Farmers’ Non-Agricultural Income, Agricultural Technological Progress, and Sustainable Food Supply Security: Insights from China

Abstract

This study explores the intricate relationships among farmers’ non-agricultural income, agricultural technological progress, and sustainable food supply security using China’s provincial panel data from 2003 to 2021. Employing multidimensional fixed effects, moderated effects, and threshold regression models, the analysis yields several key findings. First, an inverted “U” shaped relationship exists between farmers’ non-agricultural income and sustainable food supply security, where food security is initially promoted but subsequently suppressed as income increases. Agricultural technological progress significantly enhances sustainable food supply security, and mechanism analysis confirms a moderating effect of technological progress on the relationship between non-agricultural income and food security. Second, regional heterogeneity analysis shows that the inverted “U” shaped inflection point for farmers’ non-agricultural income occurs earliest in the western region and main food-producing areas, with the strongest impact observed in the eastern region and main non-food-producing areas. Agricultural technological progress more significantly promotes sustainable development in the central and western regions and main food-producing areas. Third, threshold testing reveals a non-linear double threshold effect of agricultural technological progress on sustainable food supply security, with non-agricultural income thresholds at 6.759 and 9.427. Beyond these thresholds, the promotion effect of technological progress exhibits a “flat-steep-flat” S-shaped fluctuation. These findings suggest a need for a balanced approach to managing farmers’ non-agricultural income growth, maintaining incentives for sustainable food production and increasing financial and technological investments in central and western regions to secure China’s long-term food supply security.

1. Introduction

Food security is “the most important thing in the country”, and a good grasp of stable food production and supply is an important foundation to maintain economic development and social stability. By 2023, China’s total food production will remain stable at 1.3 trillion pounds for nine consecutive years, realizing a historic “twenty consecutive bountiful years”, and the effective supply of important grain products has been fully guaranteed. Although the Chinese people’s rice bowls are now firmly in their own hands, the future security of China’s food supply still faces numerous challenges. In the long run, the trend toward below-replacement fertility and the trend toward fewer children and aging as a result of increased life expectancy have resulted in a lower agricultural labor force and a higher number of persons supported on average. In the short to medium period, continued industrialization and urbanization, as well as the effects of urban agglomeration and the increase in non-agricultural income sources, attract a lot of rural laborers to look for jobs in cities. In 2023, China’s migrant workers outside the country reached 176.58 million, up 2.7% year-on-year, which reduced the actual number of farmers and the quality of agricultural personnel, bringing about a series of problems, such as land being left to fall into disuse and arable land shrinking. Food production depends not only on the area under cultivation and the number of laborers, but the level of sophistication of agricultural production technology also plays a crucial role. In the face of the food security crisis that could result, the central “No. 1 document” in 2023 proposes stabilizing food production and striving to improve the efficiency of food production, which hinges on the improvement of the level of agricultural production technology [1].

Since the reform and opening up period, the agricultural production in China has shown obvious “agricultural machinery-biased” technology progress through the of use machinery instead of labor factors, which has been not only the core driver of the sustained improvement in the efficiency of food production but also the key to the leaps and bounds in agricultural productivity over the decades [2]. In addition to the central government’s focus on the deployment, planning, and promotion of agricultural science and technology, farmers’ families’ spontaneous renewal of agricultural production equipment and data is also an important factor in agricultural technology progress, and this process is also affected by non-agricultural income and rural labor flow. According to the Lewis–Fei–Ranis Model (1961) [3,4], the emergence of rural labor mobility implies that urban incomes are higher compared to rural incomes, so, on one hand, for part-time rural laborers primarily involved in agricultural production, the rise in non-agricultural income provides them with the means to invest in more agricultural machinery and services. These workers can channel their higher off-farm earnings into household capital, which in turn enhances agricultural productivity and increases food output. On the other hand, if non-agricultural incomes become sufficiently substantial, the continued engagement in agricultural activities may lead to a decline in income efficiency and personal welfare. In such cases, farmers are likely to opt for leaving their land uncultivated or transferring land ownership, resulting in a reduction of their investments in farming equipment and ultimately lowering food production. Therefore, with the increase in non-agricultural income, the number of rural transfer labor to buy advanced agricultural equipment first increased and then decreased, and the corresponding grain output also increased and then decreased.

The research objective of this paper is to conduct an empirical test based on provincial-level panel data from 2003 to 2021 in China, examining the current situation and further clarifying the impact of farmers’ non-agricultural income and agricultural technological progress on the sustainable development of food supply during the urbanization process. Additionally, the research seeks to investigate the underlying mechanisms connecting these variables. On this basis, this paper summarizes policy recommendations to enhance the efficiency of food production and ensure the security of sustainable food supply. The methodologies employed in this paper include, but are not limited to, the multidimensional fixed effects model, the moderating effect model, the threshold effect model, and the Instrumental Variables Two-Stage Least Squares (IV-2SLS) model. These approaches have been selected for their rationality and widespread application in analyzing the relationships between variables [5].

2. Literature Review

Scholars have conducted relatively rich research on China’s sustainable food supply security. It has been found that many factors such as land use, population size [6], natural disasters [7], freshwater resources [8], the level of mechanization [9], food prices [10], policy changes [11], and labor force shifts [12] are important factors affecting the security of food supply. In recent years, urbanization has been developing at a high speed, non-farm employment has triggered an outflow of surplus rural labor, and farmers’ non-agricultural income has increased. At the same time, along with the increase in investment in agricultural research, agricultural technology has developed rapidly; therefore, farmers’ non-agricultural income and agricultural technological progress and their effect on the security of food supply-related research have formed a new direction of academic research.

The current academic discussion regarding the relationship between farmers’ non-farm income and sustainable food supply security presents two distinct views. One side holds a positive view, emphasizing that non-farm income directly and significantly affects the decision-making behavior of farmers in productive investment in agriculture and that non-farm income can guarantee sustainable food supply security [13]. The output of grain is closely linked to the input of production factors. In Nigeria, for instance, farmers obtain non-agricultural income through the labor market and reinvest this additional income into production factors such as land and capital, which maintains the sustainability of food production and reduces poverty [14]. In China, increased non-agricultural income increases the ability of left-behind family members to introduce new production technologies, purchase more capital-intensive agricultural factors of production, and has a significant impact on the inputs of agricultural machinery, fertilizers, and pesticides [15]. In addition, the increase in off-farm incomes has accelerated the outflow of land from land-owning farmers, the formation of economies of scale of land production, the realization of economies of scale [16], which helps to promote the large-scale operation of food, as well as to achieve increased food production and income. On the other hand, however, some scholars have argued that farmers’ non-agricultural income inhibits sustainable food supply security [17]. Research conducted on the Ethiopian Plateau found that non-agricultural income reduced farmers’ motivation to engage in agricultural activities, leading to a decline in agricultural production, including food crops [18]. Scott R, J et al. (1999) concluded from the study of labor mobility in China that the increase in non-agricultural income brought about by the increase in food production was smaller than the decrease in food production brought about by the loss of labor and that farmers’ non-agricultural incomes significantly reduced crop yields [19]; Chinese scholars have found through empirical tests that the share of non-farm income in farmers’ total income is negatively correlated with the willingness to grow food, and as the level of non-farming of farmers’ income increases, farmers’ incentive to plant food decreases [20]. Based on provincial panel data, Deng Zongbing et al. (2014) empirically analyzed the reasons for the formation of and changes in China’s regional pattern of grain, and they also concluded that the increase in farmers’ non-farm income plays a significant negative role on the regional pattern of food production [21].

