How China’s Ecological Compensation Policy Improves Farmers’ Income?—A Test of Environmental Effects

Abstract

Based on the quasi-natural experiment established in China’s national key ecological function areas, this paper takes 102 counties in Hebei Province, China, from 2014 to 2018 as the research object. It uses propensity score matching and difference-in-difference methods to investigate the impact of policy implementation on farmers’ income levels and constructs a mechanism using the air quality index to examine the environmental effect. The results show that when the time and regional fixed effects are not considered, the income level of farmers in the county increased by 3.11% due to the influence of the transfer payment policy, and the policy treatment effect grew over time. Among the control variables, the degree of industrialization and agriculturalization, urbanization rate and government financial scale were all positively related to farmers’ income. Controlling the fixed effects of region and year, the impact of policy on the improvement of farmers’ income was weakened, and the regression coefficient changed from 0.2211 to 0.0366, a drop of 83.45%. This suggests that the policy is greatly affected by the city where farmers live. The “environmental effect” test results showed that transfer payments could increase the income level of farmers in counties affected by the policy. The mechanism is that the priority measure of the ecological compensation policy is to improve the ecological environment, which is conducive to improving local environmental governance and environmental productivity and increasing crop yields, and thus increasing farmers’ incomes. Because the regions where the policy is implemented overlap with highly poverty-stricken areas, it is necessary for the central government to improve transfer payment standards and enrich their content to protect people’s livelihood while promoting ecological protection. As a result, local governments will be encouraged to act ecologically, vigorously develop local ecological industries, and promote the internalization of positive externalities in ecological environmental services, further improving the level of agricultural modernization and ecological sustainability and improving the income levels of farmers and their quality of life.

1. Introduction

At the cost of burning a large amount of fossil fuels, China has achieved sustained and rapid economic growth since the reform and opening up, and the processes of modernization, industrialization, and urbanization have been continuously improved. In the 21st century, the problems of resource scarcity and ecological environment deterioration have become increasingly prominent. In the “Decision of the State Council Concerning Several Issues Concerning Environmental Protection” promulgated in 1996, it was pointed out that an economic compensation mechanism for the paid use of natural resources and restoration of the ecological environment should be established and improved. This marks the gradual establishment of law-based economic compensation for the ecological environment. The ecological compensation policy has become an important practice for China to explore the coordinated development of ecological protection and poverty eradication. The policy has already covered key fields and ecological function areas such as cultivated land, forests, grasslands, wetlands, watersheds, and oceans, and the government has adopted a series of measures such as transfer payments in national key ecological function areas, forest ecological compensation, returning farmland to forests, and grassland ecological protection subsidies. Meanwhile, academic circles have also focused on ecological compensation as a research topic.

Ecological compensation is commonly known internationally as payment for ecosystem services (PES). Wunder proposed that PES represent a combination of user fees and targeted subsidies, which is the generally accepted definition of PES [1,2]. Porras expanded on Wunder’s concept, advocating for the inclusion of taxes and non-market measures to encourage voluntary support through user fees [3]. Kemkes asserts that PES is a transparent system that provides additional environmental services by offering conditional fees to voluntary providers, alongside environmental additionality and transparency [4].

Zhang was the first to propose a definition of ecological compensation in China from an ecological perspective [5]. He suggested that ecological compensation is the return of a portion of the economic gains obtained from resource utilization, in the form of materials or energy, to sustain the ecological system’s dynamic equilibrium between input and output. Combining the theories of resources and externalities, Mao et al. proposed that the purpose of ecological compensation is to reduce the negative externalities of damaging behavior by charging those responsible for damaging the resource environment, or compensating those who protect it, thereby incentivizing protection behavior to achieve the goal of resource and environmental conservation for sustained economic growth [6]. Li et al. argued that ecological compensation is a type of economic incentive that addresses market failures and solves externality problems by motivating individuals to maintain ecosystem services operation, protecting the ecological environment and achieving environmental benefit targets while ensuring social development fairness [7]. Since then, the definition of ecological compensation in China has shifted from an ecological to an economic perspective.

Existing studies have paid more attention to the subject and object, compensation standard, and compensation method of the ecological compensation mechanism. The ecological compensation policy must not only accomplish the task of environmental protection but also take into account the protection of people’s livelihoods. Only in this way can it be sustained [8]. Therefore, it is of great significance to link the environmental effects with the effects of farmers’ incomes, especially the research on the relationship between the environmental effects of transfer payment in China’s national key ecological function areas and farmers’ incomes. Transfer payment in national key ecological function areas is the current largest ecological compensation policy in China, characterized by a wide scope, long implementation time, and large investment. The policy includes the transfer payment of the central government and the provincial government to the national key ecological function area and the provincial key function area. The transfer payment funds are distributed through each layer, which forms the transfer payment system of the national key ecological function area. From the perspective of the fund allocation standard of transfer payment, the central government’s annual transfer payment to local governments is mainly based on the financial level of the local government, the quality of the ecological environment, and the degree of poverty. In terms of the distribution scope, there are mainly key subsidies, development prohibition subsidies, and ecological forest ranger subsidies. The central government uses them to guide local governments to strengthen ecological and environmental protection, improve the basic public service guarantee capabilities of local governments in key national ecological function areas, accelerate labor mobility, and, consequently, promote agricultural progress and sustainable economic and social development, and generate positive spatial spillovers [9]. This will further people’s livelihoods and protect the environment simultaneously.

Hebei Province surrounds the capital city of Beijing, bordering Tianjin City and the Bohai Sea to the east, Shandong and Henan Provinces to the southeast and south, respectively, and Shanxi Province to the west, which is separated by the Taihang Mountains. To the northwest and north lies the Inner Mongolia Autonomous Region, while Liaoning Province is located to the northeast. The national and provincial key ecological function areas in Hebei Province cover a total area of 90,786 square kilometers and have a population of 9.4024 million, which represents 48.47% and 12.98% of the province’s total area and population, respectively. The national key ecological functional areas are located in the Bashang plateau area, mainly in the Yanshan mountainous area in northern Hebei and the Taihang mountainous area in western Hebei. The key ecological function area of Hebei Province serves as a crucial sandstorm source control and ecological restoration protection area in Beijing, Tianjin, and Hebei. It is also a key tax source protection area and ecological restoration management area in the Hailuan River Basin and an important barrier for the ecological security of Beijing, Tianjin, and Hebei. Furthermore, it is a key area for forestry and biodiversity protection in Hebei Province, a core area for the development of the eco-tourism industry, and a key development area for green energy and characteristic agriculture in Hebei Province’s ecological industries.

