Agricultural Production Services Expand Major-Grain Output Through the Sown-Area Margin: Evidence from China

1. Introduction

As the world’s largest developing country, China has experienced rapid economic growth, whereas the growth of its agricultural productivity has remained relatively modest. The Household Responsibility System addressed the weak incentives and difficulties in labor supervision associated with collective farming and substantially improved agricultural productivity in China [1,2]. However, the egalitarian allocation of farmland resulted in highly fragmented landholdings, making small-scale and dispersed farming the dominant form of agricultural production and constraining the development of larger-scale operations [3,4]. Meanwhile, rapid urbanization has accelerated the migration of young and middle-aged workers from rural areas. Nevertheless, institutional barriers associated with the household registration system have hindered their permanent urban settlement. Consequently, the agricultural workforce has become increasingly older and more engaged in part-time farming. These conditions have weakened farmers’ incentives to consolidate land and impeded land transfers. Together with the institutional legacy of the Household Responsibility System, these factors have restricted the expansion of operational farm size and the further improvement of agricultural productivity [2,5,6,7,8]. According to the Communiqué on Major Data of the Third National Agricultural Census, released by the National Bureau of Statistics of China, China had 207.4 million agricultural households at the end of 2016, of which 203.4 million were non-scale agricultural operators. In 2022, households operating less than 10 mu of farmland still accounted for 85.33% of all agricultural households. Small-scale and fragmented family farming is therefore expected to remain a dominant feature of Chinese agriculture for a considerable period [9].

Against this background, facilitating the continued transfer of rural labor to non-agricultural sectors while improving agricultural productivity under fragmented landholdings has become a central challenge in China’s agricultural modernization. Further rural labor outmigration, however, can leave farmland without sufficient labor, increase land abandonment, and threaten food security, thereby imposing substantial economic costs on developing countries, particularly China. Using a provincial-level agricultural dataset and a difference-in-differences (DID) strategy, this study evaluates China’s Agricultural Production Service Pilot Program, which was introduced in 2013 and expanded in 2016. The results show that the program reduced the dependence of grain production on local agricultural labor. Declining agricultural labor demand, an expansion of the sown area of the three major grain crops and an increase in their aggregate output occurred simultaneously following policy implementation. These findings remain robust across a series of specification and identification tests.

This finding is important for two reasons. First, it demonstrates the possibility of reconciling rural labor outmigration with agricultural production growth under constraints on operational farm size, an issue relevant to many developing countries. Second, subsidizing farmers’ access to agricultural production services is less capital-intensive for individual farmers than directly subsidizing their purchases of agricultural machinery. A policy framework combining support for service procurement with machinery-purchase subsidies can therefore generate greater agricultural production gains under a given fiscal budget than a strategy centered exclusively on machinery ownership.

Using crop- and operation-specific data on mechanized plowing, sowing, transplanting, and harvesting, we find that the pilot program significantly expanded the mechanically sown area of wheat, the mechanically transplanted and harvested area of rice, and the mechanically sown and harvested area of maize in the pilot regions. These results indicate that agricultural production services do not merely lead to a general increase in machinery input. Instead, they alleviate seasonal labor bottlenecks concentrated in the sowing and harvesting stages of grain production.

By relaxing seasonal labor bottlenecks during sowing and harvesting, agricultural production services shorten farming cycles, ease constraints associated with critical farming seasons, and improve the temporal utilization of farmland, thereby increasing the multiple-cropping index for grain. The wider adoption of agricultural machinery can also encourage farmers to cultivate crops that are more compatible with mechanized operations. Rice, wheat, and maize are field crops with a relatively high degree of suitability for mechanized production. The expansion of mechanized services can therefore induce a grain-oriented adjustment in cropping patterns. Consistent with the existing agricultural economics literature, an improved cropping structure can contribute to higher total factor productivity [10,11].

Although these findings demonstrate that agricultural productivity can be improved despite constraints on operational farm size, much of the existing literature on China continues to emphasize land transfers and farm consolidation as the principal routes to realizing economies of scale. the estimated effect on major-grain output, but that the effectiveness of agricultural production services did not depend on the prior establishment of large-scale land operations. We further examine the production-function channels through which the policy affected agricultural production. Following policy implementation, agricultural labor input declined, cultivated land use expanded, and agricultural total factor productivity (TFP) increased. The program-supported expansion of agricultural machinery use is labor-saving and directly addresses labor shortages in labor-intensive stages of field-crop production, particularly sowing and harvesting. It thereby compensates for labor losses and input constraints resulting from non-agricultural employment, improves production efficiency, reduces land abandonment, and supports machinery-driven, area-based growth in grain production.

Overall, this study contributes to two strands of literature. First, agricultural mechanization has received relatively limited attention in studies of developing countries, partly because of their limited endowments of agricultural machinery [12,13]. Existing research on the effects of mechanization has focused primarily on China and has examined land productivity [14,15,16,17,18], cultivated-land use [19], and rural labor mobility [19]. Other studies have investigated the relationship between mechanization and smallholders and found that the benefits of mechanization do not necessarily extend effectively to small farms [20]. The present study extends this literature by showing that the Agricultural Production Service Pilot Program separates machinery ownership from machinery use and aggregates the operational demand of dispersed smallholders into a large-scale service market without requiring the concentration of land ownership or contractual management rights. This arrangement reduces the cost of accessing mechanization and alleviates constraints associated with critical farming seasons. It also facilitates further rural labor outmigration, reduces labor input per unit of land, and increases the amount of land that can be operated by each agricultural worker. By reducing land idleness, increasing multiple-cropping intensity, and expanding the grain-sown area, agricultural production services generate machinery-driven, area-based growth in grain production and improve agricultural productivity. To determine whether agricultural production services can further improve efficiency under constraints on land-based economies of scale, we estimate TFP using alternative production-function specifications and assumptions regarding returns to scale. This analysis examines whether area-based grain-production growth driven by agricultural production services is accompanied by improvements in overall production efficiency. We also investigate the relationship between agricultural production services and land transfers to address the policy debate over whether the two approaches can develop jointly.

