The Impact of Rural Enterprise Park Policy on the Income of Rural Residents: Evidence from China

Abstract

Improving the income of rural residents plays an important part in sustainable rural development. Using a difference-in-difference (DID) approach and Chinese counties’ data from 2014 to 2019, this paper evaluates the effects of the rural enterprise park policy on the income of rural residents. We find that the policy significantly promotes rural residents’ income, and a series of additional tests suggest that the effects appear robust. We also find that stimulating entrepreneurial activities serves as a channel through which the policy affects rural incomes. Finally, the policy effects are mainly significant for counties with a higher level of human capital, a younger demographic age structure, and a higher level of financial development.

1. Introduction

Ever since the reform and opening up, the rapid development of the Chinese economy has been accompanied by a significant improvement in the level of urbanization. During this process, the problem of a hollow phenomenon and even recession in rural areas caused by the massive population transfer to cities has become increasingly prominent. Achieving rural revitalization has become an important issue for the Chinese government. In June 2017, the Ministry of Agriculture and Rural Affairs of China launched a policy of Rural Enterprise Parks (REPs), planning to establish 1096 enterprise parks in rural areas across the country to stimulate entrepreneurial activities and, thus, revitalize rural industries. According to China Rural Enterprise Parks Development Report (2020), the policy has achieved considerable success. By the end of 2019, 53.5% of the parks had a total output value of more than ¥ 100 million, and parks with more than 500 employees account for 48.8%, which shows, to a certain extent, the positive impact of REPs on rural economic development. The income of rural residents, as one of the core issues of rural economic development, naturally receives extensive attention (Li and Ma, 2021) [1]. Although China had planned to eliminate absolute poverty by 2020, relative poverty persisted for a long time, and it is more severe in rural areas than in urban areas (Sun et al., 2022), which will bring great challenges to sustainable development (Wang et al., 2022) [2,3]. Therefore, we are largely interested in the question of whether the REP policy promotes the income of rural residents.

Regarding the REP policy as a quasi-natural experiment, we employ a two-way fixed effect difference-in-difference (DID) approach and use the data of Chinese counties in the years 2014–2019 to evaluate the impact of the policy on rural residents’ income. Our identification strategy has two major advantages. On the one hand, the REP policy is generally exogenous for rural residents because it was issued by the central government, which greatly helps resolve the concern of reverse causality. On the other hand, with a common trend assumption, the policy effects are estimated by comparing the treated and the untreated groups, which alleviates the endogeneity problem posed by omitted variables. Our findings are as follows:

First, we found that the policy has a positive and significant effect on the income of rural residents. Since the validity of the DID approach relies on the parallel trend assumption, we conducted a parallel trend test using the event study method. To check whether the baseline regression results are driven by chance or some unobservable factors, we conducted a placebo test, which suggests that the results of the baseline regression are unlikely driven by chance. We then employed the propensity score matching (PSM)-DID method to address the concern of self-selection biases. Considering some counties have established multiple REPs, we further assessed the impact of the policy’s intensity.

Second, we investigated the underlying channel through which the REP policy affects the income of rural residents. Motivated by the policy’s objective of providing places and services for rural residents to start businesses, which makes some residents become entrepreneurs and/or creates jobs for more residents, we hypothesized that stimulating entrepreneurial activities would be the main channel through which the policy works. Exploiting the business registry data, we found that the REP policy indeed promotes entrepreneurial activities and, thus, affects rural incomes.

Finally, we explored the heterogeneous policy effects across counties. Since it has been well studied that human capital, the age structure of the population, and financial development are important for entrepreneurial activities, which is how the REP policy works, the effects of the policy on rural incomes should depend on these factors. Our heterogeneity analysis suggests that, for counties with a higher level of human capital, younger demographic age structure, and higher level of financial development, the REP policy appears to have a larger and more significant impact.

This paper contributes threefold to the literature. First, our study supplements the literature with the economic consequences of place-based policies by evaluating the effects of China’s rural enterprise park policy, a newly rural place-based policy, on the income of rural residents. The previous literature on place-based policies has mainly targeted America and Europe [4,5,6,7,8,9,10,11,12]. The scanty research on Chinese location-based policies primarily focus on various special economic zones (SEZ) [13,14,15,16]. Furthermore, almost all the above studies refer to place-based policies targeting urban areas rather than rural areas. This paper is among the first to provide solid evidence for the economic consequences of China’s rural enterprise park policy. Second, this study contributes to the literature on the determinants of the income of rural residents. The existing literature has documented that many elements would affect rural residents’ income, such as agricultural loans, digital finance, internet use, rural nonfarm sector development, road infrastructure, and so on [1,17,18,19,20,21]. Our study identifies a new determinant—China’s rural enterprise park policy—which refers to government efforts to improve economic performance (such as job opportunities, wage, and economic growth) of less developed areas. Third, this study is particularly relevant to the policymakers who want to employ place-based policies to realize poverty reduction. Our baseline results imply that the REP policy can serve as a tool to increase the income of rural residents, and the heterogeneous policy effects highlight the importance of promoting human capital accumulation and financial market development to make place-based policies effective.