The research area selected in this paper is the county-level units in Hebei Province. Specifically, the main research objects are the county-level units that were included in the national key ecological function areas in Hebei Province in 2016. The reasons for choosing this region are as follows: First, Hebei Province is an integral part of the Beijing-Tianjin-Hebei coordinated development plan. This paper aims to analyze the impact of the policy of transfer payment in national key ecological function areas on the Beijing-Tianjin-Hebei region and provide suggestions for the planning of key functional areas in the Beijing-Tianjin-Hebei region in the future. Second, the county-level units selected in this paper are newly included as national key ecological function areas by the State Council. Most of the existing research is based on the previous adjustments made by the State Council to national key ecological function areas. Therefore, this study on the latest policy adjustment can provide a theoretical basis for the study of ecological compensation policy. Third, Hebei Province has the largest number of counties newly included in national-level key ecological function areas among all provinces in this latest adjustment. This ensures that the sample requirements are met and the research is reliable to a certain extent.

2. Literature Review

At present, the transfer payment mechanism for national key ecological function areas has been gradually improved. The central government and provincial finance have invested a massive amount of funds into it, and the policy coverage has reached a relatively high level. With the progressive advancement of the policy, improving and increasing the level of national welfare has become an important implementation goal of national concern in China [10]. Some scholars contend that ecological transfer payments can significantly improve the welfare level of residents. It can objectively compensate local finance [11], increase people’s livelihood [12], significantly improve poverty [13,14], support ecosystem diversity [15,16], improve regional comprehensive welfare [17], enhance residents’ satisfaction with ecological compensation policy [18], and boost residents’ subjective welfare levels [19]. However, the effect of the policy may be affected by differences in regional characteristics, such as the environmental conditions of the previous period [20], farmers’ production and lifestyles [21], the degree of agricultural development [22], the opportunity cost of regional development [23], and the level of local financial expenditure [24].

Other scholars oppose the idea that ecological environment improvement and poverty reduction cannot be achieved at the same time or can be achieved under certain conditions. Excessive pursuit of the coordinated development of the ecological environment and poverty reduction will probably lead to the failure of ecological compensation [25]. Furthermore, an unreasonable target system, low subsidy level, and standardized compensation policy may worsen the situations of participants [26], leading to the phenomenon of “relative institutional failure” in environmental regulation [27]. While policy implementation improves the environmental quality and ecological quality of the ecological function areas, it also leads to a relative decline in the level of regional economic development and basic public service supply [28]. This may be due to the fact that transfer payment inhibits the local spending behavior to a certain extent and weakens the enthusiasm of the local government itself [29], or due to a relatively small subsidy scale and imprecise poverty reduction effect [14].

Foreign research on ecological compensation differs from current research in China, as it focuses on policy design rather than the impact of policies on others. For instance, Gastineau et al. propose a normative approach to determine the optimal location and level of environmental compensation, while maintaining social welfare unchanged [30]. Sonter et al. find that ecological compensation policies have limitations in achieving no net loss of biodiversity and ecosystem services, using spatial simulation models [31]. In Sweden, Koh et al. suggest the use of complementary ecological valuation methods, compensation pools, and social safeguards to assess the implementation of ecological compensation policies through case studies [32]. Meanwhile, Simmonds et al. propose a new framework for ecological compensation that aligns with judicial biodiversity targets by reviewing the limitations of biodiversity offsetting [33]. Maryanney et al. reviewed eight case studies of market initiatives in Latin America for carbon sequestration and watershed protection to construct a conceptual framework for market mechanisms for forest environmental services [34].

The national key ecological function areas are positioned as “mainly protecting the ecological environment, limiting large-scale and high-intensity industrialization and urbanization, and maintaining national ecological security”. The county-level units covered by the policy are highly overlapped with the national poverty-stricken counties, which contributes to the difficulty in balancing the goal of people’s livelihoods and environmental protection. Confronted with policy design standards and assessment pressures, local governments tend to directly carry out ecological construction and pollution control. For instance, they expand the scale of natural resources (such as afforestation area, water resource area, etc.), encourage local environmental pollution control, and improve local ecological environments (such as improvement of water sources, windbreak, and sand fixation, etc.). All of these are conducive to improving local environmental governance and environmental productivity, thus increasing crop yields, reducing the input costs of local farmers, and increasing farmers’ income levels, namely the “environmental effect” of policy environmental protection goals on people’s livelihood goals. By incorporating ecological indicators into the mechanism design, the policy effects of transfer payments in national key ecological function areas can be evaluated from a dual perspective of “improving the environment” and “safeguarding people’s livelihoods”.

With 102 counties in Hebei Province from 2014 to 2018 as the research object, this paper uses propensity score matching and difference-in-difference methods (PSM-DID) to establish a multiple regression model, and to analyze the impact of transfer payments in national key ecological function areas on farmers’ incomes based on the regression model results. In the mechanism test, the intermediary mechanism of “environmental effect” in the relationship between ecological compensation policy and farmers’ incomes is discussed. The heterogeneity test is carried out on the structure of the agricultural labor force and the grouping of local fiscal expenditures to analyze the differential impact of transfer payments in national key ecological function areas on farmers’ income effects under different county characteristics. Finally, policy recommendations are proposed from the scale and standard of transfer payments in national key ecological function areas, government assessment and ecological industry development, with the aim to provide theoretical and practical bases for the realization of the dual goals of ecological compensation policy “promoting environmental protection and ensuring people’s livelihood”.

