Revitalizing Agricultural Economy Through Rural E-Commerce? Experience from China’s Revolutionary Old Areas

Abstract

Many of the world’s less developed regions may not be able to improve the well-being of rural residents through agricultural revitalization because of their remoteness from agricultural markets. Using the county-level data set of China’s underdeveloped old revolutionary base areas from 2010 to 2021, this paper takes the policy planning of rural e-commerce as event intervention to investigate the driving role of the digital product market on agricultural economic development. Empirical results show that rural e-commerce planning policy has significantly promoted the agricultural added value of the pilot counties, and the digital market is the key driving factor of the agricultural economic growth in these underdeveloped areas. Both food production and livestock output have increased significantly as a result of e-commerce policies. Considering the potential bias of the bidirectional fixed effect estimators of staggered differences-in-differences (DID), this study uses heterogeneous robust estimators to verify the growth effect of the agricultural economy. Specifically, digital agricultural markets have significantly promoted agricultural mechanization and significantly improved agricultural total factor productivity. Moreover, empirical evidence does not support transmission mechanisms for off-farm employment and agricultural entrepreneurship. The findings can help less developed countries and regions develop policies to expand the agricultural markets with digital dividends, thereby promoting the development of the agricultural economy.

1. Introduction

Underdeveloped regions account for the majority of the global population, which is a huge challenge to progress towards the contemporary Sustainable Development Goals of eradicating poverty and hunger [1]. The less developed areas in China are mainly concentrated in the central and western regions, the northeast region, and some border areas. These regions have a relatively low level of economic development, uneven development of productive forces, and relatively backward technologies. Inhabitants in less developed regions rely on agriculture for income and well-being, and promoting agricultural economic development is a key tool for governments to fight poverty and hunger [2,3]. The economic, social, and environmental development of rural areas is an important aspect of achieving global sustainable development. However, agricultural economic growth suffers from a series of potential obstacles, including backward technology, low-quality labor, and decentralized management, which greatly limits the role of agricultural economy in improving the well-being of residents [4,5]. One of the hindering factors is the mismatch between the competitive mechanism of agricultural product market and the price coordination in the digital market. The information asymmetry of agricultural products leads to the contradiction between supply and demand [6]. Therefore, the markets of agricultural products are the prerequisite and key engine for promoting agricultural economic growth in underdeveloped areas.

The application of information technology in the agricultural field has provided the opportunity for agricultural economic growth, which not only promotes the improvement of production technology, but also helps to promote the development of the agricultural product market [7]. In particular, Internet technology has greatly narrowed the distance between producers and end consumers, and producers of agricultural products in underdeveloped areas can find suitable consumer markets and desirable prices [8]. Rural e-commerce is an effective channel for developing countries or less developed regions to expand product markets for agriculture [9]. Information asymmetry always occurs in underdeveloped areas due to factors such as transportation facilities, distance from cities, and scale of operation. Rural e-commerce can help farmers effectively link the final product market and avoid unsalable rural products with lower entry costs [10]. The old revolutionary base areas are mainly less developed or far away from large cities in China [11], and these areas have greatly relied on digital agricultural market means such as e-commerce to open agricultural markets and promote agricultural development [12].

The old revolutionary area is the living and military support areas established under the leadership of the Communist Party of China (CPC) during the Agrarian Revolutionary War (1927–1937) and the War against Japanese invasion (1937–1945). These areas cover more than 1300 county-level districts in 27 provinces (or autonomous regions or municipalities of the same level). During the war years, the old revolutionary base areas nourished the CPC and the people’s army under its leadership, provided the fighters, reserves, food, and funds needed to persevere in the long-term struggle, and contributed greatly to the strengthening of the revolutionary forces. These revolutionary areas overlap with hilly areas, inter-provincial adjacent areas and environmentally fragile areas, and are also economically underdeveloped and poverty-stricken areas in China [11]. Classical political economy provides an analytical pillar for the role of history and the necessity of social class analysis in the study of sustainable development. Due to the unique historical and political reasons, it is very important to promote the development of agricultural economy and sustainable development in China’s old revolutionary base areas [13]. The lack of endogenous driving force for economic growth and the poverty of residents are extremely serious problems, which pose challenges to China’s balanced regional development and common prosperity [14]. Geographic distance and infrastructure constraints have formed an island effect on the population in these poor areas, further restricting the development of revitalization industries [15]. With the embedding of information facilities and digital technologies in rural areas, the Internet has become the driving force for rural industrial development and market expansion [16], which provides the possibility for the digitization of agricultural markets to promote the poor regions to achieve leapfrog development.

The laws and characteristics of agricultural economic development are unique, and agricultural problems have always been an important field of economics and management. One of the typical characteristics of agricultural economy is information asymmetry and externality, which leads to the dependence of agricultural development on government subsidies and policy support [17]. Existing research has categorized the tools used in public policy into three main groups [18]. For China, economic and financial instruments are mainly used in public policies to support agricultural development in less developed regions by improving infrastructure and financing. There is a paradox in reality: agriculture is a vital means of economic take-off in less developed areas and improving the well-being of rural residents, but it depends on the support of scarce government financial funds [19,20]. Policy implementation needs to focus not only on the long-term goal of economic growth but also on the objectives of innovation policies, which must be aligned with the main causes of real challenges. A large number of studies have aimed to explore the driving factors and realization paths of agricultural development [21,22], and relevant research has found that the instability of the agricultural product market caused by information asymmetry may be the major challenge to agricultural economy [23]. In addition to the support of public policies, some studies have found that digital technologies such as the Internet are conducive to promoting stable agricultural prices and unhindered markets [10,24], which brings potential opportunities for agricultural revitalization in less developed regions.