“Developing agriculture by science and technology” is an important national strategy. The study of scientific and technological progress in agriculture and the security of sustainable food supply, as focal points of academic attention, have been receiving extensive attention and in-depth discussion both at home and abroad. In related studies, most scholars believe that scientific and technological progress is an important guarantee and a strong driving force to achieve increased food production, ensure food security, and help sustainable development [22]. From a theoretical point of view, according to Solow’s theory of economic growth, technological progress in a broad sense is defined as the residual part of economic growth after the exclusion of factor inputs such as labor, time, capital, natural resources, etc., which improves agricultural productivity and plays an important role in increasing and stabilizing food production [23]. Abbas A. et al. (2022) found that technological advances have changed the traditional pattern of cereal production, which in turn has contributed to increased income from cereal production [24]. Giannakas K. et al. (2001), based on unbalanced data from 100 wheat farms in Saskatchewan, Canada, similarly concluded that technological advances are the main source of productivity and wheat growth yield [25]. There is also a detailed distinction in the literature between technological progress and technological efficiency, concluding that technological progress has a more significant impact on sustainable food supply security than technological efficiency [26]. As research has progressed, other scholars have argued that technological advances have insignificant or inhibitory effects on sustainable food supply security [27]. For example, Tonukari NJ et al. (2010), through their study of biotechnology in developing countries, suggested that biotechnology can be used to improve agriculture but is not conducive to the development of food production capacity [28]. After examining the impact of agricultural technological progress bias on sustainable food supply security in China, Xu Taotao et al. (2017) found that land, labor, and capital-enhancing technological advances have a decreasing role in increasing food production in China in that order [29]. In more economically developed regions, the stronger the fit between technological progress and human capital, the more likely farmers are to plant high-yield cash crops, leading to the “non-food” structure of arable land and the planting ratio of food crops being weakened positively, which negatively affects the security of the food supply [30].

In summary, existing literature separately analyzed the impact of farmers’ non-agricultural income and agricultural technology progress on sustainable food supply security, and the studies are relatively independent. Second, in the technical progress measurement index selection, most scholars take the agricultural TFP or mechanization level as the key variable, and the selection of indicators is more solidified. Finally, from the perspective of variable relationships, existing articles have paid less attention to the dynamic and non-linear relationships of the variables of interest, among others. Therefore, this paper incorporates farmers’ non-agricultural income and agricultural technological progress into a unified research framework, and it sets farmers’ non-agricultural income as a threshold variable, aiming at exploring the non-linear “threshold” effect of agricultural technological progress on sustainable food supply security, thus putting forward high-quality suggestions so as to provide some reference for decision making to realize the security of the national food supply.

3. Analysis of the Influence Mechanism

3.1. Non-Linear Impact of Farmers’ Non-Agricultural Income on Sustainable Food Supply Security

On the one hand, farmers’ non-agricultural income is conducive to sustainable food supply security. The increased non-farm income of farmers leads to faster rural labor mobility [15] and the flow of labor to the city, resulting in the acceleration of the market-oriented flow of rural land, the outflow of labor triggers the increase in land management scale, and farmers expand their production scale through the “scale effect”, which can rapidly improve the efficiency of food production, and the effect of increasing food production and income becomes obvious [31]. In addition, an increase in farmers’ non-farm income implies an increase in the overall income level, and there is a significant positive effect of an increase in non-farm income on expenditures on agricultural machinery services [21], thus inducing farmers to increase their investment in machinery, fertilizers, and other factors of production to improve land productivity and promote sustainable food supply security. Conversely, as farmers’ incomes from non-agricultural sources continue to increase, the adverse impact on food supply security is increasingly coming to the fore as in Figure 1. Accompanied by the increase in non-agricultural income, in order to seek higher planting income, farmers are forced to invest more capital elements in modern food production again so that the cost of planting food is further expanded, and ultimately, the comparative return of planting food declines, which affects the incentive of farmers to plant food [26]. As young and middle-aged family members increasingly engage in non-agricultural industries, there is a corresponding decline in the traditional agricultural labor force. This shift results in an aging workforce, more part-time agricultural work, and lower levels of education among those remaining in agriculture, thereby reducing production efficiency and affecting food output. Additionally, with stable non-agricultural income and higher living standards, farmers gradually lose enthusiasm and focus on food production. Consequently, the number of farmers involved in food cultivation and the areas dedicated to planting continue to decrease. Incidents of land abandonment and desertion occur periodically, and individual investments in capital and technology also diminish, leading to further reductions in the number of food-growing farmers and cultivated areas. Simultaneously, as young and middle-aged family members engage in non-agricultural industries, the traditional agricultural labor force diminishes. This leads to phenomena such as an aging workforce, increased part-time employment, and low education levels [32], which reduce agricultural production efficiency and affect grain yield. Finally, along with the stable non-agricultural income and high quality of life standard, farmers gradually lose the enthusiasm and attention to produce food, the number of food-growing farmers and planting area continue to shrink, the phenomenon of land abandonment and desertion occurs from time to time, the individual capital and technical inputs are also decreasing [33], and the number of food-growing farmers and the area under cultivation continues to shrink. Ultimately, it is not conducive to sustainable food supply security. Accordingly, Hypothesis 1 of this paper is formulated as follows:

H1. 

There is an inverted “U”-shaped non-linear relationship between farmers’ off-farm income and sustainable food supply security, which is facilitated and then suppressed.

3.2. Moderating Effects of Technological Progress in Agriculture on the Relationship between Farmers’ Off-Farm Income and Sustainable Food Supply Security

Agricultural technology progress can positively regulate farmers’ non-farm income on the sustainable food supply security through an inverted “U”-shaped curve relationship, which is manifested in the steepening of the two, along with agricultural technological progress. The inverted “U” curve rising stage, that is, the early stage of the growth of farmers’ non-agricultural income. On the one hand, an increase in non-agricultural income provides financial support for farmers to adopt new production technologies and broaden their technological choices [34], which can promote the increase in food production and improve the quality of food, based on which the strengthening of agricultural scientific and technological inputs provides more means and technical support for modernized food cultivation so that the non-agricultural income, supported by the advancement of agricultural technology, has a stronger role in the promotion of sustainable food supply security. On the other hand, the growth of non-agricultural income attracts an outflow of rural labor, leading to issues such as labor shortages and an aging workforce in rural areas, as well as challenges in farmland cultivation and management. However, the promotion of agricultural technology enhances the labor productivity of those remaining in agriculture, partially alleviating the labor shortage problem caused by the rise in non-agricultural incomes. This, in turn, helps to increase food production, thereby strengthening the role of enhanced sustainable food supply security in the short term [12]. Simultaneously, advances in agricultural mechanization, improved germplasm resources, and other measures will accelerate land annexation and promote the intensification of arable land, large-scale production, and a substantial increase in food production efficiency. The downward period of the inverted “U” curve, that is, after farmers’ non-agricultural income, reaches a high level. As non-agricultural income continues to rise and farmers become more accustomed to urban life, they will increasingly divert their investments from agriculture and rural areas to higher-yielding urban and non-farm industries. Consequently, the rural labor force also migrates towards non-farm sectors, resulting in a trend of “weakening” surplus labor in rural areas [35]. Farmers gradually reduce the incentive to grow food; shift their preference toward planting higher-yield cash crops, such as fruits, flowers, etc.; and prioritise the application of advanced agricultural technology in cash crop planting, which results in the traditional food crop planting area shrinkage and the security of food supply inhibition being increased, resulting in the second stage of the inverted “U” curve tending to be relatively steep. Accordingly, this paper proposes Hypothesis 2:

H2. 