Second, our findings contribute to the broader literature on barriers to agricultural productivity growth. Existing studies have proposed numerous strategies for increasing agricultural productivity, including the adoption of high-yielding seed varieties [21,22], subsidies for modern agricultural inputs [23], farmer training and agricultural extension [24], the construction of large irrigation dams [25], investment in agricultural infrastructure [26], and the introduction of index-based agricultural insurance [27]. This study complements that literature by showing that local governments can increase agricultural productivity by providing targeted financial support to rural households and collective organizations for purchasing agricultural production services. Such support expands access to mechanized operations, promotes more intensive use of cultivated land, and improves the allocation of agricultural production factors.

The remainder of this paper is organized as follows. Section 2 describes the development of China’s agricultural production service system and the institutional background of the pilot program. Section 3 develops the theoretical framework and presents the research hypotheses. Section 4 introduces the empirical identification strategy, variable definitions, and data sources. Section 5 reports the baseline estimates and robustness tests. Section 6 examines the underlying mechanisms, focusing on labor-saving mechanization, land-use intensity, and changes in land-use patterns. Section 7 further investigates the relationships among agricultural production services, land transfers, and total factor productivity. Section 8 discusses the contributions and limitations of the study. Section 9 concludes and presents the policy implications.

2. Development of China’s Agricultural Production Service System and the Pilot Program

Following the launch of market-oriented rural reforms in the 1980s, China began developing an agricultural socialized service system to enhance agricultural productive capacity and reform the agricultural service sector. The meaning and institutional scope of this system have since been progressively refined. The concept of “socialized services” first appeared in the No. 1 Central Document issued by the State Council in 1983. In 1990, the Central Committee of the Communist Party of China and the State Council formally introduced the term “agricultural socialized service system” in the Notice on Agricultural and Rural Work for 1991. In 1991, the State Council issued the Notice on Strengthening the Development of the Agricultural Socialized Service System, proposing the establishment of a system “based on village collectives or cooperative economic organizations, supported by specialized economic and technical agencies, and supplemented by services organized by farmers themselves.” During this period, agricultural service provision shifted from tractor stations operated by rural collectives to market-oriented providers, including specialized agricultural service households and agricultural service companies. By the end of 1984, China had more than 930,000 households specializing in agricultural machinery services and over 30,000 households specializing in machinery repair. Specialized machinery-service households had consequently become the principal providers of mechanized farming services [28].

Since the beginning of the 21st century, changes in agricultural and rural development have moved the agricultural socialized service system into a new stage in which public institutions perform public-interest functions, whereas commercial services are increasingly provided through market mechanisms. This transition has placed greater emphasis on the market-oriented provision of agricultural services. Existing studies define the agricultural socialized service system as the complete set of organizations, institutional arrangements, and operating mechanisms through which socialized services are provided for agricultural production [29]. By mobilizing resources from different sectors of society, this organizational model enables relatively small agricultural producers to adapt to the market economy, overcome the disadvantages associated with their limited operational scale, and obtain the benefits of large-scale production [30]. Nevertheless, the system continued to rely primarily on public-interest services supplied by village collectives, cooperative economic organizations, and specialized economic and technical agencies, with market-based agricultural production services playing a supplementary role.

Since the 18th National Congress of the Communist Party of China, the development of agricultural producer services has gradually become an important policy instrument for cultivating new drivers of economic growth under China’s “new normal.” Various market-oriented service providers have expanded rapidly. Agricultural brokers, agricultural input retailers, and specialized machinery-service households have become increasingly important participants in rural development. Agricultural enterprises and farmers’ cooperatives have established service companies or integrated specialized service providers, while supply and marketing cooperatives, the postal system, and state-owned grain enterprises have revitalized their rural operations through the development of agricultural producer services.

Against this background, the Ministry of Finance issued the Guiding Opinions on Implementing the 2013 Pilot Program for Full-Process Agricultural Production Services. The program was initially implemented in eight provinces, including Hebei, Jiangsu, and Hunan. Through government-funded purchases of market-supplied agricultural services, the program sought to strengthen market-based service providers and promote the development of agricultural production services. Its principal measures included subsidies for service providers to purchase agricultural machinery and subsidies for farmers to purchase mechanized farming services.

Fengcheng City in Jiangxi Province provides an illustrative example. In 2013, the municipal government planned to allocate CNY 4.900 million in program subsidies. Of this amount, CNY 3.000 million was designated for service organizations to purchase medium- and large-scale high-efficiency crop-protection equipment and other agricultural machinery. The subsidy could be paid in cash upon presentation of the relevant purchase invoice or provided in the form of agricultural machinery and equipment of equivalent value. The remaining CNY 1.900 million was allocated to cover the fuel and labor costs incurred by service organizations when providing services to farmers and was paid directly to specialized service providers. In 2016, the Ministry of Agriculture and the Ministry of Finance jointly issued the Notice on Implementing the 2016 Pilot Program for Full-Process Agricultural Production Services to continue the program and expand its coverage to 17 provinces.

3. Theoretical Framework

This section develops a conceptual framework for understanding the potential effects of the Agricultural Production Service Pilot Program on local agricultural production. Agricultural production services may first affect how grain production is organized. Land fragmentation and small-scale farming make it difficult for individual farmers to bear the fixed costs of large agricultural machinery and restrict the continuous operation of machinery across small and dispersed plots. Agricultural production services separate machinery ownership from machinery use. Specialized service providers purchase and operate machinery centrally while aggregating farmers’ demand for land preparation, sowing, crop protection, and harvesting services. As the operational coverage of service providers expands, the fixed costs of machinery can be distributed across a larger number of plots. Smallholders can therefore access modern mechanized technologies without purchasing machinery themselves. This arrangement reduces labor input per unit of land, increases the area that can be cultivated by each agricultural worker, reduces land idleness, raises multiple-cropping intensity, and expands the grain-sown area. Agricultural production services can thus generate machinery-driven, area-based growth in grain production. Accordingly, we propose the following hypothesis:

H1. 

