The Impact of Rural Households’ Part-Time Farming on Grain Output: Promotion or Inhibition?

Abstract

Given the prevalence of part-time farming behaviors in rural households, studying the impact of part-time farming behaviors on grain output is of great practical significance. Using a panel dataset of 5629 Chinese national rural fixed observation point data from 2009 to 2015, this paper used the Propensity Score Matching-Difference in Differences method (PSM-DID) to examine the impact and dynamic effect of rural household’s part-time farming behavior on grain output. In addition, the paper also aims to explore the mechanism of how part-time farming affects grain output by running an OLS in an intermediary effect model. We analyzed the data from two aspects: the effect of rural households’ part-time farming decisions on grain output and the relationship between the income from rural households’ part-time farming and grain output. After accounting for the control variables, the results showed a significant positive correlation between rural households’ part-time farming behavior and grain output, with an influence coefficient of 0.304 tons increase in total grain output. While rural households’ part-time farming behavior inhibited grain output over the short term, it ultimately promoted grain output in the long run. The promotion effect increased with the duration of the rural households’ part-time farming. Moreover, the degree of part-time farming and its impact on grain output are mediated by agricultural labor inputs and agricultural technology inputs. Hence, differences in the employment times of rural households’ part-time farming can lead to varying results in grain output. To improve the efficiency of grain output, this study recommends that the government implement policies that promote orderly rural households’ part-time farming behavior.

1. Introduction

Grain output is the foundation of the global economy and influences people’s livelihoods. The United Nations has consistently placed a high priority on ensuring grain output and food security [1]. According to the Food and Agricultural Organization report on The State of Food Security and Nutrition in the World 2022, around 2.3 billion people experienced food insecurity at a moderate or severe level, and 11.7 percent of the global population was severely food insecure in 2021 [2]. Meanwhile, a series of policies have been implemented in China to increase the effectiveness of subsidies and improve the subsidy system for agricultural products, such as grain [3], which has resulted in a significant guarantee of grain output and food security. In fact, since its reform and opening, China has seen its total grain output more than double from 305 million tons in 1978 to 682 million tons in 2021. However, improvements in living standards and population growth have also led to a rise in the rigid demand for agricultural products [4]. Despite achieving the essential supply–demand balance for agricultural products, a structural grain shortage is still evident in China, and the outlook of grain output and food security is still pessimistic. With the development of China’s economy and the relaxation of restrictive regulations, such as registered residence, there is an outflow of agricultural laborers into off-farm fields in search of higher income [5], as the survival and needs of farmers cannot be sustained by agricultural production alone [6]. As a result, more and more farmers choose part-time farming models to earn income through both agricultural and non-agricultural means [7]. Thus, as the mainstay of agricultural operations, peasant households’ changes in the management mode (i.e., from full-time to part-time farming mode) will inevitably affect the balanced development of grain output and food security [8]. Therefore, it is important to investigate whether and how rural households’ part-time farming decisions affect grain output.

Rural households’ part-time farming behavior is a typical occurrence in the growth of agriculture around the world and is also an unavoidable outcome of urbanization and industrialization [9,10]. The data showed that nearly half of the rural households in industrialized countries suffer a deficit in agricultural revenue and have to rely on off-farm income to make a living. Nevertheless, developed countries have a higher proportion of households engaging in part-time farming as compared to developing countries, as the farmers in developed countries have better technical expertise and human capital than their counterparts in developing countries [11].

Numerous studies have been conducted in China on large-scale part-time farming households, and researchers have examined the influence of rural households’ part-time farming on grain production from a variety of perspectives. This paper carefully and systematically reviewed the existing comprehensive research on part-time farming and farming-related issues and mainly referenced studies conducted in China. The research included food crops, cash crops, animal husbandry, the use of chemical fertilizer, and other facets [12,13], but with no consensus in the findings. Some scholars have found that rural households’ part-time farming behavior limits grain output. According to Sukusui’s research, households’ part-time farming behavior could effectively allocate resources and raise rural households’ incomes but reduce grain output [14]. Wei discovered that the technical efficiency of grain output was lower in part-time farming households than in full-time farming households [15]. This finding may be explained by the fact that part-time farming households’ cost of capital and time invested in grain output are significantly lower than that of full-time farming households [16]. In addition, there were certain distinctions between part-time farming rural households at different levels in terms of grain output efficiency. Some academics contended that the productivity of full-time farming households was much greater than that of part-time farming households. In their opinion, external support for the grain output of part-time farming households should be strengthened, grain output subsidies should be increased, and rural households’ enthusiasm for production should be mobilized [17]. Productive services could also have a disincentive effect on grain output [18]. Some scholars contended that the majority of income from part-time farming was spent on bettering life, rather than on compensating for productivity loss caused by labor outflow, which reduced grain output [19]. Moreover, rural households’ part-time farming caused the migration of highly qualified rural laborers to urban areas, resulting in aging, feminization, and lower agricultural productivity [20]. This led to disorganized operations and a decline in labor quality, which ultimately impacted grain output [21].

On the other hand, some scholars believed that households’ part-time farming behavior would have a positive effect on grain output. They suggested that by optimizing the production structure of rural households, rural households’ part-time farming behavior increased grain output [22]. Additionally, part-time farming households have made greater investments in agricultural equipment, raising their level of mechanization to be above that of full-time farming households and increasing their grain output per unit area [23]. Simultaneously, rural households’ part-time farming was also conducive to land leases, which aided in the large-scale production of grain [24]. Rural households with higher levels of part-time farming ultimately increased grain output, as they tend to be more technically proficient [25]. Furthermore, a higher degree of part-time farming brought higher off-farm income, which could greatly alleviate the financial constraints of rural households, improve the inadequate supply of rural credit, and enable rural households to supply agricultural materials on time, thus increasing grain output [26].

Although the existing literature provides reliable theoretical support for this paper, there is still a potential expansion at rural households’ part-time farming level. First, some researchers have misunderstood the terms “agricultural labor migration” and “rural households’ part-time farming” in the literature, which examines the effect of rural households’ part-time farming on grain output [16]. They also exaggerated the detrimental impact of rural households’ part-time farming on grain output [27]. Part-time farming highlights the dual roles of farmers and workers, as part-time farmers continue to produce agricultural goods [28]. Numerous current studies assume that part-time farming time is “homogeneous” and neglect to consider the farming time factor, omitting variations in the duration of part-time farming and time nodes that affect agricultural output [29]. Different scholars have examined the effects of time duration, part-time income, the number of part-time workers, and outside employment experience on grain output. This paper contends that, in contrast to previous research literature, the impact of rural households’ part-time farming behavior on grain output is not static, and that the measures of rural households’ part-time farming behavior on grain output used in existing studies vary greatly. In addition, it is necessary to broaden and enhance the scientific evidence regarding how rural households’ part-time farming affects grain output [30].