To capture the benefits of the Internet for agricultural products, China has piloted a planning policy for the development of rural e-commerce in many counties since 2014, many of which are poor areas or old revolutionary base areas [25]. The purpose of this study is to investigate the impact of digitalization of the agricultural product market on agricultural economic growth and the mechanism of how it can influence economic growth, and to explore the special laws of digitalization of the agricultural product market in China’s old revolutionary base areas by using panel data and a staggered DID model. It is found that the digital agricultural product market significantly promotes the growth of agricultural economy in the old revolutionary base areas, and also promotes the grain output and the added value of animal husbandry. These findings support the positive effect of building agricultural markets in remote or underdeveloped areas on agricultural development. In addition, the study discusses the mechanisms by which rural e-commerce plays a role, which provides useful insights for promoting agricultural development in less developed regions.

This research aims to make the following contributions in theory and practice. Firstly, this study theoretically contributes to the insights of the digitalization realization mechanism of agricultural economy. It employs rural e-commerce as a method for the digitalization development of agricultural markets to explore the driving factors and implementation mechanisms of agricultural development in less developed areas. Secondly, this study adopts a quasi-natural experimental design with staggered DID regression to avoid the potential endogenous threat of identifying the causality between digital adoption and agricultural revitalization. In this study, robust estimators of Goodman–Bacon decomposition and heterogeneous processing effects were also used to verify the robustness of our findings. Thirdly, this study contributes to policy innovation and optimization in developing countries and less developed regions around the world. Technological and organizational innovation is an important reason for the accelerated economic prosperity [26]. It has been shown in the literature that the third framework of contemporary innovation policy focuses on environmental and social change issues and emphasizes the establishment of knowledge networks between producers and user organizations through experiments in pilot policies [27], and it is in this way that rural e-commerce policies thereby alleviate information asymmetry and the status quo of poverty in underdeveloped regions. This study takes China’s old revolutionary base areas as the research object, explores the realization mechanism of digitalization to promote agricultural development, and provides insights for realizing high-quality agricultural development in underdeveloped regions around the world.

The remaining research contents are arranged as follows. Section 2 presents the background and theoretical hypothesis of agricultural revitalization in China’s old revolutionary base areas. Section 3 introduces the study design, including samples, data, and empirical methods. Section 4 analyzes the empirical results, while Section 5 discusses the mechanisms. Section 6 is the conclusions and insights.

2. Background and Theoretical Analysis

2.1. Literature Review

Through the popularization of the Internet and the construction of digital platforms, digital technology has greatly promoted the development of the agricultural product market in developing countries [28] and become a key strategy to promote sustainable agricultural development [29]. The application of digital technologies such as the Internet of Things (IoT) and blockchain in agriculture continues to mature, resulting in a digital trend in the agricultural sector. The digital innovation of agricultural technology improves the market structure of agricultural products and drives the digital transformation of agriculture [30]. In addition, the digital development of agricultural markets cannot be achieved without the support of policies [30]. For developing countries in particular, promoting digitization of agriculture through subsidy incentives is a key factor in achieving high-quality transformational development.

Rural e-commerce is a business model rooted in information technology and Internet facilities and applied to rural industries. Through the use of digital intelligent technologies such as the IoT, rural e-commerce can improve agricultural production efficiency and sustainability [31], thus promoting agricultural economic growth. A study by Borrero and Mariscal (2022) concluded that online platforms will facilitate the development of markets for agricultural outputs, and that rural e-commerce can contribute to the effectiveness of communication in an online environment [32]. Rural e-commerce reduces the cost and threshold for farmers to do business, and also provides them with more market opportunities [7]. Digital market channels can improve the operational efficiency of agricultural management, supporting farmers to start their own businesses and develop characteristic industries [8]. The convenience of digital markets is a benefit which can help rural residents to carry out agricultural operations beyond geographical boundaries [33,34]. In addition, it also produces a factor flow effect and promotes the efficiency of resource allocation [35]. Rural e-commerce helps both the downward movement of urban industrial goods and the upward movement of rural commodities, thus reducing the cost of living and production [16,36]. The emergence and application of e-commerce are likely to benefit less developed regions in the long term due to the spillover effects of knowledge and technology from developed regions [37,38].

The significance of rural e-commerce policy is to promote agricultural economic development, rural revitalization, and farmers’ living standards. The implementation of the policy can not only optimize agricultural supply chain management, which increases the added value of agricultural products and promotes agricultural modernization, but also increase farmers’ income, promote employment and entrepreneurship, and enrich consumption choices, thereby improving quality of life and promoting cultural prosperity. Therefore, the rural e-commerce policy is an important starting point to promote the high-quality development of agriculture and rural areas and achieve rural revitalization.