Agricultural technological progress positively moderates the inverted “U”-shaped relationship between farmers’ non-agricultural income and sustainable food supply security.

3.3. Threshold Effect of Farmers’ Non-Agricultural Income

According to different farmers’ non-agricultural incomes, the threshold effect of agricultural technology progress on sustainable food supply security presents a differentiated effect. Agricultural technological progress for increased food production is a gradual process. Initially, the main source of farmers’ income was agricultural income from family business income, whereas non-farm incomes accounted for only a small share of farmers’ incomes, so the non-farming of farmers’ income was at a low level. At this time, the agricultural model became a single smallholder production, where family management and food production were mainly based on human and animal power; the farmers’ low-income levels could only maintain the basic use of fertilizers and pesticides. For agricultural machinery replacement and germplasm resources development, there has been a lack of investment in capital factors, a lack of agricultural mechanization level, and low grain production efficiency. Therefore, the agricultural technology investment at this time had a small role in promoting grain increase. As China’s industrialization and urbanization process accelerated, the contribution of household operating income to farmers’ net incomes gradually decreased, higher-yielding non-farm employment attracted rural labor to non-farm industries, farmers’ non-agricultural income increased and reached a medium-to-high level, and rural labor exodus led to fewer grain-farming farmers. As well, there was a mechanism of forcing a shortage of rural labor, which triggered the increasing input of farm households to agricultural capital elements such as machinery, fertilizers, and pesticides [36]. Meanwhile, as non-agricultural income increases, farmers will invest more in agricultural technologies [37]. This investment, coupled with the effects of the economies of scale and industrialized operations, will significantly boost agricultural production efficiency and enhance sustainable food supply security levels. However, with farmers’ non-agricultural incomes reaching high levels and breaking through the critical value, the law of diminishing marginal efficiency comes into play in the case of increasing agricultural scientific and technological inputs, along with the diminishing marginal efficiency of agricultural technological progress for food production, wherein the promotion effect of technological progress on the sustainable food supply security tends to slow down. According to the above analysis, we propose Hypothesis 3:

H3. 

With the growth of farmers’ non-agricultural income, the promotion effect of agricultural technological progress on the sustainable food supply security is characterized by a “flat-steep-flat” stretching “S” (threshold effect).

4. Variable Selection, Model Setting, and Data Source

4.1. Variable Selection

4.1.1. Sustainable Food Supply Security (Lngy)

China’s food supply mainly relies on domestic food production self-sufficiency, wherein the relationship between sustainable food supply security and food production is closely linked. The higher the food production, the greater the security of the sustainable food supply. So, this variable has been measured according to food production and logarized.

4.1.2. Farmers’ Non-Agricultural Income (Lnnfi)

This indicator refers to the income obtained by farmers in the secondary and tertiary industries, and farmers’ non-agricultural income includes non-agricultural income from family business and wage income, i.e., wage income plus family business income minus net income from the primary industry in the per capita disposable income of rural households [38]. To eliminate the inflation influence on the price factor, the CPI index has been deflated with 2003 as the base period price. To eliminate heteroskedasticity as much as possible and reduce the variable fluctuations’ impact on the estimation accuracy of the model, the indicators have been logarithmically treated.

4.1.3. Technological Progress in Agriculture (Te)

At present, there are many methods for measuring the progress of agricultural technology in academic circles, but they have differences. Some scholars focus on the discussion of agricultural technological progress to save labor, and the general literature regards the level of agricultural mechanization as a substitute variable for agricultural technological progress [39]. Of course, the more popular measure today is the use of total factor productivity (TFP), also known as the “Solow surplus”, which focuses on analyzing technological progress in agriculture across the agricultural sector. However, in the real economy, economic growth requires a variety of factor inputs, and TFP cannot completely distinguish non-technical factors, so there are certain limitations in measuring technological progress. Considering the limitations of the above indicators, this paper draws on the methodology of Ma Yikun scholars in the Journal of International Trade Issues [40], and it selects the proportion of agricultural output and the labor force engaged in agricultural production, i.e., agricultural output per capita, as an indicator for measuring agricultural technological progress; increasing output per capita reflects higher levels of technological progress in agriculture.

4.1.4. Control Variables

As the factors affecting the sustainable food supply security are more complex, this article builds upon existing research to select control variables such as the level of economic development (Lngdp), the agricultural labor force (Lnap), the irrigation area proportion (Iap), the total machinery power (Lnmach), the soil erosion control area (Lnseca), and the reservoirs number (Lnnor) as control variables. Here, the economic development level is measured in terms of GDP per capita for each region; the proportion of the irrigated area is calculated as the ratio of the irrigated area to the sown area of crops; and the agricultural labor force is represented by the year-end agricultural population. To improve the fitness of the empirical model and reduce the influence of heteroscedasticity, except for the irrigated area proportion of the control variable, all other control variables have been processed logarithmically.

4.2. Model Construction

4.2.1. Base Regression Model

Among the linear regression models for panel data, the one we most commonly use is the fixed effects model. The estimation principle of the fixed effects model is used to explain the effect of the explanatory variables on the dependent variable by introducing individual dummy variables to capture the heterogeneity among individuals and for controlling for the inherent characteristics of the individual. This paper constructs the following econometric model:

$${Lngy}_{it}=C+{\alpha }_1{Lnnfi}_{it}+{\alpha }_2{Lnnfi2}_{it}+{\beta }_1{X}_{it}+{\eta }_i+{\mu }_t+{\epsilon }_{it}$$

(1)

$${Lngy}_{it}=C+{\alpha }_3{Te}_{it}+{\beta }_2{X}_{it}+{\eta }_i+{\mu }_t+{\epsilon }_{it}$$

(2)

Among them, Lngy is the explained variable, representing food production. Te and Lnnfi are the core explanatory variables, representing agricultural technological progress and the de-agriculturalization of farmers’ incomes. Lnnfi2 is the square term of Lnnfi, which is used to verify the non-linear relationship of non-agricultural income for the explanatory variables. Subscript i represents each province; t represents each year; $\alpha$₁, $\alpha$₂, $\alpha$₃, $\beta$₁, and $\beta$₂ are parameters to be estimated; C is the intercept term; $\eta$_i, and $\mu$_t are individual fixed effects and time-fixed effects; and $\epsilon$_it is a random disturbance term.

4.2.2. Moderating Effects Model

The main way to model moderating effects is to construct interaction terms. In Stata, the variables in the regression equation were tested for moderating effects mainly on the basis of the significance of the regression coefficients of the independent variables multiplied by the moderating variables. This paper constructs the following model to test the moderating effect of agricultural technological progress [41]:

$${Lngy}_{it}=C+{\alpha }_1{Lnnfi}_{it}+{\alpha }_2{Lnnfi2}_{it}+{\alpha }_3{Te}_{it}+{\alpha }_4{Lnnfi}_{it}\times Te+{\alpha }_5{Lnnfi2}_{it}\times Te+{\beta }_3{X}_{it}+{\eta }_i+{\mu }_t+{\epsilon }_{it}$$

(3)

where ${Lnnfi}_{it}\times Te$ represents the interaction term between farmers’ non-farm incomes and agricultural technological progress; ${Lnnfi2}_{it}\times Te$ represents the interaction term between the squared term of farmers’ non-farm incomes and agricultural technological progress; and $\alpha$₄, $\alpha$₅, and $\beta$₃ are the parameters to be estimated.