The Agricultural Production Service Pilot Program increases total grain output in the pilot regions.

The mechanization effects of agricultural production services are concentrated in labor-intensive operations that must be completed within narrowly defined farming periods. Specialized machinery services substitute for household labor in operations such as sowing, rice transplanting, and harvesting and reduce the time required to complete these operations per unit of land [31]. The development of agricultural production services therefore does not require agricultural workers to remain in the agricultural sector. Instead, these services reduce the dependence of grain production on local household labor and enable a smaller agricultural workforce to cultivate a larger area [32]. Accordingly, we propose the following hypothesis:

H2. 

The Agricultural Production Service Pilot Program reduces grain production’s dependence on local agricultural labor by expanding mechanized operations in standardized production stages, thereby increasing the grain-sown area cultivated per agricultural worker.

Agricultural production is highly seasonal. Delays in sowing and harvesting not only affect production in the current season but also disrupt the coordination of successive crops and cropping seasons. In the context of rural labor outmigration and population aging, seasonal labor shortages can prevent farmland from being cultivated in a timely manner and increase the risk of land idleness and abandonment. Agricultural production services centrally allocate machinery and skilled operators, accelerate sowing and harvesting, and alleviate constraints associated with critical farming periods. They can therefore increase the intensity with which existing cultivated land is used [33,34].

Agricultural production services may also alter farmers’ crop choices. Grain crops generally involve relatively standardized production processes and are particularly compatible with mechanized services. By reducing the labor costs of grain cultivation, agricultural production services increase the relative returns to grain crops and encourage a grain-oriented adjustment in the cropping structure [35]. Accordingly, we propose the following hypotheses:

H3a. 

The Agricultural Production Service Pilot Program reduces the idleness of existing cultivated land and increases multiple-cropping intensity and land-use intensity by alleviating seasonal labor shortages and constraints associated with critical farming periods.

H3b. 

The Agricultural Production Service Pilot Program promotes a grain-oriented adjustment in the cropping structure by strengthening the comparative advantage of mechanized production for major field grain crops.

The mechanization promoted by agricultural production services constitutes a labor-saving technology. Its primary function is to increase the area that can be cultivated by a given amount of labor rather than directly alter the biological yield potential of crops. The effect of the pilot program on grain output should therefore operate primarily through an expansion of the grain-sown area, whereas its effect on grain yield per unit area should be comparatively limited. Accordingly, we propose the following hypothesis:

H4. 

The Agricultural Production Service Pilot Program increases major-grain output primarily by expanding the sown area of rice, wheat, and maize rather than by raising their average yield per unit area.

4. Empirical Strategy and Data

4.1. Empirical Strategy

We employ a staggered difference-in-differences (DID) model to estimate the effect of the Agricultural Production Service Pilot Program on grain output. The baseline specification is expressed as follows:

$$y_{it}=\alpha +\beta {ASS}_{it}+\delta X_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(1)

In the baseline model, y_it denotes the natural logarithm of grain output in province $i$ in year $t$. ${ASS}_{it}$is the treatment indicator for the Agricultural Production Service Pilot Program. If province i was included in the pilot program in year t, ${ASS}_{it}$ equals one in the year of implementation and all subsequent years and zero otherwise. In addition, α is the constant term, β is the coefficient of primary interest, and δ is the vector of coefficients associated with the control variables. $X_{it}$denotes a vector of time-varying control variables, ${\mu }_i$ and ${\gamma }_t$ represent province and year fixed effects, respectively, and ${\epsilon }_{it}$ is the random error term.

The dependent variable is the natural logarithm of major-grain output, which is measured as the aggregate provincial output of rice, wheat, and maize. The average yield of the three major grain crops is calculated as their aggregate output divided by their total sown area. The principal explanatory variable, ${ASS}_{it}$ , is constructed according to the official list of provinces included in the Agricultural Production Service Pilot Program. The program was introduced in two batches, in 2013 and 2016. For each pilot province, ${ASS}_{it}$ is assigned a value of one in the initial year of policy implementation and all subsequent years and zero before implementation. It remains zero throughout the sample period for provinces that were not included in the program.

This study examines whether agricultural production services promote labor-saving mechanization and machinery-driven, area-based growth in grain production by reducing land idleness, increasing multiple-cropping intensity, and expanding the grain-sown area. To measure changes in production methods associated with labor-saving mechanization, we use the natural logarithms of the mechanically plowed, sown, and harvested areas of wheat and maize, as well as the mechanically plowed, transplanted, and harvested areas of rice. These indicators are used to determine whether the pilot program expanded mechanized operations, transformed agricultural production methods, and increased the amount of farmland that could be cultivated by each agricultural worker.

To capture changes in the pattern and intensity of cultivated-land use, we employ the natural logarithms of the total crop-sown area, cultivated-land area, and grain multiple-cropping index, together with the land-transfer rate and the degree of grain orientation in the cropping structure. These variables allow us to examine whether agricultural production services reduced land idleness, increased multiple-cropping intensity, and expanded the grain-sown area, thereby generating machinery-driven, area-based growth in grain production.

Grain production is affected by a wide range of economic, climatic, and policy-related factors. Following the existing literature, the baseline regressions include rural residents’ income (RRI), the level of agricultural development (LAD), the level of mechanization (ML), the level of industrial development (LID), climate-related disasters (CC), government support for agriculture (GS), environmental regulation (ER), the level of urbanization (UL), and fiscal support for agricultural machinery investment (CAMI).