2.2. Policy of E-Commerce Development in China

The evolution of China’s e-commerce has experienced four stages: germination, acceleration, standardization, and internationalization [39]. The Chinese government officially issued the national strategy for the development of e-commerce in 2012, which put forward the basic requirements of improving the basic system of e-commerce and enhancing the security and technical support capacity of e-commerce. Since then, Chinese governments at all levels and their relevant departments have introduced institutional arrangements and specific measures to promote the development of e-commerce. Agriculture is an important beneficiary of e-commerce policies [39]. In order to make e-commerce more deeply embedded in rural areas, China selected 56 counties in 2014 to pilot rural e-commerce policies; since then, more than 200 pilot counties have been added each year, and 1489 counties nationwide have been included in the comprehensive demonstration project of e-commerce in rural areas, which has covered 91% of China’s counties. The rural e-commerce policy has achieved important results in infrastructure construction, e-commerce platform cooperation, and new business model development. China Post continued to promote the project of express delivery to the countryside and, since 2019, has basically realized the access of villages and towns to mail, and the outlets of major express delivery brands now cover 98% of China’s towns and villages. The rural e-commerce policy encourages e-commerce platforms to actively participate in the development of rural e-commerce, combining online sales and offline distribution through the construction of e-commerce public service centers, logistics distribution centers, and village-level e-commerce centers to connect rural agricultural products with urban consumers. In addition, in order to promote the high-quality development of rural e-commerce, China’s counties have actively created demonstration projects and summarized and promoted good experiences and practices. Shuyang, Jiangsu Province, has become a nationally recognized flower and tree hometown through e-commerce. The rapid development of rural e-commerce in Lishui, Zhejiang Province, has given birth to the nationally renowned “Suichang Model”, “Beishan Model”, and “Lishui Experience”, which have provided a model for the development of rural e-commerce and the revitalization of rural areas. It has provided a model for the development of rural e-commerce and rural revitalization throughout the country. The implementation of China’s e-commerce strategy in rural areas has achieved remarkable results and not only promoted the development of the rural economy and increased farmers’ income, but also promoted the overall progress and development of rural society. The distribution of rural e-commerce policy pilot counties between 2010 and 2021 is shown in Figure 1.

2.3. Theoretical Analysis and Hypotheses

The fundamental industry for economic takeoff in underdeveloped areas is agriculture, and some classic theories have discussed the importance and driving mechanisms of agricultural economic development [40,41]. The theory of agricultural industrialization proposes that agricultural economic development can provide a material foundation for industrialization in developing countries and regions, as well as the main support for labor transfer and product demand for industrialization [40]. Although agriculture is crucial for the economic development of underdeveloped areas, it often faces challenges such as decentralized management, weak industrial resilience, low willingness to adopt technology, and low technological content. Therefore, public subsidies or price controls have become the mainstream theory in the agricultural field. Nevertheless, market factors are the core challenge for the lagging agricultural economic development in underdeveloped areas, and market mechanisms can promote agricultural economic development through direct and indirect effects. Figure 2 shows the theoretical framework of this study, namely the nexuses of product markets, e-commerce, and growth in the agricultural sector.

The agricultural economy in the revolutionary old areas of China typically faces challenges from the accessible product markets, and the obstruction of market mechanisms leads to slow development of the agricultural economy [42]. The planning policy of rural agricultural e-commerce is essentially to build a digital market for agricultural products, which can reduce information asymmetry in the agricultural product market and match the market supply and demand of agricultural commodities, thereby solving the problem of imperfect product markets that hinder the development of agricultural economy. Hence, this study proposes a hypothesis that rural e-commerce promotes the development of agricultural economy in China’s old revolutionary areas.

Hypothesis 1. 

The planning policy of rural e-commerce significantly promotes the development of agricultural economy in China’s old revolutionary base areas.

Rural e-commerce is the construction of a digital agricultural product market, which enhances the enthusiasm of agricultural production activities to increase input and efficiency, and directly promotes agricultural economic development by influencing agricultural production activities [38]. The digital market provides a vast consumer market for agricultural producers, who are more inclined to adopt agricultural mechanization technology to increase the quantity of agricultural product supply and achieve mechanized and large-scale operations [35]. Therefore, the planning policies of rural e-commerce can significantly promote the adoption of mechanical technology and thus promote the development of agricultural economy. Both mechanized and large-scale operations would bring about an improvement in input–output efficiency, thereby enhancing agricultural total factor productivity. In addition, the stability of the product market promotes producers to improve land quality, invest in agricultural production infrastructure, and adopt agricultural technologies, all of which would promote total factor productivity in the agricultural sector. Hence, this study proposes the following two direct effects of rural e-commerce on agricultural production.

Hypothesis 2a. 

Rural e-commerce promotes agricultural development through increasing the adoption of mechanized technologies.

Hypothesis 2b. 

Rural e-commerce generates productivity effects and promotes agricultural economic development.

Rural e-commerce may also indirectly affect the operation of the agricultural sector, thereby supporting its development. The digital agricultural product market generally generates more demand for agricultural processed food, leading to the extension of the value chain of subsidiary agricultural products. The development of the agricultural processing industry and rural e-commerce promotes the transfer of agricultural employment, and more workers are turning to the secondary and tertiary industries [43]. Non-agricultural employment may have significant impacts on the development of agricultural economy. In addition, rural e-commerce makes agricultural entrepreneurship activities more attractive, such as by increasing demand for agricultural product processing and specialized agricultural services. In addition, rural e-commerce brings knowledge spillovers and industry information, which may also stimulate the entrepreneurial enthusiasm of rural residents [44]. Non-agricultural employment and an agricultural entrepreneurial spirit may indirectly promote the development of agriculture. Therefore, the following two hypotheses have been proposed.

Hypothesis 3a. 

Rural e-commerce promotes non-agricultural employment and thus affects agricultural economic development.

Hypothesis 3b. 

Rural e-commerce stimulates the spirit of agricultural entrepreneurs and affects agricultural economic development.

