The Impact of Agricultural Fiscal Expenditure on Water Pressure in Grain Production: Provincial-Level Analysis in China

Abstract

Financial support for agriculture has mainly focused on grain production, while insufficient efforts have been made to ensure water security, potentially intensifying water pressure in grain production (WPGP). This study applies the entropy weight Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) method to measure WPGP from the perspective of sustainable agricultural water use, investigating the impact of agricultural fiscal expenditure on WPGP. Our findings reveal several key points. First, there is a clear linkage between the spatial and temporal patterns of fiscal support and WPGP. Projections indicate that water pressure for grain production in China will continue to rise from 2019 to 2030, with the fastest increases in the Northeast and Huang-Huai-Hai regions, at 20.53% and 13.39%, respectively. Second, agricultural fiscal expenditure distorts the allocation of grain production factors, causing cultivation areas to expand beyond local water resource capacity and, thus, exacerbating WPGP. This effect exhibits a time lag due to the gradual nature of factor allocation. Further analysis shows that in non-major grain-producing regions, lower production efficiency and higher opportunity costs of factor use weaken the impact of fiscal expenditure on WPGP compared to major grain-producing regions. Third, in regions with advanced technical conditions for grain production, the negative impact of agricultural fiscal expenditure on WPGP is mitigated by higher irrigation technology levels, improved water allocation efficiency, and lower water demand per unit of grain. Fourth, the public good characteristics of water resources and water conservancy facilities—namely, strong externalities and non-exclusivity—along with the agronomic demonstration effect, lead to a spatial spillover effect of agricultural fiscal expenditure on WPGP.

1. Introduction

In recent years, food security has become an increasingly urgent global issue [1], due to the regional wars. Food security was an perennial theme related to national stability and people’s well-being, while ensuring sustainable water utilization was essential for realizing food security [2]. China’s water use amounted to only 26% of the global average, indicating that the country faced a severe water shortage, yet agriculture accounted for 61.15% of the total water consumption in 2019. That was far more than the combined amount of water used for other purposes. China’s agricultural production was highly dependent on irrigation, especially grain production. For example, in 2019, irrigated cultivated land accounted for 50% of the total cultivated area, which produced about 70% of the country’s food output. Water for grain production accounted for the vast majority of the total agricultural water consumption [3]. Thus, water shortage seriously restricted sustainable food production in China. Some studies have shown that the mismatch between grain production and spatial distribution of water had led to increasing water pressure in some regions, such as the Huang-Huai-Hai region and Northeast China. The massive project of “South-to-North Water Diversion” and “Grain transportation from the North to the South” was the last resort for China’s agricultural production and consumption [4]. Although the Chinese government had implemented more than 20 large-scale inter-basin water transfer projects to address the spatial mismatch between water supply and demand, these initiatives had not substantially alleviated the water pressure in grain production.

In this study, agricultural fiscal expenditure refers specifically to the financial resources allocated by the government to support agricultural activities, including but not limited to subsidies for grain production, investment in irrigation and water conservancy infrastructure, research and development in agricultural technologies, and agricultural extension services. According to relevant data, China’s fiscal expenditure on agriculture, forestry, and water affairs amounted to CNY 2.286 billion in 2019, accounting for 12.01% of national fiscal revenue and 32.44% of the GDP of the primary industry. Between 2003 and 2019, the Huang-Huai-Hai and Northeast regions received the highest proportion of annual agricultural fiscal expenditure, accounting for 41.19% of the national total. These two regions comprise 44.26% of China’s cultivated land and 54.76% of its grain planting area, yet they possess only 9% of the country’s total water resources. This spatial mismatch highlights the potential linkage between agricultural fiscal expenditure and land use intensity, particularly grain cultivation. Notably, fiscal expenditures in water-related areas have primarily focused on water extraction and utilization, with relatively limited investment in water conservation and protection. This imbalance may have contributed to increasing water pressure in grain-producing areas. Therefore, this study aims to analyze whether agricultural fiscal expenditure has exacerbated water pressure in northern China and to examine the underlying mechanisms. Clarifying the specific components and effects of fiscal support for agriculture is essential for optimizing subsidy policy and promoting sustainable agricultural development. Hence, this study aims to explore the reasons behind the increasing water pressure of regional grain production from agriculture-related financial subsidies.