Rural residents’ income is measured as the natural logarithm of per capita disposable income of rural residents. The level of agricultural development is measured as the natural logarithm of regional agricultural gross output value. The level of mechanization is measured as the ratio of total agricultural machinery power to the total crop-sown area. The level of industrial development is measured as the ratio of value added in the secondary sector to regional gross domestic product. Climate-related disasters are measured as the ratio of disaster-affected crop area to the total crop-sown area. Government support for agriculture is measured as the ratio of fiscal expenditure on agriculture, forestry, and water affairs to total local general public budget expenditure. Environmental regulation is measured as the ratio of local fiscal expenditure on environmental protection to total local general public budget expenditure. The level of urbanization is measured as the ratio of the urban population to the rural population. Fiscal support for agricultural machinery investment is measured as the natural logarithm of total fiscal expenditure on agricultural machinery investment.

4.2. Measurement of Total Factor Productivity

Total factor productivity (TFP) is not a directly observable statistical indicator. Following the stochastic frontier production function approach adopted by Chen and Gong [36], we estimate grain-production TFP using provincial data on grain output and four production inputs: labor, land, machinery, and fertilizer. Grain output is measured as the sum of wheat, maize, and rice production. Labor input is measured by employment in the primary sector; land input by the total crop-sown area; machinery input by total agricultural machinery power; and fertilizer input by the quantity of agricultural fertilizer applied on a nutrient-equivalent basis. All output and input variables are transformed into natural logarithms.

Because agricultural production is susceptible to random shocks arising from climatic conditions, natural disasters, and other external factors, we employ a panel stochastic frontier analysis (SFA) model. The composite error term of the production function is decomposed into a two-sided random disturbance and a one-sided technical inefficiency component. The baseline specification uses a Cobb–Douglas production function and incorporates year dummy variables into the production frontier to capture shifts in production technology over time. The technical inefficiency term is assumed to follow a truncated normal distribution and is allowed to vary over time.

$$Outpu{t}_{it}={\beta }_0+{\beta }_{1}L{r}_{it}+{\beta }_{2}L{d}_{it}+{\beta }_3{M}_{it}+{\beta }_4{F}_{it}+{\lambda }_t-{\mu }_{it}+{\nu }_{it}$$

(2)

Within the stochastic frontier framework, we construct log total factor productivity by combining the estimated frontier technology level, conditional on observable production inputs, with technical efficiency, as follows:

$$\ln TF{P}_{it}={\widehat{\beta }}_0+{\widehat{\lambda }}_t-{\widehat{u}}_{it}$$

(3)

where${\widehat{\beta }}_0+{\widehat{\lambda }}_t$ captures temporal shifts in the production frontier conditional on observable inputs, and u ̂_it represents the estimated degree of technical inefficiency. To assess the sensitivity of the TFP estimates to the choice of production function and the assumption regarding returns to scale, we further estimate a translog production function and recalculate TFP under both constant returns to scale (CRS) and variable returns to scale (VRS).

4.3. Data Sources

Data on mechanized operations for grain crops were obtained from the 2006–2024 editions of the China Agricultural Machinery Industry Yearbook. Data on the total area of household-contracted farmland transferred and the total area of household-contracted farmland under cultivation were collected from the 2006–2024 editions of the China Rural Management Statistics Annual Report, the China Rural Policy and Reform Statistics Annual Report, and the China Cooperative Economy Statistics Annual Report. All other data were obtained from the National Bureau of Statistics of China. Missing observations for selected variables were imputed using linear interpolation.

Table 1. Descriptive Statistics.

VarName	Obs	Mean	SD	Min	Median	Max
ln(Major-Grain Output)	558	6.74	1.70	0.00	7.15	8.87
ASS	558	0.29	0.45	0.00	0.00	1.00
RRI	558	9.20	0.62	7.65	9.28	10.67
LAD	558	6.93	1.18	3.56	7.23	8.85
ML	558	0.67	0.36	0.22	0.57	2.70
LID	558	0.42	0.08	0.15	0.42	0.62
CC	558	0.18	0.15	0.00	0.14	0.70
GS	558	0.11	0.03	0.03	0.11	0.20
ER	558	0.03	0.01	0.01	0.03	0.09
UL	558	1.78	1.68	0.27	1.26	8.60
CAMI	558	11.33	1.51	0.00	11.47	14.33

Table 1. Descriptive Statistics.

VarName	Obs	Mean	SD	Min	Median	Max
ln(Major-Grain Output)	558	6.74	1.70	0.00	7.15	8.87
ASS	558	0.29	0.45	0.00	0.00	1.00
RRI	558	9.20	0.62	7.65	9.28	10.67
LAD	558	6.93	1.18	3.56	7.23	8.85
ML	558	0.67	0.36	0.22	0.57	2.70
LID	558	0.42	0.08	0.15	0.42	0.62
CC	558	0.18	0.15	0.00	0.14	0.70
GS	558	0.11	0.03	0.03	0.11	0.20
ER	558	0.03	0.01	0.01	0.03	0.09
UL	558	1.78	1.68	0.27	1.26	8.60
CAMI	558	11.33	1.51	0.00	11.47	14.33

5. Empirical Results

5.1. Baseline Regression Results

Columns (1)–(4) of Table 2 report the estimated effects of the pilot program on the aggregate output and average yield of rice, wheat, and maize. After controlling for province and year fixed effects and a set of time-varying covariates, the pilot program increased major-grain output by an average of 46.668% and reduced the average yield of the three major grain crops by 2.737%. Both effects are statistically significant at the 1% level.

These results demonstrate that the pilot program increased total grain output, thereby supporting H1. However, the program did not increase grain yield per unit area. Instead, its estimated effect on yield was significantly negative. This finding indicates that the increase in grain output induced by the pilot program was not achieved through improvements in land productivity.

Table 3 reports the estimated effects of the pilot program on the grain-sown area, agricultural employment, and grain-sown area per agricultural worker. The results show that the program increased the grain-sown area by an average of 13.655%, reduced agricultural employment by an average of 4.972%, and increased the grain-sown area per agricultural worker by 0.101 ha per person.