3. Data Sources and Research Methods

3.1. Model Specification

As the rural e-commerce policy was implemented in 2014, while new pilot counties were added in batches year by year, the reform counties were subject to the policy impact at different points in time. This study utilizes a two-way fixed effects (TWFE) approach to construct a staggered DID model using DID_EC_it as the treatment variable to estimate the impact of rural e-commerce policies on the level of agricultural economic development, and the benchmark model is set as follows:

$${Agric\_Output}_{it}={\alpha }_0+{\beta }_1{DID\_EC}_{it}+{Controls}_{it}\gamma +{\mu }_i+{\eta }_t+{\epsilon }_{it}$$

(1)

where the policy is assumed to have persistent effects, i refers to the county and t represents the year, which is from 2010 to 2021. Agric_Outputit is the dependent variable, which means the level of agricultural economic development in the i-th county in year t. The coefficient β₁ before the core explanatory variable indicates the impact of the policy shock on the agricultural economy of the pilot counties. Controls_it denotes a set of control variables. μ_i and η_t denote county-level and year-level fixed effects, respectively.

3.2. Definitions and Descriptions for the Variables

The variable of interest of this study is agricultural economic development (Agric_Output), which is actually measured as the logarithm of the value added of the primary industry. In addition, it uses the index of agricultural production prices to deflate the nominal variable of agricultural value added. This study also uses grain production and added value of animal husbandry as proxy variables for agricultural economic development.

The core variable of interest is the planning policy of rural e-commerce development (DID_EC), which is actually an interactive term of DID design. In addition, a group dummy variable (Treated) is defined based on whether the county is in the list of rural e-commerce planning counties. If a county is on the list for the planning policy, the variable of Treated is equal to one; otherwise, it is zero. The value of the DID term is equal to one only for the Treated counties under intervention of the planning policy, and it is equal to zero in other cases.

This study also uses four mechanism variables to examine the impact of rural e-commerce on agricultural development, including agricultural total factor productivity (lnTFP), agricultural mechanization (Agric_Mech), non-agricultural employment (Non_Agric), and agricultural entrepreneurship (Agric_Enter). Agricultural mechanization is defined as the logarithm of the total power of agricultural machinery. Agricultural total factor productivity is calculated by using agricultural output and land, and machinery and labor input is calculated using the stochastic frontier model. Non-agricultural employment is the number of employees in the secondary industry and the tertiary industry accounts for the total number of employees. Agricultural entrepreneurship is measured by the number of new enterprises in agriculture, which reflects entrepreneurship in the agricultural sector.

A series of economic and social factors that may directly affect agricultural development are taken as control variables. The change in rural population quantity and structure has a great influence on the development of agricultural economy. Taking into account the impact of demographic factors, this study selected population size (lnPop) and the student dependency ratio (School_rate) as control variables, measured by the logarithm of the proportion of permanent population and primary school students in the total population, respectively [45]. Economic factors include economic development (lnRGDP), foreign direct investment (FDI), and industrial structure (Struct_Ser) measured by the logarithm of gross domestic product per capita, the logarithm of the amount of foreign direct investment, and the proportion of the size of the tertiary and secondary industries, respectively. The process of social modernization has caused a series of unsustainable problems in rural areas, which affect the development of rural economy [46]. This study selects urbanization (Urbanization) and digitization (Digitization) as social control variables, which are defined as the proportion of urban population in the total population and the logarithm of employment in the information technology service industry, respectively. Rural economic development cannot be separated from policy support; financial factors include fiscal dependence (Fiscal_Dep) and financial development (Finance_Develop). Fiscal dependence is measured by the ratio of fiscal expenditure to fiscal revenue, while financial development is measured by the proportion of the total balance of household deposits and loans to the gross domestic product. The descriptive statistics for the related variables are displayed in

Table 1. Descriptive characteristics of related variables.

Variables	Observations	Mean Value			Std. Dev.	Min	Max
		Full	Treated = 0	Treated = 1
Agric_Output	14,760	12.2237	12.3154	12.1408	0.8846	8.4645	15.0906
DID_EC	14,760	0.2178	0.0000	0.4147	0.4128	0.0000	1.0000
lnPop	14,760	3.8163	3.9023	3.7386	0.7102	0.6931	5.5910
School_rate	14,760	0.1175	0.1157	0.1191	0.0357	0.0111	0.3185
lnRGDP	14,760	10.4769	10.6725	10.3001	0.7180	8.2736	16.6860
FDI	14,760	0.0193	0.0152	0.0229	0.0276	0.0002	0.2606
Struct_Ser	14,760	1.1163	1.0298	1.1946	0.8467	0.1996	5.0214
Urbanization	14,760	0.2120	0.2276	0.1978	0.1491	0.0122	0.9310
Digitization	14,760	10.9503	11.0130	10.8937	0.6846	4.8520	16.0392
Fiscal_Dep	14,760	4.5407	3.3196	5.6446	3.9783	0.9250	24.3718
Finance_Develop	14,760	1.4643	1.4061	1.5169	0.6993	0.3254	3.9891
lnTFP	14,726	9.1361	9.2369	9.0452	0.6898	6.4336	12.1554
Agric_Mech	14,760	3.5851	3.6733	3.5053	0.8713	0.0000	5.8201
Agric_Enter	14,760	5.2913	5.2471	5.3313	1.0458	0.0000	9.2932
Non_Agric	14,760	0.5421	0.5810	0.5069	0.1820	0.0083	0.9923

.

Table 2. Analysis of multicollinearity between variables.

Variable	VIF	1/VIF
lnPop	2.95	0.33895
lnTFP	2.34	0.427106
lnRGDP	1.97	0.506358
Agric_Mech	1.84	0.543173
Digitization	1.71	0.584039
Fiscal_Dep	1.69	0.590797
Struct_Ser	1.61	0.622242
Agric_Enter	1.55	0.644294
Finance_Develop	1.48	0.67728
Non_Agric	1.44	0.694894
FDI	1.41	0.710706
DID_EC	1.27	0.790019
Urbanization	1.18	0.850402
School_rate	1.11	0.901129
Mean VIF	1.68

shows whether there is multicollinearity between the variables. The results show that the variance inflation factor (VIF) of each variable is less than 5, and the average VIF is less than 2, indicating that there is no multicollinearity between the variables.

