Sustainable Food Supply from the Perspective of Paddy Ecosystem Elasticity: Policies and Implications

Abstract

Rice is the staple food for 2.5 billion people worldwide and most farmers depend solely on rice for their livelihood. This study estimates how paddy ecosystem elasticity and external human activity affect paddy ecosystem sustainable food supply. In particular, we analyzed how sustainable food supply is affected by three key domains of external factors—ecological factors, including the proportion of paddy Area (Are), per capita cultivated land area (Lan), and annual wastewater discharge per capita (Was); economic factors, including the agricultural economy level (Inv) and urbanization rate (Urb); and social factors, including the education of farmers (Edu) and rural medical level (Med). We use ANEV, or net paddy ecosystem services value per unit area, to assess the sustainable food supply, which not only represents the food supply quantity and quality, but also the sustainability of the food supply. Results from our panel and threshold regressions suggest that Lan and Urb have a threshold effect on paddy ANEV; Are, Was, Inv, and Edu have a linear negative correlation with ANEV; and Med has a positive linear correlation with ANEV. Based on our findings, we lay out a series of recommendations that may guide future formulation of policies on paddy ecosystem protection and sustainable food supply.

1. Introduction

Focusing on food security, the 2020 Global Food Policy Report proposes the building of a healthier, more inclusive, and more resilient food system. Food security refers to a social condition where all people have ready access to adequate, safe, and nutritious food for daily food and health needs, whether from a health perspective, or from the social and economic perspectives [1]. According to the United Nations, the global population will exceed 9.8 billion by 2050, and the demand for food will increase by more than 50% [2]. Food demand [3,4] and supply [5] merit attention in the context of major global pandemic and food crisis. The deterioration of farmland quality—for instance, the proportion of middle- and low-quality cultivated land in China increased from 67.35% in 2008 to 70.51% in 2015 [6]—and the rising factor prices have squeezed the grain profits and the enthusiasm of grain producers. Natural disasters and international trade also pose a serious threat to food security. Governments have taken a series of measures to ensure food security. For example, since 2013, the EU has put forward a direct subsidy policy through compulsory food production [7]. In 2017, India increased its food reserves to ensure food supply [8]. In 2020, China issued a notice on the resolute curbing of farmland “de-agriculturalization” (Notice of the State Council General Office on resolutely stopping the practice of “de-agriculturalization” of cultivated land) and in 2021, the country stressed the red line of 1.8 billion mu of cultivate land to ensure food security (The opinions of the State Council on comprehensively promoting rural revitalization and accelerating agricultural and rural modernization). Most of the policies and measures were food output-oriented. They cannot fundamentally improve the grain and agricultural income, nor can they promote a sustainable food supply from a long-term perspective. Therefore, many of those food policies have failed to adapt to the complex environment and new challenges [9], and the problems associated with abandoned cultivated land [10] and “de-agriculturalization” [11] still exist.

To achieve food security, it is essential to promote food quality [12] and green agricultural production [13]. In recent years, food policies have started to shift the focus from food output to food quality and security. Food policy can help in shaping healthy and sustainable food systems [14]. In 2017, the Economist used food supply adequacy, quality, and security, and the resilience of natural resources as indicators to evaluate food security. In 2019, China issued the Opinions on Further Deepening Ecological and Environmental Supervision Services to Promote High-quality Economic Development, and in 2020, issued the Regulations on ecosystem protection to explore subsidies for ecosystem protection and cultivated land restoration.

Although some extant studies have examined food quality and safety from the perspective of the green agricultural production mode [15], few of them focus on sustainable food supply from the perspective of the farmland ecosystem. The development of newer, more feasible, and food security policies needs the support of theoretical research, which, from the perspectives of supply and demand, looks into food quantity security in terms of food waste and loss, grain self-sufficiency rate, etc. [16,17,18] The paddy ecosystem is the agricultural ecosystem core. Rice is one of the most important foods in the world. The world rice planting areas are mainly distributed in the tropical rain forests and monsoon areas of East, Southeast, and South Asia. China and India are the world’s main rice producing areas, accounting for more than half of the world’s total rice output. In 2020, China’s total rice output was about 21.2 million tons, accounting for 27.7 percent of the world’s total rice output (76.6 million tons), ranking first among all countries in rice production (Based on data from Food and Agriculture Organization of the United Nations, GIEWS, quarterly global report No. 4 in 2021 of the Crop Prospects and Food Situation, https://www.fao.org/giews/reports/crop-prospects, accessed on 1 December 2021). The sustainable development of the paddy ecosystem directly affects the sustainable food supply capacity (quantity and quality of grain). It is essential to fundamentally improve the quality of cultivated land used for grain production. Paddy ecosystem elasticity (E), representing the stability and anti-interference ability, is an important indicator that can be used to measure the level of sustainable development of a paddy ecosystem [19]. In the present study, we explore the key factors underlying food sustainable supply from the perspective of elasticity. Findings will offer important guidance for the development of policies on global food supply and solutions to food supply security.

2. Literature Review

2.1. Food Consumption

In the context of frequent global pandemics and the highly likely outbreak of a global food crisis, food demand [3,4] and supply [5] have received a lot of attention recently. Research in food demand mainly focuses on food consumption and food waste [4,17]. Food waste and loss (FLW) is related to consumer preferences, harvest-to-sale time, and marketing strategies [20]. Waste of food largely occurs in commercial catering settings. Food waste has drawn a lot of attention in recent years; governments have realized that it is essential to develop public policies to reduce food waste [17]. Many countries have passed laws to limit food waste. The Food and Agriculture Organization of the United Nations (FAO) believes that food loss occurs mainly along the food supply chain, from harvesting to transportation, storage, and processing, and the main measures to reduce food waste should focus on improving basic conditions such as technology, facilities, or equipment. Policies on food supply should not only target reducing waste in food consumption [21], but also aim at minimizing waste during food production and logistics processes [22]. Food waste can jeopardize food supply; serious food waste may even endanger food security, resource and environmental security, and ecological security.