The rest of the paper is structured as follows. Section 2 introduces the background of China’s rural enterprise park policy, followed by the literature review and hypothesis development in Section 3. Section 4 presents our empirical design, including estimation strategy, sample selection procedure, and variable construction. Section 5 reports the baseline regression results and additional robustness checks. Section 6 discusses the mechanism and heterogeneity. Section 7 concludes the paper.

2. A Brief Description of the Rural Enterprise Park Policy

In June 2017, the Ministry of Agriculture and Rural Affairs of China released the “National Rural Enterprise Parks Directory” for the first time and established 1096 enterprise parks in rural areas across the country to provide places and services for rural residents to start businesses. The construction of rural enterprise parks mainly provides services in the following aspects to stimulate the vitality of market entities and revitalize rural industries.

First, in terms of organizational leadership, by the end of 2017, the Ministry of Agriculture of China, together with 12 departments, including the National Development and Reform Commission, established a “coordination mechanism for promoting rural entrepreneurship” and set up offices under the agricultural departments of each province. Second, the parks have ensured the refinement and implementation of preferential policies regarding project application, entrepreneurship guidance, policy consultation, operation management, market expansion, entrepreneurship incubation, business registration, and financial services. Third, the Ministry of Agriculture of China provides training courses for leaders of rural enterprise parks to improve their service and management capabilities. In addition, activities such as rural entrepreneurship competitions, achievements display, and experience communications are also held to share entrepreneurial experience and build an entrepreneurial atmosphere.

According to China Rural Enterprise Parks Development Report (2020), the establishment of rural enterprise parks has achieved considerable success. First, by the end of 2019, 53.5% of the parks had a total output value of more than ¥ 100 million, and parks with a total output value of more than ¥ 1 billion account for 21%. Second, 52.3% of the parks included more than 10 market entities, and 20.8% of the parks had more than 50. Third, in terms of employment, 83.7% of the parks employed more than 100 people, and parks with more than 500 employees accounted for 48.8%.

To summarize, the establishment of rural enterprise parks largely facilitates entrepreneurial activities and creates a lot of jobs for rural residents. So, we expect that the policy plays an active role in the income of rural residents.

3. Literature Review and Hypothesis Development

3.1. Literature Review

Given that our study aims to examine the effects of China’s rural enterprise park policy, which is a newly developed place-based policy, on the income of rural residents, it is most closely related to the following two strands of literature:

The first strand is the literature on the economic consequences of place-based policies. The previous literature on place-based policies has mainly targeted America and Europe. Among place-based policies in the United States, the most heavily studied are Federal Empowerment Zones and State Enterprise Zones. Elvery (2009), Neumark and Kolko (2010), Ham et al. (2011), Freedman (2013), Busso et al. (2013), and Neumark and Young (2019, 2021) examine the impact of these two policies on employment, wages, and entrepreneurial activities [4,5,6,7,10,11,12]. Furthermore, there are also many papers studying place-based policies in European countries, including Regional Selective Assistance in the United Kingdom and the French Zones, Franches Urbaines, in France [22,23]. In recent years, research on Chinese location-based policies has mainly focused on various SEZs. Using a unique Chinese municipal dataset, Wang (2013) finds that the SEZ program increases foreign direct investment without crowding out domestic investment [16]. Based on the Chinese city data from 1988 to 2010, Alder et al. (2016) find that the establishment of a state-level SEZ is associated with an increase in GDP levels of about 20% [13]. According to Zheng et al. (2017) and Lu et al. (2019), establishing zones is found to have a positive effect on productivity, employment, and wages [15,24]. Using the mass closure of development zones in 2004 as a natural experiment, Chen et al. (2019) show that loss of special development zone policies led to an average loss of 6.5% of corporate TFP [25]. Lu et al. (2022) evaluate how China’s resource-exhausted city policy affects regional innovation capacity and find that the policy leads to a decrease in city-level innovation by increasing government intervention [26].