Taken together, these findings provide preliminary evidence that the increase in grain output was driven by area expansion rather than higher yield per unit area. By reducing labor input per unit of land and increasing the amount of farmland cultivated by each agricultural worker, agricultural production services helped reduce land idleness, raise multiple-cropping intensity, and expand the grain-sown area. The resulting increase in grain production therefore primarily took the form of machinery-driven, area-based growth, consistent with H2 and H4.

5.2. Robustness Tests

Parallel-Trends Test. The validity of the difference-in-differences strategy relies on the parallel-trends assumption. Specifically, in the absence of the Agricultural Production Service Pilot Program, grain output in the treatment and control groups should have followed similar trends before policy implementation. After implementation, grain output in the treated provinces may deviate from its counterfactual trajectory because of the policy intervention. We employ an event-study specification to examine the dynamic effects of the pilot program. As shown in Figure 1, the estimated coefficients for the pre-treatment periods are not statistically different from zero, indicating no systematic difference in pre-policy trends between the treatment and control groups. The parallel-trends assumption is therefore supported.

Placebo Test. Following Chen [37], we conduct a placebo test using the Stata command didplacebo to further assess the validity of the baseline estimates. Specifically, we perform 1,000 random permutations in which both pseudo-treated provinces and pseudo-treatment years are reassigned, and the baseline model is then re-estimated for each permutation. The results are presented in Figure 2. The estimated placebo effects are concentrated around zero, whereas the baseline treatment-effect estimate, represented by the vertical solid line, lies in the right tail of the placebo distribution. This result indicates that an effect comparable to the baseline estimate is unlikely to arise solely from random treatment assignment, thereby reducing concerns that the baseline findings are driven by spurious correlations or unobserved random shocks.

Propensity Score Matching and Entropy Balancing. To mitigate potential sample-selection bias arising from differences in observable characteristics between the treatment and control groups, we combine the DID framework with propensity score matching (PSM) and entropy balancing. The covariates used for matching include the level of mechanization, the level of industrial development, climate-related disasters, government support for agriculture, environmental regulation, and the level of urbanization. We first estimate the propensity score and then construct matched samples using 1:3 nearest-neighbor matching, kernel matching, and radius matching. We additionally apply entropy balancing to reweight the control observations so that their covariate distributions more closely resemble those of the treated observations. The effect of the Agricultural Production Service Pilot Program on grain output is subsequently re-estimated using each matched or reweighted sample. The estimated policy effect remains positive and statistically significant across these alternative procedures, indicating that the baseline finding is robust to improved balance in observable characteristics.

Table 4. Results of PSM-DID and Entropy Balancing Estimations.

	(1)	(2)	(3)	(4)
	Nearest-Neighbor Matching	Kernel Matching	Radius Matching	Entropy Balancing
ASS	0.234***	0.263***	0.286***	0.398***
	(0.090)	(0.069)	(0.087)	(0.139)
RRI	-0.248	-0.459	-0.603	-3.580**
	(0.694)	(0.471)	(0.591)	(1.390)
LAD	-0.006	0.040	0.130	0.972**
	(0.239)	(0.152)	(0.218)	(0.409)
ML	-1.062***	-1.123***	-1.115***	-0.649***
	(0.384)	(0.309)	(0.288)	(0.196)
LID	-2.129*	-1.150	-1.820*	-2.279***
	(1.249)	(0.702)	(1.084)	(0.736)
CC	0.086	0.018	0.035	0.373
	(0.227)	(0.170)	(0.189)	(0.296)
GS	-2.240	-1.633	-2.150	-8.198**
	(2.635)	(1.976)	(2.210)	(3.867)
ER	-4.376	-5.432	-4.247	-5.718*
	(5.118)	(3.421)	(4.267)	(3.008)
UL	-0.049	-0.037	-0.045	-0.292*
	(0.135)	(0.104)	(0.118)	(0.177)
CAMI	0.210***	0.144***	0.168***	0.194***
	(0.071)	(0.044)	(0.055)	(0.056)
_cons	8.840*	10.800***	11.472***	33.388***
	(5.221)	(3.679)	(4.357)	(10.138)
Province	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	410	520	512	558
F	2.597***	3.825***	3.258***	3.482***
r2	0.906	0.914	0.915	0.938

Table 4. Results of PSM-DID and Entropy Balancing Estimations.

	(1)	(2)	(3)	(4)
	Nearest-Neighbor Matching	Kernel Matching	Radius Matching	Entropy Balancing
ASS	0.234***	0.263***	0.286***	0.398***
	(0.090)	(0.069)	(0.087)	(0.139)
RRI	-0.248	-0.459	-0.603	-3.580**
	(0.694)	(0.471)	(0.591)	(1.390)
LAD	-0.006	0.040	0.130	0.972**
	(0.239)	(0.152)	(0.218)	(0.409)
ML	-1.062***	-1.123***	-1.115***	-0.649***
	(0.384)	(0.309)	(0.288)	(0.196)
LID	-2.129*	-1.150	-1.820*	-2.279***
	(1.249)	(0.702)	(1.084)	(0.736)
CC	0.086	0.018	0.035	0.373
	(0.227)	(0.170)	(0.189)	(0.296)
GS	-2.240	-1.633	-2.150	-8.198**
	(2.635)	(1.976)	(2.210)	(3.867)
ER	-4.376	-5.432	-4.247	-5.718*
	(5.118)	(3.421)	(4.267)	(3.008)
UL	-0.049	-0.037	-0.045	-0.292*
	(0.135)	(0.104)	(0.118)	(0.177)
CAMI	0.210***	0.144***	0.168***	0.194***
	(0.071)	(0.044)	(0.055)	(0.056)
_cons	8.840*	10.800***	11.472***	33.388***
	(5.221)	(3.679)	(4.357)	(10.138)
Province	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	410	520	512	558
F	2.597***	3.825***	3.258***	3.482***
r2	0.906	0.914	0.915	0.938

Heterogeneity-Robust DID Estimation. Under staggered policy adoption, two-way fixed-effects estimators may be biased in the presence of heterogeneous treatment effects because already-treated units can serve as controls for units treated at a later date [38,39]. The Bacon decomposition shows that potentially problematic comparisons in which earlier-treated provinces serve as controls for later-treated provinces account for only 7.9% of the total estimation weight. This relatively small share indicates that such comparisons have a limited influence on the baseline estimate.

