.0310 ***	0.0467 ***	0.1280 ***	0.1246 ***
	(0.0062)	(0.0035)	(0.0087)	(0.0203)
The total sown area of crops (thousand hectares)	0.3718 ***	0.6758 ***	0.7772 ***	0.7722 ***
	(0.0113)	(0.0058)	(0.0142)	(0.0384)
Observations	23,385	90,928	86,152	48,515
R-squared	0.9868	0.9765	0.9567	0.9554
Panel C: Chemical oxygen demand emissions from industrial wastewater (ton)
Promotion	0.0078 ***	0.0369 ***	0.0329 ***	0.0558 ***
	(0.0021)	(0.0010)	(0.0024)	(0.0049)
Chemical oxygen demand emissions from industrial wastewater (ton)	−0.0027 ***	−0.0007 ***	−0.0039 **	−0.0062 **
	(0.0003)	(0.0002)	(0.0004)	(0.0009)
Promotion × chemical oxygen demand emissions from industrial wastewater	0.0018 ***	0.0001 *	0.0030 ***	0.0060 ***
	(0.0005)	(0.0003)	(0.0007)	(0.0014)
The total power of farm machinery (thousand watts)	0.0098 **	0.1322 ***	0.1686 ***	0.0515 ***
	(0.0045)	(0.0023)	(0.0054)	(0.0122)
The total sown area of crops (thousand hectares)	0.4183 ***	0.6803 ***	0.7155 ***	0.6253 ***
	(0.0083)	(0.0037)	(0.0086)	(0.0215)
Observations	43,111	203,807	196,607	109,074
R-squared	0.9823	0.9679	0.9475	0.9497

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, and * p < 0.1.

. The different pollutant emission data were grouped into a panel sub-table, and each column in each panel had a separate regression equation. The first Panel A example illustrates the total effect of the pollutant discharge. When local officials are not promoted, increasing industrial expense discharge will reduce agricultural production by 0.39%. However, when local officials are promoted, increasing industrial wastewater discharge will reduce agricultural production by 0.05% (−0.0029 + 0.0024 = −0.0005).

4.5. Heterogeneity of Upstream and Downstream River Channels

To examine how the pollution gap of the upstream and downstream river affects agricultural production in the context of official promotions, the upstream and downstream variables for the river and the promotion of local officials were taken as interaction terms, and a regression analysis was carried out according to Model (2). The results are shown in

Table 4. Influence of promotion of officials on agricultural production.

Variables	(1) Agricultural Added Value (CNY Ten Thousand)	(2) Total Grain Production (ton)	(3) Oil Production (ton)	(4) Cotton Production (ton)
Promotion	0.0074 ***	0.0416 ***	0.0419 ***	0.0143 ***
	(0.0017)	(0.0009)	(0.0022)	(0.0043)
Upstream	0.0014 *	0.0006 *	0.0028 *	0.0158 ***
	(0.0013)	(0.0009)	(0.0022)	(0.0043)
Promotion × upstream	0.0067 ***	0.0045 ***	0.0005 *	0.0378 ***
	(0.0020)	(0.0014)	(0.0033)	(0.0062)
Total power of agricultural machinery (thousand kilowatts)	0.0047	0.1342 ***	0.1642 ***	0.0091
	(0.0037)	(0.0019)	(0.0045)	(0.0100)
Total sown area of crops (thousand hectares)	0.4368 ***	0.7165 ***	0.8335 ***	0.5350 ***
	(0.0064)	(0.0030)	(0.0071)	(0.0181)
Observations	78,593	342,836	330,958	177,545
R-squared	0.9815	0.9667	0.9439	0.9500

Standard errors in parentheses; *** p < 0.01, and * p < 0.1.

, with each being listed as a separate regression equation. The promotion of officials in the lower reaches of the rivers brings about a 0.74% increase in agricultural production. Meanwhile, the promotion of officials in the upper reaches brings about a 1.41% increase in agricultural production (0.74% + 0.67%).