4.3. Explanation of the Data Source

The sample length of this paper is from 2003–2021, and the data for the relevant variables have all been sourced from the China Rural Statistical Yearbook, the China Statistical Yearbook, and the provincial and municipal statistical yearbooks spanning the years 2004 to 2022. In the cross-section selection, the four provinces of Tibet, Ningxia, Hainan, and Yunnan were deleted from the sample due to the seriousness of some of the missing data; in addition, the share of food production in the cities of Beijing, Tianjin, and Shanghai was found to be extremely low, accounting for only 1.8% of the national share of food production (in 2022), which lacks representativeness, so they were also excluded from the treatment, and the final data contain 24 provinces (Figure 2). Some of the data for individual provinces are missing, and they were filled in using interpolation. To eliminate the price fluctuations impact, the monetary data involved in this paper, such as the rural residents per capita net income, have been deflated using the rural consumer price index. The final statistical descriptions of the variables are shown in

Table 1. Statistical description of the full sample variables.

Variable Name	Variable Meaning	Number of Samples	Max	Min	Standard Deviation	Mean Value
Lngy	Sustainable food supply security	456	8.970	4.464	0.862	7.458
Lnnfi	Farmers’ non-agricultural incomes	456	10.654	5.729	0.955	8.356
Te	Agricultural technological progress	456	7.764	0.165	1.111	1.653
Lngdp	Level of economic development	456	11.828	8.218	0.713	10.295
Lnap	Agricultural labor	456	8.111	4.234	0.668	6.801
Iap	The proportion of irrigated area	456	1.005	0.147	0.153	0.401
Lnmach	Total mechanical power	456	9.500	5.678	0.756	7.921
Lnseca	Soil erosion control area	456	9.678	4.800	0.779	8.115
Lnnor	Number of reservoirs	456	9.554	5.024	1.145	7.615

Note: The data have been calculated and compiled by Stata17.

.

5. Empirical Analysis

5.1. Benchmark Regression

The multidimensional fixed effects model (reghdfe) was used for basic regression, and the regression results are displayed in

Table 2. Basic regression results.

Explanatory Variables	Model 1	Model 2	Model 3	Model4	Model 5
Lnnfi	0.339 *** (0.045)	0.320 *** (0.044)	1.379 *** (0.116)
Lnnfi2			−0.074 *** (0.008)
Te				0.070 *** (0.013)	0.050 *** (0.012)
Lngdp		−0.390 *** (0.053)	−0.387 *** (0.048)		−0.311 *** (0.057)
Lnap		0.210 *** (0.053)	0.044 (0.050)		0.305 *** (0.055)
Iap		−0.352 *** (0.120)	−0.141 (0.110)		−0.583 *** (0.121)
Lnmach		0.138 *** (0.037)	0.039 (0.035)		0.100 ** (0.041)
Lnseca		−0.161 *** (0.023)	−0.146 *** (0.021)		−0.127 *** (0.023)
Lnnor		0.216 *** (0.074)	0.046 (0.069)		0.278 *** (0.077)
-cons	4.874 *** (0.315)	5.827 *** (0.734)	5.013 *** (0.667)	7.203 *** (0.026)	6.291 *** (0.762)
time fixed	yes	yes	yes	yes	yes
area fixed	yes	yes	yes	yes	yes
N	456	456	456	456	456
R2	0.483	0.668	0.730	0.451	0.640

Note: ***, ** indicate significance at the significance levels of 1%, 5%, respectively.

. Model 1 yielded a positive coefficient on farmers’ non-agricultural incomes (Lnnfi) after controlling for time fixing and region fixing, indicating that the growth in farmers’ non-agricultural incomes significantly contributes to food supply security. In Model 2, after the inclusion of control variables, the core explanatory variable coefficient (Lnnfi) remained significant, and the conclusion that farmers’ non-agricultural incomes positively affect the sustainable food supply security was found to be robust after controlling for other factors affecting the sustainable food supply security. Model 3 added the square term coefficient of farmers’ non-agricultural incomes (Lnnfi2) for regression, and the results show that Lnnfi was positive and Lnnfi2 was negative, with both at the 1% significant level. This indicates that the impact of farmers’ non-agricultural incomes on the sustainable food supply security initially increased but eventually decreased, exhibiting an inverted “U”-shaped non-linear relationship. The reason for this is that the early increase in farmers’ non-agricultural income accelerated the large-scale operation of agriculture, and the market-based transfer of land can prompt the transfer of land from farmers with smaller marginal outputs to those with larger marginal outputs, optimize rural land resources allocation, and improve the efficiency of sustainable land use. And farmers’ non-agricultural incomes contributed to a marked increase in spending on agricultural machinery service expenditure, which has helped to strengthen farmers’ investment ability in agricultural production, as well as increase machinery investment, fertilizers, and other factors of production to raise the agricultural mechanization level, thereby improving the food production efficiency and maintaining sustainable food supply security. When farmers’ non-agricultural incomes exceeded 10,938.02 RMB, a turning point occurred in the curve, and the increase in non-agricultural incomes caused households to begin to reconfigure their labor resources, with a large number of male laborers migrating to the non-agricultural sector to work, with women, children and the elderly—who are not as productive—staying behind to run the land, thus exacerbating the “Low quality” of the rural labor force. The second point is that as farmers’ non-farm incomes rose further, accounting for a larger share of household income; as a result, farmers could rely on non-agricultural income to maintain household expenses and achieve a quality standard of living’ accordingly, farmers become less motivated toward traditional food production, and they may reduce the financial and technological inputs to food production or even curtail the cultivation of traditional food crops, which in turn will have a negative impact on the security of the food supply, validating Hypothesis 1. The results of Models 4 and 5 show that agricultural technological progress (Te) positively contributed to sustainable food supply security when controlling for time and region fixation, and after adding control variables, the effect between the two remained positive and significant. As the level of agricultural technology improves, it leads to an increase in agricultural labor productivity and land productivity, which in turn has a positive impact on sustainable food supply security.

As for the other control variables, the level of economic development had a coercive mechanism on the food production system, mostly at the expense of the arable land resources of the food production system, which therefore had a negative impact. Soil and water erosion control was mainly realized using intensive afforestation and revegetation, and the increase in the area under control led to a further reduction in the area planted with food crops. The increase in the agricultural labor force quantity and quality has raised the level of human capital in rural areas, thereby providing the necessary conditions for enhanced agricultural production. Additionally, the rise of machinery total power has effectively substituted labor, thereby improving the efficiency of food cultivation and production. The increase in the number of reservoirs has improved the capacity to ensure the replenishment of irrigation water sources, and the ability of food production to withstand natural disasters such as droughts and floods has been gradually strengthened, resulting in a steady increase in food production.

5.2. Robustness Test

The robustness test was mainly considered from three aspects: substituting variables, the shortening period, and an endogeneity test. As shown in

Table 3. Robustness test.