To focus on the impact of households’ part-time farming behavior on grain output over time, this paper profoundly divides households’ part-time farming into short-term and long-term dimensions, followed by an investigation of the effects of the degree of part-time farming on grain output based on existing research. Starting with the Propensity Score Matching method (PSM), the full-time farming households (control group) and part-time farming households (treatment group) in the sample are matched one to one. Consequently, this prevents the endogenous issue of an incomplete parallel trend in which the control group and the treatment group differ prior to the rural households’ part-time farming behavior in the double-difference model. To minimize estimation error in earlier research methods, this paper tests the net impact of rural households’ part-time farming on grain output using the double-difference model. If rural households’ part-time farming behavior significantly facilitates or inhibits grain output, further research will be carried out on the persistence of this effect and the specific mechanisms of action.

2. Theoretical Analysis Framework and Research Hypothesis

With the purpose of maximizing profit, part-time farming households transfer production factors that were initially invested in agriculture to off-farm fields. Farming households have to optimize their consumption of agricultural staples and market-purchased commodities while facing production constraint (land and technology), time constraint (where time is allocated to leisure, on-farm production, and off-farm employment), and cash constraint (based on anticipated prices of output, market prices, and market wage) [31,32].

According to the new economic theory in development economics, the redistribution of production factors between agricultural and off-farm fields is the nature of rural households’ part-time farming [33]. Theoretically, rural households’ part-time farming behavior will have both a “labor loss effect” and an “investment effect,” with rural households responding to a labor loss by scaling back agricultural production or boosting agricultural inputs by replacing labor with labor-saving factors of production [33].

In recent years, urbanization, cultivated land resources, agricultural technology, and agricultural income have gradually become more significant factors affecting grain output, which serves as a vital reason for the emergence of rural households’ part-time farming. First, with the existence of an urban–rural dual economic structure and the progression of urbanization and industrialization, wage income has gradually become the main source of income for rural households [34]. Second, rural households must rely on the non-agricultural sector for income growth, as depending solely on agricultural income is insufficient to achieve sustainable income growth due to the low income and relatively narrow high-profit space in agriculture [35]. Third, regarding cultivated land resources, in rural areas where there are some insufficient arable land resources for utilizing the surplus laborers, rural households can transition to part-time farming, which will not entirely disengage them from agricultural production [36]. Lastly, the progress of agricultural technology has reduced the time cost of agricultural input and increased the time input available for off-farm activities for rural laborers [37].

As a result of the rural households’ part-time farming decision, which has a short-term inhibition effect on grain output [38], the issues of “labor quality decline and agricultural input decrease” in the grain industry are becoming more and more apparent [39]. Nonetheless, as long as there is continuation of rural part-time farming, off-farm income can help expand rural households’ inputs in agricultural production [40]. In addition, rural households’ part-time farming behavior can promote the application of agricultural technology, improve the efficiency of agricultural production, and bolster the grain output capacity of part-time farming households [41]. It also gradually optimizes the industrial structure, improving the product service system of the grain industry and thereby promoting grain output [42]. In this regard, it can be concluded that the longer the rural households’ part-time farming behavior is in place, the more positive an impact it has on grain output.

In short, the growth of rural households’ part-time farming can be attributed to the urbanization process, agricultural income, age of laborers, number of laborers, cultivated land resources, and agricultural technology. The paper theorizes that rural households’ part-time farming ultimately affects grain output through the phenomenon of short-term decreases in agricultural labor input and subsequent increases in agricultural technology input after the persistence of rural part-time farming behavior. As a result, the following hypotheses are proposed in this paper:

Hypothesis (H1): 

There is a positive correlation between rural households’ part-time farming and grain output.

Hypothesis (H2): 

The higher the non-farm income of part-time farming rural households, the greater the promotion effect part-time farming has on grain output.

To further analyze the differences in the impact of rural households’ part-time farming on grain output, this study will also use panel data to compare how the effects of rural households’ part-time farming on grain output differ in the short and long run. Additionally, it will examine how agricultural labor input and agricultural technology input play an intermediary role in the process of rural households’ part-time farming, affecting grain output.

3. Model Construction and Variable Description

3.1. Data Source and Survey Description

3.1.1. Data Source

The study data was drawn from the “China Rural Fixed Observation Point Data”, covering the years between 2009 and 2015. The sample survey was based on farm households, including 23,000 farm households and 360 administrative villages dispersed among 31 Chinese provinces. The sample obtained continuous data through long-term follow-up surveys of fixed towns and households, using random sampling, household surveys, and other methods. The number of farmers, the state of the land, and the planting of agricultural products are topics covered in the survey, creating a comprehensive and representative sample. Based on the validation requirements of this paper and the exclusion of missing and abnormal data samples, a final valid balanced panel dataset of 19,927 observations was determined. To evaluate the impact of rural households’ part-time farming on grain output before and after part-time farming, this paper used the year 2012 as the node, when China increased the employment subsidy (a policy to increase farmers’ employment income) for part-time farming households, and selected 2431 rural households that began part-time farming in 2012 and have been part-time farming ever since as the treatment group. Rural households with part-time farming before 2012 were removed from the remaining sample, and the Propensity Score Matching (PSM) model was applied to match rural households with part-time farming behavior in the range of values. A total of 5629 sample observations were obtained after matching 2429 rural households with part-time farming behavior to the equivalent control sample (5629 is the number of samples in one year).

3.1.2. Definition of the Concept of Rural Households’ Part-Time Farming

According to the criteria of China’s Third Agricultural Census, a farming household classified as a full-time farming household is one whose household income is entirely derived from agricultural income, while a farming household defined as a part-time farming household is one whose household income is partly derived from non-agricultural income. Hence, this paper divided the sample farmers into two groups: full-time farming households and part-time farming households [43].

