The Impact of a Full-Cost Insurance Policy on Fertilizer Reduction and Efficiency: The Case of China

Abstract

Excessive fertilizer input and inefficient utilization in agricultural production have caused significant negative environmental impacts. Based on provincial panel data in China from 2005 to 2021, this study adopts the full-cost insurance pilot launched in 2018 and uses the DID method to empirically analyze its impact on fertilizer application intensity and utilization efficiency. The study reveals the following findings: (1) Implementing full-cost insurance can reduce fertilizer application intensity by 21.761% and increase utilization efficiency by 1.915%. (2) Full-cost insurance reduces fertilizer application intensity and improves fertilizer utilization efficiency by expanding the land scale and reducing the agricultural labor force. (3) Full-cost insurance significantly improves fertilizer utilization efficiency in high-risk and low-risk areas. Nevertheless, while the policy significantly reduces fertilizer application intensity in high-risk areas, its effect on low-risk areas is not apparent. (4) Full-cost insurance has an environmental protection effect. It can significantly reduce 11.593% of nitrogen pollution emissions, 2.577% of phosphorus pollution emissions, and 35.400% of equivalent pollution emissions. The implementation of full-cost insurance plays an important role in reducing fertilizer use and improving utilization efficiency. So, we should continue to intensify the promotion of full-cost insurance policy to fully leverage the advantages of agricultural insurance and promote sustainable agricultural development.

1. Introduction

Chemical fertilizers are a vital component of agricultural production and play a crucial role in ensuring food security [1]. By the end of 2021, global fertilizer consumption had increased to 195.400 million tons (Data from FAO Statistical Database). Asia accounts for 69% of global fertilizer consumption, of which China accounts for about half [2]. As depicted in Figure 1, the issue of excessive fertilizer inputs in China remains relatively serious. Even though China’s fertilizer inputs in 2021 were 319.110 kg/ha, which is significantly higher than the world average of 118.620 kg/ha and the Asian average of 176.270 kg/ha, the fertilizer utilization rates in developed countries range from 50% to 60% (Data from the Ministry of Agriculture and Rural Affairs of China), which is considerably higher than China’s rates. The reduction of fertilizers and the improvement of efficiency are pressing concerns in the context of sustainable development in agriculture. At the same time, in 2015, the government issued the Zero Growth Action Plan for Fertilizer Use by 2020, aimed at addressing the issue of excessive fertilizer use at its source. In 2024, the government also emphasized the importance of “continuing to promote the reduction of chemical fertilizers and pesticides while enhancing efficiency”. Therefore, the issue of fertilizer reduction will continue to receive academic attention, and the continuous promotion of fertilizer reduction and efficiency enhancement will be crucial in the future of China’s ecological environment management and rural environmental protection.

Agricultural insurance as a tool for mitigating agricultural production risks, and its promotion and popularization, will affect the fertilizer application behavior of agricultural operators [3]. In 2018, relevant state departments issued the “Circular on the Pilot Work of Carrying Out Full-cost Insurance and Income Insurance for the Three Major Food Crops.” Full-cost insurance (FCI) extends coverage to include service, labor, and land costs based on traditional yield insurance [4]. This not only enhances farmers’ incentives to grow grain and mitigates the likelihood of losses from agricultural risks, but also influences agricultural production behavior by altering the expected marginal returns from agricultural production [5]. Therefore, guiding farmers to adopt rational fertilization practices through policy-based agricultural insurance has significant practical implications for promoting sustainable agricultural development.

In recent years, numerous studies have focused on the role of agricultural insurance in agricultural production and socioeconomic aspects, such as improving crop production structure [6], facilitating agricultural technology progress [7], enhancing food production [8], and decreasing income inequality [9]. Regarding fertilizer application, one perspective suggests that agricultural insurance increases fertilizer application and reduces farmers’ adoption of environmentally friendly technologies as the insurance coverage increases [10]. Fertilizer and pesticide application tend to increase risk significantly, and yield insurance may improve identification of the type of risk [11,12,13]. Moreover, the increase in agricultural insurance coverage level has a more pronounced impact on the income effect than on the substitution effect, thereby elevating the level of environmental pollution [14]. Another perspective suggests that agricultural insurance promotes fertilizer use reduction. Some studies argue that agricultural insurance presents a clear moral hazard. Agricultural operators may regard premiums as production costs, potentially leading to a crowding-out effect on the inputs of other factors [15]. As a result, agricultural insurance may reduce fertilizer and pesticide inputs, thereby increasing the substitution or supplementation with green technology applications [16,17,18]. The third perspective posits that agricultural insurance has little impact on agricultural fertilizers and other chemical inputs [19], primarily because the impact of agricultural insurance on fertilizer application varies depending on the study object, crop type, and type of insurance.

Based on the studies above [11,12,14,16,18], we find that there is no agreement on the impact of agricultural insurance on fertilizer inputs. The possible reasons are as follows. First, the variability in the impact of agricultural insurance on fertilizer application depends on the study object, crop type, and type of insurance. Unlike yield or income insurance in the United States, China’s agricultural insurance primarily focuses on production costs. Existing studies lack policy evaluations that examine the impact of agricultural insurance on fertilizer application intensity and utilization efficiency from an overall or macro perspective. Second, there are differences in theoretical models. Horowitz and Lichtenberg’s 1993 study [20] examined the impact of agricultural insurance on inputs based on income insurance. However, building on the 2000 study by He et al. [21], we set a model to analyze the effect of full-cost insurance on fertilizer application in China from the perspective of cost insurance.

Based on the panel data of 25 provinces spanning 2005 to 2021, this study treats China’s full-cost insurance as a quasi-natural experiment and employs the DID model to evaluate its effect on fertilizer application intensity and utilization efficiency, and examines the transmission paths of its impact on fertilizer application to provide insights for promoting sustainable agricultural development. This research contributes in several ways. First, the study assesses the impact of China’s full-cost insurance on fertilizer application using a theoretical model, providing policy insights for further promotion and design of the full-cost insurance system. Second, the analysis examines the impact of full-cost insurance on both fertilizer application intensity and fertilizer utilization efficiency, thereby offering empirical evidence for the role of policy-based agricultural insurance in ensuring food security in China. Third, the study explores the pathways through which full-cost insurance affects fertilizer application. Specifically, the investigation focuses on three main pathways: land scale, agricultural labor force, and agricultural technological progress.