3. Empirical Design

3.1. Data Sources

Due to the delays in data collection, the data used in this article mainly emanate from various official statistical yearbooks for the years 2015–2019. These yearbooks include the “Hebei Provincial Economic Yearbook”, the “Hebei Provincial Rural Statistical Yearbook”, and the “China County Statistical Yearbook”, as well as yearbooks and county-level bulletins from cities within Hebei Province. The data related to environmental quality was obtained from the official website of the Department of Ecology and Environment of Hebei Province. Additionally, data regarding transfer payments for national key ecological function areas were sourced from the “Disclosure by Application” system, accessible via the portal website of the Hebei Provincial Government. Finally, Stata 17.0 software was used to process and analyze the collected data.

In this paper, the samples were divided into a treatment group and a control group based on whether the sample was included in the national key ecological function areas. The county-level units included in the transfer payment of national key ecological function zones are the treatment group, and the county-level units that are not included are the control group. Hebei Province contains a total of 167 county-level units, including 91 counties, 6 autonomous counties, 47 municipal districts, and 21 county-level cities. Of these, prefecture-level cities and municipal districts are also county-level administrative units. However, as there are obvious differences between them in terms of economic structure and autonomy [34], they are not included in the research scope of this paper together with counties and county-level cities. Nevertheless, considering the completeness of the samples in the treatment group, some municipal districts in the treatment group were retained. This paper will exclude them in the subsequent robustness test to verify the stability of the model. There were two batches of establishment of national key ecological function zones in Hebei Province, the first in 2008 and the second in 2016. Considering the availability of data and the effect of policy processing, this paper uses the second batch of selected county-level data in 2016, and excludes some county-level units with serious missing data (hereinafter collectively referred to as counties). Finally, 102 samples were screened out, with 22 in the treatment group and 80 in the control group.

Table 1. List of counties included in the national key ecological function areas in Hebei Province.

National Key Ecological Function Area (2008 year)	Fengning Manchu Autonomous County, Weichang Manchurian Autonomous County, Zhangbei County, Kangbao County, Guyuan County, Shangyi County
National Key Ecological Function Area (2016 year)	Lingshou County, Zanhuang County, Qinglong Manchu Autonomous County, Xingtai County, Fuping County, Laiyuan County, Yi County, Quyang County, Shunping County, Xuanhua District, Yu County, Yangyuan County, Huaian County, Wanquan District, Huai Lai County, Zhuolu County, Chicheng County, Chongli District, Chengde County, Xinglong County, Luanping County, Kuancheng Manchu Autonomous County

Note: The content comes from the “Reply of the State Council on Approving the Incorporation of Some Counties (Cities, Districts, and Banners) into National Key Ecological Functional Areas” and “Hebei Province’s Main Functional Area Planning”.

is a list of two batches of county-level units in Hebei Province that have been included in the national key ecological function areas.

3.2. Variable Description

3.2.1. Explained Variables

Selection of explained variables: The explained variables in this paper mainly describe the impact of transfer payments in national key ecological function areas on farmers’ incomes. The logarithm of the per capita disposable income of farmers indicates the income level of farmers. Taking the logarithm of income can help account for the fact that income distributions are often heavily skewed, and the logarithmic transformation can improve the fit and interpretability of regression models.

3.2.2. Core Explanatory Variables

This paper selects the difference-in-differences variable (DIDit) of the interaction term of the policy dummy variable (DZit) and the time dummy variable (DTit) as the core explanatory variable. It reflects the effect of transfer payments on farmers’ incomes in the treatment and control groups in national key function areas.

3.2.3. Control Variables

Referring to the studies of Qi [35], Miao [28], and Zhu [36], this paper selects the degree of industrialization, the degree of agriculturalization, the scale of the local government, the level of education, the rate of urbanization, the level of medical care, per capita fiscal revenue, per capita fiscal, and the agricultural labor force structure as control variables. The descriptive statistics of the above variables are shown in

Table 2. Descriptive of each variable.

Variable Name	Symbol	Definition	Unit
Farmers’ income level	Income	The logarithm of the per capita disposable income of farmers	/
Level of industrial development	Lncp	The logarithm of the number of industrial enterprises above designated size	/
Green level	Pforest	Afforestation area/total population at the end of the year	square kilometers per 10,000 people
Air quality level	Lnpm2.5	Logarithm of fine particulate matter concentration (PM2.5) in the air	/
Time dummy variable	DT	Take 0 before 2016, take 1 after 2016	/
Policy dummy variable	DZ	The area that enjoys the transfer payment takes 1, otherwise takes 0	/
Difference-in-differences variable	DID	The product of the time dummy variable and the policy dummy variable	/
Degree of industrialization	Indus	County primary industry GDP/county total GDP	%
Agriculturalization degree	Agri	County secondary industry GDP/county total GDP	%
Government financial scale	Govs	County government fiscal expenditure/county GDP	%
Education level	Edu	The logarithm of the number of primary and secondary school students	/
Urbanization rate	Urban	County non-rural population at the end of the year/county total population at the end of the year	%
medical level	Med	Total medical beds in counties/total population at the end of the year in counties	medical beds per 10,000 people
Per capita fiscal expenditure	Pfex	County fiscal expenditure/county total population at the end of the year	yuan per person
Per capita fiscal revenue	Pfre	County fiscal revenue/county total population at the end of the year	yuan per person
Agricultural Workforce	Labor	Employment in agriculture, forestry, animal husbandry And fishery/county total population at the end of the year	%

.