4. Analysis and Discussion of Empirical Results

4.1. Analysis of Baseline Results

In this study, Stata software was used for data analysis (StataMP 17, https://www.stata.com/, accessed on 7 October 2024)

Table 3. Results of benchmark DID regression.

Variables	Traditional DID				Two-Way Fixed Effect DID
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
DID_EC	0.0593 ***	0.0392 ***	0.0426 ***	0.0292 **	0.0343 ***	0.0366 ***	0.0303 ***	0.0368 ***
	(0.0172)	(0.0146)	(0.0157)	(0.0139)	(0.0115)	(0.0104)	(0.0113)	(0.0103)
Treated	−0.0486 ***	−0.0367 ***	0.0334 ***	0.0059
	(0.0118)	(0.0101)	(0.0110)	(0.0097)
lnPop	0.8740 ***	0.8123 ***	0.8652 ***	0.8298 ***	−0.0470	0.0294	0.1227 *	0.0574
	(0.0093)	(0.0107)	(0.0099)	(0.0104)	(0.0477)	(0.0554)	(0.0638)	(0.0577)
School_rate	−2.1601 ***	−0.8336 ***	−2.4110 ***	−1.2115 ***	−1.2538 ***	−0.4073 *	−1.0831 ***	−0.2709
	(0.1896)	(0.2243)	(0.1652)	(0.1995)	(0.2433)	(0.2408)	(0.2300)	(0.2356)
Controls	No	No	Yes	Yes	No	No	Yes	Yes
County FE	No	No	No	No	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Pro FE × Year FE	No	Yes	No	Yes	No	Yes	No	Yes
City FE	No	No	Yes	Yes	No	No	No	No
R-Squared	0.5071	0.6685	0.5974	0.7094	0.3437	0.5443	0.3921	0.5581
Observations	14,760	14,760	14,760	14,760	14,760	14,760	14,760	14,760

Notes: The values in brackets are standard errors, and significance is indicated by asterisks, with *** (1%), ** (5%), and * (10%). Columns (5) to (8) cluster errors at the county level.

shows the results of the benchmark DID regression. In particular, columns (1)–(4) control for group dummy variables without county fixed effects, but control for time fixed effects. Columns (5) to (8) conduct the TWFE regression for DID design. Since the development level of China’s provinces is quite different due to regional reasons, columns (2), (4), (6), and (8) fixed the interaction terms between provinces and years to further avoid endogeneity problems. Only two variables, population size and student dependency ratio, are controlled in columns (1), (2), (5), and (6), which are exogenous relative to the planning policy of rural e-commerce. Control variables are added to the remaining columns.

The empirical results consistently support that planning policies for agricultural e-commerce significantly contribute to regional agricultural economic growth and that markets for agricultural products are crucial factors for agricultural development. The results show that the DID_EC coefficients in columns (1)–(8) are significantly positive at the 1% level. It proves that rural e-commerce significantly contributes to value addition in the primary sector and rural e-commerce plays a positive role in agricultural economic growth. The empirical results also have economic implications; the coefficient in column (7) is 0.0303, indicating that the implementation of rural e-commerce policy has increased the level of agricultural economic development by 3.03%. Table 3 actually shows the comprehensive effect of rural e-commerce planning policies on rural economic development in China. The results of DID regression consistently show that Hypothesis 1 is valid.

After adding a series of control variables, the coefficient of DID_EC remains significant and positive at the one percent level, and the coefficient values are not much different from the corresponding values in columns (1) to (4), proving once again that rural e-commerce promotes agricultural economic growth. As for the control variables, the coefficients of FDI, Urbanization, and Finance_Develop are significantly negative. This indicates that foreign direct investment plays a negative role in the prosperity of the agricultural sector, probably due to the existence of the overall attraction of investment in the process of attracting investment is small, the availability of funds is low, the advantages of attracting investment do not exist, and the investment environment is not perfect enough and other problems. The increase in urbanization rate and financial development promotes the trend of farmers moving to cities, which has a negative effect on the development of agricultural economy. The coefficient of economic growth rate is significant and positive, implying that the faster the economic growth, the better the agricultural economic development of the region, in line with reality.

4.2. Empirical Results of Parallel Trend Test

The assumption of parallel trends is a prerequisite for the design of the DID method, which means that there is no systematic difference in the level of agricultural economic development between these two different groups of counties prior to the implementation of the policy. The parallel trend test is a prerequisite for the use of the DID model, meaning that there is no trend difference between the two groups before the event intervention. For this reason, this paper investigates whether the parallel trend condition is satisfied, and the model is set up as follows.

$${Agric\_Output}_{it}={\alpha }_0+{\sum }_{k=-6}^{-1}{{\beta }_{k}before}_{i,t+k}+{\sum }_{k=1}^6{{\beta }_{k}after}_{i,t+k}+{Controls}_{it}\gamma +{\mu }_i+{\eta }_t+{\epsilon }_{it}$$

(2)

where the core explanatory variables are before_it and after_it, which are dummy variables indicating the year from the implementation of the rural e-commerce policy. If the pilot area is in year t before the policy implementation, before_it takes the value of 1, and otherwise, it is 0. The value of after_it is similarly known. Based on the time of policy release and the lag of data collection, this study took the year when the comprehensive demonstration policy of e-commerce entering rural areas was promulgated as the dividing line, and chose 6 years before and after the implementation of the policy as the leads and lags, respectively, to conduct parallel trend tests for the group without the control variables and the group with the control variables. As can be seen from Figure 3, the confidence intervals of the estimated coefficients before the policy pilot are intersected with the horizontal axis, and the trend of change is flat. It indicates that there is no systematic difference in the level of agricultural economic index between these two types of counties before the interference of rural e-commerce policy. After the implementation of the policy, the index of agricultural economic development of pilot counties is significantly higher than that of non-pilot counties, which is consistent with the core hypothesis that digital market drives agricultural prosperity.