2. Empirical Setting: Full-Cost Insurance in China

The government of China issued the “Notice on the Pilot Work of Full-cost Insurance and Income Insurance for Three Major Crops” in 2018. Given the limited number of revenue insurance pilots, this analysis focuses exclusively on full-cost insurance. In 2018, the government designated Inner Mongolia, Heilongjiang, Anhui, Shandong, Henan, and Hubei as pilot provinces, later expanding to include the 13 major grain-producing provinces. Full-cost insurance covers various perils, including major natural disasters, pests and diseases, and accidents. The insured amount covers the cost of capital and services, labor, and land. The insurance targets three major grain crops: rice, wheat, and corn. To effectively avoid moral hazard, full-cost insurance coverage is no higher than 80% of the output value of the corresponding crop variety. Additionally, the self-payment ratio is set at no less than 30%, to ensure farmer participation. Subsidies are distributed according to regional disparities, with the central, western, and eastern regions receiving 40%, 40%, and 35%, respectively.

Full-cost insurance shares similarities with traditional yield insurance regarding its role in economic compensation and redistributive function. However, full-cost insurance can effectively enhance risk protection and optimize premium subsidy structures compared to traditional yield insurance [22]. With the support and promotion of both central and local governments, the pilot policy has yielded significant results. Firstly, the insured areas increased from 861.313 thousand hectares in 2019 to 941.273 thousand hectares in 2020, reflecting a gradual rise in farmers’ enthusiasm for insurance. Secondly, post-disaster compensation expenditures notably rose from 472 million yuan in 2019 to 796 million yuan in 2021, thereby strengthening the efficacy of post-disaster compensation mechanisms [23].

3. Conceptual Framework

3.1. The Impact of Agricultural Insurance on Fertilizer Application

The assumption is that farmers are rational economic agents who make investment decisions in various factors to maximize profits under certain risk conditions and market risks. Consider a representative farmer who owns one hectare of arable land. Its production function is $y=f\left(x, \omega , \phi ,m\right)$, and $\mathrm{y}>0$, where x is fertilizer inputs, ω is the uncertain production environment, and $m$ is other related inputs. If ωg (·) follows the distribution of $\left[{\omega }_{min},{\omega }_{max}\right],$ when ω is close to ${\omega }_{max}$, there are no natural disasters or crop pests during the production season; when ω is close to ${\omega }_{min}$, there are more severe natural disasters or crop pests during the production season. $\phi$ is a random variable.

In the context of the FCI policy, it is assumed that the insurance company requires an output level of $\overline{f}$ (based on the previous year’s production). The compensation will be required if the actual production y is less than $\overline{f}$. If the loss rate is greater than 20% and not more than 80%, the FCI provides for compensation of the total input costs of the damaged areas; if the loss rate is greater than 80%, the total loss is compensated based on the entire planted area; and no compensation is paid for the loss if the loss rate is less than 20%. If the output y of the insured area is lower than 0.8$\overline{f}$, the insurance company is required to pay $\left(1-{\displaystyle \frac{y}{\overline{f}}}\right)\tilde{p}x$, where $\tilde{p}$ is the price of production inputs. There exists a trigger state ${\omega }^{*}=\omega \left(x,\overline{f}\right)$ such that, for $f\left(x,{\omega }^{*},\phi ,m\right)<0.8\overline{f}$, the farmers will receive compensation.

On this basis, the expected utility function of farmers in the agricultural insurance framework is:

$$\begin{gathered} EU={\int }_{{\omega }_{min}}^{{\omega }_{max}}u\left(\pi \right)g\left(y\right)dy, where u^{\prime }\left(\pi \right)>0, u^{″}\left(\pi \right)<0, \pi =\left\{\begin{array}{c}{\pi }_1, {\omega }_{min}<y<{\omega }^{\mathrm{*}}\\ {\pi }_2, {\omega }^{\mathrm{*}}<y<{\omega }_{max}\end{array}\right. \\ {\pi }_1=p_{c}y-\tilde{p}x+\theta \left(1-{\displaystyle \frac{y}{\overline{f}}}\right)\tilde{p}x, {\pi }_2=p_{c}y-\tilde{p}x \\ EU={\int }_{{\omega }_{min}}^{{\omega }^{\mathrm{*}}}u\left[{\pi }_1\right]g\left(\omega \right)d\omega +{\int }_{{\omega }^{\mathrm{*}}}^{{\omega }_{max}}u\left({\pi }_2\right)g\left(\omega \right)d\omega \end{gathered}$$

(1)

Then, the first order condition is:

$$\begin{gathered} {\displaystyle \frac{\phi EU}{\phi x}}={\int }_{{\omega }_{min}}^{{\omega }^{\mathrm{*}}}{u}^{\prime }\left[{\pi }_1\right]\left[p_c{y}_x^{\prime }-\tilde{p}+\theta \left(1-{\displaystyle \frac{y}{\overline{f}}}\right)\tilde{p}+\theta \left(1-{\displaystyle \frac{y_x^{\prime }}{\overline{f}}}\right)\tilde{p}x\right]g\left(\omega \right)d\omega -u[p_c{y}^{\mathrm{*}}-\tilde{p}x+\theta \left(1-{\displaystyle \frac{y^{\mathrm{*}}}{\overline{f}}}\right)\tilde{p}x]g\left({\omega }^{\mathrm{*}}\right){\displaystyle \frac{d{\omega }^{\mathrm{*}}}{dx}}+ \\ {\int }_{{\omega }^{\mathrm{*}}}^{{\omega }_{max}}{u}^{\prime }\left({\pi }_2\right)\left(p_c{y}_x^{\prime }-\tilde{p}\right)g\left(\omega \right)d\omega -u(p_c{y}^{\mathrm{*}}-\tilde{p}x)g\left({\omega }^{\mathrm{*}}\right){\displaystyle \frac{d{\omega }^{\mathrm{*}}}{dx}} \end{gathered}$$

(2)

where $u^{\prime }$ is the partial derivative of relative expected utility, θ is the loss rate in the insured areas, and $p_c$ is grain sales price. On the other hand, FCI can effectively cover loss by adding coverage for labor and land costs based on traditional yield insurance. To obtain the maximum expected utility, we set ${\displaystyle \frac{\phi EU}{\phi x}}=0$ and focus on the relationship between x and $\theta$.