3.3. Model Building

As a public policy carried out in some counties of various provinces in China, transfer payments in national key ecological function areas provide a good “quasi-natural experiment” for this paper. As an effect evaluation method widely used after policy implementation, the DID can be used to evaluate the difference between the specific behaviors of the control group and the treatment group before and after the implementation of relevant policies. Furthermore, it can reflect “policy treatment effects” to a certain extent [37]. This paper uses the DID method to divide the county-level samples of Hebei Province into two groups: one group is the county that was first included in the national key ecological function area in 2016 and received the central transfer payment, and the other group is the county that was never included in the national key ecological function areas. This paper needs to construct a time dummy variable DT = {0, 1}. If it is in the year when the national key ecological function area transfer payment can be obtained (after 2016), DT = 1; otherwise, DT = 0. Then, it constructs the policy dummy variable DZ = {0, 1}. When the county belongs to the national key ecological function area, DZ = 1; otherwise DZ = 0. According to the two binary dummy variables defined above, an interaction term DID = DT * DZ is defined to describe the “policy treatment effects” after a county becomes a national key ecological function area. The purpose of this paper is to analyze the income level difference between farmers who received transfer payments in national key ecological function areas (the treatment group) and farmers who did not receive transfer payments (the control group). Income_it is defined as the impact effect of farmers’ income in county_i in period t and serves as the final explained variable of this paper. Therefore, the basic model is assumed to be:

Income_it = α₀ + α₁ DID_it + ε_it

(1)

The DID model assumes that the treatment group and the control group have the same “time effect” trend, and that the changes in the results before and after the experiment are purely caused by the policy treatment effect [38]. However, whether a county can be included in the national key ecological function area may not be a random event. In this paper, we use the PSM method to effectively solve the problem of selection bias. Combining PSM with a DID analysis can help address potential confounding factors that may influence the outcome variable and affect the validity of the estimated treatment effect. Matching the treatment and control groups based on their propensity scores, PSM can help to reduce the bias and increase the precision of the estimated treatment effect. This is because it can balance the distribution of observed covariates between the treatment and control groups, which can help to isolate the effect of the treatment from other factors that may affect the outcome variable. The main idea of this method is to construct a control group that has not been included in the national key ecological function area and, through matching, maintain the basic characteristics of the counties in the treatment group before inclusion. After matching, the two groups only differ in whether they receive national key ecological function transfer payment. Combining PSM-DID not only solves the sample selection bias, but also avoids the endogenous problem caused by the omission of dependent variables, allowing us to obtain the policy treatment effect.

The steps for PSM are as follows: First, design the main characteristic variables before matching to ensure consistency of the basic characteristics of the treatment and control groups that will affect whether a county can obtain transfer payment areas for national ecological function areas. We use Control_it to express the main characteristic variables. Second, construct a regression model where the explanatory variable is a binary dummy variable, with a value of 1 for the treatment group and 0 for the control group. Third, calculate the propensity score (P), i.e.,

P = Pr{DZ_it } = φ{Control_it }

(2)

The fourth step is to select a specific matching method based on the propensity score value. For each county_i in the treatment group, we find other counties with similar propensity scores in the corresponding control group. The fifth step is to remove unmatched samples, and perform regression analysis on the newly matched control group and the original treatment group. The corresponding estimation model is:

Income_it = α₀ + α₁ DID_it + Control_it + ε_it

(3)

Here, Control_it represents the characteristic variable that affects farmers’ incomes and also influences whether the county can obtain transfer payments for the national key ecological function areas. This model serves as a benchmark to evaluate the impact of transfer payments in national key ecological function areas on farmers’ incomes.

4. Results

4.1. Descriptive Analysis

A simple descriptive analysis was first performed on the treatment group and the control group without propensity score matching.

Table 3. Descriptive statistics of main variables.

Variable	Full Sample (N = 510)			Treatment (N = 110)		Control (N = 400)
	Mean	SD	Range	Mean	SD	Mean	SD
Pforest	94.16	173.32	2625.35	257.68	303.09	49.19	61.86
Lnpm2.5	4.17	0.62	2.67	4.11	0.48	4.46	0.33
Lncp	4.17	0.62	2.52	3.62	0.57	4.32	0.54
Income	9.25	0.31	1.45	9.03	0.28	9.31	0.29
Indus	42.73	11.49	56.35	39.87	12.62	43.52	11.05
Agri	18.38	8.4	39.43	20.04	8.20	17.92	8.41
Urban	43.75	9.03	56.70	40.19	8.98	44.73	8.81
Edu	10.77	0.47	2.81	10.56	0.49	10.82	0.45
Govs	0.21	0.11	0.87	0.28	0.15	0.19	0.08
Med	0.38	0.16	1.31	0.41	0.14	0.37	0.16
Pfre	0.19	0.21	2.07	0.18	0.10	0.19	0.23
Pfex	5.82	3.26	31.93	7.19	3.40	5.44	3.12
Labor	22.24	7.19	39.36	26.26	5.78	21.14	7.15
DT	0.60	0.49	1.00	-	-	-	-
DZ	0.22	0.41	1.00	-	-	-	-
DZ * DT	0.06	0.23	1.00	-	-	-	-

describes the basic characteristics of the main variables in both the treatment group and the control group. The value of N, which represents the total number of county-year observations included in the study, is derived by multiplying the number of counties in each group by the duration of the study period, which spans 5 years. Specifically, the treatment group, consisting of 22 counties, yields a value of N equal to 110, while the control group, consisting of 80 counties, yields a value of N equal to 400. It was found that Pforest of counties in the treatment group was much higher than that of the control group; in contrast, Lnpm2.5, Lncp and Income of the treatment group were slightly lower than those in the control group.

In terms of control variables, Agri, Govs, and Edu in the treatment group counties were higher than those in the control group; Pfex and Labor were also higher in the treatment group. Meanwhile, Indu, Urban, Edu, and Pfre were lower in the treatment group compared to the control group. Overall, significant differences exist between the treatment group and the control group, indicating that the counties that received transfer payments in national key ecological function areas have different characteristics in different regions. Therefore, it is effective to select the above control variables as matching characteristic variables.

Table 3 displays the descriptive statistics of all sample-related variables. The data shows that Pforest and Indu are relatively large, and Pforest, Indu, Agri, Urban, and Labor have a large extreme deviation. This suggests that the overall levels of some variables change greatly during the period.

4.2. PSM and Inspection

First, the PSM method was used to match the treatment group to an appropriate control group. The specific steps were as follows: Select the estimation model—the Logit model was chosen to estimate the propensity score and Formula (2) was used to calculate the propensity score. Choose the matching method drew on the study of Abadie [39] and adopted the “k-nearest neighbor matching” method with k = 4. Finally, 96 county-level samples of the treatment group and 397 county-level samples of the control group were obtained.

In order to ensure the reliability of the matching results, both the common value test and balance test were carried out before the panel regression.