The second strand is the literature about the influencing elements of the rural residents’ income. Wang and Liu (2016) use quantile regression to examine the impacts of fiscal expenditure and agricultural loans on rural residents’ income using cross-sectional data from 853 counties of 11 western Chinese provinces in 2010 [20]. They find that the relationship between agricultural loans and rural income follows an inverted U-shape. Wang and Hu (2018) examine the influences of trade liberalization on rural poverty reduction in China and find that trade liberalization reduces rural poverty incidence and boosts income for poor rural populations [27]. Rural financing is a significant factor in the rise of farmer incomes. Sun et al. (2018) investigate the relationship between the social capital and farm households’ capacity to secure formal and informal loans and point out that friendship and kinship social capitals help farm families obtain formal and informal loans, respectively [28]. Li and Ma (2021) employ panel regression models to examine the effect of digital financial inclusion on the income of rural residents, utilizing a panel data set of 1624 counties on the Chinese mainland from 2014 to 2019 [1]. They show that digital financial inclusion considerably increases the income of rural residents, and industrial structure, education level, and financial development level will amplify this effect. Then, He et al. (2022) reach similar conclusions based on 2011–2019 panel data from 31 provinces in China [18]. Moreover, using the 2016 CFPS data, multiple linear regression analysis, and instrumental variable model, Zhou et al. (2020) explore the effects of internet use on rural residents’ income growth [21]. The results show that internet use can not only directly increase the income of rural residents, but also increase the income of rural residents by increasing entrepreneurship or non-agricultural employment. Han et al. (2021) study the impacts of rural nonfarm sector development on rural resident income and confirm a positive relationship [17]. Employing China’s “Broadband Village” (B&V) initiative as a quasi-natural experiment, which is a considerable investment in rural internet infrastructure in six western provinces, Liu et al. (2021) utilize the regression discontinuity design to evaluate the effect of the B&V project on rural residents’ income [29]. They find that the B&V project can significantly improve the income of rural residents, while the policy effects decrease with the levels of the rural residents’ income. In more recent contributions, Lu et al. (2023) confirm that road infrastructure leads to a rise in the wage and business income of rural residents; Xu et al. (2023) show that the establishment of nature reserves can increase rural residents’ income [19,30].

In summary, while rural economies also contribute immensely to the employment and prosperity of the country, most of the place-based policies studied in the aforementioned papers focus on urban areas rather than rural areas [31]. Meanwhile, to the best of our knowledge, the existing literature also has not documented the impacts of place-based policies targeting rural areas on the rural residents’ income. Further inspired by the fact that rural place-based policies such as enterprise hubs may serve as the channel of promoting rural economic development and thus contributing to the increase in rural residents’ income, it is relevant to explore the effects of rural place-based policies on the income of rural residents, especially in the context of China [31]. To be more specific, the world’s largest developing country, China, eradicated absolute poverty in 2020, but this does not mean that poverty alleviation and reduction are over; relative poverty will persist for a long time, and it is more severe in rural areas than in urban areas [2]. The effects of rural place-based policies on rural residents’ income would be of significant interest to China’s policymakers. Hence, our study attempts to fill this research gap and thus contribute to the above two strands of the literature by evaluating the impacts of China’s rural enterprise park policy, a newly rural place-based policy, on the income of rural residents.

3.2. Hypothesis Development

As we all know, China’s rural economic development has long been hampered by geographical isolation, a lack of capital, and restricted market access [31]. These challenges make it difficult to start businesses and thus provide nonfarm employment opportunities in rural areas, which is crucial for raising the income of rural residents. Consequently, resolving these obstacles and fostering economic activity is essential for sustainable rural development and rural incomes.

The establishment of enterprise parks in China’s rural areas provides supporting places and services, such as project application, entrepreneurship guidance, policy consultation, operation management, market expansion, entrepreneurship incubation, business registration, and financial services. The REP policy largely lowers the entry barrier to entrepreneurship.

By starting businesses, some rural residents become entrepreneurs or self-employed, which directly increases their incomes. Entrepreneurial activities create employment opportunities for more residents, which indirectly promotes the income of rural residents. Moreover, the agglomeration of enterprises increases the demand for the labor force and thus generates wage premiums, by which rural residents further enjoy higher income [23].

To summarize, the REP policy is expected to promote the income of rural residents, and stimulating entrepreneurial activities plays a key role in this income-increasing effect of the policy. Based on the above analysis, we propose the following two hypotheses:

Hypothesis 1. 

The rural enterprise park policy has a positive effect on the income of rural residents.

Hypothesis 2. 

Stimulating entrepreneurial activities serves as the mechanism through which the policy affects rural residents’ incomes.