Water pressure refers to the degree of pressure on water quality or quantity caused by human and socio-economic activities in specific natural and social environments that exerted influences on water resource utilization beyond the capacity of the water environment. The water pressure index not only can be used to assess regional water scarcity but also provides data subsidies for influencing factors of water pressure. Some scholars decomposed the research perspective of water pressure into water quantity pressure, water quality pressure [5], water ecological environment pressure [6], economical water pressure [7], water technology pressure, and water population pressure [8], etc. However, most of the previous studies focused on the quantity and quality of water and paid special attention to calculating the water pressure index [9,10,11,12,13]. The water resource pressure index method was also widely used in agricultural production research. The existing method of measuring water pressure index had laid a solid foundation for the measurement of indicators in this study. However, the measurement process lacked weight assignments for the differences in endowments among different regions, which reduced the accuracy of the measurement results.

Scholars also analyzed the influence on regional water resource pressure from the perspectives of crop types [14], planting structure adjustment, and planting crop layout [15]. Changes in the regional water pressure index were also explained from the perspective of the water footprint of crop growth [16]. Additionally, the water index method has also been used to explore the increasing pressure of regional water and unsustainable food production caused by the spatial mismatch between food production and water and show that the spatial mismatch between food production and water in China was decreasing, especially in Huang-Huai-Hai region and Northeast China. Research on water pressure in grain production needed accurate measurement methods and empirical support.

The research on agricultural fiscal expenditure was mainly carried out from three aspects: the influence of agricultural fiscal expenditure on farmers’ income [17], rural economic growth [18,19,20,21], and farmers’ consumption [22]. These studies should pay greater attention to resource conservation rather than solely focusing on benefits.

From the above analysis, it is evident that previous research on agricultural fiscal expenditure and water pressure, as well as their influencing factors, has provided a rich theoretical and empirical foundation for this study. However, important gaps still remain. First, existing studies on agricultural fiscal expenditure have primarily focused on benefit-oriented outcomes, such as impacts on farmers’ income, rural economic growth, and consumption behavior. Few studies have examined the potential environmental consequences of such expenditures, particularly from a resource conservation perspective. In contrast, this study shifts the analytical lens by exploring how agriculture-related financial subsidies may inadvertently intensify water pressure in grain-producing regions, thereby contributing to unsustainable agricultural practices.

Second, while earlier studies on water pressure in agriculture have largely centered on factors such as crop species, planting structure adjustment, climate change, and the water footprint of crop growth, they have seldom integrated fiscal policy as a potential driver. This study fills that gap by using agricultural fiscal expenditure as the analytical entry point to investigate the mechanisms through which government financial support influences the spatial mismatch between grain production and water availability in China.

Third, with regard to methodological advancements, previous research has often employed the 40% threshold of sustainable water use in agriculture to measure water pressure, which may oversimplify regional differences. To address this limitation, we introduce water resources carrying capacity as a weighted index in our evaluation of water pressure in grain production (WPGP). This improved index better reflects interregional heterogeneity in water resource endowments and offers a more scientific and accurate measure of the sustainability of grain production [23].

Therefore, our study not only extends the existing literature on agricultural fiscal expenditure by incorporating an environmental sustainability perspective, but it also deepens the understanding of water pressure dynamics in grain production through a more refined empirical approach.

2. Methods and Materials

2.1. Methods

2.1.1. Water Pressure in Grain Production

The water resource pressure index calculation mainly included the single index method, supply-demand ratio method, comprehensive evaluation method, and water footprint method. This study followed operational and scientific principles and referred to the existing literature [24]. According to the research needs, the water pressure index was calculated by combining a single index method and a comprehensive evaluation method. This is primarily because economic, social, and ecological conditions differ greatly across various regions of China. In these areas, the intensity of water use related to grain cultivation is shaped not only by the natural availability of water resources but also by regional economic structures, societal dynamics, and environmental considerations. Only 40% of water endowment was used to judge whether agricultural water was sustainable in previous studies. The influence of the above factors on regional water resources carrying capacity was not considered. We included water resources, allowing the ability to reflect regional differences in water endowment, economic, social, and ecological aspects. To represent the pressure variable associated with water use in grain production, we employed the ratio of water consumed for grain cultivation to the region’s water carrying capacity. This metric offers a scientifically sound and precise reflection of the degree of water consumption pressure experienced in each area. The specific expression was as follows.

$$Wpg{p}_{it}={{\displaystyle \frac{FW{C}_{it}}{TW{C}_{it}•WRCC}}}_{it}$$

(1)

In Formula (1), Wpgp_it represented the water pressure index of grain production in region i in year t; FWC_it represented the water consumption of grain production in region i in year t, and TWC_it meant the total water consumption in region i in year t. WRCC_it represented the water resources carrying capacity of region i in year t.

$$FW{C}_{it}={\displaystyle \frac{GS{A}_{it}•AW{C}_{it}}{AS{A}_{it}}}$$

(2)

In Formula (2), GSA_it represented the grain sown area in region i in year t; ASA_it represented the planted area of crops in region i in year t, and AWC_it meant the agricultural water consumption in region i in year t.