We can obtain ${\displaystyle \frac{dx}{d\theta }}=-{\displaystyle \frac{\frac{dA}{d\theta }}{\frac{dA}{dx}}}$ if ${\displaystyle \frac{\phi EU}{\phi x}}=A$, where ${\displaystyle \frac{dA}{dx}}$ is the second-order derivative of expected utility, so $u^{″}\left(\pi \right)<0$. To explore the relationship between x and $\theta$, we can examine the sign of ${\displaystyle \frac{dA}{d\theta }}$ as follows:

$$\begin{gathered} {\displaystyle \frac{dA}{d\theta }}={\int }_{{\omega }_{min}}^{{\omega }^{\mathrm{*}}}\left[u^{\prime }\left(p_{c}y-{\displaystyle \frac{y}{\overline{f}}}\tilde{p}x\right)\left(p_c{y}_x^{\prime }-{\displaystyle \frac{y_x^{\prime }}{\overline{f}}}\tilde{p}x-{\displaystyle \frac{y}{\overline{f}}}\tilde{p}\right)-u^{\prime }\left(p_{c}y-\tilde{p}x\right)\left(p_c{y}_x-\tilde{p}\right)\right]g\left(\omega \right)d\omega \left[A\right]+ \\ \left[u\left(p_c{y}^{\mathrm{*}}-{\displaystyle \frac{y^{\mathrm{*}}}{\overline{f}}}\tilde{p}x\right)-u\left(p_c{y}^{\mathrm{*}}-\tilde{p}x\right)\right]g\left({\omega }^{\mathrm{*}}\right){\displaystyle \frac{d{\omega }^{\mathrm{*}}}{dx}}\left[B\right] \end{gathered}$$

(3)

In Equation (3), when $\omega$ is in the state ${\omega }_{min}<w<{\omega }^{\mathrm{*}}$, ${\displaystyle \frac{y^{\mathrm{*}}}{\overline{f}}}<0.8$, and ${u^{\prime }\left(\pi \right)>0,u}^{″}\left(\pi \right)<0$. As a result, $u^{\prime }\left(p_{c}y-{\displaystyle \frac{y}{\overline{f}}}\tilde{p}x\right)<u^{\prime }\left(p_{c}y-\tilde{p}x\right)$. Let $F\left(w\right){=p}_c{y}_x^{\prime }-{\displaystyle \frac{y_x^{\prime }}{\overline{f}}}\tilde{p}x-{\displaystyle \frac{y}{\overline{f}}}\tilde{p}, T(\mathrm{w})=p_c{y}_x-\tilde{p}$. When T(w) < F(w) < 0 or T(w) < 0 < F(w), the conclusion of Equation (3) [A] is positive; when F(w) < T(w) or 0 < T(w) < F(w), the conclusion is negative. So, the conclusion of Equation (3) [A] can be either positive or negative. When the state of nature is ${\omega }^{\mathrm{*}}$, $f\left(x,{\omega }^{\mathrm{*}},\phi ,m\right)=0.8\overline{f}$; as a result, $u\left(p_c{y}^{\mathrm{*}}-{\displaystyle \frac{y^{\mathrm{*}}}{\overline{f}}}\tilde{p}x\right)-{\displaystyle \frac{y^{\mathrm{*}}}{\overline{f}}}\tilde{p}x>u\left(p_c{y}^{\mathrm{*}}-\tilde{p}x\right)$, so the conclusion of Equation (3) [B] is negative. Therefore, this paper deduces that the sign of ${\displaystyle \frac{dA}{d\theta }}$ is uncertain. To determine the positive or negative relationship between x and $\theta$ depends on the effect of the factor input x on expected income. If a factor input x positively affects expectations under different risk conditions, then ${\displaystyle \frac{dx}{d\theta }}$ < 0.

So, in the following section, we discuss the effect of insured farmers on different production factors. We assume that farmers in uninsured areas should expect the following utility function:

$$E{\pi }_3=\left(1-\theta \right)p_{c}f\left(x,w_1,\phi \right)+\theta p_{c}f\left(x,w_2,\phi \right)-\tilde{p}x$$

(4)

where θ is the risk probability, and $w_1$ and $w_2$ are the risk-free and risky states, respectively.

The expected utility function of farmers in the insured areas is as follows:

$$E{\pi }_4=\left(1-\theta \right)p_{c}f\left(x,w_1, \phi \right)+\theta p_{c}f\left(x,w_2,\phi \right)+\theta \left(1-{\displaystyle \frac{f\left(x,w_2,\phi \right)}{\overline{f}}}\right)\tilde{p}x-\tilde{p}x$$

(5)

If the expected utility function maximization condition is ${\displaystyle \frac{\phi E\pi }{\phi x}}$ = 0, we can obtain:

$${\displaystyle \frac{\partial E{\pi }_4}{\partial x}}-{\displaystyle \frac{\partial E{\pi }_3}{\partial x}}=-\theta \tilde{p}\left[1-{\displaystyle \frac{f\left(x, w_2, \phi \right)}{\overline{f}}}+x\left(1-{\displaystyle \frac{1}{\overline{f}}}{\displaystyle \frac{\partial f\left(x, w_2, \phi \right)}{\partial x}}\right)\right]$$

(6)