4. Empirical Design

4.1. Estimation Strategy

Considering the limitation of data availability, the county is used as the unit of our empirical analysis. To accurately identify the effects of China’s rural enterprise parks policy on the income of rural residents, we undertake a standard DID approach by estimating the following econometric model:

$${\mathrm{Rincome}}_{\mathrm{it}}=\mathsf{\alpha }+\mathsf{\beta }{(\mathrm{Treat}}_{\mathrm{i}}\times {\mathrm{Post}}_{\mathrm{t}})+\mathsf{\gamma }{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

where $\mathrm{i}$ indexes county and $\mathrm{t}$ indexes year. ${\mathrm{R}\mathrm{i}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}}_{\mathrm{i}\mathrm{t}}$ refers to the income of rural residents of county $\mathrm{i}$ in year $\mathrm{t}$. ${\mathrm{Treat}}_{\mathrm{i}}\times {\mathrm{Post}}_{\mathrm{t}}$ is the key variable of interest, which is a dummy variable, and its coefficient $\mathsf{\beta }$ reflects the policy effects on the income of rural residents. Specifically, ${\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}}_{\mathrm{i}}$ is a treatment indicator, and ${\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}_{\mathrm{t}}$ is a post-treatment indicator. ${\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}}_{\mathrm{i}}$ equals one if a rural enterprise park was established in 2017, and zero otherwise. ${\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}_{\mathrm{t}}$ is assigned to one if $\mathrm{t}\ge 2017$, and zero otherwise. ${\mathrm{X}}_{\mathrm{i}\mathrm{t}}$ represents a vector of control variables. We include year-fixed effects ${\mathsf{\lambda }}_{\mathrm{t}}$ and county-fixed effects ${\mathsf{\mu }}_{\mathrm{i}}$ to account for time-specific shocks and time-invariant unobservable county characteristics. ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is a random disturbance term, and the standard errors are clustered at the county level in all regressions.

4.2. Sample Selection

First, considering the REP policy targeting Chinese rural areas and the particularity of municipalities, Beijing, Tianjin, Shanghai, and Chongqing are not included in the sample. Second, because of the difficulty of obtaining the per capita disposable income data of rural residents at the county level in Inner Mongolia, Liaoning, Heilongjiang, Hainan, Tibet, Shanxi, Qinghai, and Xinjiang, the counties of these eight provinces are also dropped from the sample. Third, due to the high level of urbanization among county-level administrative units, urban districts are excluded from our sample. Based on the above considerations, our sample finally consists of 1384 counties (including county-level cities) across nineteen provinces, which accounts for 73.58% of all 1881 counties and county-level cities (excluding municipal districts) in mainland China.

As the urban-rural integration household survey reform was implemented at the end of 2013, the indicator of the annual per capita income of rural residents began to be calculated as the annual per capita disposable income of rural residents in 2014. Before the reform, it was the annual per capita net income of rural residents, which was incomparable with that after the reform. Therefore, this paper selects 2014–2019 as the sample period, which, on the one hand, ensures the timeliness of data and the consistency of statistical caliber; on the other hand, covers the first and last three years of the implementation year of the policy (2017), meeting the data requirements of our empirical research.

4.3. Variables and Data

Dependent variable. Our Dependent variable is the income of rural residents (Rincome), which is defined as the annual per capita disposable income of a county’s rural residents, and the data comes from Provincial Statistical Yearbooks from the 19 provinces in our sample.

Control variables. Based on the existing literature, we collect a set of control variables (${\mathrm{X}}_{\mathrm{i}\mathrm{t}}$) that may affect rural residents’ income, including Pgdp, Pop, Indstru, Fiscex, and Infra. Pgdp is the gross domestic product per capita of a county. Pop is a county’s population. Indstru is the proportion of a county’s added value of the tertiary industry in its GDP. Fiscex is the ratio of a county’s fiscal expenditure to its GDP. Infra is the ratio of a county’s road lengths to its land area. The data of these control variables all come from China County Statistical Yearbook.

Other data. We also use other data for mechanism and heterogeneity analysis, including the business registry data from the State Administration for Industry and Commerce of China and information on county-level population from the China Population Sampling Survey 2015, which will be elaborated on later in this paper.

All variables are winsorized at the 1st and 99th percentiles to eliminate the impact of outliers.

Table 1. Descriptive statistics.