Drawing from four key dimensions—namely, the water, economic, social, and ecological subsystems—we identified 23 indicators (as detailed in

Table 1. Water carrying capacity measurement index.

Target Layer	Rule Layer	Index Layer	Calculation	Category
Water resource capacity	Water subsystem (B1)	Runoff modulus (C1)	Total water resources/regional area	Negative
		Proportion of surface water supply (C2)	Surface water resources/water supply	Negative
		Water supply modulus (C3)	Water supply/total area	Negative
		Total domestic water (C4)	Statistical yearbook	Negative
		Water resource utilization per capita (C5)	Water consumption/total population	Negative
		Per capita water resources (C6)	Total water/total population	Positive
		Utilization rate of water development (C7)	Water supply/total water	Negative
		Average annual precipitation (C8)	Statistical yearbook	Positive
	Economic subsystem (B2)	GDP (C9)	Statistical yearbook	Negative
		Per capita GDP (C10)	Statistical yearbook	Negative
		Proportion of output value of secondary and tertiary industries (C11)	(Secondary industry output value + tertiary industry output value)/total output value	Negative
		Proportion of output value of primary industry (C12)	(Primary industry output value)/total output value	Positive
		Freshwater production (C13)	Statistical yearbook	Negative
	Social subsystem (B3)	Urbanization rate (C14)	Statistical yearbook	Negative
		Urban water population (C15)	Statistical yearbook	Negative
		Population density (C16)	Total population/area	Negative
		Natural population growth rate (C17)	Statistical yearbook	Negative
		Water consumption per capita (C18)	Total domestic water consumption/total population	Negative
	Ecological subsystem (B4)	Forest coverage (C19)	Statistical yearbook	Positive
		Water consumption rate of ecological environment (C20)	Total ecological water use/total water resources	Negative
		Agricultural fertilizer application amount (C21)	Statistical yearbook	Negative
		Afforestation area (C22)	Statistical yearbook	Positive
		Sewage discharge (C23)	Statistical yearbook	Negative

) to assess the water resources carrying capacity (WRC) across 31 provincial-level regions in China over the period 2003 to 2018. For the comprehensive evaluation, we adopt the entropy-weighted TOPSIS method to measure the water pressure in grain production (WPGP) across Chinese provinces. The rationale for selecting this method lies in its compatibility with the multidimensional and regionally heterogeneous nature of the data. The TOPSIS method evaluates alternatives based on their relative distance to the ideal and negative-ideal solutions, which is particularly suitable for sustainability assessments where optimal and suboptimal conditions are conceptually well-defined. To enhance the objectivity of this process, entropy weighting is applied to determine indicator weights based on the degree of dispersion in the data, thereby reducing the subjectivity associated with expert scoring or uniform weighting. Furthermore, the entropy-TOPSIS approach allows for the integration of diverse indicators across water, ecological, economic, and social subsystems, capturing the complex nature of agricultural water pressure. It also facilitates interprovincial comparisons over time, which is essential for understanding regional patterns and policy implications. Therefore, the entropy-weighted TOPSIS method provides a robust and theoretically grounded framework that aligns with the study’s goals of evaluating sustainability and informing targeted policy interventions [13]. This study followed the method of Peng et al. [25] and was modified according to our research situation. The calculation process for the entropy weight TOPSIS method refers to Li et al. [26] and Li et al. [27].

2.1.2. Prediction Method of Water Pressure in Grain Production

The autoregressive integrated moving average model (ARIMA) model, known for its classical structure, remains one of the most extensively applied approaches in time series prediction. In this study, ARIMA was employed to predict the future trend of the water pressure in grain production (WPGP) index based on historical data. The modeling process followed a standard three-step procedure: model identification, parameter estimation, and diagnostic checking.