2.2. Food Supply

In addition to cutting food waste, maintaining a sufficient level of food production capacity is also crucial to securing the food supply. In the post-epidemic era, food supply is more essential, and it points directly to food security. Current research on food supply mainly focuses on quantity [23,24]. Relevant topics include food self-sufficiency rate, import and export [17], etc. The level of food self-sufficiency varies greatly from country to country. While Australia and France have a self-sufficiency rate of over 150%, Japan and South Korea have a rate lower than 50%. China is facing a growing food self-sufficiency problem; its food self-sufficiency rate has continuously declined to below 85% since 2015, and further down to 83.8% in 2017 [9]. Australia enjoys a high level of agricultural development; France produces a variety of quality wheat; Japan and Korea lead industrialization but have insufficient land resources; Africa has vast land resources, but a low agricultural development level [25]. A high rate of food self-sufficiency is not equivalent to food security [8]. Although food production is important for food security [26]—especially for large population areas [9]—other factors, including food quality and food supply capacity, are also crucial for food security. The existing literature largely emphasizes food quantity and pays little attention to food quality. Little research has been dedicated to the problem of food supply capacity, and the resource and environment problem for sustainable food supply [27].

2.3. Ecosystem Elasticity and Sustainable Food Supply

The current COVID-19 epidemic has serious implications for food supply, health, and the environment [28]. The pursuit of food quantity security inevitably increases the frictions between agricultural development and ecological environment protection [29]. As the food production environment deteriorates, the sustainability of food supply capacity will face severe challenges. It is thus necessary to shift agricultural production from quantity-driven to sustainability-driven [5]. Food security policies should also center on supply-side medium- to long-term strategies in the post-epidemic era [30].

According to FAO (2013), the food security system can be divided into four subsystems: food availability, economic and physical access to food, food utilization, and stability. Improving the paddy ecosystem sustainable food supply is one of the primary tasks to ensure global food security. The sustainable food supply is essential to the fulfilment of Sustainable Development Goals [31]. However, there is no unified opinion on how to ensure a sustainable food supply (quantity, quality, ecosystem service supply) of a paddy ecosystem [32]. E is not a simple biological factor such as soil organic matter and cation exchange quantities [33], but a variable affected by these external factors. E measures the extent of such effects, and it reflects the anti-interference ability and adaptability of a paddy ecosystem, as well as the total dynamic balance and recovery of production, ecosystem service, and social security. E has been proven to effectively measure paddy ecosystem sustainability, adaptability, and the anti-interference capacity of sustainable food supply [19]. As suggested by Yang et al. (2021), ANEV is the net value of paddy ecosystem services per unit area, which is influenced by ecological, economic, and social factors. A comprehensive evaluation of paddy ecosystem sustainability is conducive to strengthening agricultural ecosystem management [34] and promoting food supply. However, the research does not further analyze the mechanism of ecosystem elasticity and cannot provide meaningful guidance on policies for paddy ecosystem protection and sustainable food supply capacity.

While a lot of meaningful research has been conducted on food supply over the years, there are clearly some gaps that need to be filled. Firstly, extant research mainly focuses on the quantity of food supply rather than other dimensions such as sustainable food supply, quality analysis, and ecosystem protection. Secondly, how the grain ecosystem is affected by external factors is yet to be examined in depth. Thirdly, most policies on food security concentrate on cultivated land area, yield, prices, subsidies, and barely stress ecosystem protection and sustainable supply capacity from the perspective of sustainable development. This study fills these research gaps by proposing the use of E and ANEV to measure food supply capacity and exploring key external factors underlying sustainable food supply capacity of the paddy ecosystem from the perspective of ecological, economic, and social factors.

Food security policy should address the vulnerabilities [35] and resilience (2020 Global Food Policy Report) of paddy farming. From the perspective of long-term sustainability of food supply, this study focuses on sustainable food supply policy formulation based on E and ANEV. In this study, food supply does not only refer to food quantity and quality, but also the sustainable food supply capacity. Taking the Chinese paddy ecosystem as an example, we first analyze how ANEV is affected by the external ecological, economic, and social factors. Then we evaluate sustainable food supply capacity based on paddy ecosystem elasticity [19]. Finally, from the perspective of E and sustainability, we put forward targeted policy suggestions on the sustainable food supply of the paddy ecosystem.

3. Theoretical Analysis and Hypothesis Development

Ecosystem services include: (a) provisioning services such as food, by-products, fuel, and other harvestable goods; (b) regulating services such as pest and disease control, pollination, soil retention; (c) cultural services such as recreational and cultural benefits; and (d) supporting services, such as nutrient cycling, hydrological flow, and soil fertility that maintain favorable conditions for life on Earth [36]. The paddy ecosystem is the core agriculture ecosystem. Rice is a staple food feeding almost half of the world population. Rice accounts for nearly one third of the total croplands in China, making it one of the biggest producers of rice in the world [37]. With the rapid development of urbanization and industrialization, the problems of chemicals [38], soil deterioration [39] and water pollution are becoming increasingly serious [40], which is jeopardizing food quality and quantity and paddy ecosystem sustainability.

An appropriate sustainability index is essential while developing feasible policies that safeguard a sustainable food supply. E is the change and stability of ANEV; it estimates the paddy ecosystem sustainability and plays a major role in food supply. The net paddy ecosystem service value (NEV) is the difference between the paddy ecosystem service value (EV) and the paddy ecosystem services cost (EC); it is the net contribution provided by the paddy ecosystem. ANEV measures not only the food quantity and quality of the paddy ecosystem, but also evaluates the ability and stability of food supply [19]. E is the extent to which ANEV is affected by external ecological, economic, and social factors. It measures the adaptability and anti-interference of the paddy ecosystem, and evaluates the stability and sustainability of the paddy ecosystem [35,41]. Based on the ANEV, we evaluate the sustainable food supply capacity, the ability and sustainability of provisioning services (including food production, raw material ecology, and water supply), supporting, regulating, and cultural services [36] of the paddy ecosystem, and propose policy recommendations for paddy ecosystem protection and sustainable food supply.