Explanatory Variables	Substitution Variable		Shorten the Period		Endogenous Test
	(1)	(2)	(1)	(2)	(1)	(2)
Lnnfi	1.533 *** (0.124)		1.357 *** (0.124)
L.Lnnfi					1.476 *** (0.113)
Lnnfi2	−0.084 *** (0.008)		−0.074 *** (0.008)
L.Lnnfi2					−0.078 *** (0.008)
Te		0.041 *** (0.013)		0.036 *** (0.013)
L.Te						0.034 ** (0.014)
control variable	control	control	control	control	control	control
Anderson canon. Corr. LM Statistics					261.211 (0.000)	286.462 (0.000)
Cragg–Donald Wald F Statistics					292.886 (7.030)	755.822 (16.380)
time fixed	yes	yes	yes	yes	yes	yes
area fixed	yes	yes	yes	yes	yes	yes
N	456	456	408	408	432	432
R2	0.745	0.647	0.716	0.614

Note: ***, ** indicate significance at the significance levels of 1%, 5%, respectively.

.

First, we replaced the explained variable. According to “China Rural Statistical Yearbook” statistical indicators, seasonal grain production has mainly included summer grain, early rice, and autumn grain. Among them, the quality of autumn grain is higher and accounts for the highest proportion of the total annual grain production. Therefore, the output of autumn grain was used as the substitution variable of the explained variable food production; we further tested the consistency of the effects of farmers’ non-farm income growth rates and agricultural technological progress on grain production. Regression results with the substituted dependent variable indicate that as farmers’ non-farm incomes increases, their impact on sustainable food supply security exhibits an inverted “U-shaped” pattern—initially promoting and then inhibiting it. Furthermore, agricultural technological progress exhibited a positive contribution to food supply security and was significant at the 1% level of significance. This was in the same direction as the significant results of the underlying regressions, and the conclusions were highly robust.

Second, we shortened the period. We restricted the sample to 2004–2020 to re-run the regression. The regression results with reduced sample size show that farmers’ non-agricultural incomes were significant at the significance level of 1%, and the relationship between farmers’ non-agricultural incomes and food supply security was non-linear, showing an inverted “U”-shaped trend of facilitating and then inhibiting; Agricultural technological progress had a positive contribution to food supply security under the significance level of 1%. This result is likely the result of the basic regression, which further shows that the result of the basic regression of the research is stable.

Third, we considered the existence of endogenous problems. Since the core explanatory variable and the explained variable may have a reverse causal relationship, to solve the possible endogenous problems, we used farmers’ non-farm incomes and lagged one-period agricultural technological progress as instrumental variables to conduct two-stage least squares estimation (2SLS) in order to address potential endogeneity issues. By testing the Anderson canon. Corr. LM statistic and the Cragg–Donald Wald F statistic, the selected instrumental variables were found to be reasonable and effective. There existed an inverted “U-shaped” relationship between farmers’ non-farm incomes and sustainable food supply security, while technological progress significantly promoted sustainable food supply security. These findings indicate the robustness of the results.

5.3. Heterogeneity Test

5.3.1. Regional Heterogeneity Test

Referring to the approach of Zhao Wei et al. [42], this article divided the sample data into three major economic belts of “East, Central, and West” according to regional divisions is shown in

Table 4. Regression results of samples by region (Eastern, Central, and Western regions).

	East Region		Central Region		Western Region
	(1)	(2)	(3)	(4)	(5)	(6)
Lnnfi	1.741 *** (0.458)		1.135 *** (0.235)		1.466 *** (0.232)
Lnnfi2	−0.099 *** (0.028)		−0.063 *** (0.015)		−0.086 *** (0.015)
Te		0.008 (0.029)		0.039 *** (0.013)		0.101 *** (0.025)
Lngdp	0.328 ** (0.164)	0.381 ** (0.177)	−0.317 *** (0.064)	−0.384 *** (0.060)	−0.121 (0.107)	−0.186 * (0.096)
Lnap	−0.056 (0.120)	0.293 *** (0.091)	−0.087 (0.055)	−0.032 (0.058)	0.495 *** (0.095)	0.652 *** (0.096)
Iap	−0.270 (0.236)	−0.535 ** (0.241)	0.809 *** (0.176)	0.579 *** (0.187)	−0.556 *** (0.187)	−0.403 * (0.205)
Lnmach	−0.107 (0.126)	−0.126 (0.122)	0.035 (0.030)	0.038 (0.033)	0.182 * (0.094)	0.105 (0.102)
Lnseca	−0.282 *** (0.065)	−0.273 *** (0.068)	−0.051 (0.040)	−0.164 *** (0.039)	−0.148 *** (0.035)	−0.116 *** (0.034)
Lnnor	−0.356 * (0.185)	−0.420 ** (0.181)	0.092 (0.075)	0.210 *** (0.072)	0.130 (0.161)	0.714 *** (0.149)
N	133	133	152	152	171	171
R2	0.678	0.631	0.932	0.923	0.819	0.791

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% significance levels, respectively.

. (The eastern region includes Hebei, Jiangsu, Liaoning, Shandong, Zhejiang, Fujian, and Guangdong; the central region includes Shanxi, Jilin, Jiangxi, Heilongjiang, Anhui, Henan, Hubei, and Hunan; the western region includes Neimenggu, Sichuan, Guangxi, Chongqing, Qinghai, Guizhou, Shaanxi, Gansu, and Xinjiang.). The results show that there was an inverted U-shaped relationship between farmers’ non-agricultural incomes and food supply security in the eastern, central, and western regions, it passed the 1% significance level test, and the regression coefficients of Lnnfi and Lnnfi2 in the eastern region were 1.741 and −0.099, respectively, which suggests that farmers’ non-farm incomes have the strongest impact on sustainable food supply security in the eastern region.

The farmers’ non-agricultural incomes and sustainable food supply security relationship in the three regions is shown in Figure 3. According to the calculations, the inflection points of the inverted “U”-shaped curve emerged when the non-agricultural income levels of farmers in three regions surpassed 6634.24 RMB, 8103.08 RMB, and 4914.77 RMB, respectively, with the occurrence of these inflection points following the sequence of the western, eastern, and central regions. Notably, the eastern region exhibited the strongest impact on grain supply security among these areas. This is due to the higher level of economic development in the seven provinces in the eastern region, where farmers have more non-farm employment opportunities and can easily earn more non-agricultural income, as well as the relatively advanced planting techniques and planting concepts in the eastern region, where farmers have the flexibility to invest more of their non-agricultural income in mechanized, large-scale, high-quality cultivation of the land. At the same time, farmers in the eastern region are the least dependent on traditional land food cultivation: once the non-agricultural income exceeds the inflection point value, through non-farm employment pathway to increase other incomes to meet the basic needs of life, farmers will reduce the cultivation of traditional food crops, so the eastern region of the growth of non-agricultural income on the degree of impact and sensitivity to food production ended up being the strongest. Compared to the eastern region, the western region, with its relatively underdeveloped economy, exhibits a high degree of dependence on traditional food cultivation, and it has fewer ways of obtaining non-agricultural income; it therefore yielded the smallest value of the turning point of non-agricultural income: once the non-agricultural income crosses the “turning point” of 4914.77 RMB, traditional farmers believe that replacing traditional farming with non-farm employment will secure their incomes while appropriately freeing up their family’s labor force. It was only when farmers cut back on growing food that the “tipping point” in the west occurred first.

The impact of agricultural technological advancements on sustainable food supply security also exhibited heterogeneity. The impact of technological progress in agriculture on the security of sustainable food supply was not significant in the eastern region, but it showed a positive contribution in the central and western regions, with a more pronounced impact in the western region in particular. This can be attributed to the fact that the central and western regions lag behind the eastern region in economic development, education, and technological advancement. Consequently, accelerated technological advancements can more effectively promote the improvement of grain production efficiency, thereby driving an increase in grain production and income. The western region is a traditional agricultural area, and the level of education and science and technology is more backward than that of the central region. Technological progress brings greater marginal benefits to food production in the western region, so the promotion effect is more obvious in the western region.