Variables	N	Mean	Sd	Min	Max
Rincome	8304	12.88	5.264	2.541	43.90
Treat	8304	0.243	0.429	0	1
Pgdp	8304	38.92	25.84	5.102	242.6
Pop	8304	48.89	32.31	1.050	212.1
Indstru	8304	0.408	0.102	0.0988	0.864
Fiscex	8304	0.297	0.259	0.00495	3.944
Infra	8304	292.3	114.2	71.79	533.4

reports the descriptive statistics of the variables. The mean value of $\mathrm{r}\mathrm{i}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}$ is ¥ 12.88 thousand, with a minimum of ¥ 25.41 thousand and a maximum of ¥ 43.9 thousand. The mean value of $\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}$ is 0.243, indicating that 24.3% of counties in the sample have established rural enterprise parks.

5. Empirical Results

5.1. Baseline Regression

The estimated results of the average effects of the rural enterprise parks policy on the income of rural residents are reported in

Table 2. Baseline regression.

	(1)	(2)
Dependent Variables	Rincome	Rincome
Treat × Post	0.183 **	0.194 **
	(0.092)	(0.082)
Pgdp		0.037 ***
		(0.006)
Pop		0.071 ***
		(0.017)
Indstru		1.060 **
		(0.414)
Fiscex		0.470 ***
		(0.176)
Infra		0.003 ***
		(0.001)
Constant	12.858 ***	6.477 ***
	(0.011)	(0.983)
County FE	YES	YES
Year FE	YES	YES
Observations	8304	8304
R-squared	0.976	0.978

Notes: Robust standard errors are clustered at the county level and are reported in parentheses. *** and ** represent significance levels of 1% and 5%, respectively.

. With only controlling for county and year fixed effects in column (1), the regression coefficient of Treat × Post is 0.183, which is significantly positive at the 5% level. Column (2) shows the estimation results with a set of control variables, and the regression coefficient of Treat × Post is still significantly positive at the 5% level. The baseline regression results suggest that the implementation of the REP policy can significantly increase the income of rural residents. Therefore, Hypothesis 1 is confirmed.

Our results are similar to other studies evaluating the impact of China’s place-based policies on the residents’ income. Using a municipal dataset and comparing the changes in the municipalities that created special economic zones with those that did not, Wang (2013) finds that the establishment of special economic zones significantly increases workers’ wages [16]. In addition, based on firm-level data, Lu et al. (2019) also conclude that establishing special economic zones has a positive effect on wages [15]. Unlike their studies targeting place-based policies in urban areas, this paper confirms the income-increasing effects of China’s rural enterprise park policy, which further complements this strand of the literature.

5.2. Robustness Checks

5.2.1. Parallel Trend Test

The validity of the above DID approach relies on the parallel trend assumption that treated counties and those in the control group have a common trend before the implementation of the policy shocks. We use the event study method to test the parallel trend assumption with the following specification:

$$\begin{array}{cc}\hfill {\mathrm{Rincome}}_{\mathrm{it}}& =\mathsf{\alpha }+{\mathsf{\beta }}_{-3}{\mathrm{Before}}_{\mathrm{it}}^{-3}+{\mathsf{\beta }}_{-2}{\mathrm{Before}}_{\mathrm{it}}^{-2}+{\mathsf{\beta }}_0{\mathrm{Current}}_{\mathrm{it}}^0\hfill \\ & + {\mathsf{\beta }}_1{\mathrm{After}}_{\mathrm{it}}^1+{\mathsf{\beta }}_2{\mathrm{After}}_{\mathrm{it}}^2+\mathsf{\gamma }{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}\hfill \end{array}$$

(2)

where ${\mathrm{Before}}_{\mathrm{it}}^{-\mathrm{k}}(\mathrm{k}= -3, -2)$ is a dummy variable that equals one if the observation is k years before the implementation year for a treated county and zero otherwise. ${\mathrm{Current}}_{\mathrm{it}}^0$ is a dummy variable that equals to one if the observation is in 2017 for a treated county and zero otherwise. ${\mathrm{After}}_{\mathrm{it}}^{\mathrm{k}}(\mathrm{k}= 1, 2)$ is a dummy variable that equals one if the observation is k years after the implementation year for a treated county and zero otherwise. We define the year before the implementation year as the benchmark year, that is, ${\mathsf{\beta }}_{-1}$ is normalized to zero. Hence, the coefficient ${\mathsf{\beta }}_{\mathrm{k}}(\mathrm{k}= -3, -2, 0, 1, 2)$ presents the policy effects relative to the period prior to the implementation year.

The estimated values of ${\mathsf{\beta }}_{\mathrm{k}}$ and the corresponding 95% confidence intervals from Equation (2) are shown in Figure 1. The estimated coefficients of the pre-policy indicators (${\mathsf{\beta }}_{-3}$, ${\mathsf{\beta }}_{-2}$) are not significant in statistics, which indicates that treated counties and those in the control groups have a common time trend before the implementation of the REP policy.