① Common value test: after comparing the kernel density maps of the propensity score values of the treatment group and the control group after “k-nearest neighbor matching”, as shown in Figure 1 and Figure 2, it was found that there was almost no overlap between the treatment group and the control group before matching, and the score trends were also inconsistent. If the assumption of common values was met, subsequent matching could be carried out. After matching, the treatment group and the control group tended to overlap in most areas and tended to be consistent overall.

② The matching balance test. The matching balance test hypothesis requires that:

DZ_it ⊥ Control _it | P (Control _it)

It means that under the condition that the probability that a county can obtain the central transfer payment is P (Control _it), whether the county can obtain the central transfer payment and its control variables are independent of each other [38]. Judging whether the balance test is satisfied after matching can be observed from two aspects: the change of t statistics before and after matching, and the change of standard deviation before and after matching. As shown in

Table 4. Matching balance test results.

Variable	Matching	Mean		STD (%)	t-Test
		Treatment	Control		t	p > |t|
Indus	Before	39.871	43.518	−30.80	−2.97	0.003
	After	40.174	39.068	9.30	0.66	0.510
Agri	Before	20.040	17.925	25.50	2.35	0.019
	After	20.409	22.104	−20.40	−1.45	0.150
Urban	Before	40.188	44.731	−51.10	−4.77	0.000
	After	40.028	39.807	2.50	0.19	0.849
Edu	Before	10.562	10.824	−56.10	−5.34	0.000
	After	10.635	10.574	13.10	0.96	0.340
Govs	Before	0.280	0.188	76.30	8.66	0.000
	After	0.241	0.245	−3.30	−0.33	0.746
Med	Before	0.412	0.369	27.90	2.49	0.013
	After	0.403	0.408	−3.70	−0.33	0.739
Pfre	Before	0.181	0.189	−4.90	−0.39	0.700
	After	0.166	0.153	7.70	0.99	0.326
Pfex	Before	7.193	5.439	53.80	5.12	0.000
	After	6.274	5.949	10.00	1.28	0.204
Labor	Before	26.257	21.138	78.70	6.91	0.000
	After	25.995	26.584	−9.10	−0.65	0.515

, if the t-statistic becomes insignificant after matching, it means that the matching is valid. Most of the control variables changed from significant before matching to non-significant after matching, while some variables were not significant before matching and remained insignificant after matching. It indicates that all variables passed the test of t statistics. From the perspective of the change of standard deviation, if the absolute value of the standard deviation after matching is smaller, it means that the matching effect is better [39]. If the absolute value of the standard deviation after matching is lower than 20%, it means that the matching is effective. The absolute value of the standard deviation after matching of most variables is lower than 10%, and only the standard deviation of Agri after matching is 20.4%, slightly higher than 20%. This matching basically satisfies the balance assumption, and the “k-nearest neighbor matching” method chosen in this paper is effective.

4.3. Basic Regression

Table 5. Basic regression empirical results of the relationship between policy and Income.

Variable	Model 1	Model 2	Model 3	Model 4
	Income	Income	Income	Income
DID	0.2211 ***	0.0366 ***	0.0518 ***	0.0311 ***
	(0.0302)	(0.0061)	(0.0135)	(0.0059)
Indus			−0.0018 **	0.0011 ***
			(0.0008)	(0.0004)
Agri			−0.0043 ***	0.0017 **
			(0.0013)	(0.0008)
Urban			0.0245 ***	0.0030 ***
			(0.0014)	(0.0011)
Edu			0.1494 ***	0.0027
			(0.0301)	(0.0161)
Govs			0.2324 *	0.1209 **
			(0.1245)	(0.0608)
Med			0.3314 ***	0.0323
			(0.0556)	(0.0313)
Pfre			−0.2027 ***	−0.0435
			(0.0784)	(0.0342)
Pfex			0.0108 ***	−0.0028
			(0.0042)	(0.0018)
Labor			−0.0008	−0.0002
			(0.0012)	(0.0006)
Constant	9.2245 ***	9.0015 ***	6.5394 ***	8.7642 ***
	(0.0270)	(0.0983)	(0.3331)	(0.2117)
Region fixed effect	No	Yes	No	Yes
Year fixed effect	No	Yes	No	Yes
Observation	493	493	493	493

Notes: Standard errors are in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

depicts the benchmark model where the explained variable is Income, and this model is the result of full sample estimation based on Formula (3). Among them, Model 1 verifies the impact of the core explanatory variable DID on Income, Model 2 considers the fixed effects of region and year on the basis of Model 1, and Model 3 adds control variables on the basis of Model 1. By contrast, Model 4 considers the fixed effects of region and year while adding the control variables.

The regression results show that from Model 1 to Model 4, after the county enjoyed the transfer payment in the national key ecological function areas, the Income had been significantly improved. Moreover, all coefficients were significant at the 1% level and the regression results of the model were all positive, indicating that the Model is stable. After adding control variables—region and time fixed effects—the regression coefficients changed to a certain extent. Specifically, after controlling for the region and year fixed effects and comparing Model 1 with Model 2, we found that the impact of policies on improving income had been weakened. The regression coefficient changed from 0.2211 to 0.0366 and the regression coefficient decreased by 83.45%. This shows that the county was greatly affected by the city where it is located and the time. After comparing Model 1 with Model 3 and adding control variables, the positive impact of the transfer payment on the county was also weakened, and the regression coefficient dropped from 0.2211 to 0.0518, with a 76.57% decrease in the regression coefficient. This shows that the selection of control variables is effective. According to the decline of the regression coefficient, it can be preliminarily judged that the inhibitory effect of the time and region fixed effects are slightly higher than those of the control variables.

The benchmark Model 4 is used for specific illustration. After the county was affected by the transfer payment policy, the income of farmers in the county increased by 3.11 percentage points. The high degree of overlap between national key ecological function areas and poverty-stricken areas shows that the implementation of transfer payment policy in the national key ecological function areas had also played a role in poverty reduction to a certain extent. Table 5 reports the yearly effect of policy implementation on income. It was found that, as time went by, the effect of policy treatment was getting larger. In other words, the effect of improving farmers’ income levels brought about by the policy became more successful. Among the control variables, the regression coefficients of Indu, Agri, Urban, and Govs are all significant, and the regression coefficients are positively related to farmers’ incomes. This indicates that the increase in each of these control variables is sufficient to improve farmers’ income levels to a certain extent.