If the production factor inputs cause the marginal output in the risky state to be greater than in the risk-free state, and there is ${\displaystyle \frac{\partial f\left(x, w_2, \phi \right)}{\partial x}}>{\displaystyle \frac{\partial f\left(x, w_1, \phi \right)}{\partial x}}$ and ${\displaystyle \frac{\partial f\left(x, w_2, \phi \right)}{\partial x}}<0$, the production factor is defined as a strong risk-increasing factor. If ${\displaystyle \frac{\partial f\left(x, w_2, \phi \right)}{\partial x}}>{\displaystyle \frac{\partial f\left(x, w_1, \phi \right)}{\partial x}}$, we can define the production factor as a weak risk-increasing factor. If ${\displaystyle \frac{\partial f\left(x, w_2, \phi \right)}{\partial x}}<{\displaystyle \frac{\partial f\left(x, w_1, \phi \right)}{\partial x}}$, the production factor is a risk-reducing factor [18]. Generally, insured farmers are more inclined to the inputs of strong risk-increasing factors and decrease the weak risk-increasing and risk-reducing factors, whereas fertilizer has a risk-reducing and yield-increasing effect [24]. Full-cost insurance, by offering broader coverage compared to traditional insurance, may reduce the intensity of fertilizer application while enhancing fertilizer utilization efficiency. Therefore, we propose H1:

H1. 

Full-cost insurance reduces fertilizer application intensity and increases fertilizer utilization efficiency.

3.2. The Theoretical Mechanism

Building on the previous discussion, the theoretical mechanism by which full-cost insurance affects fertilizer application is elaborated through the lenses of land scale, agricultural labor force, and agricultural technological progress. First, agricultural insurance can expand the cultivated land area by mitigating risks of insured crops and increasing crop compensation [25,26,27,28]. Moreover, the scale effect can achieve economies of scale to improve agricultural production efficiency. This can effectively reduce fertilizer application intensity per unit area and improve utilization efficiency by matching crops with more precise factor inputs and implementing scientific field management practices [29]. Second, the new economic labor transfer theory suggests that the interaction between labor and capital influences production decisions. Insured farmers are highly susceptible to faithless or opportunistic behavior, such as reallocating agricultural input time to non-agricultural activities. Fertilizer application intensity may decrease as non-farm labor increases [30]. Additionally, the rise in non-farm income promotes the substitution of fertilizer with machinery, thus enhancing fertilizer utilization efficiency. Third, full-cost insurance provides a high level of protection, enhancing farmers’ risk resilience ability and stabilizing the expected agricultural returns [31]. Consequently, the development of agricultural insurance promotes the use of agricultural machinery and the adoption of high-risk and high-yield technologies, fostering agricultural technological progress [32,33]. Ultimately, agricultural insurance may incentivize adoption of environmentally friendly production techniques [34], increasing the application of novel fertilizers and thus reducing traditional fertilizer inputs. In summary, we propose H2:

H2. 

Full-cost insurance affects fertilizer application intensity and fertilizer utilization efficiency by influencing land scale, agricultural labor force, and agricultural technological progress.

4. Materials and Methods

4.1. Data Resources

This study selects panel data from 25 provinces in China spanning the years 2005 to 2021 for empirical analysis. To ensure data availability and consistency, regions such as Qinghai, Tibet, and Hong Kong are excluded from the analysis. The relevant data utilized in this paper are sourced from the following publications: the China Statistical Yearbook (CSY), China Rural Statistical Yearbook (CRSY), China Insurance Yearbook (CIY), and provincial statistical yearbooks. These sources provide comprehensive and reliable data necessary for examining the impact of full-cost insurance on fertilizer application intensity and utilization efficiency. Figure 2 illustrates the 25 provincial regions, as well as the six pilot provinces.

4.2. Variables

The dependent variable is the fertilizer application status. Fertilizer application intensity and utilization efficiency are used to measure the fertilizer application status, which is conducive to reflecting fertilizer application in general and its intensity. Regarding fertilizer application intensity, most studies use pure quantity per unit area to measure it [34,35]. However, due to data limitations, some studies use the physical quantity or expenditure amount of fertilizer per unit area [36]. Considering that using expenditure amount might be affected by fertilizer prices and cause significant errors, this paper measures fertilizer application intensity using pure quantity per unit area. In terms of fertilizer use efficiency, existing studies have measured fertilizer use efficiency using the SFA method [37], while agronomic studies often use indicators such as fertilizer consumption per unit yield (the ratio of fertilizer application per unit area to yield per unit area) and agronomic efficiency of fertilizer (the ratio of crop yield to fertilizer application) to measure fertilizer use efficiency [38]. Therefore, this paper chooses the ratio of crop yield to fertilizer application to measure fertilizer use efficiency.

The independent variable is the full-cost insurance policy. In this paper, we set a double difference term ${DID}_{it}={Post}_i\times {Treat}_t$. If ${DID}_{it}$ = 1, the FCI policy is implemented. Otherwise, it is not implemented. ${Treat}_t=1$ for the six pilot provinces of Inner Mongolia, Heilongjiang, Anhui, Shandong, Henan, and Hubei, and ${Treat}_t=0$ for the control group. When the policy started in 2018 and after, ${Post}_i=1$; ${Post}_i=0$ otherwise.

Referring to relevant studies [39,40], the control variables selected for this study include agricultural industry structure (A_industry), urbanization, income, arable land irrigation (Ir_land), financial support (F_support), and agricultural production value (p_value). The details are shown in

Table 1. Description and summary statistics of variables.

Variables	Definition	Mean	SD	Min	Max
Dependent variables
Fer_Total	Fertilizer application per hectare (kg/ha)	362.505	117.756	172.509	610.998
Fer_Efficiency	The ratio of crop yield to fertilizer application (kg/kg)	9.736	4.613	0.923	31.551
Independent variable
FCI	Has the province implemented a full-cost insurance policy? Yes = 1, No = 0	0.056	0.231	0	1
Control variables
A_industry	Primary sector output value/total GDP × 100%	13.087	4.895	4.800	39.400
Urbanization	Urban population/resident population	51.291	10.176	31.020	74.630
Income	Farmers’ disposable income (ten thousand yuan)	0.967	0.535	0.233	2.679
Ir_land	Irrigated area of arable land (thousand hectares/take log)	14.442	0.652	12.033	15.872
F_support	Agriculture-related expenditures/general expenditures	0.125	0.388	0.007	8.049
p_value	Agricultural production value per capita (yuan/take log)	8.317	0.814	4.939	13.188
Mediator variables
Land scale	Agricultural planted area (thousand ha/take log)	7.891	0.869	5.154	9.325
Agricultural labor force	Primary sector employment/Agricultural planted area(people/ha)	3.728	2.951	0.101	26.563
Agricultural technological progress	Total machinery power (billion KW)	0.369	0.280	0.027	1.335

.