3.1.3. PSM-DID Method

Propensity Score Matching (PSM) and Difference-in-Differences (DID) are two statistical techniques combined in the PSM-DID model (DID). The DID is used to determine the impacts of policy shocks, whereas the PSM is used for the screening control of subjects for the treated subjects. In this paper, PSM-DID is employed to analyze how rural households’ part-time farming behavior affects grain output. In the literature, Propensity Score Matching (PSM) has commonly been applied in similar research [44]. In the process of analysis, Propensity Score Matching (PSM) was not selected because the PSM approach can only create “quasi-natural experiment” conditions by matching the dimension of observable variables and alleviating the problem of missing variables due to bias in modeling visible variables. However, PSM does not control for unobservable systematic differences between the “treatment group” and the “control group” when these differences exist. Hence, the PSM-DID model is appropriate. PSM-DID can control for unobservable, but not time-varying, between-group differences.

3.2. Model Construction

In order to verify the impact of rural households’ part-time farming on grain output, the level of grain output before and after rural households’ part-time farming can be directly compared separately. However, there may be some selection errors in this direct comparison, as numerous variables might affect rural households’ part-time farming. Therefore, this paper employs a double-difference model to test the net impact of rural households’ part-time farming on grain output for samples within the common value range after matching the corresponding treatment group and control group through the Propensity Score Matching (PSM) model to control for any possible selective bias.

To test hypotheses H1 and H2, the following model formula was used to verify the net effect of rural households’ part-time farming on grain output:

$$Y_{it}={\beta }_0+{\beta }_K\cdot Treat\cdot T^K+\sum {\beta }_X\cdot Control+r_i+{\epsilon }_i$$

(1)

In Formula (1), Y_it is the explained variable that measures the grain output level, and i and t represent the i th farmer and tth year, respectively. Treat is used to distinguish the treatment group and control group, T is used to distinguish before and after the test, and Treat·T is the interaction term used as the core explanatory variable to measure whether farmers have rural households’ part-time farming behavior. Control represents the control variables, including cultivated land resources, agricultural income, age of laborers, number of laborers, agricultural input, and agricultural technology. r_i represents individual fixed effects unaffected by time changes, and ε_it is a random interference term.

To further analyze the dynamic effect of the impact of rural households’ part-time farming on grain output, building on Formula (1), Formula (2) is as follows:

$$Y_{it}={\beta }_0+{\beta }_1\cdot Treat\cdot T^{}+\sum {\beta }_X\cdot Control+r_i+{\epsilon }_{it}$$

(2)

In Formula (2), Treat·T ^k is a dummy variable of the kth year for rural households with part-time farming behavior since 2012 and is assigned a value of 1. The year before rural households’ part-time farming behavior occurs is assigned a value of 0. K is used to measure the effect of rural households’ part-time farming on grain output in the kth year after rural households’ part-time farming.

To further analyze whether agricultural labor time input and agricultural technology input play a mediating role in the process of rural households’ part-time farming affecting grain output, the model formula is as follows:

$$\begin{gathered} Y_i=V_1+a_i{T}_i+b_{1i}{X}_{1i}+{\epsilon }_{1i} \\ {M}_i=V_2+c_i{T}_i+b_{2i}{X}_{2i}+{\epsilon }_{2i} \\ {Y}_i=V_3+d_i{T}_i+e_i{M}_i+b_{3i}{X}_{3i}+{\epsilon }_{3i} \end{gathered}$$

(3)

In Formula (3), Y represents grain output, T represents the part-time farming behavior of rural households, M represents agricultural labor input and agricultural technology input variables, X is the control variable, and V is a constant term. The three formulas are regressed using the OLS model.

The following model formula is used to test hypothesis H1 and the mechanism of rural households’ part-time farming behavior on grain output:

$$Control={\beta }_0+{\beta }_j\cdot Treat\cdot T^K+{\epsilon }_{it}$$

(4)

In Formula (4), to test the mechanism and factors of rural households’ part-time farming behavior that affect grain output, the control variable is taken as the explained variable, and the variable Treat·T^k is regressed.

3.3. Variable Selection

Explained variable. Y_it is the explained variable. The National Bureau of Statistics states that grain output consists of a sample survey of the sown area and unit area output. Therefore, to measure farm households’ grain output level, total grain output—which refers to the total amount of grain output in a year and per capita grain output—is selected as the explanatory variable. Grain is categorized into summer grain, early rice, and autumn grain according to the harvesting season, while cereals, beans, and potatoes are categorized according to crop varieties.

Core explanatory variables. The core explanatory variable, Treat·T, indicates whether a rural household exhibited part-time farming behavior or not. Treat is a dummy variable; if the sample rural household established part-time farming in 2012, it is assigned a value of 1. Otherwise, a value of 0 is given. T is a dummy variable to indicate the time period and is assigned 1 during and after 2012, and 0 before 2012. The twofold difference estimate, β₁, indicates the net impact of rural households’ part-time farming on grain output. If the i th rural household displayed rural part-time farming behavior in 2012, and t ≥ 2012, the Treat·T interaction term is assigned to 1. In contrast, it is assigned to 0.

Control variables. This paper uses off-farm income to reflect the degree of rural households’ part-time farming behavior in 2012 and beyond. In China, each household member is registered under either rural or urban residency. The urbanization process, an essential factor in determining whether rural households should establish part-time farming, is reflected in the share of the ratio of urban registration to total registration in a household [45]. As urbanization progresses, more rural households will form part-time farming behaviors, affecting grain output [46]. To reflect the agricultural income of various rural households, this paper selects the annual agricultural net income of a household of rural households. One of the critical drivers of part-time farming behavior is the agricultural income of rural households [47]. When agricultural income cannot sustain their daily needs, rural households frequently relocate from the agricultural field to the off-farm field, impacting grain output [48]. This paper chooses farmland areas managed by households to reflect the situation of cultivated land resources. One of the factors causing the development of part-time farming behavior is the direct relationship between the size of the cultivated land and grain output. Insufficient arable land resources leave a large number of idle laborers in rural areas, resulting in a waste of human resources [49]. In addition, the shortage of arable land resources often triggers a shortage in grain output. In this paper, the number of machines per hm² (hectare) of a household is selected to reflect agricultural technology, where the number of machinery includes both households owned as well as households leased. Within a unit area of 1 hm², the machinery consists of planting type machinery, harvesting machinery, etc. The efficiency of grain output is significantly influenced by the level of agricultural technology, and as agricultural technology advances, grain output efficiency increases [50]. According to the principle of supply and demand balance, the arable land resources are directly proportional to the number of machineries, as the general farmers use the machinery for agriculture, forestry, animal husbandry, and fishery in the form of leasing [51]. To reflect the average age of the laborers in a household and the corresponding number of laborers, this paper uses the age and number of laborers. In this study, people aged 18–65 with working ability, including hired workers and excluding students, are counted as laborers. The age and number of laborers are essential determinants of grain output efficiency. This paper uses grain planting inputs, such as labor, fertilizer, and seed inputs, to represent agricultural input [52]. Agricultural inputs also have a positive relationship with grain output [53].