5.2.2. Placebo Test

To check whether the baseline regression results are driven by chance or some unobservable factors, we conduct a placebo test by randomly assigning policy treatment status to counties. The rationale of this test is that if the implementation of the REP policy indeed promotes the income of rural residents, the effect should only exist in real treatment counties. Specifically, we randomly choose the implementation year of the policy and randomly assign treatment and control counties by which we build false policy shocks. Then, we can estimate the effects of the false policies on the income of rural residents using Equation (1). We repeat the above procedure 500 times to ensure the effectiveness of this placebo test, and the distributions of estimated coefficients of the false policy shocks are shown in Figure 2. These coefficients roughly follow the zero-mean distribution, while the actual regression coefficients of the real policy (see column (2) in Table 2) are located in the entire distribution. That is, the false policy shocks have no significant effect on rural residents’ income, suggesting that our main findings of baseline regression are unlikely driven by chance.

5.2.3. PSM-DID

One concern is that treatment counties may not be comparable to control counties, resulting in self-selection biases. Hence, our baseline regression results could be driven by the differences in counties’ characteristics rather than the implementation of the rural enterprise park policy. We employ the PSM-DID method to address this concern. Selecting control variables in the baseline regression as covariates, a logit model is conducted to generate the propensity scores, by which we perform the one-to-one nearest-neighbor matching to match the treatment and control groups. Figure 3 reports the results of the propensity score matching balance test, and the covariates show an apparent reduction in the difference between the two groups, indicating that the balancing assumption is satisfied. We estimate Equation (1) using the matched samples, and column (1) in

Table 3. Robustness checks.

	(1)	(2)
Dependent Variables	Rincome	Rincome
Treat × Post	0.197 **
	(0.082)
Nrep × Post		0.079 **
		(0.031)
Pgdp	0.037 ***	0.037 ***
	(0.006)	(0.006)
Pop	0.080 ***	0.071 ***
	(0.018)	(0.018)
Indstru	1.052 **	1.083 ***
	(0.415)	(0.413)
Fiscex	0.578 ***	0.468 ***
	(0.219)	(0.177)
Infra	0.003 **	0.003 ***
	(0.001)	(0.001)
Constant	6.029 ***	6.480 ***
	(1.000)	(0.986)
County FE	YES	YES
Year FE	YES	YES
Observations	8277	8304
R-squared	0.979	0.978

Notes: Robust standard errors are clustered at the county level and are reported in parentheses. *** and ** represent significance levels of 1% and 5%, respectively.

reports the regression results. The regression coefficient of Treat × Post is significantly positive at the 5% level. Further excluding the problem of self-selection biases, the estimated results still support the conclusion from baseline regression.

5.2.4. The Impact of the Policy’s Intensity

In the baseline regression strategy, the treatment indicator (${\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}}_{\mathrm{i}}$) is a binary dummy variable, which only reflects whether a county is affected by the REP policy. Some counties, however, have established multiple enterprise parks, and the binary dummy variable can not reflect the intensity of the policy. Hence, drawing lessons from Qian (2008) and Chen et al. (2020) [32,33], we use the number of enterprise parks established in a county (${\mathrm{N}\mathrm{r}\mathrm{e}\mathrm{p}}_{\mathrm{i}}$) as a continuous measure of the policy intensity and replace the treatment indicator (${\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}}_{\mathrm{i}}$) in the baseline regression equation, Equation (1). The estimated results are reported in column (2) of Table 3. The regression coefficient of Nrep × Post is 0.079 and significant at the 5% level, re-confirming the conclusion from baseline regression that the REP policy positively affects rural incomes.

6. Mechanism and Heterogeneity

So far, we have presented empirical evidence suggesting that the implementation of the rural enterprise park policy has a positive effect on the income of rural residents. In this section, we explore the channel through which the establishment of rural enterprise parks affects the income of rural residents and the REP policy’s heterogeneous effects across counties.

6.1. Mechanism Analysis

Since the purpose of the rural enterprise park policy is to provide places and services to promote local entrepreneurial activities, which makes some residents become entrepreneurs on the one hand and creates jobs for more residents on the other hand, we argue that stimulating entrepreneurial activities would be the main channel through which the policy affects the income of rural residents.