4.6. Robustness Test to Exclude Downstream Disturbance Caused by River Pollution

Through model regression analysis, it was concluded that officials in the upstream region pursue promotion incentives to conduct stricter environmental pollution control. This means that the agricultural output in the upstream region is higher than that in the downstream region. However, we are still concerned about the accumulation of downstream pollutants when upstream pollutants spread downstream along the river, as this affects the robustness of the conclusions of this study. Therefore, we have drawn a scatter plot, using the pollutant emissions of upstream enterprises and the agricultural output of downstream rivers. This allows us to examine whether upstream pollution affects the downstream agricultural output, using the fitting line of the scatter plot and the interpretation strength of the fitting line. The scatter plot is shown in Figure 6. First, the scatter and fitting lines of all the sub-graphs show that there is no significant positive correlation between upstream emissions and downstream agricultural output, and that they are more like independent relationships. Second, the R-Squares for all the sub-graph fitting lines are relatively small. (An R-Squared above 0.7 would generally be seen as showing a high level of correlation.) This indicates that the strength of the interpretation of the scatter plot fitting lines is relatively weak. The possibility of upstream pollution causing downstream agricultural output reductions is relatively low. Therefore, in the quasi-natural experiment in which river-water-quality-monitoring stations are used to collect data from upstream and downstream, there is no problem from the upstream river pollutants accumulating downstream and affecting the downstream agricultural output. Compared with the situation in the downstream area, it can be concluded that the promotion behavior of officials pursuing environmental achievements in the upstream area leads to a greater increase in agricultural output in the area.

5. Discussion

In this study, we used river pollution control as an example to assess how the promotion of environmental performance by local officials affects agricultural output in China. We further aimed to quantify the impact of heterogeneous behavior of officials in upstream and downstream areas on agriculture. Under a decentralized environmental governance system, as policymakers and executives, the central government and local governments are subject to the system design and incentive mechanism for their direct interaction [26]. The central government has set targets for local governments to motivate them to clean their environment. Local officials have reduced pollution from companies along the riverbanks to implement the central government’s environmental targets for promotion. This increases agricultural production along the river. The single promotion incentive policy for environmental performance in the decentralization system is the key limiting factor. However, the existing decentralized environmental governance system fails to realize the compatibility between contracts and incentives fully. When local officials pursue environmental achievements for promotion, they fail to implement the central government’s environmental governance goals. Local officials pay more attention to the pollution discharge from upstream enterprises that can be monitored by river-monitoring stations, while neglecting to supervise the pollution discharge from downstream enterprises. As a result, agricultural yields were significantly higher upstream than downstream.

Much of the literature focuses solely on the effects of water pollution on agriculture [1,2,27,28,29]. However, this study further reveals the institutional mechanism behind the impact of water pollution on agriculture. We used the promotion of local officials in China as a natural experiment. The promotion of local officials was used as the standard to divide the experimental and control groups. We then used the pollutant discharge from enterprises along rivers as the regulating variable and quantified how the incentive mechanism of local officials’ environmental performance influenced agriculture through environmental governance. In this context, local officials promoted better environmental performance, increased supervision of polluters, reduced river pollution, and increased agricultural output. There are similar studies in the existing literature that have examined how the actions of officials affect environmental pollution [7,8,9,30]. However, this study focused on the differentiated impact of such behavior on agriculture.

On this basis, we have also used water-quality-monitoring stations as the boundary to divide rivers upstream and downstream and have further investigated the heterogeneity of the promotion incentive mechanism of local officials in China on agriculture. Previous studies on the effects of upstream and downstream heterogeneity have predominantly focused on how upstream pollution diffusion affects downstream areas [31]. The entire river was also studied by dividing it geographically into upstream and downstream sections. However, the current study considers the division of river-water-quality-monitoring stations into upstream and downstream rivers as the regulating variable. This can be used to analyze the heterogeneity of the impact of the incentive mechanism of official promotion. Our analysis of the differentiated impact of pollution on agriculture between upstream and downstream areas is predominantly based on the mechanism of the difference in the promotion behavior of officials, in contrast with previous research based on the impact of pollutants flowing down the river [32]. The central government can only monitor the performance of upstream local governments by using data from river-water-quality-monitoring stations. Therefore, downstream local officials relax their environmental monitoring in the pursuit of economic achievements, and differentiated pollution levels in the upstream and downstream areas have different impacts on agriculture.