Intermediate variable. Intermediary effect analyses are used in studies to investigate and study the mechanisms underlying an observed relationship [54]. In this study, agricultural labor input and agricultural technology input are used as mediating variables to investigate the mechanism by which part-time farming affects grain output. They are represented by the share of time spent by agricultural laborers in agricultural labor in the total labor time of rural households and the share of agricultural technology expenditures (including breeding technology, irrigation technology, and machinery technology) in total expenditures among part-time farming households, respectively [55].

Instrumental variable. Considering relevant research, “the distance between the village and the county” for rural households’ part-time farming was chosen as the instrumental variable [56].

Variable definitions and descriptive statistics are shown in

Table 1. Variable Definition.

Variable Name	Variable	Variable Definition
Total grain output of a household	TGH	Total grain output of a rural household (unit: ton)
Per capita grain output of household	PGH	Per capita grain output of a rural household (unit: ton)
Whether rural households’ part-time farming	WPH	Whether rural households engage in part-time farming
Degree of part-time farming	DP	Annual non-farm income of a household (unit: 10,000 yuan)
Urbanization process	UP	Ratio of urban registration to total registration in a household
Agricultural income	AI	Annual agricultural net income of a household (unit: 10,000 yuan)
Cultivated land resources	CLR	Farmland area managed by a household (unit: hm2)
Agricultural technology	AT	Number of machines per hm2 of a household (unit: set/hm2)
Age of laborers	AL	Average age of laborers in a household, the sample population is people aged 18–65 who are able to work (excluding students) (unit: years)
Number of laborers	NL	Number of laborers in a household, including hired workers (unit: person)
Agricultural input	AIT	Cost of grain planting input, include labor, fertilizer, and seed inputs (unit: 10,000 yuan)
Agricultural labor input	ALI	Agriculture labor input (hour/day)
Agricultural technology input	ATI	Agricultural technology investment, including investment in breeding, irrigation, machinery technologies, etc. (unit: 10,000 yuan)

and

Table 2. Descriptive Statistics.

Variable Name	All Farmers				Full-Time Farming Households				Part-Time Farming Households
	Mv	Sd	Max	Min	Mv	Sd	Max	Min	Mv	Sd	Max	Min
TGH	4.66	0.08	20.31	0.22	4.80	0.07	20.31	2.19	4.23	0.06	15.33	11.24
PGH	2.31	0.05	13.26	0.19	2.45	0.02	8.22	1.27	1.85	0.09	6.27	1.08
DP	1.24	0.11	10.29	0.37	0.00	0.00	0.00	0.00	2.12	0.19	10.29	0.37
UP	0.38	0.01	0.80	0.00	0.30	0.03	0.60	0.00	0.49	0.02	0.80	0.25
AI	1.73	0.15	5.31	0.13	2.03	0.08	5.31	1.26	1.49	0.09	2.34	0.13
CLR	0.39	0.01	1.34	0.02	0.42	0.01	1.34	0.12	0.33	0.01	1.02	0.02
AT	0.12	0.01	0.41	0.03	0.10	0.00	0.30	0.03	0.14	0.00	0.41	0.07
AL	50.22	0.77	72.22	32.67	52.56	0.56	72.22	49.33	65.24	0.44	68.30	32.67
NL	2.15	0.33	6.21	1.36	2.14	0.31	5.21	1.67	2.37	0.29	6.21	1.36
AIT	0.51	0.14	3.15	0.14	0.53	0.17	3.15	0.14	0.44	0.10	2.98	0.17
ALI	0.38	0.05	0.77	0.23	0.48	0.02	0.77	0.29	0.29	0.07	0.66	0.23
ATI	0.28	0.03	0.49	0.17	0.25	0.07	0.36	0.17	0.35	0.03	0.49	0.21

Note: 1. See Table 1 for definitions of the variables; 2. Mv is the mean value; 3. Sd is the standard deviation; 4. Max is the maximum; 5. Min is the minimum.

.

3.4. Descriptive Statistical Analysis of Variables

In this paper, we employed group analysis to examine the mean value, standard deviation, and maximum and minimum values of the same variables of different groups. According to Table 1 and Table 2, the average values of full-time farming households in terms of total grain output (TGH) and per capita grain output (PGH) of rural households were 4.80 tons and 2.45 tons, respectively, which were higher than 4.23 tons and 1.85 tons for part-time farming households. Regarding the degree of part-time farming (DP), the average annual off-farm income of rural households with part-time farming behaviors was CNY 21,200, and the maximum annual off-farm income of part-time farming of CNY 102,900 was far higher than the annual maximum agricultural income of CNY 53,100. Regarding the urbanization process (UP), the mean ratio of urban registration to total registration in a full-time farming household was 30%, which was lower than the 49% mean ratio for part-time farming households. In terms of agricultural income (AI), the average annual agricultural net income of full-time farming households was CNY 20,300, which is higher than the CNY 14,900 of part-time farming households. With regard to cultivated land resources (CLR), the average area of farmland managed by full-time farming households was 0.42 hm², larger than the 0.33 hm² of part-time farming households. Regarding agricultural technology (AT), the average number of machines per hm² of full-time farming households was 0.10 sets/hm², lower than the 0.14 sets/hm² of part-time farming households. Regarding the age of laborers (AL) and the number of laborers (NL), the average age and number of laborers in full-time farming households were 52.56 and 2.14, respectively, while the average age and number of laborers in part-time farming households are 65.24 and 2.37, respectively. In terms of agricultural input (AIT), full-time farming households had an input of CNY 5300, which was higher than the part-time farming households’ input of CNY 4400. Finally, on agricultural technology input (ATI), the average investment in agricultural technology by full-time farming households was CNY 2500, lower than the CNY 3500 of part-time farming households.