Economic, ecological, and social factors are more systematic to the development of green agriculture. Our analysis of extant literature using ROST NAT, a text analytics software, reveals that the high frequency words of the ecosystem sustainability research include industry, resources, discharge, soil, industrial wastewater, chemical fertilizer and discharge, agricultural economy level, urbanization, GDP, population, education, and rural medical level. Given these high frequency words, data availability, and the nature of this study, we focus on ecological indicators, including the proportion of paddy area (Are), per capita cultivated land area (Lan), annual wastewater discharge per capita (Was); economic indicators, including agricultural economy level (Inv) and urbanization rate (Urb); and social indicators, including farmer education expenditure (Edu) and rural medical level (Med).

3.1. Ecological Factors Underlying ANEV

3.1.1. Proportion of Paddy Area and ANEV

Proportion of paddy area (Are) refers to the proportion of paddy to the total sown area. According to the China Statistical Yearbook, the national paddy accounts for 25.9% of the total sown area of crops in 2017. The crops mainly include food, cotton, oil, Chinese herbal medicine, vegetables and edible fungi, melons and strawberries, feed, and flowers. Grain mainly includes cereal, beans, and potatoes; cereal mainly includes rice, wheat, and corn, among which rice has the largest sown area. Based on the data on the paddy area and NEV in 30 provinces of China from 2008–2017, the paddy area is positively correlated to the NEV [42]. The niche priority occupation model (NPM) suggests that the population of the dominant species is proportional to the number of resources they possess [43]. Therefore, it is reasonable to argue that the proportion of paddy is positively related to the resources dedicated to it. The higher the proportion of paddy, the more NEV it provides, and the greater is the ANEV. In addition, the difference in the land coverage rate [44] and the grain planting structure affect its NEV. The increase of the paddy promotes rice production and the services supply (food production and raw material production). With the above analysis, we posit:

Hypothesis 1.

The proportion of paddy area has a positive impact on net ecosystem service value of paddy per unit area.

3.1.2. Per Capita Cultivated Land Area and ANEV

The proportion of cultivated land area to population (Lan) is an important factor likely to influence ANEV. Land is key to human survival and the quantity and quality of land represents its capacity to ensure a sustainable food supply. Based on the results from the second land survey, Lan in China is less than half of the world average (3.38 mu/person), the quality of land is bad, and the reserve land resources are insufficient [45]. The quantity and quality of cultivated land are not only related to food security, but also closely related to the health and stability of the Earth ecosystem. On the one hand, the land ecosystem provides the material support of farmland resources and water resources for grain production, but on the other, farmers’ excessive cultivation of the land and overuse of pesticides and fertilizer to meet the increasing food demand have caused the loss of ecosystem service values.

Given a certain level of Lan, excessive fertilizer application to promote grain production leads to soil degradation and foundation fertilizer decline. In the following year, farmers have to apply more pesticides and fertilizer to maintain the original production level, thus producing “dependence” on the pesticide and fertilizer, further deteriorating land quality and reducing the overall utilities of land ecosystem service. The excessive use of natural resources beyond the ecosystem capacity inevitably results in the decline of the ecological environmental quality [46].

As Lan increases, the yield can better meet people’s basic needs, and the application of pesticides and fertilizers will be reduced, weakening the positive impact of Lan on ANEV. Thus, we predict that Lan has a positive effect on ANEV, but as Lan increases, this effect diminishes.

Hypothesis 2.

Per capita cultivated land area has a positive and nonlinear impact on net paddy ecosystem service value per unit area.

3.1.3. Wastewater Discharge Per Capita and ANEV

Another important factor underlying ANEV is wastewater discharge per capita (Was), or the ratio of total wastewater discharge to population (tons/person). The paddy ecosystem demands a large volume of water resources, and the wastewater discharge leads to extensive pollution of water resources, which affects the supply, regulation, support, and culture services of the paddy ecosystem. Wastewater directly irrigates farmland, resulting in decreased soil quality and damage to agricultural production, which in turn leads to reduced production and quality of food [40]. Firstly, wastewater negatively impacts ANEV by reducing regulating and support services of the paddy ecosystem. Increased wastewater discharge not only leads to the decrease of harvest per unit area, but also affects environmental purification, soil conservation, nutrient circulation, and biodiversity functions. Secondly, wastewater discharge reduces the paddy supply service. Soil pollution resulting from wastewater discharge directly affects the quality of rice, leading to “cadmium rice” that contains high levels of unhealthy, heavy metals. Rice is the largest cultivated grain crop with the highest yield per unit area in China [47], and irresponsible wastewater discharge thus has even more serious implications: It contaminates the soil on a broad scale with heavy metals and causes soil compaction, which in turn causes the decrease of paddy ecosystem service values. Thirdly, wastewater discharge increases the paddy ecosystem and pollutes the air, water, and soil, thus reducing the ANEV. Based on the above analysis, we predict:

Hypothesis 3.

Wastewater discharge per capita has a negative impact on net paddy ecosystem service value per unit area.

3.2. Economic Factor Underlying ANEV

3.2.1. Agricultural Economy Level and ANEV

Agricultural economy level (Inv) refers to the percentage of overall GDP accounted for by agriculture. Land protection policies are different in the developed and developing countries [48]. The relationship between ecosystem and economic development is a fundamental area of study in regional development, spatial layout optimization, and sustainable development [49]. Based on the environmental Kuznets inverted U curve theory, economic growth is often achieved at the expense of natural resources, especially in the developing stage. With the advancement of economic level, agriculture’s proportion will gradually decrease. According to the contribution of agriculture to national GDP across the world in 2016, India ranked the highest, at 16.28%. The contribution rate of China’s agriculture to its GDP ranked second, at 8.56%, and the world average level was 3.55% (the data comes from the network). China’s agriculture has not been truly modernized, and the overall level of agricultural development is low [50]. Kuznets inverted U theory assumes that the Inv has a negative impact on ANEV.

Hypothesis 4.

Agricultural economy level has a negative impact on net paddy ecosystem service value per unit area.