To observe the relationship between full-cost insurance (FCI) and fertilizer application intensity and fertilizer utilization efficiency, we divided the provinces into pilot and non-pilot provinces based on the collected data. The means of fertilizer application intensity and fertilizer utilization efficiency are shown in Figure 3. Following implementation of the FCI policy in 2018, the fertilizer application intensity in pilot provinces was lower than in non-pilot provinces. This indicates that the FCI policy effectively reduced fertilizer application intensity in pilot provinces. Additionally, fertilizer utilization efficiency was higher in pilot provinces than in non-pilot provinces. Moreover, after 2018, the gap between pilot and non-pilot provinces gradually widened.

4.3. Empirical Strategy

4.3.1. The DID Model

This part adopts the DID model to verify the impact of the FCI policy on grain production inputs. The DID model can effectively examine the difference between the treatment group and the control group before and after the policy pilot and determine the impact of the FCI pilot. The details are as follows:

$${Fer\_Total}_{it}=\alpha +\beta {DID}_{it}+\gamma {control}_{it}+u_t+{\delta }_i+{\epsilon }_i$$

(7)

$${Fer\_Efficiency}_{it}={\mathsf{\alpha }}^{\prime }+{{\beta }^{\prime }DID}_{it}+{\gamma }^{\prime }{control}_{it}+u_t^{\prime }+{\delta }_i^{\prime }+{\epsilon }_i^{\prime }$$

(8)

In Equations (7) and (8), the dependent variable ${Fer\_Total}_{it}$ is fertilizer application intensity per hectare in region $i$ in year $t$, and the dependent variable ${Fer\_Efficiency}_{it}$ is fertilizer utilization efficiency in region $i$ in year $t$. The independent variable ${DID}_{it}$ is whether to implement the full-cost insurance policy in region $i$ in year $t$. control_it is a series of control variables, including agricultural industry structure, urbanization, income, arable land irrigation, financial support, and agricultural production value; and $u_t, {\delta }_i, \mathrm{a}\mathrm{n}\mathrm{d} {\epsilon }_{it}$ are the time-fixed effect, the individual fixed effect, and a random error term.

4.3.2. The GMM Model

To address potential issues of omitted variables and reverse causality between full-cost insurance and fertilizer application, which leads to endogeneity, we used dynamic panel models and the generalized method of moments (GMM) to mitigate endogeneity. Two-order lagged terms of both dependent and independent variables were selected as instrumental variables. The model parameters were then estimated using the GMM method [41]. The dynamic panel model is specified as follows:

$${Fer}_{it}=a+a_0{Fer}_{i,t-1}+a_1{DID}_{it}+\partial {control}_{it}+u_t+{\delta }_i+{\epsilon }_i$$

(9)

In Equation (9), the lagged period of fertilizer application (${Fer}_{i,t-1}$) is used as an independent variable. The fertilizer application (${Fer}_{it}$) is used as a dependent variable.

4.3.3. The Fixed Effects Model

In terms of the environmental effects of FCI, the excessive and inefficient application of chemical fertilizers has detrimental effects on the environment. Irrational fertilizer application is one of the primary contributors to agricultural non-point source pollution in China. Building on the finding that FCI can reduce fertilizer application intensity and improve utilization efficiency, this section further investigates the environmental effects of full-cost insurance. Based on the studies by Luo et al. (2020) [42] and Niu et al. (2022) [43], pollution sources are identified as nitrogen fertilizers, phosphorus fertilizers, and compound fertilizers. The levels of pollution emissions for these three types of fertilizers are calculated using the following formula:

$$E=\sum T_i\times {\rho }_i\times {\mu }_i$$

(10)

In model (10), $E$ is fertilizer non-point source pollution; $T_i$ is the discounted application amount of the $i$ types of chemical fertilizer; and ${\rho }_i$ is the pollution coefficient. The nitrogen production coefficients of nitrogen, phosphorus, and compound fertilizer are 1, 0, and 0.33, respectively, while the phosphorus production coefficients for the same fertilizers are 0, 0.44, and 0.15. ${\mu }_i$ is the pollution emission coefficient. To measure pollution emission coefficients, this study references three specific metrics based on previous studies [42,43]: nitrogen pollution emissions per unit area (TN, kg/ha), phosphorus pollution emissions per unit area (TP, kg/ha), and equivalent pollution emissions per unit area (EPS, m³/ha). The EPS metric is derived by converting nitrogen and phosphorus emissions into equivalent pollution units. Specifically, EPS is calculated as the pollution emissions per unit area divided by water quality standards. The water quality standards for nitrogen and phosphorus are 1 mg/L and 0.2 mg/L, respectively [44].

5. Results

5.1. Baseline Model Results

This section examines the effect of the full-cost insurance (FCI) policy on fertilizer application.

Table 2. Effects of FCI on fertilizer application.

Variables	Fer_Total		Fer_Efficiency
	Equation (1)	Equation (2)	Equation (3)	Equation (4)
FCI	−19.727 *** (7.108)	−21.761 *** (7.086)	1.797 *** (0.426)	1.915 *** (0.419)
A_industry		−0.181 (0.384)		0.030 (0.023)
Urbanization		−0.762 * (0.458)		0.049 * (0.027)
Income		−30.781 *** (9.656)		−0.206 (0.572)
Ir_land		−7.318 (6.575)		−1.625 *** (0.389)
F_support		−4.193 (3.493)		0.137 (0.207)
p_value		−2.818 (2.533)		0.260 * (0.150)
Constant	312.578 *** (5.309)	482.947 *** (97.854)	9.552 *** (0.318)	28.423 *** (5.792)
Time-fixed effect	YES		YES
Province-fixed effect	YES		YES
N	425	425	425	425
R2	0.430	0.456	0.297	0.347

Note: ***, **, and * indicate significance at 1%, 5%, and 10%. N is the sample size. R² stands for model fit.

presents the regression results, with Equations (1) and (3) showing results without control variables and Equations (2) and (4) incorporating control variables.