4. Empirical Analysis

4.1. Average Effect of Rural Households’ Part-Time Farming on Grain Output

Since it is impossible to observe the difference in grain output between the same rural household when there is part-time farming and without part-time farming, this paper first selected urbanization process (UP), agricultural income (AI), cultivated land resources (CLR), age of laborers (AL), number of laborers (NL), and agricultural input (AIT) as matching variables. The adjacent matching method in PSM matched the data according to a 1:1 matching in 2012, with a maximum distance of 0.05 between the test group and the matching group. The difference in differences (DID) for the matched samples was then calculated with a common trend assumption. As seen in Figure 1, the nuclear density curves of the treatment group and the control group differed notably before PSM matching. After matching, the nuclear density curves of the two groups of samples coincided, proving that the characteristic variables of the rural households in the two sample groups were relatively similar.

According to the matching balance test results in

Table 3. Balance test results of the treatment group and control group.

Variable		Mv		Sd	Ti
		Pg	Cg		T	p
UP	Before matching	0.30	0.49	33.9	5.24	0.000 ***
	After matching	0.32	0.57	2.1	0.48	0.639
AI	Before matching	2.03	1.49	13.9	−0.07	0.946
	After matching	2.05	1.52	−4.2	−1.71	0.087 *
CLR	Before matching	6.29	4.91	15.7	6.33	0.00 ***
	After matching	6.15	4.83	8.3	1.83	0.061 **
AT	Before matching	1.56	2.02	17.8	1.83	0.069 **
	After matching	1.51	2.05	6.3	0.39	0.692
AL	Before matching	52.56	65.24	26.3	2.69	0.008 ***
	After matching	52.33	65.22	2.9	0.58	0.617
NL	Before matching	2.14	2.37	−5.6	−0.88	0.413
	After matching	2.23	2.51	−9.6	−2.19	0.031 **
AIT	Before matching	0.53	0.44	17.7	1.56	0.109
	After matching	0.51	0.43	4.2	0.68	0.471

Note: 1. See Table 1 for definitions of the variables; 2. Mv is the mean value; 3. Pg is the processing group; 4. Cg is the control group; 5. Sd is the standard deviation; 6. Ti is the T inspection; 7. T is the T value; 8. p is the p value; 5. ***, ** and * are significant at 1%, 5% and 10% levels respectively

, the standard deviation values of the primary variables in the post-matching treatment group and control group were less than 10%, and the T-test results did not reveal any discernible differences in the matching variables between the two groups of samples after matching. The above analysis showed that the matching variables were reasonably selected. An ideal data sample was chosen for the following double-difference regression, which helped obtain the overall effect of rural households’ part-time farming on grain output.

To test research hypothesis H1, the total grain output of a household (TGH) and per capita grain output of household (PGH) were used as explained variables to estimate the average effect of rural households’ part-time farming on grain output, according to the model in Formula (1).

In

Table 4. Test results of the average effect of rural households’ part-time farming on grain output.

	TGH		PGH
	(1)	(2)	(3)	(4)
WHP	0.983 *** (2.571)	0.304 ** (1.973)	0.570 *** (4.257)	0.274 ** (3.183)
UP		−0.654 ** (−2.112)		−0.425 ** (−2.036)
AI		0.191 ** (2.486)		0.127 ** (2.043)
CLR		0.434 *** (36.451)		0.176 *** (22.472)
AT		0.134 *** (2.381)		0.010 *** (2.511)
AL		−0.196 *** (−8.716)		−0.115 *** (−6.922)
NL		0.133 ***(3.125)		0.124 ***(3.125)
AIT		0.974 *** (3.662)		0.456 *** (6.118)
Constant term	2.121 *** (9.081)	1.519 *** (9.901)	0.894 *** (7.890)	0.454 *** (4.885)
Number of observations	5926	5926	5926	5926
R2	0.113	0.671	0.104	0.453

Note: 1. See Table 1 for definitions of the variables; 2. **, and *** are significant at the 5%, and 1% levels, respectively; 3. Figures in brackets are t values.

, the OLS was run on the panel data with a fixed effect model, and the control variables were not included in the regression results of columns (1) and (3). For comparison, the regression findings in columns (2) and (4) include control variables. When the total grain output of a household (TGH) and per capita grain output of household (PGH) were taken as the explained variables, the coefficient of the interaction term “whether rural households’ part-time farming” (WHP) was significant at the 1% and 5% levels and was positive. The total grain output of a rural household (TGH) and per capita grain output of a rural household (PGH) of the treatment group after part-time farming were 0.983 tons and 0.570 tons, respectively, higher than those of the control group, according to the regression results in columns (1) and (3), without controlling for other variables that affect grain output. From the regression results in columns (2) and (4), with the addition of other variables that affect grain output, the total grain output of a rural household (TGH) and per capita grain output of rural household (PGH) of the treatment group after part-time farming were 0.304 tons and 0.274 tons, respectively, higher than those of the control group. The data allowed the conclusion that there was a positive correlation between rural households’ part-time farming and grain output, thus verifying research hypothesis H1.

When other control variables affecting grain output were accounted for, it could be seen that the greater the share of the ratio of urban registration to total registration in a household, the lower the total grain output of a household and per capita grain output of a household, i.e., the urbanization process (UP) negatively impacted grain output to some extent. Thus, if agricultural income (AI) was low, rural households would engage in non-agricultural activities through rural households’ part-time farming to boost household income, negatively affecting grain output in the short term. Regarding cultivated land resources (CLR), the farmland area managed by a household had a substantial positive effect on grain output, suggesting that the larger the area of farmland managed by a household, the greater the grain output. Regarding agricultural technology (AT), the number of machines per hm² of a household had a considerable positive impact on grain output, proving that a higher level of mechanization promotes grain output. The average age of laborers (AL) had a significant negative effect on grain output, whereas the number of laborers (NL) had a significant positive effect on grain output. Concerning agricultural input (AIT), grain output could be improved by increasing the grain planting inputs since the higher the grain planting input, the higher the grain output. In general, the effects of the control variables are higher for the total grain output of a rural household (TGH) than those of the per capita grain output of a rural household (PGH). This may be explained by the fact that the rural household is considered a unit and may consist of non-laborers, such as the elderly and children.