3.2.2. Urbanization Rate and ANEV

Urbanization rate (Urb) refers to the percentage of total urban population at the end of a given year. Rapid urban expansion in most developing countries consequentially contributes to environmental change [51]. China’s rapid urbanization has accelerated the consumption of high-quality farmland [52] and sounded the alarm over long-term food security [48]. China’s urbanization advances with the synchronous or coordinated development of big cities and small towns, and it is marked by obvious government intervention—local governments use low-cost land expropriation to obtain large land resources for urban development [53]. The rapid urbanization, which demands construction land [11], is often done at the expense of agricultural land resources. In recent years, due to the advancement of urbanization and agricultural structure adjustment, China’s arable land has diminished [26]. Studies [54] have shown an inverted U-type curve relationship between urbanization and the ecosystem services value.

The impact of Urb on the paddy ecosystem may be complex. On the one hand, urbanization reduces farmland and rural labor [55], causes land degradation, water shortages, the aggravation of agricultural and rural environmental pollution, and thus has an inhibitory effect on food security. On the other hand, urbanization promotes economic and technological advancement, accelerates agricultural mechanization, variety improvement, and scientific fertilization and irrigation, and therefore helps improve the efficiency and stability of the paddy ecosystem. Based on the above analysis, we predict:

Hypothesis 5.

Urbanization rate has a nonlinear impact on net paddy ecosystem service value per unit area.

3.3. Social Factors Underlying ANEV

3.3.1. Farmer Education Expenditure and ANEV

Edu refers to the proportion of rural expenditure on culture, education, and entertainment in overall consumption expenditure. Generally speaking, the higher the education one receives, the more likely one would be to value, and financially support, environmental protection [56]. In rural areas, however, those who receive education are the farmers’ children, not farmers themselves. Therefore, the increase in expenditure on farmers’ education would contribute little to environmental awareness. Instead, because farmers bear the burden of paying for the education, the increase in education expenditure may eventually force them to use more pesticides and fertilizers in order to improve production and increase their income. The paddy ecosystem costs incurred by the excessive application of pesticides and fertilizers may exceed the ecosystem service values. Based on the above analysis, we posit:

Hypothesis 6.

Farmer education expenditure has a negative impact on net paddy ecosystem service value per unit area.

3.3.2. Rural Medical Level and ANEV

Agricultural production is associated with health awareness [13]. Physical health may foster environmentally friendly behavior—people who value their physical health tend to be more environmentally conscious [56]. In this study, we measure rural medical level (Med) with the number of doctors per 1000 people. With a higher Med, farmers would have more opportunities to receive necessary medical treatment and regular physical examinations, which in turn may enhance their health awareness. People with higher health awareness would naturally value the quality of food and the protection of the ecological environment. Based on that, we predict:

Hypothesis 7.

Rural medical level has a positive impact on net paddy ecosystem service value per unit area.

4. Materials and Methods

4.1. Panel Data Model

With panel data on paddy ecosystems from 30 Chinese provinces from 2008–2017, we examine the influence of ecological, economic, and social factors on sustainable food supply. In particular, we construct a panel data model to verify the effects of Are, Lan, Was, Inv, Urb, Edu, and Med on ANEV:

$${\mathrm{ANEV}=\mathsf{\beta }}_0{+\mathsf{\beta }}_1{\mathrm{Are}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\beta }}_2{\mathrm{Lan}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\beta }}_3{\mathrm{Was}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\beta }}_4{\mathrm{Inv}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\beta }}_5{\mathrm{Urb}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\beta }}_6{\mathrm{Edu}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\beta }}_7{\mathrm{Med}}_{\mathrm{i},\mathrm{t}}{+\mathsf{\upsilon }}_{\mathrm{i},\mathrm{t}}$$

(1)

where ANEV is the explained variable, and Are_i,t, Lan_i,t, Was_i,t, Inv_i,t, Urb_i,t, Edu_i,t, Med_i,t are the explanatory variables; i and t represent each region and year respectively; υ is a composite error item.

4.2. Variable Selection

4.2.1. Explained Variable

ANEV is the net value of ecosystem services provided by a paddy ecosystem per unit area, which reflects the sustainable food supply capacity. ANEV not only accounts for the values of food and other material supply services of a paddy ecosystem, but also considers the values of adjustment, support, and landscape services, as well as the negative impact generated in the process of artificial planting of paddy. Based on data availability, we define ANEV as:

$${\mathrm{ANEV}}_{\mathrm{i},\mathrm{t}}=\frac{{\mathrm{EV}}_{\mathrm{i},\mathrm{t}}-{\mathrm{EC}}_{\mathrm{i},\mathrm{t}}}{{\mathrm{s}}_{\mathrm{i},\mathrm{t}}}$$

(2)

ANEV_i,t is the net value of paddy ecosystem services per unit area for the t year of i area; EV_i,t represents the paddy ecosystem services value in the i area in year t; EC_i,t refers to the t-year cost of paddy ecosystem services in the i area; s_i,t is the paddy area expressed in ha. ANEV, measuring the sustainable paddy supply capacity, is essential in paddy ecosystem protection and sustainable food supply.

4.2.2. Paddy Ecosystem Elasticity

According to the elasticity theory proposed by Wang et al. [57], the paddy ecosystem elasticity represents the extent to which ANEV is affected by external factors. The following is the formula for calculating e [19]:

$$\mathrm{e}={\displaystyle \sum _{\mathrm{n}=1}^2\left[{\displaystyle \sum _{\mathrm{i}=1}^{10}\frac{\left|{(\mathrm{ANEV}}_{\mathrm{i}+1}-{\mathrm{ANEV}}_{\mathrm{i}})\right|{/\mathrm{ANEV}}_{\mathrm{i}}}{{(\mathrm{V}}_{\mathrm{i}+1}-{\mathrm{V}}_{\mathrm{i}}{)/\mathrm{V}}_{\mathrm{i}}}}\right]}·{\mathrm{W}}_{\mathrm{Vn}}$$

(3)

E indicates the change of ANEV influenced by external factors; i is the year; V are the ecological–economic– social factors; W represents the direct and indirect effects of ecological–economic–social factors on a paddy ecosystem, W_V1 is the proportion of direct effect of V index on unit ecosystem of paddy, and W_V2 is the proportion of indirect effect of V index. A bigger e value indicates that external factors have greater impact of on ANEV and the paddy ecosystem is less stable [19]. The paddy ecosystem elasticity evaluates the sustainability of a paddy ecosystem and assess the sustainable food supply capacity.