5.3.2. Tests for Heterogeneity in Functional Agricultural Regions

Referring to the approach of Zhao Wei et al. [42], the study provinces were categorized into grain-producing and non-grain-producing regions is shown in

Table 5. Regression results of samples by region (Grain-Producing Regions and Non-Grain-Producing Regions).

	Grain-Producing Regions		Non-Grain-Producing Regions
	(7)	(8)	(9)	(10)
Lnnfi	0.622 *** (0.134)		1.749 *** (0.147)
Lnnfi2	−0.038 *** (0.009)		−0.098 *** (0.010)
Te		0.029 *** (0.009)		0.016 (0.023)
Lngdp	−0.350 *** (0.045)	−0.280 *** (0.054)	−0.360 *** (0.074)	−0.168 * (0.095)
Lnap	−0.018 (0.051)	0.148 *** (0.052)	0.124 * (0.064)	0.406 *** (0.078)
Iap	0.860 *** (0.144)	0.856 *** (0.148)	−0.260 * (0.146)	−1.156 *** (0.166)
Lnmach	−0.017 (0.029)	0.003 (0.029)	−0.037 (0.058)	−0.016 (0.076)
Lnseca	−0.160 *** (0.031)	−0.233 *** (0.031)	−0.094 *** (0.029)	−0.011 (0.034)
Lnnor	−0.261 *** (0.068)	0.360 *** (0.065)	−0.247 *** (0.093)	−0.046 (0.119)
N	209	209	247	247
R2	0.927	0.922	0.747	0.565

Note: ***, * indicate significance at the 1%, 10% significance levels, respectively.

. It was found that farmers’ non-agricultural incomes promoted sustainable food supply security at the 1% significance level, the coefficient of the squared term of farmers’ non-agricultural incomes suppressed sustainable food supply security, and the non-agricultural income of non-food-producing regions had the strongest effect on sustainable food supply security.

As shown in Figure 4, the inverted “U”-shaped curve turned when farmers’ non-agricultural incomes in grain-producing and non-grain-producing regions exceeded 3640.95 RMB and 7331.97 RMB, respectively, with the greatest impact in the non-food-producing areas. The degree of intensification and mechanization of food production in non-main food-producing regions was lower than in main food-producing regions, and food incomes accounted for a smaller proportion of farmers’ incomes in these regions. Therefore, on the left side of the “turning point”, along with the growth of non-agricultural income, farmers strived to increase capital investment in technology, mechanization, fertilizers, pesticides, and other factors of production, which contributed to the rapid growth of food production. When non-agricultural income crossed the “turning point”, the income of farm households in non-grain-producing regions reached a high level, which triggered a massive exodus of rural laborers and the phenomenon of part of the land being left fallow. It also led to a change in the structure of arable land to planting higher-yielding cash crops, thus inhibiting grain production. Compared with major grain-producing areas, the contribution of non-agricultural income growth to grain production was even lower in non-major grain-producing area.

The estimation results of models (8) and (10) show that agricultural technological progress in main grain-producing regions had a significant contribution to sustainable food supply security, and agricultural technological progress in non-grain-producing regions had a non-significant impact on sustainable food supply security. This is due to the fact that the main grain-producing regions are more intensive and large-scale than the main non-food-producing areas, which makes them more conducive to the application of advanced agricultural technology, the promotion of management experience, and the concentration of inputs of factors of production, thus facilitating the efficient transformation of technological inputs and the realization of increased grain yields and increased incomes.

6. Analysis of the Moderating Effects of Technological Progress in Agriculture

To study the role of farmers’ non-agricultural income in the sustainable food supply security mechanism, this paper used the basic regression of agricultural technological progress as a moderating variable, and it used farmers’ non-agricultural incomes to construct the interaction term for empirical testing. Drawing on Haans [43] and other studies, the moderating effect of the inverted “U” curve was reflected in two aspects, i.e., the influence of the moderating variables on the shape of the inverted “U” curve and the location of the inflection point. We simplify and construct the quadratic function based on model (3) as follows:

$${Lngy}_{it}={\beta }_0+{\beta }_1{Lnnfi}_{it}+{\beta }_2{Lnnfi2}_{it}+{\beta }_3{Lnnfi}_{it}\times Te+{\beta }_4{Lnnfi2}_{it}\times Te+{\beta }_5{Te}_{it}$$

(4)

① The influence of the adjustment variable on the curve shape mainly depends on the coefficient $\beta$₄ of the interaction between the adjustment variable and the quadratic term. For the inverted “U” curve, when $\beta$₄ > 0, the curve shape is flatter; when $\beta$₄ < 0, the curve shape is steeper. The interaction coefficient between the adjustment variable and the quadratic term was significantly negative ($\beta$₄ = −0.024 < 0), indicating that agricultural technological progress positively moderates the inverted “U”-shaped relationship between farmers’ non-agricultural incomes and sustainable food supply security, making the inverted “U”-shaped curve steeper. Thus, Hypothesis H2 has been verified. ② The effect of the moderator variable on the turning point of the inverted “U” curve. When $\beta$₁$\beta$₄ −$\beta$₂$\beta$₃ < 0, the curve inflection point shifts to the left and vice versa.

Table 6. Empirical results of moderating effects.

Explanatory Variables	Model 6	Model 7	Model 8
Lnnfi	1.335 *** (0.099)	1.368 *** (0.113)	0.851 *** (0.180)
Lnnfi2	−0.080 *** (0.007)	−0.073 *** (0.007)	−0.042 *** (0.012)
Te	0.038 *** (0.011)	0.050 *** (0.010)	−1.701 *** (0.463)
Lnnfi×Te			0.406 *** (0.102)
Lnnfi2×Te			−0.024 *** (0.006)
Lngdp		−0.327 *** (0.048)	−0.278 *** (0.050)
Lnap		0.079 (0.050)	0.052 (0.051)
Iap		−0.155 (0.108)	−0.150 (0.108)
Lnmach		−0.018 (0.036)	−0.030 (0.036)
Lnseca		−0.164 *** (0.021)	−0.185 *** (0.022)
Lnnor		0.008 (0.068)	0.020 (0.067)
-cons	1.817 *** (0.392)	5.083 *** (0.649)	7.135 *** (0.821)
time fixed	yes	yes	yes
area fixed	yes	yes	yes
N	456	456	456
R2	0.626	0.745	0.756

Note: *** indicate significance at the 1% significance levels.

shows that $\beta$₁$\beta$₄ − $\beta$₂$\beta$₃ = −0.002196 (<0), indicating that the adjusted effect of agriculture technological progress has caused the inverted “U” curve to shift to the left and appear earlier.

This result is consistent with the theory: the initial period of growth in non-agricultural income has increased investment in agricultural technology, and technological progress has improved agricultural labor productivity, promoted land transfer and large-scale operation of arable land, and had a positive effect on sustainable food supply security. In the late stage of non-agricultural income growth, non-agricultural income reached a high level, the rural labor force outflow was serious, the surplus labor force for food cultivation enthusiasm was reduced, and increasing investment of advanced technologies in cash crops that yield greater economic returns—driving higher economic returns that made the grain area shrink—which is not conducive to the security of the food supply, thus strengthening the inhibitory effect of farmers’ non-agricultural incomes.