4.2. The Impact of the Degree of Rural Households’ Part-Time Farming on Grain Output

In this paper, 2429 rural households with part-time farming behavior were selected from 5629 sample observations to examine the effect of the degree of rural households’ part-time farming on grain output. In

Table 5. Test results of the average effect of the degree of rural households’ part-time farming on grain output.

	TGH	PGH
DP	0.127 *** (4.163)	0.216 ** (3.458)
Control variable	Yes	Yes
Number of observations	2429	2429
R2	0.513	0.629

Note: 1. See Table 1 for definitions of the variables; 2. **, and *** are significant at the 5% and 1% levels, respectively; 3. Figures in the parentheses are t values.

, the total grain output of a household (TGH) and the per capita grain output of a household (PGH) were taken as the explained variables. The regression results demonstrated that part-time off-farm income promoted the household’s total grain output and per capita grain output at the 1% and 5% levels, respectively, proving that the higher the degree of rural households’ part-time farming, the greater the promotion it has on grain output. Accordingly, hypothesis H2 was verified. Moreover, the promotion effect is higher for per capita grain output compared to total grain output. This may be due to higher efficiency and investment in agricultural technology inputs using off-farm income.

4.3. Dynamic Effect and Mechanism Analysis of Rural Households’ Part-Time Farming on Grain Output

The results in Table 4 show the average effect of rural households’ part-time farming on grain output but not the impact of rural households’ part-time farming on grain output over time. Therefore, Formula (2) was used in this paper to further analyze the dynamic effect of rural households’ part-time farming on grain output.

Table 6. Test results of the dynamic effects of households’ part-time farming on grain output.

	TGH		PGH
	(1)	(2)	(3)	(4)
Treat·T1	−0.494 *** (−3.017)	−0.176 *** (−2.272)	−0.241 *** (−2.046)	−0.172 *** (−2.388)
Treat·T2	−0.217 *** (−3.161)	−0.108 *** (−3.138)	−0.199 *** (−3.459)	−0.103 *** (−2.160)
Treat·T3	0.748 *** (5.691)	0.347 *** (3.854)	0.351 *** (6.501)	0.191 *** (5.289)
Treat·T4	0.524 *** (8.021)	0.426 *** (4.130)	0.436 *** (7.292)	0.262 *** (7.785)
Control variable	No	Yes	No	Yes
Constant term	2.178 *** (10.030)	1.315 *** (7.941)	0.883 *** (6.146)	0.423 *** (4.177)
Number of observations	5629	5629	5629	5629
R2	0.203	0.542	0.167	0.710

Note: 1. See Table 1 for definitions of the variables; 2. *** are significant at the 1% levels; 3. The number in the bracket is the t value; 4. Treat. T^k refers to the kth year after households’ part-time farming; 5. For convenience of comparison, the regression results without other control variables affecting grain output are also reported in the table.

shows the regression results of the dynamic effects of rural households’ part-time farming on grain output.

As observed in Table 6, the coefficients of the interaction terms Treat.T¹ and Treat.T² were negative and significant at the 1% level when the total grain output of a household (TGH) and the per capita grain output of a household (PGH) were used as the explained variables, regardless of whether the control variables were introduced or not. It could be proven that part-time farming behavior would inhibit grain output in the early stage of rural households’ part-time farming. The coefficients of the interaction terms Treat.T³ and Treat.T⁴ were both positive and significant at the 1% level, showing that the effect of part-time farming behavior on grain output changes from inhibition to promotion after part-time farming behavior persists for a period, and the promotion effect enhances as time increases. However, when other control variables affecting grain output were introduced, the dynamic effect of rural households’ part-time farming behavior on grain output was weaker. Notably, in the dynamic case, the effects of rural household’s part-time farming behavior were higher for the total grain output of a household compared to the per capita grain output of a household in the first four years.

To investigate the differences between the short-term and long-term effects of rural households’ part-time farming, Formula (2) was used to evaluate the impact of rural households’ part-time farming on grain output and its mechanism.

Table 7. Test results of the impact of a households’ part-time farming on the grain output mechanism.

	UP	AI	CLR	AT	AL	NL	AIT
Treat·T1	0.249 (0.441)	−0.008 (−0.036)	0.140 (0.013)	0.073 (1.162)	0.037 *** (4.892)	0.059 (0.135)	−0.253 *** (−3.679)
Treat·T2	0.440 (1.061)	−0.037 (−1.041)	0.424 (0.028)	0.135 *** (3.062)	0.058 * (1.842)	0.061 (0.153)	−0.272 (−0.984)
Treat·T3	0.530 *** (3.391)	−0.046 *** (−3.574)	0.477 (0.523)	0.204 *** (3.449)	0.061 * (1.821)	0.079 (0.132)	−0.306 (−0.121)
Constant term	1.002 *** (19.007)	0.089 *** (18.293)	0.727 (1.259)	0.335 *** (24.126)	0.083 *** (14.337)	0.678 (1.138)	0.314 *** (14.259)
Number of observations	5629	5629	5629	5629	5629	5629	5629
R2	0.278	0.182	0.223	0.305	0.173	0.129	0.166

Note: 1. See Table 1 for definitions of the variables; 2. *, *** are significant at the 10% and 1% levels, respectively; 3. The number in the bracket is the t value; Treat. T^k refers to the kth year after the farmer started part-time farming.