To further assess the robustness of our findings to heterogeneous treatment effects, we apply the imputation-based DID estimator implemented through did_imputation [40]. The results are reported in Table 5. The estimated effect of the Agricultural Production Service Pilot Program remains positive and statistically significant, consistent with the baseline findings. These results indicate that the estimated increase in grain output is not driven by the inappropriate comparisons inherent in conventional two-way fixed-effects estimation under heterogeneous treatment effects.

Winsorization. To examine whether the estimates are sensitive to extreme observations, we winsorize all continuous variables at both tails using the 1% and 5% thresholds and re-estimate the baseline model using the winsorized data. The results are reported in Columns (2) and (3) of Table 5. The estimated coefficients of the Agricultural Production Service Pilot Program on grain output remain broadly consistent with the baseline estimate in terms of sign, statistical significance, and magnitude. These findings indicate that the baseline results are not driven by extreme observations.

Grain-Producing-Region-by-Year Fixed Effects. Major grain-producing regions and other regions differ systematically in their agricultural resource endowments, grain-production objectives, fiscal support, and exposure to agricultural policies. Moreover, the two groups may experience different common shocks over time. To account for these differences, we augment the baseline specification with major-grain-producing-region-by-year fixed effects. These fixed effects allow major grain-producing regions and other regions to follow separate, nonparametric annual trajectories and absorb year-specific policy, market, and agricultural production shocks shared by provinces within each group.

After controlling for major-grain-producing-region-by-year fixed effects, the estimated effect of the Agricultural Production Service Pilot Program on grain output remains positive and statistically significant. This result indicates that the estimated increase in grain output cannot be explained by differential annual trends between major grain-producing regions and other regions.

6. Mechanism Analysis

6.1. The Agricultural Production Service Pilot Program and Labor-Saving Mechanization

Table 6 reports the estimated effects of the Agricultural Production Service Pilot Program on the mechanized production of wheat, rice, and maize. The results show that the pilot program increased the mechanically sown area of wheat by an average of 33.109%, the mechanically transplanted area of rice by 168.854%, the mechanically harvested area of rice by 19.842%, the mechanically sown area of maize by 64.872%, and the mechanically harvested area of maize by 111.488%.

The estimated effects are concentrated in sowing, transplanting, and harvesting—operations that are labor-intensive and subject to strict seasonal constraints. By expanding access to mechanized services during these production stages, the pilot program substituted machinery services for household labor and reduced the operating time required per unit of land. Agricultural production services therefore reduced grain production’s dependence on local household labor and enabled a smaller agricultural workforce to cultivate a larger area. These results provide support for H2.

6.2. Effects of the Agricultural Production Service Pilot Program on Land-Use Patterns and Intensity

Table 7 reports the estimated effects of the Agricultural Production Service Pilot Program on land-use patterns and land-use intensity. The results show that the pilot program increased the cultivated-land area in the pilot regions by an average of 4.081%, the total crop-sown area by 7.681%, the degree of grain orientation in the cropping structure by 3.252%, and the grain multiple-cropping index by 9.090%.

Agricultural production is highly seasonal. Delays in critical operations such as sowing and harvesting not only reduce returns in the current season but also disrupt the timing and coordination of different crops and successive cropping seasons. Continued rural labor outmigration and the aging of the agricultural workforce have intensified seasonal labor shortages. Consequently, some cultivated land cannot be brought into production within the appropriate farming period, increasing the risk of land idleness and abandonment.

Agricultural production services address these constraints through the centralized allocation of machinery and specialized operating teams. By accelerating sowing and harvesting, these services ease binding constraints associated with critical farming periods, improve the utilization of existing cultivated land, and increase land-use intensity. More complete utilization of cultivated land expands the total crop-sown area. Compared with other crops, grain crops have more standardized production processes and are better suited to specialized mechanized services. Consequently, a greater proportion of the additional sown area is allocated to grain crops, producing a grain-oriented shift in the cropping structure. Meanwhile, faster completion of agricultural operations widens the available farming window and facilitates an increase in regional multiple-cropping intensity. These findings support H3a and H3b.

7. Further Discussion

7.1. Agricultural Production Services and Land Transfers

Agricultural production services can generate service-based economies of scale by conducting operations across multiple farms, even when contractual land-management rights remain dispersed. Land transfers and agricultural production services therefore represent two distinct pathways toward larger-scale agricultural operations. Land transfers increase the area controlled by an individual farm operator, whereas agricultural production services expand the operational coverage of machinery, technology, and specialized labor.

The two pathways are not necessarily interdependent. On the one hand, agricultural production services reduce the labor, machinery-purchase, and production-supervision costs faced by farmers who acquire additional land. As their operational land area expands, these farmers do not need to proportionally increase family labor input or purchase a complete set of machinery. Instead, they can outsource land preparation, sowing, crop protection, and harvesting to specialized service providers. This arrangement lowers the cost of farm expansion and strengthens farmers’ capacity and willingness to acquire additional land.

On the other hand, agricultural production services separate machinery ownership from machinery use, allowing smallholders to purchase mechanized services for specific production stages. By operating across households and plots, specialized service providers distribute the fixed costs of machinery over a larger service area. Service-based economies of scale can therefore emerge without the prior consolidation of contractual land-management rights.