4.4. Robustness Test

The robustness of the conclusions was tested by parallel trend tests, changing propensity score matching methods, screening samples, and changing matching characteristic variables.

4.4.1. Parallel Trend Test

This article used the parallel trend test to determine whether the treatment group and the control group met the parallel trend before receiving treatment. Figure 3 shows the average growth trend of Income. Before the implementation of the policy, the growth trend of Income for the treatment group and the control group was basically the same from 2014 to 2016. After the implementation of the policy, the slope of Income in the control group increased year by year. Figure 4 shows that the treatment group and the control group satisfied the parallel trend test.

4.4.2. Changing the PSM Matching Method

Considering that different PSM matching methods might affect the result of the final regression equation, this article used two matching methods: caliper matching and kernel matching. In caliper matching, we used a caliper with a value less than or equal to ¼ of the propensity score standard deviation, and the absolute distance of the score was limited to 0.1 times the propensity score sample standard deviation [40]. Then, we observed whether the regression results would change after the updated matching method screened the samples.

Table 6. Balance test results of caliper matching and kernel matching.

Variable	Matching	Caliper Matching		Kernel Matching
		STD (%)	p > |t|	STD (%)	p > |t|
Indu	Before	−30.8	0.003	−30.8	0.003
	After	−3.6	0.806	−2.8	0.847
Agri	Before	25.5	0.019	25.5	0.019
	After	9.3	0.532	8.2	0.565
Urban	Before	−50.1	0.000	−51.1	0.000
	After	−10.0	0.485	−13.1	0.329
Edu	Before	−56.1	0.000	−56.1	0.000
	After	0.2	0.986	−10.1	0.484
Govs	Before	76.3	0.000	76.3	0.000
	After	1.4	0.892	14.2	0.179
Med	Before	27.9	0.013	27.9	0.013
	After	0.4	0.976	−17.3	0.219
Pfre	Before	−4.9	0.700	−4.9	0.700
	After	−15.5	0.392	−8.3	0.564
Pfex	Before	53.8	0.000	53.8	0.000
	After	−14.3	0.353	−4.5	0.738
Labor	Before	78.7	0.000	78.7	0.000
	After	−5.0	0.741	−11.9	0.428

reports the balance test results of caliper matching and kernel matching. According to the judgment criteria of k-nearest neighbor matching, the absolute value of the standard deviation after matching was not more than 20%, and the t statistics after matching were not significant. This indicates that both the caliper and kernel matching have passed the balance test.

Table 7. Comparison of regression results after caliper, kernel, and k-nearest neighbor matching.

Variable	Caliper Matching	Kernel Matching	K-Nearest Neighbor Matching
DID	0.0274 ***	0.0285 ***	0.0311 ***
	(0.0061)	(0.0062)	(0.0059)
Constant	−0.0064	8.9444 ***	8.7642 ***
	(0.0156)	(0.2104)	(0.2117)
Control	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes
Observation value	429	501	493

Notes: Standard errors are in parentheses; *** p < 0.01.

reports the regression results obtained by selecting some explanatory variables according to Formula (3) after caliper and kernel were screened. After changing the method, although the regression coefficient and significance level of the core explanatory variables in the regression results had changed slightly, the overall trend was basically consistent with the above regression results.

4.4.3. Further Screening of Samples

The samples were subsequently screened as follows: As mentioned above, there are significant differences between the municipal districts and counties and autonomous counties in terms of economic structure and governance power, which can affect the regression results. Therefore, this paper excluded the municipal districts that were previously included in the treatment group. Since there are only a small number of transfer payments for provincial-level key ecological function areas in Hebei Province, this paper excluded the counties included in the provincial-level key ecological function areas in Hebei Province from the total sample to avoid other policy interference.

In summary, a total of 70 samples were eliminated from the sample.

Table 8. Comparison of regression results before and after removing samples.

Variable	After	Before
DID	0.0351 ***	0.0311 ***
	(0.0061)	(0.0059)
Constant	8.6143 ***	8.7642 ***
	(0.2464)	(0.2117)
Control	Yes	Yes
Regional fixed effect	Yes	Yes
Year fixed effect	Yes	Yes
Observation value	423	493

Notes: Standard errors are in brackets; *** p < 0.01.

reports the results of the regression model estimated according to Formula (3) and the original benchmark regression model for some of the explained variables (k-nearest neighbor is used here). After removing the samples from the treatment group, although the regression results of the core explanatory variable DID change slightly, the overall influence trend remains unchanged.

4.4.4. Replacement of Characteristic Variables

PSM aims to match the treatment group and control group based on characteristic variables in order to find similar individuals. Except for the dummy variable indicating whether they received national key ecological function areas transfer payment, the remaining characteristic variables should be kept as consistent as possible. When the selected feature variables differ, the results and regression outcomes also vary. This paper follows the method of Huang [41] and verifies the stability of the regression results by eliminating characteristic variables. With reference to Li’s [14] classification method of characteristic variables, this paper divides them into the comprehensive level of counties (Indus, Agri, Edu, Med), county population and employment structure (Urban, Labor), and county fiscal level (Govs, Pfre, Pfex).

Table 9. Regression results after replacing variables.

Variable	Remove Comprehensive Feature	Remove Population and Employment	Remove Financial Level
DID	0.0309 ***	0.0233 ***	0.0376 ***
	(0.0059)	(0.0062)	(0.0062)
Indus		0.0011 ***	0.0008 **
		(0.0004)	(0.0004)
Agri		0.0018 **	0.0018 **
		(0.0008)	(0.0008)
Edu		−0.0052	−0.0122
		(0.0197)	(0.0165)
Med		0.0027	0.0220
		(0.0248)	(0.0266)
Govs	0.0416	0.0793
	(0.0739)	(0.0718)
Pfre	−0.0634	−0.0326
	(0.0484)	(0.0658)
Pfex	−0.0012	−0.0014
	(0.0029)	(0.0033)
Urban	0.0025 *		0.0022 **
	(0.0013)		(0.0011)
Labor	0.0001		−0.0004
	(0.0006)		(0.0006)
Constant	8.9111 ***	8.9229 ***	8.9576 ***
	(0.1210)	(0.2384)	(0.2102)
Regional fixed effect	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes
Observation value	459	423	507

Notes: Standard errors are in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the regression results of some explanatory variables after removing the three types of indicators, demonstrating that all regression coefficients are significant. Although the regression coefficients have changed to a certain extent, the overall trend remains consistent with the above.