We obtain the registration data of new market subjects including enterprises and individual businesses from the State Administration for Industry and Commerce of China and aggregate the entrepreneurial activity data at the county level. Then, we measure a county’s entrepreneurial activities by the number of new enterprises registered ($\mathrm{N}\mathrm{e}$) and the number of new individual businesses registered ($\mathrm{N}\mathrm{i}\mathrm{b}$). Using these two variables as the dependent variables in Equation (1), columns (1)–(2) in

Table 4. Mechanism analysis.

	(1)	(2)
Dependent Variables	Ne	Nib
Treat × Post	0.694 ***	0.580 ***
	(0.208)	(0.149)
Pgdp	0.052 **	0.031 **
	(0.025)	(0.013)
Pop	0.174 **	0.096 **
	(0.079)	(0.040)
Indstru	−1.080	−0.898
	(0.898)	(0.696)
Fiscex	0.703	0.231
	(0.559)	(0.307)
Infra	0.002	0.004 *
	(0.003)	(0.002)
Constant	−6.121	−3.120
	(3.873)	(2.074)
County FE	YES	YES
Year FE	YES	YES
Observations	8304	8304
R-squared	0.795	0.810

Notes: Robust standard errors are clustered at the county level and are reported in parentheses. ***, ** and * represent significance levels of 1%, 5%, and 10%, respectively.

report the policy effects on $\mathrm{N}\mathrm{e}$ and $\mathrm{N}\mathrm{i}\mathrm{b}$, respectively. Both the coefficients of Treat × Post on $\mathrm{N}\mathrm{e}$ and $\mathrm{N}\mathrm{i}\mathrm{b}$ are significantly positive at the 1% level, indicating that the REP policy indeed promotes entrepreneurial activities. Hence, Hypothesis 2 is verified.

Our conclusion is similar to other studies evaluating the impact of China’s place-based policies on entrepreneurship. Lu et al. (2019) and Tian and Xu (2022), using China’s industry business performance database and China’s new firm registration data, respectively, find that the special economic zone policy has a significantly positive impact on entrepreneurial activities [15,34]. By assessing the effect of China’s rural enterprise park policy, our paper contributes to this field of research.

6.2. Heterogeneity Analysis

The REP policy can promote the income of rural residents by boosting entrepreneurial activities, and entrepreneurial activities would be affected by the level of human capital, the demographic age structure, and the level of financial development according to existing research. Hence, the impact of the REP policy on the income of rural residents would be stronger and more significant for counties with a higher level of human capital, a younger demographic age structure, and a higher level of financial development. This part provides a discussion and additional estimation results to further examine the heterogeneous effects of the policy.

6.2.1. Heterogeneity Analysis Based on Human Capital

This paper uses the average years of education of a county’s population as an indicator to measure the level of the human capital of the county. The data comes from the China Population Sampling Survey 2015, which was conducted before the implementation of the policy. According to the median level of human capital in 2015, counties are divided into high and low levels. Then, we estimate Equation (1) using these two subsamples, respectively, and the results are reported in

Table 5. Heterogeneity analysis based on human capital.

	High	Low
	(1)	(2)
Dependent Variables	Rincome	Rincome
Treat × Post	0.223 *	0.132
	(0.120)	(0.100)
Pgdp	0.032 ***	0.048 ***
	(0.007)	(0.010)
Pop	0.115 ***	0.019
	(0.020)	(0.021)
Indstru	0.584	1.488 ***
	(0.597)	(0.540)
Fiscex	0.311	0.218
	(0.490)	(0.171)
Infra	0.006 ***	−0.001
	(0.002)	(0.001)
Constant	4.274 ***	8.693 ***
	(1.267)	(1.206)
County FE	YES	YES
Year FE	YES	YES
Observations	4134	4170
R-squared	0.980	0.977

Notes: Robust standard errors are clustered at the county level and are reported in parentheses. *** and * represent significance levels of 1% and 10%, respectively.

. The coefficient of Treat × Post is only statistically significant in counties with a high level of human capital, and its value is nearly twice that of counties with a low level of human capital. That is to say, the magnitude of the REP policy’s effects on rural residents’ income depends on the level of human capital, and the policy effects are mainly significant for counties with a high level of human capital.

6.2.2. Heterogeneity Analysis Based on Demographic Age Structure

We measure a county’s demographic age structure by the proportion of the population aged 15–64 in total, which is calculated based on the China Population Sampling Survey 2015. Then, we divide counties into high-proportion and low-proportion groups according to the median proportion of the population aged 15–64 and estimate Equation (1) using these two subsamples, respectively. The estimated results are reported in

Table 6. Heterogeneity analysis based on demographic age structure.