7. Threshold Utility Analysis of Farmers’ Non-Agricultural Incomes

The conclusion that farmers’ off-farm incomes affect sustainable food supply security through the moderating effect of technological progress in agriculture has been well studied in the previous section. Considering this further next, is the impact of technological progress in agriculture on sustainable food supply security a simple linear relationship? Does it show a non-linear relationship as farmers’ non-farm incomes increase? To further delve into the distinct mechanisms through which agricultural technological advancements impact food security under varying levels of non-farm income, the following empirical study has been conducted using a threshold model.

7.1. Construction of Threshold Regression Model

The threshold panel data model is a tool used to analyze non-linear relationships between variables, especially in cases where one variable reaches a specific critical value (threshold) and then has a significant effect on another variable. This type of model helps to reveal structural breakpoints in economic phenomena, leading to a more accurate understanding of the interactions between economic variables. To study the impact of technological progress in agriculture on sustainable food supply security at differentiated levels of farmers’ non-agricultural income, we drew on Hansen’s threshold panel data model method [44] and used farmers’ non-agricultural incomes (Lnnfi) as the threshold variable, and the model was set as follows:

$${Lngy}_{it}={\theta }_0{X}_{it}+{\alpha }_1{Te}_{it}\times I\left({Lnnfi}_{it}\le \omega \right)+{\alpha }_2{Te}_{it}\times I\left({Lnnfi}_{it}>\omega \right)+{\mu }_i+{\gamma }_t+{\lambda }_{it}$$

(5)

Among them, Lngy_it is the explained variable, $X_{it}$ is a set of control variables that affect the explained variables—including six variables of the level of economic development, agricultural labor force, the proportion of irrigated area, the total mechanical power, the soil erosion control area, and the number of reservoirs—${\theta }_0$ is the corresponding coefficient vector, Te_it represents the progress of agricultural technology, $I\left(·\right)$ is the indicator function— Lnnfi_it represents the non-agricultural income level of farmers—which is the threshold variable—$\omega$ is the specific threshold value, ${\mu }_i$ is the regional effect, ${\gamma }_t$ is the time effect, and $\lambda$_it represents the random disturbance term.

The above model assumes that there is only a single threshold, considering that there may be double or even multiple thresholds in reality. Taking the existence of a dual threshold as an example, the dual threshold is set as follows according to Formula (6):

$${Lngy}_{it}={\theta }_0{X}_{it}+{\alpha }_1{Te}_{it}\times I\left({Lnnfi}_{it}\le {\omega }_1\right)+{\alpha }_2{Te}_{it}\times I\left({\omega }_1<{Lnnfi}_{it}<{\omega }_2\right)+{\alpha }_3{Te}_{it}\times I\left({Lnnfi}_{it}>{\omega }_2\right)+{\mu }_i+{\gamma }_t+{\lambda }_{it}$$

(6)

In Formula (6), ${\gamma }_1$ and ${\gamma }_2$ are double thresholds, where ${\gamma }_1$<${ \gamma }_2$. Other variables are the same as those in Formula (5).

7.2. Threshold Test

First, the threshold effect test was performed on the Formula (6) model. According to the F statistical value in the table and the p-value obtained by the bootstrap method, the model was judged to have several thresholds. According to the analysis of the test results, Lnnfi passed the 5% significance test for the effects of the single threshold and double threshold of Te, but the triple threshold effect failed the significance test. According to the F statistic values and the p-value in the table to determine the existence of several thresholds in the model. According to the analysis of the test results, it was concluded that Lnnfi passed the 5% significance test for both the single and double threshold effects of te, but the triple threshold effect failed the significance test. Therefore, there were two threshold values in the model. This article chose the double threshold model for quantitative analysis.

7.3. Results of Model Parameter Estimation

Based on the threshold test, this article used dual thresholds for regression analysis, and they are shown in

Table 7. Threshold value and its confidence interval.

	Threshold Estimate	95% Confidence Interval
First threshold value	6.7593	[6.5976, 6.7907]
Second threshold value	9.4276	[9.3748, 9.4476]

.

As seen from

Table 8. Threshold affect test results.

Threshold Variable	Model	F Value	p Value	Bootstrap Frequency	Critical Value
					1%	5%	10%
De-agriculturalization of farmers’ income	Single threshold	36.22 **	0.0220	500	39.7733	26.5612	21.3931
	Double threshold	22.38 **	0.0320	500	25.5713	19.3881	16.6882
	Triple threshold	26.01	0.4000	500	68.5946	50.7198	43.9733

Note: ** indicate significance at the significance levels of 5%.

and

Table 9. Threshold regression results.

Variable	Coefficient	Standard Error
Lngdp	0.0995 ***	(0.0224)
Lnap	0.4979 ***	(0.0796)
Iap	−0.1835	(0.1137)
Lnmach	−0.0355	(0.0364)
Lnseca	−0.1766 ***	(0.0237)
Lnnor	0.0960	(0.0722)
${\mathrm{Lnnfi}}_{\mathrm{it}} \le {\mathsf{\omega }}_1$	0.0086	(0.0343)
${\mathsf{\omega }}_1 < {\mathrm{Lnnfi}}_{\mathrm{it}} < {\mathsf{\omega }}_2$	0.1370 ***	(0.0150)
${\mathrm{Lnnfi}}_{\mathrm{it}} > {\mathsf{\omega }}_2$	0.0839 ***	(0.0138)
_cons	3.8794 ***	(0.7947)
R2	0.645

Note: *** indicate significance at the significance levels of 1%.

, the farmers’ non-farm income levels exhibited a double-threshold characteristic in relation to sustainable food supply security. When farmers’ non-farm incomes fell below the first threshold of 6.7593, the impact of agricultural technological progress on sustainable food supply security failed to pass the significance test. As farmers’ non-farm incomes surpassed this first threshold (6.7593–9.4276), the regression impact coefficient became 0.1370, indicating a positive correlation at a 1% significance level. Once farmers’ non-farm incomes exceeded the second threshold of 9.4276, the influence coefficient of agricultural technological advancements on sustainable food supply security decreased to 0.083 yet still maintained a positive correlation at the 1% significance level. From the changing trend of the influence coefficient, from crossing the first threshold to crossing the second threshold, the influence coefficient changed from 0.0086 to 0.1370 to 0.0839. The coefficient increased and then became small again, and the impact trend was from gradual to steep and then to gradual, showing a stretched “S”.

When farmers’ non-agricultural incomes had not surpassed the first threshold, the impact was not significant between agricultural technological progress and grain production growth. The reason is that before farmers’ non-agricultural incomes reach a certain level, the level of non-agriculturalization of farmers’ incomes is relatively low, and farmers are in a backward traditional agricultural production method. Moreover, agricultural production activities mainly rely on human and animal power, and the main proportion of farmers’ household income is agricultural income at a very low level. Lack of financial investment in agricultural science and technology by households and low use of advanced technology and equipment in agricultural production. Therefore, during this period, although agricultural technological progress has a positive impact on food supply security, the effect is not significant. When farmers’ non-agricultural incomes crossed the first threshold but did not reach the second threshold, agricultural technology progress had a significant positive promoting effect on grain production growth, and the promotion effect increased significantly (from 0.0086 to 0.1370). The reasons are as follows: With the continuous increase in non-agricultural income, farmers will increase capital investment in agricultural technology and improve agricultural production efficiency to form scale effects and industrialized operations, thereby promoting grain production and sustainable food supply security. When the non-agricultural income of farmers crossed the second threshold, technological progress still had a positive effect on increasing grain production, and the promotion effect was significantly weakened (from 0.1370 to 0.0839) but was still more effective than before crossing the first threshold (0.0839 > 0.0086). With the non-agricultural income exceeding a certain level, the degree of de-agriculturalization of farmer income is extremely high. Farmers’ non-farm incomes increased as a share of household income and became the main source of total household income, and farmers relied on non-agricultural income to achieve a better standard of living. At this time, their attention to agricultural production was significantly reduced, affecting the input of agricultural technology so that the effect on the promotion of food production was much less effective than the previous stage of the threshold. At the same time, fertilizer and pesticides, biotechnology, and mechanization efficiency grew to at a higher level at this time, so the effect after the second threshold was much stronger than before the first threshold.