From Table 2, Equations (1) and (2) indicate that the FCI policy negatively affects fertilizer application intensity at a 1% level. Specifically, Equation (2) shows that the FCI policy reduces fertilizer inputs per unit area by 21.761%. The data also reveal that the average fertilizer inputs were 349.920 kg per hectare between 2018 and 2021. The FCI policy is expected to decrease fertilizer inputs by 76.146 kg per hectare. Additionally, urbanization and income significantly and negatively affect fertilizer application intensity. This suggests that accelerated urbanization has led to a high degree of off-farming, resulting in lower fertilizer inputs. The increase in disposable income could encourage farmers to adopt high-risk, high-yield technologies and increase inputs of novel fertilizers while decreasing traditional fertilizer inputs.

In Equations (3) and (4), the FCI policy significantly contributes to fertilizer utilization efficiency. Specifically, Equation (4) shows that the FCI policy increases fertilizer utilization efficiency by 1.915%, indicating that implementation of the FCI pilot can enhance fertilizer utilization efficiency in the pilot provinces. Furthermore, urbanization and agricultural product value (p_value) promote fertilizer utilization efficiency. In contrast, arable land irrigation (Ir_land) negatively affects fertilizer utilization efficiency at a 1% significance level.

5.2. Identification of the Hypothesis Test

5.2.1. Parallel Trend Test

The parallel trend hypothesis means that, before the implementation of the FCI policy, there existed a parallel trend in the change of fertilizer application between the treatment and control groups. Following existing research [45], the event study method is utilized for analysis. To avoid the interference of other policies, data from two years before the policy implementation to two years after the policy implementation are used for the test. As shown in Figure 4, regarding fertilizer application intensity, the treatment and the control groups exhibited a similar trend in the year before the policy. In the year after policy implementation, the pilot provinces began to reduce fertilizer inputs, although this reduction did not pass the significance test due to a lag in the implementation and promotion of the policy. Fertilizer inputs in pilot provinces exhibited a significant decrease in the second year of policy implementation, indicating that this paper passes the parallel trend test. Regarding fertilizer utilization efficiency, there was no significant difference between the treatment and control groups in the two years preceding the policy implementation. However, fertilizer utilization efficiency increased significantly in the pilot provinces after the policy implementation, further indicating that the parallel trend test is satisfied for this aspect as well.

5.2.2. Placebo Test

A placebo test was conducted to further robust-type analysis to investigate unobservable factors’ impact on disparities in both pre-policy and post-policy implementations. The following procedures were employed to obtain spurious estimated coefficients and plot their distribution: randomly selecting provinces as the pseudo-treatment and pseudo-control groups from the total sample, randomizing the policy pilot time, and repeating this process 500 times. As shown in Figure 5, the spurious estimated coefficients are mainly concentrated around 0, indicating no significant policy effect for the treatment group when randomly assigned. This confirms that the observed policy effect is genuine and not driven by unobservable factors. Hence, the study passes the placebo test.

5.3. Endogenous Problems

The regression results are shown in

Table 3. System GMM analysis methods.

Variables	Equation (5) Fer_Total	Equation (6) Fer_Efficiency
FCI	−4.368 * (2.403)	1.240 ** (0.577)
Control variables	YES	YES
Time-fixed effect	YES	YES
Province-fixed effect	YES	YES
Dependent variables lagged during one period	1.000 *** (0.013)	1.155 *** (0.054)
Sargan test	0.233	0.120
AR (1) test	0.005	0.019
AR (2) test	0.160	0.558
N	400	400

Note: ***, **, and * indicate significance at 1%, 5%, and 10%. N is the sample size. R² stands for model fit.

. The p-value of the Sargan test is greater than 0.1, indicating the absence of an over-identification problem in the model. The p-values of the AR (2) test are 0.160 and 0.558, which indicates no second-order autocorrelation in the models. Based on these results, Equations (5) and (6) accept the assumptions of “no autocorrelation of disturbance terms” and “valid instrumental variables”. This section employs system GMM regression to demonstrate the statistically significant effect of FCI in reducing fertilizer application intensity and promoting fertilizer utilization efficiency.

5.4. Heterogeneity Analysis

The frequency and severity of natural disasters reflect differences in arable land resource endowment. Thus, under the same coverage levels, regions with varying levels of natural disaster risks may experience divergent impacts of agricultural insurance on fertilizer inputs [22]. This study references Liang’s 2010 research [46], which categorizes provinces based on the impact of natural disasters on food security. The cluster analysis was used to classify risk. The provinces were then classified into high-risk and low-risk areas, with the regression results shown in

Table 4. Heterogeneity analysis.

Variables	Fer_Total		Fer_Efficiency
	High-Risk	Low-Risk	High-Risk	Low-Risk
FCI	−28.854 *** (9.135)	−6.816 (15.695)	1.493 ** (0.637)	2.063 *** (0.638)
Control variables	YES		YES
Time-fixed effect	YES		YES
Province-fixed effect	YES		YES
N	237	187	237	187
R2	0.458	0.530	0.364	0.410

Note: ***, **, and * indicate significance at 1%, 5%, and 10%. N is the sample size. R² stands for model fit.

.

Table 4 shows that the FCI pilot can significantly reduce fertilizer application intensity in high-risk areas. However, the effect on fertilizer application intensity in low-risk areas could not be apparent. After farmers in high-risk areas are insured, their expected losses decrease significantly. Consequently, agricultural insurance motivates farmers to adjust their production and management decisions, leading to a reduction in fertilizer application intensity. In contrast, farmers’ expected losses are lower in low-risk areas, resulting in a limited welfare effect of agricultural insurance. Therefore, the reduction in fertilizer application intensity due to FCI in low-risk areas is not significant. Regarding fertilizer utilization efficiency, FCI promotes fertilizer utilization efficiency in high-risk areas at a 5% level and in low-risk areas at a 1% significance level. This indicates that the FCI policy significantly improves fertilizer utilization efficiency in high-risk and low-risk areas, with the improvement being more pronounced in low-risk areas.