	High	Low
	(1)	(2)
Dependent Variables	Rincome	Rincome
Treat × Post	0.269 **	0.090
	(0.124)	(0.082)
Pgdp	0.039 ***	0.036 ***
	(0.007)	(0.009)
Pop	0.170 ***	−0.031 ***
	(0.018)	(0.011)
Indstru	0.894	1.384 ***
	(0.601)	(0.498)
Fiscex	1.224 **	0.013
	(0.479)	(0.158)
Infra	0.012 ***	0.002
	(0.002)	(0.001)
Constant	0.086	11.085 ***
	(1.191)	(0.840)
County FE	YES	YES
Year FE	YES	YES
Observations	4134	4170
R-squared	0.981	0.979

Notes: Robust standard errors are clustered at the county level and are reported in parentheses. *** and ** represent significance levels of 1% and 5%, respectively.

. The coefficient of Treat × Post is only statistically significant in counties with a high proportion of the population aged 15–64, and its value is almost three times that of counties with a low proportion of the population aged 15–64. These results suggest that the positive effects of the REP policy are more pronounced for counties with a younger demographic age structure.

6.2.3. Heterogeneity Analysis Based on Financial Development

We measure the level of financial development of a county by the ratio of loans to GDP, and the data come from the China County Statistical Yearbook 2016. Then, we divide counties into high-ratio and low-ratio groups according to the median ratio of loan to GDP and estimate Equation (1) using these two subsamples, respectively. The regression results are reported in

Table 7. Heterogeneity analysis based on financial development.

	High	Low
	(1)	(2)
Dependent Variables	Rincome	Rincome
Treat × Post	0.264 **	0.101
	(0.129)	(0.088)
Pgdp	0.075 ***	0.014 **
	(0.009)	(0.006)
Pop	0.129 ***	0.007
	(0.022)	(0.008)
Indstru	1.461 **	0.645
	(0.615)	(0.517)
Fiscex	1.321 ***	−0.193
	(0.300)	(0.213)
Infra	0.002	0.005 ***
	(0.002)	(0.001)
Constant	2.571 **	10.718 ***
	(1.206)	(0.790)
County FE	YES	YES
Year FE	YES	YES
Observations	4158	4146
R-squared	0.979	0.982

Notes: Robust standard errors are clustered at the county level and are reported in parentheses. *** and ** represent significance levels of 1% and5%, respectively.

. The coefficient of Treat × Post is only statistically significant in counties with a high level of financial development, and its value is more than twice that of counties with a low level of financial development. These results indicate that the positive effects of the REP policy on the income of rural residents are primarily significant for counties with a high level of financial development.

7. Conclusions

The income gap between urban and rural inhabitants has grown as industrialization has speeded up, and it is now one of the major issues facing developing nations worldwide and has become a significant barrier to high-quality development in China [1,35]. The key to narrowing this income gap may be to increase the income of rural residents because the relative poverty in rural areas is more serious than that in urban areas [2]. Hence, how to improve rural residents’ income would be of significant interest to China’s policymakers. This paper explores a prominent place-based policy in China, rural enterprise parks, which plans to establish 1096 enterprise parks in rural areas across the country in 2017. Employing a DID approach and the data of China’s counties from 2014 to 2019, we find that the policy has a positive effect on the income of rural residents. A set of additional tests suggest that the effects appear robust. Our mechanism analysis suggests that stimulating entrepreneurial activities serves as the mechanism through which the policy affects rural incomes. We further examine the heterogeneous effects of the policy and find that the policy effects on rural residents’ income appear more significant for counties with a higher level of human capital, younger demographic structure, and higher level of financial development.

Our findings provide some important policy implications. First, our results suggest that the rural enterprise park policy has a significantly positive impact on local entrepreneurial activities and thus promotes the income of rural residents, which confirms the effectiveness of the policy. Hence, policymakers can establish more enterprise parks in other rural areas to benefit residents. Second, given that the policy has a more significant income-increasing effect on counties with a higher level of human capital, younger demographic structure, and higher level of financial development, the subsequent establishment of the rural enterprise parks should be located more in areas with these characteristics to further enhance the efficiency of the policy implementation. Additionally, for areas without those characteristics, more efforts, such as providing high-quality training and financial services, should be made by local governments to ensure the effects of the policy.

The limitations of this paper mainly reside in the dataset. We use county-level data in this paper, which is considered a limited dataset, with data availability limited. Further research needs to explore the micro-level data, such as the household-level data, to more accurately identify the effects of the rural enterprise park policy. Since the policy has an income-increasing effect on rural residents, whether it narrows the urban-rural income gap becomes a meaningful question. We acknowledge the limitations of this paper and propose an idea for future research.