The research samples selected in this study include almost all the rivers in China, and strong evidence can be obtained from a larger scope. Most previous studies have considered one or two rivers as research objects [2,31,33,34,35]. Our study used more sophisticated data with a wide range of years and spatial match of rivers, water-quality-monitoring stations, regional agricultural output, and the promotion of district officials. Therefore, the promotion incentive of officials can be used as a quasi-natural experiment on the exogenous impact of pollution on agriculture. This means that the conclusion is robust within the scope of China and has reference significance for developing countries.

We also conducted a series of robustness tests to ensure the robustness of the exogenous impact as a quasi-natural experiment of official promotion incentives. We constructed a dynamic model with four periods before and after the promotion of local officials in China and then investigated the dynamic influence of official promotion behavior on agriculture. We also tested the parallel trend between the control group and the experimental group before the promotion of officials, which is the condition of the differential model. We then conducted a placebo test by constructing a random virtual official promotion as an exogenous impact. The mean distribution of parameters obtained by using 500 regressions was 0. This indicated that the effect of virtual official promotion on agriculture was zero. The results from the regression analysis by changing the multiple explained variables to measure agriculture are also robust. The cumulative effects of river pollution were excluded. The scatter diagram of upstream river pollution in downstream agriculture shows that the relationship is independent and that the cumulative effect of pollution does not interfere with the causal relationship between the actions of government officials and agriculture.

With the development of digital information technology and digital dynamic monitoring, the central government can more closely supervise whether local officials have fulfilled environmental-protection targets. This may help improve the effectiveness of local governments in implementing these targets of the central government. Therefore, the influence of digital information technology on the causal relationship between local government behavior and agriculture may be a direction worthy of future research. The application of digital information technology in the field of environmental-protection monitoring can obtain local pollution-monitoring data in a more accurate, detailed, and timely manner. This can help the central government grasp the local environmental governance situation and take corresponding pollution-protection measures. Social welfare has improved substantially. Therefore, the monitoring capability of digital information technology is likely to play an important role in public management, and the relevant topics are worth further study.

This study has some limitations. This study quantified how the promotion behavior of local government officials affected agriculture through natural experiments. However, due to data limitations, we could not accurately obtain the pollution data from upstream and downstream river-water-quality-monitoring points; instead, we could only use the pollution data from enterprises along the river. Only three indicators were obtained for corporate emissions, which failed to examine the specific impact of pollution on agriculture.

6. Conclusions

Under the political system of local environmental decentralization, Chinese local government officials in pursuit of promotion are biased toward meeting the targets set by the central government to reduce river pollution. This can exert a different impact on agriculture along the river. Based on the quasi-natural experimental method of local government official promotion, this study quantified the causality test between China’s environmental governance incentive system and agriculture. The results have shown that to get promoted, local officials implemented stricter supervision of upstream polluters that affected river-water-quality-monitoring data. Meanwhile, there was lenient supervision of enterprises located downstream of river-water-quality-monitoring stations. This leads to the differentiation of agricultural output in the upper and lower reaches of the river. Promotion-seeking officials, encouraged by their environmental achievements, have increased agricultural output by tackling corporate pollution. In contrast, officials in the downstream regions have allowed companies to pollute, leading to a decline in agricultural output compared to the output upstream of river-monitoring stations. The conclusion of this study indicates that promoting environmental governance through political incentives may lead to deviations in the implementation of policies, which will have a differentiated impact on agricultural output.

Four conclusions can be drawn from the regression. First, the main effect of the local officials’ promotion variable is positive. This indicates that the promotion of officials brings about growth in agriculture. Second, the main effect of the pollutant-emission variables is negative. This shows that agricultural output will decrease with an increase in pollutant emissions. Third, the interaction variable coefficient of the local officials’ promotion variable and pollutant-emission variable is negative, and the interaction variable coefficient and main effect are added to the total effect. Third, the coefficient of local officials’ promotion variable is significantly positive, and the promotion of local officials strongly promotes agriculture. Fourth, the dummy variables upstream and downstream are also positive. This indicates that the agricultural output from upstream is higher than that from downstream when the other conditions remain unchanged.

This study has shown that substantial progress is still required in terms of sustainable agricultural production in China. Industrial wastewater pollution is still a considerable problem but is also subject to the administrative factors of local officials’ treatment behavior. Through system reforms, local government officials can effectively meet the pollution-reduction targets set by the central government, as they are conducive to implementing sustainable agricultural production, improving social welfare, and maintaining social stability.