5.5. How Does FCI Affect Fertilizer Application?

The baseline results indicate that the full-cost insurance (FCI) policy effectively reduces fertilizer application intensity while enhancing fertilizer utilization efficiency. The mechanisms through which FCI exerts its influence on fertilizer application can be elucidated as follows. Firstly, FCI may contribute to the expansion of crop cultivation areas and the promotion of large-scale production and specialized operations. This expansion can lead to a reduction in fertilizer application intensity by achieving economies of scale and improving fertilizer utilization efficiency through the implementation of more precise input management and scientific field practices. Secondly, FCI may incentivize off-farm employment, thereby decreasing fertilizer application intensity. However, the resultant increase in non-farm income can facilitate the substitution of fertilizer with machinery, thus enhancing fertilizer utilization efficiency. Lastly, FCI supports the advancement of agricultural technology, encouraging the use of innovative agricultural fertilizers and reducing dependency on traditional fertilizer inputs. To further explore these mechanisms, this study examines the impact of FCI on key factors such as the agricultural labor force, land scale, and agricultural technological progress.

The outcome in Equation (7) of

Table 5. The effect of FCI on the mediator variables.

Variables	Equation (7) Agricultural Labor Force	Equation (8) Land Scale	Equation (9) Agricultural Technological Progress
FCI	−1.167 * (0.640)	0.200 ** (0.095)	0.012 (0.016)
Control variables	YES	YES	YES
Time-fixed effect	YES	YES	YES
Province-fixed effect	YES	YES	YES
N	425	425	425
R2	0.169	0.083	0.459

Note: ***, **, and * indicate significance at 1%, 5%, and 10%. N is the sample size. R² stands for model fit.

demonstrates a significant positive impact of full-cost insurance on the land scale, indicating that full-cost insurance can reduce fertilizer application intensity by achieving economies of scale. Concurrently, the expansion of land scale fosters specialized management, thereby enhancing fertilizer utilization efficiency. Furthermore, the results from Equation (8) indicate that full-cost insurance negatively affects the agricultural labor force at a 10% significance level, suggesting that full-cost insurance influences fertilizer application by reducing agricultural labor inputs. This finding supports H2. Lastly, according to the results of Equation (9), the effect of full-cost insurance on agricultural technological progress is not significant, potentially due to the slower development of agricultural technology. Consequently, during the pre-pilot period of full-cost insurance, the advancements in agricultural technology may not have been fully realized.

5.6. Robustness Checks

5.6.1. Shortening the Sample Years

The base regression of this paper utilizes data from 2005 to 2021. Considering that the FCI pilot year was 2018, a robustness test was conducted using data from 2010 to 2021. The results, shown in

Table 6. Robustness checks.

Variables	(10) Shorten the Sample Years		(11) Interference of Other Policy		(12) Joint Fixed Effect
	Fer_Total	Fer_Efficiency	Fer_Total	Fer_Efficiency	Fer_Total	Fer_Efficiency
FCI	−22.052 *** (3.670)	1.749 *** (0.304)	−20.051 *** (7.020)	1.896 *** (0.419)	−20.143 *** (7.051)	1.891 *** (0.421)
Agricultural catastrophe insurance			19.053 *** (6.693)	−0.587 (0.400)
Control variables	YES		YES		YES
Time-fixed effect	YES		YES		YES
Province-fixed effect	YES		YES		YES
Province × time fixed effect	NO		NO		YES
N	300		425		425
R2	0.571	0.571	0.467	0.349	0.467	0.349

Note: ***, **, and * indicate significance at 1%, 5%, and 10%. N is the sample size. R² stands for model fit.

(10), are consistent with the base regression, supporting the initial conclusions.

5.6.2. Interference of Other Policies

China has adopted a gradual approach to promoting policy-based agricultural insurance. To avoid the influence of other policy-based insurance on fertilizer application, the dummy variable for agricultural catastrophe insurance is added to the basic regression. As seen in Table 6 (11), after including agricultural catastrophe insurance, there is a reduction in the absolute value of the FCI coefficients, but the results remain significant, indicating that the findings are robust.

5.6.3. Joint Fixed Effect

During the implementation of FCI, each province designs a program tailored to its specific conditions. These programmatic differences can have varying effects on grain production inputs across provinces. To account for these differences, a province × time fixed effect was added to the base model. The results in Table 6 (12) reveal that the model outcomes are more stable, further confirming the robustness of the findings.

6. Additional Analysis

A line graph was plotted to depict pollution emissions per unit area from 2005 to 2021, based on the average values of TN, TP, and EPS. As illustrated in Figure 6, the trends for all three types of pollution emissions show an initial increase followed by a subsequent decrease, with 2016 identified as a turning point. This shift is likely attributable to the issuance of the Zero Growth Action Plan for Fertilizer Use by 2020 by the government in 2015, thereby confirming the effectiveness of policy implementation. Post-2016, the rates of decrease in pollution emissions varied, with TN decreasing the most, followed by EPS and then TP.

Furthermore, relevant regression analyses were conducted to explore the environmental effects of FCI. Due to the large base of EPS, we logarithmize it in the regression analysis. The results, presented in

Table 7. Environmental effects of full-cost insurance.

Variables	TN	TP	Ln.EPS
Full sample	−11.593 *** (4.256)	−2.577 *** (0.969)	−0.354 * (0.019)
Wheat	−4.712 (7.344)	−1.932 (1.736)	−0.035 (0.033)
Rice	−12.056 (7.966)	−5.599 *** (1.692)	−0.063 * (0.033)
Corn	−12.223 ** (4.808)	−1.935 * (1.126)	−0.038 (0.024)

Note: ***, **, and * indicate significance at 1%, 5%, and 10%. N is the sample size. R² stands for model fit.