Table 8 examines both the effect of the Agricultural Production Service Pilot Program on the land-transfer rate and whether initial land-transfer conditions strengthen the program’s effect on grain output. Column (1) shows that the pilot program increased the land-transfer rate by an average of 2.220 percentage points. In Column (2), initial land transfer is measured as the average land-transfer rate during 2010–2012 and is interacted with the pilot-program indicator. The estimated coefficient of the interaction between the pilot program and the initial land-transfer rate is negative but statistically insignificant.

The positive effect on the land-transfer rate is consistent with the argument that agricultural production services reduce the labor, machinery-purchase, and production-supervision costs associated with expanding operational farm size. However, the interaction estimate provides no evidence that the grain-output effect of the pilot program was stronger in provinces with higher initial land-transfer rates. A high initial level of land transfer thus does not appear to be a prerequisite for agricultural production services to increase grain output. By aggregating the mechanized operational demands of dispersed farmers, agricultural production services can generate service-based economies of scale before contractual land-management rights are extensively consolidated, thereby alleviating the constraints imposed by land fragmentation on mechanized production.

7.2. Agricultural Production Services and Total Factor Productivity

The preceding analysis shows that agricultural production services substitute specialized mechanized operations for dispersed household labor, alleviate seasonal labor constraints during sowing and harvesting, and expand the grain-sown area. However, an expansion in the scale of grain production does not necessarily imply an improvement in overall production efficiency. If machinery, labor, and land inputs are not effectively coordinated, increases in machinery capital and expanded land use may result in underutilized production factors and fail to generate sustained productivity growth. It is therefore necessary to examine whether the area-based growth in grain production promoted by agricultural production services is accompanied by improvements in total factor productivity.

Under small-scale and fragmented farming, the limited operational area of individual farmers makes it difficult to distribute the fixed costs of large agricultural machinery over a sufficiently large area. Land fragmentation also increases the costs of moving machinery between plots, coordinating operations, and supervising labor, thereby reducing the efficiency with which machinery, labor, and land are allocated. Agricultural production services separate machinery ownership from machinery use. Specialized service providers purchase and allocate machinery centrally while aggregating farmers’ dispersed demand for land preparation, sowing, and harvesting. As the area covered by service operations expands, machinery fixed costs can be distributed across a larger number of plots. This arrangement reduces machinery idleness and duplicative machinery purchases, thereby lowering the cost of mechanized services per unit of land.

Specialized service providers can also allocate machinery and operators according to crop calendars and local farming conditions, improving the timeliness of sowing and harvesting and reducing temporal mismatches among labor, machinery, and land across different production stages. Moreover, specialization reduces the information, coordination, and supervision costs incurred when individual farmers organize production independently, allowing existing machinery, land, and labor inputs to be used more efficiently. The effects of agricultural production services may therefore extend beyond changes in the quantities of machinery and land inputs to include an increase in output conditional on aggregate inputs, ultimately improving total factor productivity.

Table 9 reports the estimated effect of the Agricultural Production Service Pilot Program on TFP. To reduce the sensitivity of the TFP estimates to the functional form of the production frontier and the assumption regarding returns to scale, we estimate TFP using both Cobb–Douglas and translog production functions under constant returns to scale (CRS) and variable returns to scale (VRS). Across all four TFP measures, the estimated coefficient of the pilot program is positive and statistically significant. The corresponding estimates indicate that the program increased TFP by an average of 12.862%–40.779%. Thus, the positive effect on TFP is robust across alternative production-function specifications and returns-to-scale assumptions.

Combined with the preceding results, these findings show that agricultural production services reduced dependence on local agricultural labor and expanded mechanized operations and the grain-sown area while also improving overall production efficiency. The area-based growth in grain production induced by the program was therefore not driven exclusively by increases in the quantities of production factors; it was also accompanied by greater specialization and improved factor-allocation efficiency.

8. Discussion

Using provincial-level panel data and the staggered rollout of China’s Agricultural Production Service Pilot Program, this study examines how agricultural production services sustain the production of major grain crops under fragmented land operation and continued rural labor outmigration. The evidence points to an organizational mechanism rather than a simple increase in machinery stocks: by separating machinery ownership from use and aggregating demand across farms, service providers allow small-scale landholdings to be operated with large-scale mechanized services.

8.1. Grain-Production Growth Driven by the Extensive Margin

The central finding is that the increase in major-grain output was driven primarily by an expansion of the sown area rather than higher yield per unit area. Whereas previous research has emphasized intensive-margin growth through improved varieties, modern inputs, soil-fertility management, and technological learning [21,22,24], mechanized services operate mainly by relaxing labor and timing constraints. They enable a given amount of labor to cultivate more land, allow field operations to be completed within narrow farming windows, and facilitate successive cropping seasons.

The decline in average yield should not be interpreted as direct evidence of technological regression. Additional production may involve marginal land or cropping seasons with lower yields, while changes in crop composition can reduce the regional average. Because the provincial data cannot separately identify these sources, we interpret the yield decline as a composition change accompanying extensive-margin expansion rather than as evidence of lower technical efficiency.

8.2. Service-Scale Mechanization and Labor Substitution

The expansion of mechanized operations was concentrated in wheat sowing, rice transplanting and harvesting, and maize sowing and harvesting. These standardized, labor-intensive operations face stricter seasonal constraints than plowing, which was already comparatively mechanized. The results are consistent with evidence that mechanization services substitute for agricultural labor, improve factor allocation, and affect farm productivity [14,16,18]. They also show that policy effects vary across production stages: service support is most effective where labor bottlenecks and the cost of delayed operations are greatest.

The provincial employment data do not establish that the program caused individual workers to enter non-agricultural employment, nor do they distinguish household, hired, and service-provider labor. The evidence instead shows that lower primary-sector employment, expanded mechanized operations, and a larger major-grain sown area per worker occurred together. We therefore interpret the results as reduced dependence on local agricultural labor, not as direct micro-level evidence of labor migration.