4.5. Mechanism Test of “Environmental Effects”

As restricted development areas, key ecological function counties undertake the dual missions of ecological protection and economic development [42]. Based on these two policy objectives and with reference to the method of Ding [43], this paper uses Lnpm2.5 as the representative variable for “environmental effect” to construct a mediation effect model to verify the transmission mechanism of the impact of transfer payments in national key ecological function areas on farmers’ income.

The mediation model constructed is shown in Formulas (4)–(6), and the specific steps are as follows: the first step is to regress the explained variable and the explanatory variable; the second step is to regress the intermediary variable and the explanatory variable; finally, the explanatory variables are regressed with the explanatory and mediator variables.

The intermediary model of the “environmental effect” of transfer payments in national key ecological function areas consists of the following equations:

$${Income}_{it}=f({DID}_{it},{Control}_{it})$$

(4)

$${Lnpm2.5}_{it}=f({DID}_{it},{Control}_{it})$$

(5)

$${Income}_{it}=f({DID}_{it},{{Lnpm2.5}_{it},Control}_{it})$$

(6)

Models 5–7 report the test results of “environmental effect” in

Table 10. “Environmental effect” mechanism test.

Variable	Model 5	Model 6	Model 7
	Income	Lnpm2.5	Income
DID	0.0311 ***	−0.0675 *	0.0277 ***
	(0.0059)	(0.0372)	(0.0063)
Lnpm2.5			−0.1864 ***
			(0.0144)
Constant	8.7642 ***	4.8527 ***	8.9734 ***
	(0.2117)	(0.3799)	(0.2102)
Control	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes
Observation value	493	493	493

Notes: Standard errors are in parentheses; *** p < 0.01, * p < 0.1.

. The coefficient of the core explanatory variable DID in Model 5 is positive and has passed the significance test. It indicates that transfer payments in the national key ecological function area can significantly improve the income levels of farmers in the counties affected by the policy. The core explanatory variable DID in Model 6 is negative and has passed the significance test. It shows that the county can reduce PM2.5 and improve the ecological environment after enjoying the transfer payment in the national key ecological function areas. The intermediary variable Lnpm2.5 coefficient in Model 7 is negative and has passed the significance test. It indicates that improving the ecological environment can promote the increase in farmers’ incomes. The reason is that improving the ecological environment is conducive to improving local environmental governance and environmental productivity as well as increasing crop yield and farmers’ incomes. Models 5–7 verified the “environmental effects” mechanism of transfer payments in national key ecological function areas on farmers’ income levels. The results indicate that the impact of transfer payments in national key ecological function areas on farmers’ incomes can be realized by affecting the ecological environment.

4.6. Heterogeneity Analysis

Regression analysis shows that the impact of policies on farmers’ incomes in counties is greatly affected by regional fixed effects.

Table 11. The results of heterogeneity test.

Variable		N	DID	Control	Regional Fixed Effect	Time Fixed Effect	Constant
Proportion of agricultural labor force	Low	165	0.0088 (0.0333)	Yes	Yes	Yes	6.4839 *** (0.6370)
	Medium	164	0.0442 (0.0227)	Yes	Yes	Yes	7.3525 *** (0.4700)
	High	164	0.0542 *** (0.0169)	Yes	Yes	Yes	6.6273 *** (0.5633)
Local fiscal expenditure	Low	165	0.1214 *** (0.0301)	Yes	Yes	Yes	5.9214 *** (0.6175)
	Medium	164	0.0174 (0.0235)	Yes	Yes	Yes	6.5243 *** (0.6298)
	High	164	0.0204 (0.0244)	Yes	Yes	Yes	6.8348 *** (0.4809)

Notes: Standard errors are in parentheses; *** p < 0.01.

reports the policy effects of different regional characteristic groups from two aspects: regional agricultural labor force structure and fiscal expenditure level.

The agricultural labor force structure is closely associated with the economic and ecological development of counties. Counties with different labor force structures may have different impacts on farmers’ incomes after enjoying ecological compensation in national key ecological function areas. According to the proportion of the sample agricultural, forestry, animal husbandry, and fishery labor force in the total population, the agricultural labor force is divided into three parts: “low proportion of agricultural labor force”, “medium proportion of agricultural labor force”, and “high proportion of agricultural labor force” in this paper. Explanatory variables of farmers’ incomes were subjected to multiple regression.

The transfer payment standards for counties with key ecological functions are mainly based on the local fiscal revenue and expenditure gap. At the same time, as an important structural policy, fiscal and tax policies play a critical role in regulating and promoting regional coordinated development [44]. Generally speaking, the higher the level of local fiscal expenditure, the stronger the local self-development ability [45]. Therefore, under different levels of fiscal expenditure, there may be differences in the policy effects of counties after enjoying transfer payments. In this paper, according to the amount of local fiscal expenditure, it is divided into three intervals: “low local fiscal expenditure”, “medium local fiscal expenditure”, and “high local fiscal expenditure”. The income of farmers was subjected to multiple regression.

The heterogeneity analysis shows that only in counties with a higher proportion of agricultural labor force and lower local fiscal expenditure, the implementation of the policy has a significant impact on the income levels of farmers and can increase the incomes of local farmers. The high proportion of agricultural labor force and low local fiscal expenditure both reflect the poverty characteristics of counties. It shows that transfer payment policy in the national key ecological function areas has a significant poverty alleviation function to some extent.