4.3. Empirical Results of Placebo Test

A placebo test was further constructed to exclude the effect of other randomization factors. Among the 1230 districts in the research sample of this paper, 100 districts are randomly selected and set as the experimental group, and the other districts are set as the control group. Assuming that the year of e-commerce policy implementation is 2013, 1000 regressions are conducted on the basis of the double difference model. As shown in Figure 4, the coefficients generated by the spurious policy shocks are centrally distributed around 0, and the p-values are mostly distributed above 0.1, which is in line with the expectation of the placebo test and further proves the robustness of the model.

4.4. Results of Alternative Explanatory Variables

In order to avoid the influence of the choice of explanatory variables on the results of the benchmark regression and to further ensure the robustness of the regression results, this paper refers to the relevant literature to regress grain production and livestock value added as the explanatory variables. As the results show in

Table 4. Empirical results using proxy variables.

Variables	Alternative Variables: Grain Output			Alternative Variables: Value Added of Graziery
	(1)	(2)	(3)	(4)	(5)	(6)
DID_EC	0.0297 ***	0.0044	0.0249 **	0.0644 ***	0.0389 **	0.0644 ***
	(0.0105)	(0.0097)	(0.0103)	(0.0167)	(0.0154)	(0.0166)
lnPop	−0.1041 *	−0.0463	−0.1040	−0.2308 ***	−0.1071	−0.2749 ***
	(0.0600)	(0.0695)	(0.0677)	(0.0761)	(0.0917)	(0.0901)
County FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Pro FE × Year FE	No	Yes	No	No	Yes	No
Controls	No	No	Yes	No	No	Yes
R-Squared	0.0238	0.2285	0.0365	0.4138	0.5214	0.4311
Observations	14,760	14,760	14,760	14,760	14,760	14,760

Notes: The values in brackets are standard errors, and significance is indicated by asterisks, with *** (1%), ** (5%), and * (10%). Each column clusters errors at the county level.

, the estimated coefficient of DID_EC is still significantly positive at the 1% level, and the rural e-commerce policy significantly contributes to the growth of the agricultural economy in the old revolutionary areas, with the scale of grain production and the value added of livestock significantly increasing. It indicates that the regression results are still relatively robust after replacing the measurement of the explanatory variables.

4.5. Empirical Results of Goodman–Bacon Decomposition

Time-varying DID estimates obtained using ordinary least square (OLS) may be biased when there are multiple policy shock time points. In this paper, we refer to the existing literature and use the robust estimation method of Goodman–Bacon decomposition to verify the robustness of the research results. Figure 5 shows the bias due to different shock time points with time-varying effects for rural e-commerce policies. This difference arises from the comparison of Earlier Treatment with Later Treatment. Because the implementation year of the policy was 2014, there were no “always-treated” counties in this study. The estimates for Later Treatment and Earlier Treatment are positive (0.0002), as are those for Treatment and Never Treated (0.048). However, the estimate for Treatment and Later Treatment is negative, which explains the overall bias in the estimates, but accounts for a smaller proportion of the more stable regression results.

4.6. Empirical Results of CSDID Estimator

To alleviate the problem of possible bias in the TWFE estimation of DID treatment effects and to ensure robustness of the intervention effect of temporal heterogeneity, the DID regressions proposed by Callaway and Sant’Anna (2021) (CSDID) are conducted based on the relevant literature [47].

Table 5. Empirical results using CSDID estimator.

Variables	Comparison: Never Treated Counties		Comparison: Not Yet Treated Counties
	(1)	(2)	(3)	(4)
DID_EC	0.0208 *	0.0407 ***	0.0193 *	0.0363 ***
	(0.0114)	(0.0107)	(0.0109)	(0.0101)
County FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Controls	No	Yes	No	Yes
R-Squared	0.0238	0.2285	0.0365	0.4138
Observations	14,760	14,760	14,760	14,760

Notes: The values in brackets are standard errors, and significance is indicated by asterisks, with *** (1%), * (10%).

demonstrates the regression results, with columns (1)–(2) using Never Treated counties as the control group and columns (3)–(4) using Not Yet Treated counties as the contrast counties. In this case, columns (1) and (3) regression results without control variables, while columns (2) and (4) include covariates for the regression. In addition, this study uses the idea of the event study method to distinguish different groups according to the length of the shock period to estimate the dynamic effect of the event shock. The regression results are shown in

Table 6. Dynamic effect analysis of CSDID.