8.3. Land-Based Scale Operations and Service-Based Scale Operations

Land transfers and agricultural production services represent related but distinct routes to scale. Land transfers expand the area managed by an individual operator, while service providers aggregate demand across farms and distribute machinery fixed costs over a wider operating area. The pilot program increased the land-transfer rate, but the interaction with the pre-policy land-transfer rate was statistically insignificant. Thus, the data provide no evidence that higher initial land transfer strengthened the program’s effect on major-grain output.

Service-based scale operations can complement land consolidation by reducing the machinery, labor, and coordination costs faced by expanding farms. They can also function where land-management rights remain dispersed by organizing operations across household and plot boundaries. Their viability nevertheless depends on sufficient operating coverage [12], and they do not eliminate physical land fragmentation or replace land-market reform. Their contribution is to lower the threshold for accessing large machinery and specialized technologies before land rights are extensively consolidated.

8.4. Total Factor Productivity and Its Implications

The positive estimates across alternative production functions and returns-to-scale assumptions indicate that the expansion of major-grain output was accompanied by improvements in the aggregate input–output relationship. Specialized providers can reduce duplicated machinery purchases and equipment idleness, improve the timing of operations, and coordinate machinery, labor, and land more effectively. This extends previous evidence on mechanization services, factor allocation, land productivity, and farm efficiency [14,16,18] from farm-level adoption to a province-level policy intervention.

These results require a bounded interpretation. TFP is a residual measure that combines movements in the production frontier with changes in technical efficiency and is sensitive to functional form, returns-to-scale assumptions, and input measurement. The estimates therefore support an improvement in overall production efficiency but do not identify the separate contributions of specialization, machinery sharing, scale efficiency, or technological progress.

8.5. Theoretical Contributions and Scope Conditions

This study contributes in three respects. First, it distinguishes land-based scale from service-based scale: mechanization can extend across farms even when individual landholdings remain small. Second, it distinguishes intensive-margin growth from extensive-margin growth, showing how labor-saving services expand the sown area and intensify land use without increasing average yield. Third, it shows that declining local agricultural employment can coexist with expanding major-grain output when specialized services reduce production’s reliance on household labor.

These conclusions apply most directly to field crops with standardized operations, mature mechanized technologies, and sufficiently concentrated production. They cannot be directly generalized to regions with highly fragmented terrain, diverse cropping systems, or thin service markets. Moreover, because the pilot combined service-purchase subsidies, machinery-acquisition support, and service-organization development, the estimates capture the policy package rather than any single instrument.

8.6. Limitations and Directions for Future Research

Three limitations remain. First, provincial data do not record service quantities and prices, household labor hours, plot fragmentation, land abandonment, or operating costs; the mechanism results are therefore supporting regional evidence rather than direct micro-level mediation tests. Second, the data cannot fully decompose sown-area growth into reduced abandonment, additional cropping seasons, statistical changes, and crop-composition adjustments. Matched household–provider data, plot-level records, and remote-sensing evidence would allow these channels to be distinguished.

Third, the estimates capture average post-treatment effects rather than the long-term sustainability of extensive-margin growth. Higher multiple-cropping intensity and the cultivation of marginal land may increase pressure on soil, water, and the environment. Future research should examine input-use efficiency, carbon emissions, cultivated-land quality, and farm income to determine whether service-scale mechanization can jointly support food security, household welfare, and environmental sustainability.

9. Conclusions

9.1. Main Conclusions

This study uses China’s provincial panel data and a staggered difference-in-differences design to estimate the effects of the Agricultural Production Service Pilot Program on the aggregate output of rice, wheat, and maize. The baseline estimate indicates an average increase of 46.668%, and the result remains robust across alternative identification and specification tests.

The increase was driven primarily by the extensive margin. The major-grain sown area increased by 13.655%, primary-sector employment declined by 4.972%, and the sown area per worker increased by 0.101 ha, while average yield did not increase. Operation-specific results show that mechanization expanded mainly in sowing, rice transplanting, and harvesting, where labor demand is concentrated and timing constraints are strict. The program also increased cultivated-land use, total crop-sown area, grain orientation, and multiple-cropping intensity. Together, these findings indicate that specialized mechanized services enabled fewer local workers to complete operations over a larger area.

The pilot increased the land-transfer rate by 2.220 percentage points, but there is no evidence that a higher pre-policy land-transfer rate strengthened its effect on major-grain output. Service-based economies of scale can therefore operate alongside land transfers and can also emerge before land-management rights are extensively consolidated. In addition, all four SFA-based productivity measures increased significantly, although these aggregate measures do not identify the contribution of individual efficiency channels.

Overall, agricultural production services reconcile fragmented landholding with large-scale mechanized operations by separating machinery ownership from use and aggregating demand across farms. This organizational form offers a route to sustaining major-grain production under rural labor outmigration without making prior land consolidation a necessary condition for access to modern mechanized technologies.

9.2. Policy Implications

Policy support should target standardized operations with the strongest labor and timing constraints, particularly sowing, rice transplanting, and harvesting, and should strengthen providers’ capacity to coordinate machinery and operators across regions. Subsidies should be directed toward identified production bottlenecks rather than distributed uniformly across operations.

Agricultural support should combine machinery-acquisition assistance with subsidies for service use. For smallholders, purchasing large machinery can create idle capital and duplicated investment; purchasing mechanized services enables access without requiring machinery ownership and can raise equipment-utilization rates.

Service systems should be adapted to regional crop structures, terrain, and farming calendars, with stronger platforms for matching supply and demand and monitoring service quality. Land-transfer institutions and service-based scale operations should be developed jointly: land-market reform can support expanding farms, while village collectives, cooperatives, and specialized providers can organize unified operations where land rights remain dispersed.

Finally, policy evaluation should consider the sustainability of extensive-margin growth. Expansion of the sown area and multiple cropping can support food supply, but it may also increase the use of marginal land and pressure on soil and water. Service policies should therefore be combined with cultivated-land protection, improved varieties, integrated machinery and agronomic practices, and environmentally sustainable input management.