4.2.3. Explanatory Variables

The explanatory variables are ecological, economic, and social factors that comprehensively influence the stability and sustainability of the paddy ecosystem. The definitions of the factors are included in

Table 1. Explanatory variable definition.

Type	Variable Name	Code	Variable Definition
Ecological	Proportion of paddy area	Are	Proportion of paddy to total sown area
	Per capita cultivated land area	Lan	Proportion of cultivated land area to population
	Annual wastewater discharge per capita	Was	Proportion of total wastewater discharge to population (tons/person)
Economic	Agricultural economy level	Inv	Percentage of GDP in primary industry
	Urbanization rate	Urb	Percentage of total urban population
Social	Education of farmers	Edu	Proportion of rural expenditure on culture, education and entertainment to consumption expenditure
	Rural medical level	Med	Number of doctors per 1000 population

.

4.3. Data

We collected paddy ecosystem data from 30 provinces, autonomous regions, and municipalities in China from 2008 to 2017. We exclude Qinghai, Hong Kong, Macao, and Taiwan from our analysis because the required data are not available. The ANEV data are from the accounting results of the “Evaluation of the net value of paddy ecosystem services in China” [42]. E data comes from “An Ecosystem Elasticity Perspective of Paddy Ecosystem Sustainability Evaluation: The Case of China” [19]. Other data are mainly from the China Statistics Yearbook from 2009 to 2018 [58], China Rural Statistics Yearbook [59], and China Statistics Yearbook of environment [60].

Table 2. Descriptive statistics of the main variables.

Variables	Obs	Mean	Std. Dev.	Min	Max
ANEV	300	3379	1376	1359	6660
Are	300	0.19	0.18	0.0002	0.73
Lan	300	0.11	0.08	0.01	0.42
Was	300	47.97	17.73	11.67	107.79
Inv	300	0.101	0.079	0.003	1.12
Urb	300	0.55	0.14	0.22	0.90
Edu	300	0.09	0.03	0.01	0.15
Med	300	5.67	1.79	2.37	15.46
e	300	7.94	15.50	0.22	86.07

reports the descriptive statistics of the explained variable and explanatory variables, including number of observations, mean, standard deviation, minimum and maximum of each variable.

5. Results

5.1. Panel Regression Results

We use panel regression to estimate the impact of each factor on ANEV. The results are reported in

Table 3. Panel regression results of each indicator on ANEV in 30 provinces.

ANEV	Coef.	Robust Std. Err	t	p > |t|
Are	−2176.97	649.65	−3.35	0.002 ***
Lan	774.05	269.33	2.87	0.008 ***
Was	−2.73	1.15	−2.37	0.025 **
Inv	−275.00	281.92	−1.78	0.086 *
Urb	1740.08	467.86	3.72	0.001 ***
Edu	−1557.57	645.20	−2.41	0.022 **
Med	60.88	15.51	3.93	0.001 ***

Sign. *** ≤ 0.01, ** 0.01 < p ≤ 0.05, * 0.05 < p ≤ 0.1.

.

As indicated in Table 3, all seven factors have significant effects on ANEV. Because the explained variable ANEV is an absolute value, in order to smooth the data variance and reduce the effect of heteroscedasticity, we also run the regression using the logarithmical transformations of the explained and explanatory variables. The use of logarithmical transformations may make the analysis and interpretation of the impact of each factor on ANEV more intuitive—the regression coefficient represents the change percentage of ANEV for each unit of each factor. Results using logarithmical transformations of the variable are reported in

Table 4. Panel regression results of indicators of ANEV after logarithm is taken in 30 provinces.

LnANEV	Coef.	Robust Std. Err	t	p > |t|
lnAre	−0.08	0.12	−2.60	0.015 **
lnLan	0.02	0.19	0.97	0.338
lnWas	−0.08	0.00	−3.33	0.002 ***
lnInv	−0.05	0.01	−2.40	0.023 **
lnUrb	0.03	0.11	0.37	0.715
lnEdu	−0.03	0.11	−2.22	0.034 **
lnMed	0.09	0.00	3.58	0.001 ***
_cons	8.06	0.06	42.50	0.000
rho	0.9959

Sign. *** ≤ 0.01, ** 0.01 < p ≤ 0.05, * 0.05 < p ≤ 0.1.

:

The results in Table 3 and Table 4 illustrate that there is a significant linear relationship between Are, Was, Inv, Edu, Med, and ANEV. Lan and Urb have significant nonlinear effects on ANEV. Are has a significant negative effect on ANEV at a 5% significance level, with an estimated coefficient of −0.08, which does not support Hypothesis 1. The result may be the outcome of one of the two situations where the negative impact on ANEV is possible: (1) the EC of paddy exceeds its EV, and (2) the EC of other crops exceeds its EV, which is not the focus of this study. There may be three reasons for the first situation. Firstly, food prices deviate from the market mechanism due to international crowding and government intervention. Secondly, farmers pursue greater economic benefits at the cost of ecological value [44]. The improvement of grain output per unit area mainly relies on the application of pesticides, fertilizers, and other inputs. This agriculture production method aggravates non-point source pollution [61], and the increment of grain output from the input of unit fertilizer and pesticide continues to decrease. At the same time, paddy planting is more affected by artificial interference, and excessive planting behavior decreases the organic matter, beneficial organisms, and air quality, thus reducing the ANEV. Thirdly, the paddy consumes more water, which leads to an increase in water consumption, and lowers the EV and ANEV.

Table 3 suggests that Lan has a significant positive effect on ANEV (β = 774.05; p ≤ 0.01). Table 4 shows that the relationship between the log transformations of Lan and ANEV is insignificant. The results of Table 3 and Table 4 show that Lan has a significant but not linear relationship with ANEV, offering support for Hypothesis 2. To further confirm the impact of Lan on ANEV, we will conduct a threshold regression analysis (refer to the next section of the paper).

As shown in Table 4, Was has a significant negative effect on ANEV (β = −0.08; p ≤ 0.01), offering support for Hypothesis 3. The result is expected: paddy has high demand for water resources. The increase of wastewater directly leads to broader pollution to water resources, which reduces the ANEV.