Variables	Comparison: Never Treated Counties		Comparison: Not Yet Treated Counties
	(1)	(2)	(3)	(4)
Pre_avg	0.00942 ***	0.00793 **	0.00732 ***	0.00620 **
	(0.00304)	(0.00332)	(0.00244)	(0.00245)
Post_avg	0.0282 *	0.0491 ***	0.0270	0.0460 ***
	(0.0170)	(0.0163)	(0.0166)	(0.0159)
Tm9	−0.000751	−0.00436	−0.000751	−0.00436
	(0.0118)	(0.0118)	(0.0118)	(0.0118)
Tm8	0.0187 *	0.0156	0.0185 *	0.0156 *
	(0.00981)	(0.00951)	(0.00969)	(0.00940)
Tm7	0.0110 **	0.00519	0.00928 *	0.00340
	(0.00537)	(0.00617)	(0.00533)	(0.00588)
Tm6	0.00787 *	0.00136	0.00562	0.00112
	(0.00430)	(0.00739)	(0.00403)	(0.00536)
Tm5	0.0225 ***	0.0206 ***	0.0203 ***	0.0185 ***
	(0.00436)	(0.00520)	(0.00420)	(0.00463)
Tm4	0.0133 ***	0.0108 **	0.00904 **	0.00686 *
	(0.00390)	(0.00442)	(0.00362)	(0.00381)
Tm3	0.000287	0.000103	−0.00410	−0.00255
	(0.00448)	(0.00481)	(0.00428)	(0.00456)
Tm2	0.00874 **	0.0128 ***	0.00716 *	0.0118 ***
	(0.00414)	(0.00418)	(0.00395)	(0.00373)
Tm1	0.00317	0.00924 **	0.000857	0.00530
	(0.00442)	(0.00420)	(0.00415)	(0.00397)
Tp0	0.00668	0.0154 ***	0.00681	0.0109 **
	(0.00509)	(0.00526)	(0.00479)	(0.00473)
Tp1	0.0113	0.0255 ***	0.00973	0.0193 ***
	(0.00765)	(0.00740)	(0.00705)	(0.00678)
Tp2	0.0209 *	0.0422 ***	0.0186 *	0.0359 ***
	(0.0114)	(0.0110)	(0.0107)	(0.0104)
Tp3	0.0325 **	0.0625 ***	0.0293 *	0.0574 ***
	(0.0159)	(0.0152)	(0.0153)	(0.0146)
Tp4	0.0284	0.0585 ***	0.0266	0.0564 ***
	(0.0186)	(0.0176)	(0.0181)	(0.0172)
Tp5	0.0328	0.0603 ***	0.0319	0.0598 ***
	(0.0235)	(0.0227)	(0.0233)	(0.0226)
Tp6	0.0406	0.0656 **	0.0406	0.0656 **
	(0.0290)	(0.0291)	(0.0290)	(0.0291)
Tp7	0.0525	0.0626	0.0525	0.0626
	(0.0564)	(0.0565)	(0.0564)	(0.0565)
County FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Controls	No	Yes	No	Yes
Observations	14,760	14,760	14,760	14,760

Notes: The values in brackets are standard errors, and significance is indicated by asterisks, with *** (1%), ** (5%), and * (10%).

. The coefficients of the empirical results are all positive and relatively close to those of the benchmark regression results, and consistently the planning policies supporting agricultural e-commerce significantly contribute to the regional agricultural economic growth, suggesting that the problem of bias in the TWFE estimation is not serious, and further support the core hypothesis that the digital market drives the revitalization of agricultural economy.

5. Discussion on the Mechanisms of E-Commerce Affecting Agricultural Economy

In order to further explore the mechanism of rural e-commerce policies on agricultural economic development, this paper constructs the following regression model using county-level data from 2010 to 2021 in China’s less developed old revolutionary base areas.

$${Agric\_Output}_{it}={\beta }_0+{\beta }_1{DID\_EC}_{it}+{Controls}_{it}\gamma +{\mu }_i+{\eta }_t+{\epsilon }_{it},$$

(3)

$${MED}_{it}={\gamma }_0+{\gamma }_1{DID\_EC}_{it}+{Controls}_{it}\gamma +{\mu }_i+{\eta }_t+{\epsilon }_{it},$$

(4)

$${Agric\_Output}_{it}={\delta }_0+{\delta }_1{DID\_EC}_{it}+{{\delta }_{2}MED}_{it}+{Controls}_{it}\gamma +{\mu }_i+{\eta }_t+{\epsilon }_{it}$$

(5)

where MED_it is the mediating variable, and the meaning of the rest of the variables is consistent with the Section 4 will not be repeated. Model (3) mainly reflects the impact of rural e-commerce on the mediating variable, and Model (4) mainly examines the impact of the mediating variable on agricultural economic development. The significance of the variables and the size of the comparative coefficients can determine whether the mediating effect is played.

In order to test Hypotheses 2 and 3, this paper examines the mechanism of action affecting agricultural economic development in terms of the mechanization effect, productivity effect, and non-agricultural employment and entrepreneurship mechanisms. The following mediating variables are constructed: agricultural total factor productivity (lnTFP), agricultural mechanization (Agric_Mech), non-agricultural employment (Non_Agric), and agricultural entrepreneurship (Agric_Enter).

5.1. Empirical Results of Mechanization Effects and Productivity Effects

Table 7. Empirical results on productivity effects and mechanization effects.

Variables	Mechanism: Productivity Effect			Mechanism: Mechanization Effect
	(1) lnTFP	(2) lnTFP	(3) Agric_Output	(4) Agric_Mech	(5) Agric_Mech	(6) Agric_Output
DID_EC	0.0441 ***	0.0220 **	0.0160	0.0384 ***	0.0361 ***	0.0286 ***
	(0.0082)	(0.0099)	(0.0108)	(0.0067)	(0.0136)	(0.0051)
lnTFP			0.3622 ***
			(0.0294)
Agric_Mech						0.0294 ***
						(0.0066)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
R-Squared	0.7173	-	0.4524	0.0898	-	0.3866
Observations	14,726	14,726	14,726	14,760	14,760	14,760

Notes: Columns (1) and (3) cluster errors at the county level. Columns (2) and (5) use the robustness estimator for heterogeneous intervention effects proposed by Callaway and Sant ‘Anna [47], while the other columns use the TWFE estimator with robustness standard error. The values in brackets are standard errors, and significance is indicated by asterisks, with *** (1%), ** (5%).