shows the regression results with grain output-related variables as the explained variables. Treat·T^k referred to the impact of rural households’ part-time farming on the various grain output-related variables in the kth year after rural households’ part-time farming. Table 7 shows that the Treat·T^k coefficients for the ratio of urban registration to total registration in a household and the annual agricultural net income of a household were not significant in the first two years after part-time farming but became so in the third year, indicating that part-time farming had a little immediate impact on the urbanization process (UP) and agricultural income (AI). The effect on the farmland area managed by a household was insignificant within the first three years after rural households’ part-time farming. One explanation for this would be that while the existing cultivated land resources (CLR) could unilaterally affect rural households’ part-time farming decisions, part-time farming could not alter the farmland area in this timeframe. The effect on the number of machines per hm² of a household was not significant in the first year. Still, it became significant and increasingly positive in the second and third years, indicating that the level of farm mechanization increased in the two to three years following rural households’ part-time farming. This increase is likely attributable to the farmers’ investment in their off-farm part-time income in agricultural technology, which in turn affected grain output. This is consistent with the overall higher average agricultural technology input (ATI) by part-time farming households, as shown in Table 2. The effect on the average age of the laborers (AL) during the three years was significant and increasingly positive. As shown in Table 7, the coefficient was the smallest in the first year, and in addition to the natural aging aspect, the phenomenon of youth labor shifting from agriculture to non-agricultural fields due to part-time farming persisted in the short term, leading to a lower quality of labor and ultimately affecting grain output. This is also consistent with the overall higher average age of laborers (AL) in part-time farming households, as shown in Table 2. In the first, second, and third years, the effect on number of laborers (NL) was insignificant, presumably because there would be no change in the number of household laborers in the short term. The effect on agricultural input (AI) was significant and negative in the first year, which proves that the issue of lower grain planting input occurred during the short term in part-time farming. However, after a period of part-time farming, part of the off-farm income was converted to grain planting input.

To sum up, rural households’ part-time farming behavior had a dynamic effect on grain output. In the short term, reducing agricultural labor input inhibited grain output. Nevertheless, through long-term investment in agricultural technology, it promoted grain output.

4.4. Instrumental Variable Analysis

This paper utilized instrumental variable analysis to resolve the endogenous problems brought on by missing variables and reverse causality. Considering relevant research, this paper selected “the distance between the village and the county” as the instrumental variable for rural households’ part-time farming. The cost of rural households engaging in part-time farming decreased with the proximity between the county and village. After accounting for other control variables, “the distance between the village and the county” had no direct impact on grain output. Hence, taking “the distance between the village and the county” as an instrumental variable was a suitable approach. In

Table 8. Estimation of the instrumental variable of rural households’ part-time farming and grain output (2SLS).

Variable	TGH	PGH
Distance between village and county	0.314 *** (0.957)	0.238 *** (1.152)
Control variable	Yes	Yes
Number of observations	5629	5629
F value	3.633 ***	2.863 ***

Note: 1. See Table 1 for definitions of the variables; 2. *** are significant at the 1% levels; 3. The number in the bracket is the t value.

, the impact of the total grain output of a household and the per capita grain output of a household were both still positive and significant at the 1% level, proving the credibility of this conclusion.

4.5. Parallel Trend Test

In addition to rural households’ part-time farming behavior, other aspects may also contribute to the differences in grain output. Moreover, the generation of this difference may not be linked to rural households’ part-time farming behavior, producing results inconsistent with previous research findings. To rule out this potential effect, a counterfactual parallel trend test was conducted, which advanced the rural households’ part-time farming behavior in the treatment group by one to two years. The specific operation added the time trend item based on Formula (2). To reflect the time trend, 2009 was taken as year 1, 2010 as year 2, and so forth. With 2012 as the boundary, the samples were divided into the kth year before rural households’ part-time farming, (before^k), and the kth year after rural households’ part-time farming, (after^k). At the same time, before^k, Current, and after^k, were used to replace the variable Treat·T^k in Formula (2). If the observed samples were from the first two years before rural households’ part-time farming behavior, the values before² and before¹ were allocated to 1, whereas the rest was given 0. If the observed sample was from the year rural households started to engage in part-time farming, the Current value was assigned as 1. Otherwise, a value of 0 was assigned. after¹ and after² were set to 1 if the observed sample was the data collected in the first one or two years after the rural households started part-time farming, respectively. Otherwise, they were assigned to 0. In

Table 9. Results of the parallel trend test.

	TGH		PGH
	(1)	(2)	(3)	(4)
Before2		0.003 (0.276)		0.011 (0.38)
Before1	0.025 (0.291)	0.037 (0.649)	0.014 (0.030)	0.018 (0.717)
Current	0.035 (0.751)	0.041 (0.574)	0.051 (0.714)	0.091 (0.320)
After1	−0.059 ** (2.135)	−0.054 * (1.748)	−0.068 * (1.721)	−0.066 * (1.803)
After2	0.089 *** (3.107)	0.072 *** (2.989)	0.093 ** (2.035)	0.088 ** (2.478)
Time	0.106 *** (14.290)	0.121 *** (13.741)	0.114 *** (16.231)	0.135 *** (15.277)
Control variable	Yes	Yes	Yes	Yes
Constant term	2.554 *** (9.007)	1.462 *** (8.293)	0.702 *** (4.259)	0.335 *** (3.261)
Number of observations	5629	5629	5629	5629
R2	0.667	0.667	0.667	0.667

Note: 1. See Table 1 for definitions of the variables; 2. *, **, and *** are significant at the 10%, 5%, and 1% levels, respectively; 3. The number in the bracket is the t value; 4. Before^k refers to the year k before the farmer started part-time farming, Current refers to the year in which the farmer started part-time farming, and after^k refers to the year k after the farmer started part-time farming.

, columns (1) and (3) indicate the assumed rural households’ part-time farming one year in advance, while columns (2) and (4) indicate the assumed rural households’ part-time farming two years in advance. The results in Table 9 show that the before^k coefficients were insignificant when rural households’ part-time farming was shifted one to two years in advance. Initially, it was established that the sample data’s change in grain output was mainly caused by the influence of rural households’ part-time farming behavior, excluding other factors. It was proven that, before 2012, when the treatment group was set as rural households with part-time farming, there was an insignificant difference in grain output change between the treatment group and the control group, meeting the parallel assumption. Once the treatment group was set as rural households with part-time farming, the after^k coefficients were significant, indicating that rural households’ part-time farming behavior, which started in 2012, significantly impacted grain output. The results passed the parallel trend test with robust conclusions.

4.6. Analysis of Intermediary Inspection Results

To investigate the underlying mechanism on how part-time farming affects grain production, we conducted an intermediary effects test according to Formula (3) using agricultural labor time input and agricultural technology input as the intermediate variables. The test results are shown in

Table 10. Regression results of the intermediary effect test.

Variable	ALI	ATI	TGH	PGH	TGH	PGH
DP	−0.131 *** (2.691)	0.369 *** (3.276)	0.264 *** (2.875)	0.199 *** (6.381)	0.253 *** (2.874)	0.192 *** (6.381)
ALI			0.133 ** (2.035)	0.057 ** (1.717)
ATI					0.208 *** (5.035)	0.048 * (1.699)
Control variable	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.304	0.213	0.514	0.413	0.511	0.408

Note: 1. See Table 1 for definitions of the variables; 2. *, **, and *** are significant at the 10%, 5%, and 1% levels, respectively; 3. The number in the bracket is the t value.