Table 4 reveals that Inv has a significant negative impact on ANEV (β = −0.05; p ≤ 0.05), suggesting support for Hypothesis 4. The finding indicates that, during the study period (2008–2017), the agricultural economy grew largely at the cost of the ecosystem services.

The results in Table 3 and Table 4 show that Urb has a significant positive but nonlinear effect on ANEV, providing support for Hypothesis 5. To further examine the effect of Urb on ANEV, we also analyze the specific effect of Urb at different e levels through threshold regression (refer to Section 5.2).

As shown in Table 4, Edu has a significant negative effect on ANEV (β = −0.03; p ≤ 0.05), supporting Hypothesis 6. The costs incurred by the increase of the Edu force farmers to increase the application amount of pesticides and fertilizers to make ends meet, resulting in a reduction of ANEV.

Table 4 suggests that Med has a significant positive effect on ANEV (β = 0.09; p ≤ 0.01), supporting Hypothesis 7.

5.2. Threshold Regression Results

In order to further analyze the relationship between Lan, Urb, and ANEV, we apply the Hansen threshold estimation and test method [62] to explore the threshold effect of Lan and Urb on ANEV under the different e level in each region. Threshold effect refers to an effect in the explained variable that does not occur until an independent variable reaches a certain threshold value. The form of the separation function can be expressed as:

$$\begin{array}{l}{\mathrm{y}}_{\mathrm{i},\mathrm{t}}{=\mathsf{\alpha }}_0{+\mathsf{\alpha }}_1{\mathrm{x}}_{\mathrm{i}}{+\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}{, \mathrm{if} \mathrm{e}}_{\mathrm{i}}\le \mathsf{\gamma }\\ {\mathrm{y}}_{\mathrm{i},\mathrm{t}}{=\mathsf{\alpha }}_0{+\mathsf{\alpha }}_2{\mathrm{x}}_{\mathrm{i}}{+\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}{, \mathrm{if} \mathrm{e}}_{\mathrm{i}}> \mathsf{\gamma }\end{array}$$

(4)

γ is threshold value, e is the threshold variable, x_i is an explanatory variable not associated with the perturbation term ε_i. Merging Formula (3), we get:

$${\mathrm{y}}_{\mathrm{i}}{=\mathsf{\alpha }}_1{\mathrm{x}}_{\mathrm{i}}\times Ø{(\mathrm{e}}_{\mathrm{i}}\le {\mathsf{\gamma })+\mathsf{\alpha }}_2{\mathrm{x}}_{\mathrm{i}}{\times Ø(\mathrm{e}}_{\mathrm{i}}>\mathsf{\gamma })+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(5)

where Ø (·) is an indicator function—it takes the value of 1 when the conditions in parentheses are met and otherwise, 0. The regression model constraint corresponding to the threshold value is α₁ = α₂. Distinguished by the confidence probability of the Wald statistics, if α₁ and α₂ are significantly different, there is a threshold effect; otherwise, there is no evidence for threshold effect.

In order to determine the threshold values, the thresholds and values are estimated by threshold regression. After obtaining a single threshold effect, we continue to find and test for a double and triple threshold. E is selected as the threshold variable, which estimates the influence of Lan and Urb on ANEV and sustainable food supply.

Table 5. Threshold regression of Lan and Urb on ANEV.

Threshold Variables	Threshold	Th-e	Coef.	F.stat	Prob	Different Significance-Level Critical Values
						Crit10	Crit5	Crit1
Lan	Single	55.07	0.02\0.01	32.62 ***	0.00	13.21	15.62	22.63
	Double	——		3.01	0.86	9.46	13.03	19.92
	Triple	——		2.88	0.83	11.04	13.96	20.82
Urb	Single	38.90	0.07\−0.02	26.10 ***	0.00	9.80	11.98	15.92
	Double	——		23.29	0.13	26.25	31.97	46.28
	Triple	——		14.40	0.30	30.55	37.46	47.79

Sign. *** ≤ 0.01, ** 0.01 < p ≤ 0.05, * 0.05 < p ≤ 0.1; p and critical values are obtained from threshold estimates of STATA.

reports threshold regression test results.

According to Table 5, when e is the threshold variable, Lan and Urb pass the single threshold test at a 1% significance level, and the double and triple threshold is rejected.

The extent to which Lan affects ANEV varies across e ranges. When e < 55.07, the influence coefficient of Lan on ANEV is 0.02. When e ≥ 55.07, it decreases to 0.01. With the increase of e, the stability of the paddy ecosystem gets worse, the anti-interference ability becomes weaker, and the positive effect of Lan on ANEV is thus weakened. The main reason is that, when e is in a low stage, the paddy ecosystem is more stable and sustainable, and the influence coefficient of Lan on ANEV is large; when the e increases, the paddy ecosystem is more fragile, and the positive influence of Lan on ANEV is reduced by the system vulnerability. Therefore, e should be controlled within the target point (55.07) when making paddy protection policy.

Urb also has different effects on ANEV at different e levels. When e < 38.90, the influence coefficient of Urb on ANEV is 0.07, and when e ≥ 38.90, it becomes −0.02. Thus, Urb has a positive effect on ANEV within e interval 0~38.90; when e is greater than 38.90, Urb negatively affects ANEV. As e increases, the paddy ecosystem has poorer stability and weaker anti-interference ability, which turns the positive influence of Urb on ANEV negative. Therefore, the e should be controlled within the threshold of 38.90 to ensure the large positive effect of Lan on ANEV, and the positive effect of Urb on ANEV. From the perspective of stability, sudden changes in external factors and ecological costs are not good for maintaining the sustainable development of the ecosystem.

5.3. Threshold Regression Results

As reported in Table 5, the threshold value of the e corresponding to Lan is 55.07, and the threshold value of the e corresponding to Urb is 38.90. Lan and Urb affect ANEV simultaneously, and the influence of Lan and Urb on ANEV changes when e changes, as shown in the threshold value and influence coefficient in

Table 6. Changes of influence of Lan and Urb on ANEV.