5. Conclusions

5.1. Research Conclusions

In the implementation process of China’s ecological compensation policy mechanism, the transfer payment in national key ecological function areas is the largest ecological compensation policy involving the most areas. In 2016, in the “Reply of the State Council Concerning the Approval of Adding Some Counties (Cities, Districts, and Banners) to National Key Ecological Functional Areas” issued by the State Council, a considerable number of county-level units were included in the national key ecological functional areas. In this paper, the counties in Hebei Province that were included in the national key ecological function areas in 2016 were used for quasi-natural experiments. Moreover, the county sample data of Hebei Province from 2014 to 2018 and the PSM were used to screen out a suitable control group for the counties newly included in national key ecological function areas. The DID model was used to identify the cause and effect and to evaluate the impact of the transfer payment policy on the incomes of local farmers. Furthermore, the reliability of the regression results was ensured through a variety of stability tests.

The study’s results highlight that the national key ecological function zone’s transfer payment policy has a positive impact on the income of local farmers, increasing it by 3.11%. The ecological compensation policy achieves this by generating “environmental effects”, which improve the environmental conditions of counties and, in turn, boost the income levels of local farmers. This finding contradicts Wen et al.’s research, which found that ecological compensation had a negative effect on the income of agricultural species but a positive effect on the participation and income of non-agricultural self-employment [46]. The discrepancy could be attributed to the fact that, at this stage, the primary source of income for farmers may no longer be agricultural planting but non-agricultural operating income. As Liu et al.’s research found, compensation programs significantly impact farmers’ participation in off-farm activities and lead to significant livelihood changes [47]. Additionally, the different geographic locations and policy implementation environments selected in the two studies likely contributed to the disparate findings.

Moreover, the study suggests that as the policy implementation period extends, the positive impact on farmers’ income levels will become more significant, though it will be heavily influenced by regional characteristics. This conclusion is in line with the findings of Liu et al. regarding the delayed impact of policies on livelihoods and regional disparities [47]. However, the spatial disparities posited by Zhang’s research [48], which purportedly exhibit a tendency towards gradual contraction, have yet to be substantiated. One plausible explanation for the lack of regional convergence in farmers’ income levels could be attributed to the fact that the primary impetus and modus operandi of ecological compensation in national key functional areas lies in incentivizing local governments to augment their investments in environmental protection and monitoring [49], rather than elevating farmers’ incomes. Inadequate consideration for farmers’ income levels in the policy implementation coupled with the absence of any discernible trend towards regional convergence in this regard further bolsters this notion. Since the Ministry of Finance issued the “2012 Measures for Transfer Payments from the Central Government to Local National Key Ecological Functional Areas” in 2012, until 2021 the central government’s transfer payments to local key ecological functional areas have accounted for no more than 2% of the total general transfer payments from the central government to local governments in that year [50]. The compensation scale is small, and the actual compensation value is often lower than the expectations of farmers.

Income factors exert a significant impact on farmers’ attitudes toward and implementation of ecological compensation policies, and farmers hold high expectations for compensation income [51]. This phenomenon underscores the considerable influence of economic compensation on farmers’ acceptance of ecological policies. Heterogeneous analysis demonstrates that the poverty alleviation effect of the national key ecological function zone transfer payment policy is more pronounced in poorer counties. National key ecological function zones exhibit significant overlap with impoverished areas, where a preexisting gap in economic development exists compared to non-ecological function zones. These regions often face dual constraints of inherent underdevelopment and insufficient policy compensation. Therefore, it is incumbent upon the central government to expand the scale of transfer payments to national key ecological function zones, and future strategic considerations must account for regional characteristics when designing policies for poverty reduction and local rural community revitalization.

5.2. Policy Recommendations

The development of national ecological compensation policies requires further refinement to create corresponding plans that align with compensation objectives and targets. To achieve the goals of ecological conservation and livelihood protection, the support of ecological compensation policies and government actions is essential [52]. However, significant differences exist in the various methods used to calculate ecological compensation standards. On the one hand, maximizing ecosystem service supply and economic incentives for farmers is crucial [53]. On the other hand, the potential for a “marginal effect” must be considered to avoid exceeding the needs of regional development after compensation standards have reached a certain threshold [54]. Therefore, only through the scientific planning of ecological compensation methods and standards can the dual effects of ecological conservation and increased income for farmers be achieved [55].

However, relying solely on transfer payments led by the central government is difficult to meet the funding needs of local governments for the construction of ecological environment quality and improvement of people’s livelihoods. Moreover, it is even more difficult to make up for the opportunity cost lost due to the positive externalities of ecological services. Therefore, it is possible to use the benefit-sharing of the green financial market to attract social capital into the ecological industry so as to improve the quality of the ecological environment and promote the coordinated development of economic construction and ecological construction through “environmental effects”. Meanwhile, local government departments, as policy executors and direct users of funds, must improve policy execution and establish sound and effective assessment and incentive mechanisms. The impact of the ecological transfer payment system is multifaceted, and a single indicator system cannot reflect the efforts of local governments [56]. The study found that there is a time trend effect in the transfer payment of national key ecological functional areas and that the quality of the early ecological environment has a significant role in promoting the ecological effect. Therefore, government departments can consider introducing the effects of the previous period or even multiple periods of policy implementation into the assessment mechanism to comprehensively and fairly evaluate the contribution of local governments.

5.3. Prospective

This article examines the construction of the National Key Ecological Function Zone in Hebei Province, China as a quasi-natural experiment, introducing ecological protection mechanisms to explore the impact of policy implementation on the incomes of local farmers. The study provides theoretical verification of the existence of “environmental effects” and demonstrates that ecological protection and improvement can be achieved simultaneously. With its unique geographical location adjacent to China’s capital city, Hebei Province’s achievements in ecological governance and livelihood security have important implications for China and other regions worldwide, offering valuable lessons for future policy implementation.

This paper employs official panel data to enhance the robustness of the regression results. However, despite this effort, some shortcomings remain, and there is ample scope for future development. To achieve more dynamic outcomes, the time dimension of panel data should be maximally extended. Owing to the challenges of accessing and assessing local ecological index data at the county level, the mechanism test estimation based on a single ecological environment index in this paper may be subject to bias. Thirdly, the increasing diversity of farmers’ income structure implies that analysis of the per capita disposable income data may overlook important income-related information. Hence, forthcoming research will endeavor to address these three issues.