, where the dependent variables are presented in the first row. While the degree of part-time farming was positively related to agricultural technology input (ATI), it was negatively correlated with agricultural labor time input (ALI). Both passed the significance test at the 1% level. When the agricultural labor time input and agricultural technology input were used as intermediary variables, the degree of part-time farming passed the significance test at the 1% level for the total grain output of a household and the per capita grain output of a household. At the 5% level, there was a significant and positive correlation between the agricultural labor input (ALI) and the households’ total grain output and per capita output. As for agricultural technology input (ATI), it was significantly and positively related to the total grain output of a household and per capita grain output of household at the 1% and 10% levels, respectively. The results demonstrated that the agricultural labor input (ALI) and agricultural technology input (ATI) played an intermediary role in rural households’ part-time farming, affecting grain output. This implies that, as rural households adopt part-time farming, the decrease in labor force may affect grain output in the short term, but the increase in mechanization level may lead to the increase of grain output in the long term.

5. Conclusions and Discussion

5.1. Conclusions

This study examined the long- and short-term effects on grain output along with the relationship between rural households’ part-time farming and grain output using panel data of 5629 samples from the Chinese national rural fixed observation point data from 2009 to 2015. The intermediary effect model was also used in this research to explore the mechanism of rural households’ part-time farming behavior on grain output. First, the influence of part-time farming on grain output was promoted, as rural households’ part-time farming behavior was positively related to grain output. Second, the process through which rural households’ part-time farming dynamically affected grain output involved short-term inhibition, long-term promotion, and eventually, the promotion of grain output. Third, the higher degree of part-time farming (i.e., employment in off-farm fields) had a favorable impact on grain output. Fourth, the impact of the degree of part-time farming and grain output was mediated by agricultural labor input and agricultural technology input. Part-time farming inhibited grain output in the short run due to a reduction in agricultural labor input time and promoted grain output in the long run through the enhancement of agricultural technology input.

Since its reform and opening up, rural China has experienced significant changes due to the quick rise in non-farm activity. Even though farming is still the primary source of income for rural households, non-farm income is becoming more and more prominent in their overall revenue. Compared to families that only engage in agricultural activities, households that engage in off-farm activities have much higher average incomes. Given that the social security system for urban–rural integration is still being built, farmers will inevitably engage in part-time farming to meet their essential needs. By taking on part-time jobs, peasant households can maintain their farmland and the ability to ensure their survival while increasing their overall family income.

Part-time farming, however, also plays a significant role in the low level of land-intensive use in rural areas and the lack of rural transformation. The findings from this study can provide some guidance for future agricultural policy formulations to optimize labor and land resources and propel agriculture toward modernization and mechanization. First, policymakers should focus on agricultural technology training and the promotion of mechanization to improve agricultural production efficiency to reduce the wastage of low land usage in part-time farming households. This will tap into the promotion effect of agricultural technology input on grain output and bring the allocation of labor and land resources closer to the Pareto optimal state. Second, the government should also provide employment training and increase financial support for non-agricultural skills training for farmers. Broadening the off-farm employment prospects of farmers with targeted practical training and education could increase the income and resources of part-time farming households, which will, in turn, promote grain output in the long term. Next, policymakers must consider the formulation of land policy in a way that complements the current practice of part-time farming. One way is to relax the restrictions on land transfer and provide legal protection for land transfer to free up the surplus of rural agricultural labor that may be shackled to their land. This will promote the practice of part-time farming and the flow of labor to higher-income, non-agricultural production sectors for households to maximize their income and grain output in the long run. A second way is to encourage non-farming and part-time farming households to lease or contract out their land to full-time farming households to realize the scale effects of agricultural mechanization and large-scale production. Overall, optimizing labor and land resources and increasing investment in the building of agricultural infrastructure are necessary to advance the development of China’s agricultural sector.

5.2. Discussion

The results of this study have a broader focus and go beyond exploring the positive or negative effects of rural households’ part-time farming on grain output. Instead, they provide a long-term continuum of observations on how rural households move from full-time farming to part-time farming and examine the degree of part-time farming. There are contributions in the following two main areas:

Primarily, the impact of rural households’ part-time farming on grain output was analyzed using the PSM-DID method. As a result, it was discovered that rural households’ part-time farming is positively related to grain output and will increase grain output, which is consistent with the research findings of Zhong [57]. Hence, part-time farming is aligned with the government’s long-term objective of promoting grain output. However, this study found that the degree and length of part-time farming will have varied effects on the final outcomes. Consequently, while formulating grain output policies, the government should consider the impact of the degree and employment time of part-time farming.

Second, it was found that there is a mediating effect between agricultural labor input and agricultural technology input on the impact of rural households’ part-time farming on grain output. This study innovates by examining the impacts of agricultural labor input and agricultural technology input on rural households’ part-time farming in the context of China’s actual situation at different points in time. For instance, grain output will be negatively impacted by a reduction in agricultural labor input time in the short term due to rural households’ part-time farming. As time passes with rural households’ continuing part-time farming practices, rural households will gradually shift from suppressing grain output to promoting grain output by investing in agricultural technology. Consequently, the crucial role of technological inputs in increased grain output should be emphasized.

Although this study examined the mechanism and dynamic effects of rural households’ part-time farming income on grain output, potential research gaps still exist. First, different dimensions, such as the period, occupation, and location of off-farm part-time work, should also be used to measure the degree of part-time farming. The impact of part-time farming on distinct grain types may vary. However, due to data limitations, it only allowed exploration from the perspective of part-time farming income. Finally, numerous studies have been conducted to directly compare the impact on grain output before and after rural households’ part-time farming. However, since this direct comparison does not address endogeneity problems, such as self-selection bias and Omitted Variables Bias, the results are often subjected to large estimation errors, and the exact net effect of households’ part-time farming behavior on grain output could not be ascertained. Additionally, the dynamic effects of part-time farming were rarely considered in previous studies. As a result, future research should investigate the effects of rural households’ part-time farming on grain output from the perspectives of different part-time farming dimensions and varying grain types so that more detailed and profound conclusions can be obtained.