Th-e	Threshold Variables	Coef.
0 ≤ e < 38.90	Lan	0.02
	Urb	0.07
38.8928 ≤ e < 55.07	Lan	0.02
	Urb	−0.02
e ≥ 55.07	Lan	0.01
	Urb	−0.02

below:

Table 6 shows that there are two thresholds and three intervals, 1: 0 ≤ e < 38.90, 2: 38.90 ≤ e < 55.07, and 3: 55.07 ≤ e. The three intervals of e represent the corresponding target points for sustainable food supply policies. In the interval 1: 0 ≤ e < 38.90, the influence coefficients of Lan and Urb on ANEV are 0.02 and 0.07, respectively; In the interval 2: 38.70 ≤ e < 55.07, the influence coefficients of Lan and Urb on ANEV becomes 0.02 and −0.02, respectively. Compared to interval 1, the influence coefficient of Lan in interval 2 remains unchanged, but the influence coefficient of Urb changes from positive 0.07 to negative −0.02; in interval 3: e ≥ 55.07, the influence coefficients of Lan and Urb on ANEV change to 0.01 and− 0.02, respectively. Compared to interval 2, the positive influence coefficient of Lan in interval 3 is reduced from 0.02 to 0.01, and the negative influence coefficient of Urb remains unchanged.

Based on the results of the e from 2008–2017 and the Table 6 results, the e of 30 provinces is in the interval 1: 0 ≤ e < 38.90. For that period, the influence coefficient of Lan on ANEV is 0.02, and the influence coefficient of Urb on ANEV is 0.07.

Lan has a positive effect on ANEV, but it decreases when e is greater than 55.07. As seen from

Table 7. The e exceeding 38.8928 in 2008–2017.

Provinces	Year	e	Provinces	Year	e
Heilongjiang	2009	66.77	Jiangxi	2016	75.57
Heilongjiang	2010	74.21	Hubei	2016	75.24
Tianjin	2013	55.07	Hunan	2016	61.04
Jilin	2016	86.07	Guangdong	2016	72.59
Heilongjiang	2016	85.59	Guangxi	2016	82.58
Jiangsu	2016	43.09	Chongqing	2016	58.81
Anhui	2016	55.34	Guizhou	2016	63.66

, e in Tianjin, Jilin, Heilongjiang, Anhui, Jiangxi, Hubei, Hunan, Guangdong, Guangxi, Chongqing, and Guizhou all exceeded the threshold in 2009 and 2010. The positive effect of Lan on ANEV is weakened by the large e when the paddy ecosystem is fragile.

The Urb threshold for ANEV is 38.90. In the interval 0 ≤ e < 38.90, Urb has a positive impact on ANEV, indicating that Urb promotes the sustainable food supply. Urb typically leads to technological innovation, which helps enhance the value of food and raw material production. A few provinces have an e value over 38.90 during the study period, as shown in Table 7, although the average e in the 30 provinces is below 38.8928. For those few provinces, the impact of Urb on ANEV changes from positive to negative, and the negative impact coefficient is large (−0.24), suggesting that Urb significantly reduces the food supply capacity, and thus affects the sustainable development of the paddy ecosystem. As shown in Table 7, the stability of the paddy ecosystem in Heilongjiang Province is relatively weak, which is consistent with the calculation [19]. E exceeded 38.90 in 2009, 2010, 2013, and especially in 2016, in eleven provinces, including Jilin, Heilongjiang, Jiangsu, Anhui, Jiangxi, Hubei, Hunan, Guangdong, Guangxi, Chongqing, and Guizhou, indicating the deterioration of the paddy ecosystem.

6. Discussion

Global food security is facing an increasingly grim situation [2], highlighting the importance of a sustainable food supply. Food-related policies span multiple domains including environmental, social, ethical, and economic dimensions [63,64]. Current research on agricultural policy has focused primarily on increasing farmers’ income and the agricultural economy. Few policies are focused on food ecosystems security and sustainability, perhaps due to the lack of comprehensive framework validated by empirical data [65]. EV provides valuable inputs for land management decision-making and serves as a reference for future studies [43]. NEV more accurately measures the net services provided by the ecosystem. Paddy protection policies and sustainable food supply policies with ANEV as a target metric ultimately help the country achieve sustainable development.

6.1. Policy Recommendations for Ecological Factors

The results of the panel regression reported in Table 4 show that Are and ANEV are negatively associated, with a regression coefficient of −0.08. If the negative effect of Are on ANEV is caused by higher ecological costs, we recommend the following remedies. Firstly, we need to reduce the ecological costs. The amounts of pesticides and chemical fertilizer applied in paddies should be reduced through agricultural pest control technologies and innovations. Overfarming should be controlled via mechanisms such as paddy rotation that promote the integrated utilization of soil organic matter [66]. Education and training advocating green and ecological agriculture are also needed to improve farmers’ awareness of environmental protection, Secondly, we need to increase the EV. Measures include improving the values of products through the improvement of rice varieties, increasing the provisioning (food and raw material), regulating, supporting, and cultural service values of the paddy ecosystem, and reducing water resource consumption of paddy through channel seepage prevention technology, drip irrigation technology, and sprinkler irrigation technology. Thirdly, we should adjust the crop planting structure and maintain an appropriate area of paddy. Larger paddy area is not necessarily better. Fourthly, we should increase agricultural income by labor cost reduction and mass land cultivation to increase productivity and cut the costs of production. Differentiation strategies should be put forward to improve food quality and promote food product value appreciation [9]. The increase in farmers’ income may help increase their morale, which in turn contributes to the sustainability of the food supply capacity of the paddy ecosystem.

According to the threshold regression results in Table 5, Lan has a positive and significant threshold effect on ANEV. When e is under 55.07, the influence coefficient is 0.02; when e is over 55.07, it is reduced to 0.01. The positive effect of Lan on ANEV weakens when e exceeds the threshold value. The cultivated land in northern China with large Lan and sparse population has a profound impact on the whole ecosystem. It is essential in regulating and supporting service and maintaining the sustainable food supply capacity. However, the land utilization rate is low in these sparsely populated areas. For example, Heilongjiang, Inner Mongolia, and Jilin have the largest Lan, and are also the main area for abandoned land [10]. Abandoned land reduces the total supply service value, leading to the weakening of sustainable food supply capacity and food security. In terms of policy, firstly, we should strictly prohibit non-agricultural use of cultivated land. Establishing and strictly enforcing policies that prevent non-agricultural use of farmland is key to ensuring a sustainable supply of food. Secondly, we should strive to improve farmland utilization efficiency through planning, land classification and grain rotation, and agricultural technological innovations. Thirdly, the gradient scale management should be put to use based on the capital and technical level. An appropriate paddy planting scale can better ensure grain supply. Government support policies should be put forward to reduce cultivated land abandonment, which seriously affects the food supply [10]. Agricultural technologies and variety improvement should be encouraged to increase paddy production and food supply, particularly in places where the Lan is large.