, demonstrate that FCI negatively affects TN, TP, and EPS. Specifically, in comparison to non-pilot areas, the pilot regions resulted in average reductions of 11.593% in nitrogen emissions per unit area (TN), 2.577% in phosphorus emissions per unit area (TP), and 35.400% in equivalent pollution emissions per unit area (Ln.EPS). In terms of planting crop types, full-cost insurance can significantly reduce TP and EPS in the major rice-producing area; while full-cost insurance can significantly reduce TN and TP in the major corn-producing area; however, it is difficult for full-cost insurance to create a better environmental effect on the major wheat-producing area.

7. Discussion

The more effective use of fertilizer is of great significance for addressing the contradiction between “increasing production and reducing pollution” in China. In this context, the government’s full-cost insurance policy can effectively enhance farmers’ enthusiasm for grain production, thereby influencing their production and management behaviors and ultimately guiding rational fertilization practices. Building on the work of Horowitz & Lichtenberg (1993) [20], this paper hypothesizes that full-cost insurance can significantly reduce fertilizer application intensity and improve fertilizer use efficiency. Using the quasi-natural experiment of China’s pilot full-cost insurance policy for the three major grain crops, a DID model was employed to test this hypothesis.

There is a negative relationship between full-cost insurance and fertilizer application intensity, like the findings of Ma et al. (2021) [41] and Mishra et al. (2005) [47], indicating that full-cost insurance can reduce fertilizer application intensity. The data show that the average fertilizer input from 2018 to 2021 was 349.920 kg per hectare, while the full-cost insurance policy reduced fertilizer input by 76.146 kg per hectare. It is important to note that the relationship between agricultural insurance and fertilizer application intensity may vary depending on the type of insurance, the regional policies, and the subjects of the study. This research primarily focuses on the macro performance of China’s full-cost insurance policy on fertilizer application intensity.

Full-cost insurance has a significant positive impact on fertilizer use efficiency. The study of Wang et al. (2023) [48] also points out that agricultural insurance not only reduces the intensity of fertilizer application but also improves the efficiency of fertilizer application. However, it is noteworthy that fertilizer alone does not completely contribute to improved efficiency. Farmers may adopt additional technologies or improve soil fertility to enhance fertilizer use efficiency, which explains why the impact of full-cost insurance on fertilizer use efficiency is relatively small. Our study also concludes that agricultural insurance not only reduces fertilizer application per unit area but also increases crop yields per unit of fertilizer input, i.e., the fertilizer reduction effect brought about by the development of agricultural insurance does not come at the cost of crop yield loss, but rather is achieved by increasing the efficiency of fertilizer use.

This study also has some limitations. Firstly, in research on fertilizer use efficiency, factors such as farmers adopting pest control technologies and applying more fertilizer, or other endogenous variables, might influence efficiency. Future research could incorporate these factors to gain a comprehensive understanding of the relationship between agricultural insurance and fertilizer use efficiency. Secondly, due to data limitations, it is challenging to use micro-level data to validate the hypotheses of this paper. In the future, field surveys will be conducted to obtain more accurate and detailed data to address these limitations.

8. Conclusions

Using provincial panel data from 2005–2021, we analyze the impact of the full-cost insurance policy on fertilizer application intensity and fertilizer utilization efficiency. The main conclusions are as follows. First, the implementation of the full-cost insurance policy resulted in a 21.761% reduction in fertilizer application intensity and a 1.915% increase in fertilizer utilization efficiency. Second, the policy’s mechanism indicates that it reduces fertilizer application intensity and improves fertilizer utilization efficiency by expanding land scale and reducing the agricultural labor force. However, the progress of agricultural technology does not show a significant mediating effect on fertilizer application. Third, heterogeneity analysis suggests that full-cost insurance significantly enhances fertilizer utilization efficiency in both high-risk and low-risk areas. Furthermore, the policy significantly reduces fertilizer application intensity in high-risk areas, although its effect on low-risk areas is less pronounced. Lastly, the full-cost insurance policy has notable environmental benefits, reducing nitrogen pollution emissions by 11.593%, phosphorus pollution emissions by 2.577%, and equivalent pollution emissions by 35.400%.

These findings demonstrate that full-cost insurance not only supports national security goals and stabilizes farmers’ incomes but also has significant environmental benefits. Based on these conclusions, the following insights and recommendations are provided. First, given that the policy reduces agricultural fertilizer application, it is essential to enhance its promotion. Emphasizing the “high subsidies, high payouts” characteristic, efforts should be made to promote the transmission and diffusion of full-cost insurance in both major and non-major grain-producing provinces, especially expanding coverage in high-risk areas. This will increase the structure of food cultivation and encourage agricultural operators to grow grain. Additionally, expanding insurance coverage in high-risk areas will promote fertilizer reduction and efficiency, thereby supporting sustainable agricultural development. Second, differentiated rates and mechanisms for subsidizing agricultural premiums should be implemented. Differentiation of regional risks can lead to different fertilizer input behaviors. The government can adopt differentiated agricultural insurance rates to ensure the development of agricultural insurance in different regions. Specifically, the government can differentiate provinces into different production risk classes, which is mainly based on the actual risk of each region for the rate determination. For areas with higher natural risks and greater agricultural production risks, the government should increase policy support to enhance risk protection. Third, the publicity work on fertilizer reduction and efficiency should be strengthened. For farmers, relevant training should be carried out to improve farmers’ knowledge of fertilizer application, guide farmers to grasp the reasonable amount of fertilizer to be applied, and give full play to the subjective initiative of farmers in fertilizer reduction. At the same time, the government wants family farms, farmers’ cooperatives, and other large-scale organizations to provide fertilizer reduction technology, and advanced fertilizer reduction technology to guide small farmers, giving full play to the subjective initiative of small farmers. Then, farmers should be encouraged to introduce non-farm income into agricultural production investment and be guided to promote the ecologization of arable land while raising their income. In addition, farmers should be guided to promote ecological cultivation methods while raising their incomes. At the same time, the government should promote the transfer of agricultural land and appropriate scale operations and facilitate land operators to reduce the application of chemical fertilizers to protect the long-term productivity of the land. These recommendations aim to optimize the implementation of full-cost insurance policies to ensure that their economic and environmental benefits are maximized.