Table 10. Threshold passing status in 24 provinces in 2003 and 2021.

Threshold Interval	Province (2003)	Province (2021)
Areas with low non-agricultural income for farmers (Lnnfiit ≤ 6.7593)	Neimenggu, Jilin, Heilongjiang, Guizhou, Shaanxi, Xinjiang, Gansu, Qinghai
Areas with medium-level non-agricultural income of farmers (6.7593 < Lnnfiit < 9.4276)	Hebei, Shanxi, Liaoning, Jiangsu, Zhejiang, Guangdong, Anhui, Fujian, Guangxi, Jiangxi, Shandong, Henan, Hunan, Hubei, Chongqing, Sichuan	Anhui, Shanxi, Neimemggu, Liaoning, Jilin, Henan, Hubei, Heilongjiang, Hunan, Guangxi, Chongqing, Guizhou, Shaanxi, Gansu, Qinghai, Xinjiang
Areas with higher-level non-agricultural income for farmers (Lnnfiit > 9.4276)		Hebei, Zhejiang, Jiangsu, Jiangxi, Fujian, Shandong, Guangdong, Sichuan

describes the regional distribution pattern of farmers’ non-agricultural income levels in China’s 24 provinces in 2003 and 2021. It can be seen from the table that in 2003, eight provinces had not crossed the first threshold, which were mainly distributed in the northwest, northeast, and southwest regions. A total of 16 provinces had crossed the first threshold and not crossed the second threshold, mainly in the central and eastern regions. And no province crossed the second threshold. Over time, the number of provinces with lower levels of farmers’ non-agricultural incomes decreased. By 2021, all provinces had crossed the first threshold, and the provinces that had crossed the first threshold and did not cross the second threshold had increased to 16. The provinces that had passed the second threshold had increased to eight. According to the current results, the regions with high levels of non-agricultural income for farmers are mainly in the eastern coastal regions. Most provinces in the central, western, and northeastern regions are still in the middle-level regions of non-agricultural income for farmers. This shows that farmers’ non-farm incomes in most provinces of China plays a significant role in enhancing sustainable food supply security and that promoting the non-farming of farmers’ incomes is an effective way to maintain national sustainable food supply security.

8. Conclusions and Recommendations

This paper conducted an in-depth analysis of the influence exerted by farmers’ non-agricultural incomes and agricultural technological advancements on food supply security in China. Our study employed provincial panel data spanning from 2003 to 2021, encompassing 24 provinces in the country. Utilizing multidimensional fixed-effects models, moderating effects, and threshold tests, the following key findings are drawn:

From a national perspective, the role of farmers’ non-agricultural incomes in sustainable food supply security is first promoted and then suppressed, and progress in agricultural technology has a significant role in promoting sustainable food supply security. The mechanism test found that between farmers’ non-agricultural incomes and food supply security, there was an adjusted effect from agricultural technological progress. In terms of geographic heterogeneity, farmers’ non-agricultural incomes and the security of sustainable food supply security have an inverted “U”-shaped relationship in three different geographic regions, but the inverted “U”-shaped inflection point occurs in the order of western, eastern, and central regions. Progress in agricultural technology has significantly contributed to food supply security only in the central and western regions but shows no significant impact to food supply security in the eastern region. An analysis of the heterogeneity of agricultural functional regions yielded that the non-agricultural income in the main non-grain-producing regions had the strongest effect on food production and in the main grain-producing regions that progress in agricultural technology significantly contributed to food supply security.

Through the threshold model test, it is concluded that the non-agricultural income has significant dual threshold characteristics. The impact regarding agricultural technological progress on sustainable food supply security shows a “slow–steep \–slow” stretched “S”-shaped fluctuation with the increase in non-agricultural income. For the 24 provinces studied in this article, as of 2021, all provinces have crossed the first threshold, 16 provinces are between the first and second thresholds, and 8 provinces have crossed the second threshold.

Certainly, this study also has certain limitations, which are mainly reflected in two aspects: First, this research was conducted from a macro perspective based on provincial data, without considering the micro perspective of individual households, which has certain limitations. Further research from the micro perspective of households is worthy of our attention. Second, this paper mainly used grain output as an indicator to measure the sustainability of food supply security. However, the measurement of sustainable food supply security should not only include the indicator of grain output, but it must also encompass aspects such as food quality, variety, grain market trade, and ecological protection. In future research, these aspects can be fully considered, and the related research can be improved.

Based on the above research results and combined with the actual status quo of China’s agricultural development, this paper puts forward relevant suggestions as follows:

• Increase the intensity of scientific research investment in grain production and promote modern agricultural technologies. With the central goal of “Science and technology for food revitalization”, the in-depth development of green, high-yield, and high-efficiency grain and model research can be carried out to implement the strategy of “harvesting grain in technology”. First, increase investment in scientific research in agricultural cultivation, improve the innovation system of agricultural science and technology for increasing grain production, and focus on overcoming the technical bottlenecks that affect the increase in yield, quality, efficiency, and environmental improvement. Second, according to the needs of specific agricultural land, develop different advanced and applicable agricultural machinery types, and increase the testing and demonstration of advanced machinery and technology. Third, expand the scope of the implementation of green manure planting and straw that return to the field, promote measures such as reducing the use of pesticides and fertilizers, and effectively improve soil fertility. Fourth, accelerate the development of grain germplasm resources and the modern crop seed industry, build demonstration bases for new varieties of grain crops, and increase the coverage of improved varieties.
• Based on the diverse needs of different regions, attention should be paid to the differences in agricultural technology input. In the more advanced agricultural areas of the eastern region, the application of modern information technologies such as the Internet of Things and big data in agricultural production should be promoted. Measures should be formulated to support the matching of improved varieties, better farming methods, suitable land, and advanced machinery, creating favorable conditions for large-scale production and the development of high-level mechanization. In the relatively less-developed central and western regions, agricultural infrastructure such as irrigation, roads, and power grids should be improved. The promotion of water-saving irrigation, formula-based fertilization, and other applicable agricultural technologies will enhance the stable development and sustainable growth of grain production.
• Ensure the optimal number of labor forces for rural grain production. In line with the trends of urbanization and industrialization, we should adapt to the shift of surplus rural labor and ensure the optimal quantity of labor for grain production. We encourage the development of family farms and agricultural production cooperatives, the cultivation of new types of professional farmers, the strengthening of technical training and educational guidance, and providing a talent guarantee for technological advancement and grain production.
• To achieve the optimal allocation of factors in labor production, it is necessary to reform the urban–rural system and promote the integration of urban and rural areas. By facilitating the free flow of labor factors between urban and rural areas, as well as among different industries, we can realize optimized distribution. Encouraging increased income for farmers will effectively provide financial support for the increase in grain production factors. Additionally, through policy support, we should enhance the intensity of agricultural machinery subsidies for farmers, optimize the agricultural machinery service system, and strengthen the capacity for grain production.