Table 4 shows that the influence coefficient of Was on ANEV is −0.08. Wastewater affects the output of paddy ecosystem service through irrigation and has a great negative impact on the paddy ecosystem regulating service. The regulating service of the paddy ecosystem can be divided into regulating functional value to soil, air, and water. Wastewater directly increases the accumulation of heavy metals in the soil and the regulating function of air, and thus may cause a rapid decrease in water resources. Our recommendations include the following: Firstly, we need to control the discharge and treatment of wastewater, and strengthen sewage management, especially in the rural industrial areas that are easily overlooked. More and more factories turn to rural areas to save costs, leading to increasing industrial pollution. Secondly, the government should come up with wastewater regulations on accountability, rewards, and punishments. For instance, investment in industrial drainage testing technology and treatment technology research and development should be encouraged and rewarded, while industrial wastewater discharge that has a significant impact on the environment should be outlawed and punished.

6.2. Policy Recommendations of Economic Factors

Based on the results reported in Table 4, the influence coefficient of Inv on ANEV is −0.0451. Farmers’ income is generally low in developing countries. They often increase production by increasing the application of pesticides and fertilizers. Thus, we recommend that: Firstly, we should appropriately expand the land sown and increase agricultural income through the scale effect. Measures include mechanizing farming processes, improving labor production efficiency, and improving the utilization rate of resources. Secondly, we should shift quantity-oriented farming to value-oriented farming. Instead of focusing only on output quality, we should turn our attention to increasing the added value of grain, and the integration of the primary, secondary, and tertiary industries to promote food supply security while maintaining environmental friendliness and sustainability.

As reported in Table 5, Urb has a threshold effect on ANEV. When e is less than 38.8928, Urb has positive influence on ANEV and the influence coefficient is 0.07; when e is greater than 38.90, influence of Urb on ANEV changes from positive to negative, and the influence coefficient is −0.02. High quality Urb should include Urb quality and efficiency indicators, such as the greening coverage rate and SO₂ emissions, GDP, social consumption, etc., emphasizing the coordinated development of the population, economy, resources, and environment [67]. On the one hand, the improvement of Urb feeds agricultural development when the paddy ecosystem is sustainable. For instance, economic level, agricultural technology progress, and agricultural production efficiency can promote ANEV by increasing land utilization. However, mechanized operation and industrial pollution have negative impacts on the paddy ecosystem when e is high. The effect of Urb on ANEV varies with different e levels. On the other hand, from the perspective of social development, continuous urbanization reduces the values of paddy ecosystem services, especially supply services and regulating services, and harms the environment. The e value also exceeds the critical value (refer to Table 7), although the average e of paddy ecosystems in 30 provinces from 2008–2017 is below 38.90.

We therefore recommend the following measures: Firstly, we should strictly control the use of cultivated land, especially grain land, in the process of urbanization, and put an end to the urbanization that occupies grain land resources. Secondly, e should be controlled within 38.90 to ensure the positive impact of Urb on ANEV and to maintain the anti-interference capacity of the paddy ecosystem. Thirdly, we need to improve the quality of urbanization and pursue the coordinated development of population, economy, resources, and the environment, rather than blindly pursue economic benefits and growth data.

6.3. Policy Recommendations of Social Factors

Table 4 shows the significant influence of Edu on ANEV (β = −0.03). The finding is not surprising. With the increase in education expenditure, farmers are compelled to pursue a production increase of paddy at the cost of natural resources, in order to fund such education. Unfortunately, most of farmers’ education expenses are used for their children, and such education plays little role in increasing the farmers’ own environmental awareness. For that reason, we recommend that the government should increase investment in rural education. Such investment should not only be in the construction of educational facilities, but in items such as books, to minimize farmer’s financial burden related to education. A government fund should also be secured to educate farmers themselves, so as to increase their awareness of the environment and environment-related laws and regulations.

The results in Table 4 reveal the significant positive influence of Med on ANEV (β = 0.09). Since the outbreak of COVID−19, people have paid more attention to health. The finding has important policy implications. New policies should channel more medical resources to rural areas to improve the training and pay of rural healthcare workers, and to construct community-level health service facilities in rural areas. We should further improve the new rural cooperative medical care system to provide affordable and effective health care to farmers.

7. Conclusions

ANEV is the net contribution to human beings provided by the paddy ecosystem. ANEV reflects the ability of the paddy ecosystem to provide ecosystem services, which include provisioning services, such as the supply of food and raw materials, as well as supporting, regulating, and cultural services. E represents the change rate of ANEV affected by external factors, and it estimates the stability and sustainability of a paddy ecosystem. Extant research on food security mainly focuses on quantitative aspects [22,23], and research on food supply quality and sustainable supply capacity is lacking. Based on the e and ANEV, we analyze the influence of ecological, economic, and social factors on ANEV through panel and threshold regressions, and put forward some policies and implications for sustainable food supply. The findings of our research provide meaningful guidance for formulating policies pertaining to sustainable food supply, paddy ecosystem protection, and the sustainable development of food.

We must point out a few limitations of this study. Firstly, when studying the influence of Lan and Urb on ANEV, we use the average number of e in the threshold regression model—we wish we could obtain data enabling us to calculate the e threshold value and interval in each province. Secondly, this study uses the data of 30 provinces in China. While China is the main rice producer and accounts for a large proportion of paddy farmland in the world, caution is still warranted when generalizing the findings of this research to other countries. Comparative research involving other countries would be useful to validate the findings of this research in different cultural, regulatory, and legal settings.