illustrates the results of the productivity effect as well as the mechanization effect. Columns (1) and (2) present that the estimated coefficients of DID_EC are significantly positive at the 1% and 5% levels, respectively, indicating that rural e-commerce policies promote agricultural total factor productivity. Columns (4) and (5) show that the estimated coefficients of DID_EC are significant and positive at the 1% level, indicating that rural e-commerce policies significantly promote agricultural mechanization. Columns (3) and (6) show that the estimated coefficients of lnTFP and Agric_Mech are significantly positive. Meanwhile, the coefficients of DID_EC are all smaller than the coefficient size of the baseline result (0.0303), indicating that the digital development of agricultural markets significantly promotes agricultural mechanization and agricultural total factor productivity, which in turn promotes the growth of the agricultural economy in the old revolutionary base areas. The results of different tables consistently prove that Hypothesis 2a and Hypothesis 2b hold.

5.2. The Role of Agricultural Entrepreneurship and Non-Farm Employment

Table 8. Empirical results on agricultural enterprises and non-agricultural employment.

Variables	Mechanism: Agricultural Enterprise			Mechanism: Non-Agricultural Employment
	(1) Agric_Enter	(2) Agric_Enter	(3) Agric_Output	(4) Non_Agric	(5) Non_Agric	(6) Agric_Output
DID_EC	0.1025 ***	−0.0292	0.0301 ***	0.0038	−0.0046	0.0298 ***
	(0.0245)	0.0293	(0.0113)	(0.0023)	(0.0053)	(0.0051)
Agric_Enter			0.0020
			(0.0050)
Non_Agric						−0.0236
						(0.0191)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
R-Squared	0.4481	-	0.3921	0.1511	-	0.3857
Observations	14,760	14,760	14,760	14,760	14,760	14,760

Notes: Columns (1) and (3) cluster errors at county level. Columns (2) and (5) use the robustness estimator for heterogeneous intervention effects proposed by Callaway and Sant’Anna [47], while the other columns use the TWFE estimator with robustness standard error. The values in brackets are standard errors, and significance is indicated by asterisks, with *** (1%).

shows the results for agricultural entrepreneurship and non-agricultural employment. Column (1) shows that the estimated coefficient of DID_EC is significantly positive at the 1 percent level, but the estimated coefficient of DID_EC in column (4) is not significant, implying that rural e-commerce policy significantly promotes agricultural entrepreneurship, and positively but insignificantly affects non-agricultural employment. Columns (3) and (6) show that the estimated coefficients of Agric_Enter and Non_Agric are not significant, implying that the mechanism of agricultural entrepreneurship and non-farm employment on the development of the agricultural economy has not been proved. Empirical results do not support Hypothesis 3a and Hypothesis 3b.

6. Conclusions and Enlightenments

The digital revolution represented by e-commerce and the Internet has facilitated the market for connected products in remote areas, which has helped underdeveloped areas improve residents’ well-being through the characteristic industries [48]. This study examines how institutional arrangements to promote rural e-commerce in China affect agricultural development in old revolutionary base areas, and examines the mechanization effect, productivity effect, and mechanisms of non-agricultural employment and entrepreneurship. Through robust estimators of the overlapping DID design and its heterogeneous treatment effects, the following important findings were made in this study.

It is found that the policy intervention of rural e-commerce has significantly promoted the growth of agricultural economy in the old revolutionary base areas, and the scale of grain output and added value of animal husbandry have been significantly improved. Since the planning of rural e-commerce is arranged at different time nodes, the Goodman–Bacon decomposition and robust estimators of heterogeneous treatment effects are also adopted, but the conclusions obtained are consistent with the TWFE estimators. This theoretically contributes to the hypothesis of the market theory of agricultural economic development, that the market for agricultural products is a crucial driver of agricultural economy. In addition, the digital development of the agricultural product market significantly promoted agricultural mechanization and agricultural total factor productivity, and thus promoted agricultural economic growth in the old revolutionary base areas. Rural e-commerce significantly promoted agricultural entrepreneurship, but the impact on non-agricultural employment was positive but not significant. The intermediary effect model has not proved the mechanism of agricultural entrepreneurship and non-agricultural employment. It implies that the dividends of the construction of a digital agricultural product market can be further released, especially in agricultural entrepreneurship and large-scale management.

These findings have implications for less developed countries and regions including the old revolutionary base areas of China, which should take advantage of the dividends of digital transformation in the field of agricultural economy to drive the revitalization of the agricultural sector and improve the well-being of rural residents through market forces [49]. Further research should be based on the geographical and political characteristics of China’s old revolutionary base areas, and explore the promotion methods of agricultural market digitization in accordance with local conditions in the old revolutionary base areas. This study makes the following policy recommendations. Firstly, the government needs to formulate policies to construct the market mechanism of agricultural commodities enabled by digital technology, resolve the information asymmetry hindrance of commercial commodity trade, and provide a market guarantee mechanism for agricultural development [50]. Secondly, less developed regions need to encourage the application of e-commerce and the Internet in rural agriculture through institutional arrangements, improve total factor productivity, and significantly release the digital dividend in the agricultural sector. Moreover, the government needs supply-side structural reform to encourage and support large-scale agricultural operations in the era of digital economy [51], improve the level of agricultural mechanization, and promote the revitalization and prosperity of agriculture. Finally, the government should further develop the role of agricultural entrepreneurship in the high-quality development of agriculture, and encourage more residents to innovate and start businesses in the agricultural field through digital agricultural product markets.