First, the augmented Dickey–Fuller (ADF) test was used to assess the stationarity of the original time series. If the series was non-stationary, differencing was applied to achieve stationarity. Second, the optimal order of the autoregressive (p) and moving average (q) components was determined based on autocorrelation function (ACF) and partial autocorrelation function (PACF) plots, supplemented by model selection criteria such as the Akaike information criterion (AIC) and Bayesian information criterion (BIC). Finally, residual diagnostics were conducted to ensure the white noise properties of residuals, including Ljung–Box Q tests for autocorrelation. The model was used to predict the future [28]. The expressions of the ARIMA model were as follows.

$$A(f){(1-h)}^d{X}_t=B(f){\epsilon }_t$$

(3)

In Formula (3), X_t was the time series value, f was the backward moving algorithm of time t, h = αf, α was the coefficient, and ε_t was the error or impact value.

$$A(z)=1-{\displaystyle \sum _{j=1}^p{a}_j}{z}^j\ne 0(|z|\le 1),B(z)=1+{\displaystyle \sum _{j=1}^q{b}_j}{z}^j\ne 0(|z|\le 1)$$

(4)

In Formula (4), A(Z) was the expansion of A(f); z also represented the lag operator, a_j was the autoregression coefficient, b_j was the moving average coefficient, p was the number of autoregression terms, d was the number of differences (order) made to make it a stationary series, q was the number of moving average terms, and j was a positive integer (1, 2, 3, …, N). The model was abbreviated as ARIMA (P, D, and Q). Based on the ARIMA model, the water pressure in grain production in 12 provinces of northern China in the next 10 years was forecasted and analyzed.

2.1.3. Model Setting of the Impact of Agricultural Fiscal Expenditure on the Water Pressure in Grain Production

Based on the above theoretical analysis, appropriate variables were selected for model design. The details are as follows:

Dependent variable: water pressure in grain production (WPGP). The entropy-weighted TOPSIS method was used to calculate the carrying capacity of water. Finally, based on Yu [23], the water pressure index of grain production was calculated from the sustainable utilization of agricultural water, which effectively characterizes water pressure in grain production.

Independent variable: agricultural fiscal expenditure (Feag). Among the former researchers, Huang et al. [29] selected the four agricultural subsidies from 2004; Zhou [30] chose the agricultural subsidies and protection subsidies after 2016. Our study considered that fiscal transfer payments to agriculture were generally counted as expenditure items for agricultural, forestry, and water affairs. Therefore, the provincial-level expenditure on agriculture, forestry, and water affairs was used as a proxy of agricultural subsidies to a certain extent. Agricultural fiscal expenditure was measured by the index of agriculture-related budget spending in the statistical yearbook of each province. We selected this budget category of 31 provincial administrative regions in China from 2003 to 2018 as the measurement index of agricultural fiscal expenditure.

Moderation variable: grain production technology (Gpte). This research considered that the technical background of grain production could reflect the efficiency of regional grain production and determine the water consumption per unit yield of grain to enhance or weaken the aggravating effect of financial and agricultural subsidies on the WPGP [31]. The technical environment of grain production was closely related to the output of grain-related technological achievements. Therefore, it was modified by referring to the study of Ma et al. [32] and measured by the number of grain-related patents in each provincial year.

Control variables: Based on previous research results on influencing factors of water pressure in grain production, our study selected 31 provincial administrative regions in Mainland China from 2003 to 2018. Grain sown area (Grsa), proportion of wetlands (Wetl), grain net profit (Mpmg), nature reserve proportion (Rese), irrigation facilities (Wafa), and soil erosion control level (Waso) were the control variables of this research model [33], and the specific indicators are shown in

Table 2. Control variable measurement index.

Variable	Index	Calculation Method	Unit
Grain sown area	Grsa	Statistical yearbook	×106 ha
Proportion of wetlands	Wetl	Wetland area/provincial land area × 100%	%
Grain net profit	Mpmg	[(Early indica rice net profit + medium indica rice net profit + late indica rice net profit)/3 + Wheat net profit + Corn net profit)]/3	RMB/ha
Nature reserves proportion	Rese	Nature reserve area/provincial territory area × 100%	%
Irrigation facilities	Wafa	Effective irrigated area/sown area of crops × 100%	%
Soil erosion control level	Waso	Soil erosion control area/provincial land area × 100%	%

.

Based on the above theoretical analysis and variable selection, the Tobit model was designed as follows: Equation (5) was the basic model of this study, which was used to test the impact of agricultural fiscal expenditure on the WPGP. Considering the grain production, farmers’ decision making might be affected by a year of direct or indirect subsidies. The agricultural fiscal expenditure for regional grain production and water pressure conduction might need a specific time; to join the first-order lag of agricultural fiscal expenditure, agriculture-related budgetary expenditures on the impact of grain production water pressure hysteresis effect were included (Equation (6)).

$$Wpgp=\left\{\begin{array}{ll}{{Wpgp}^{\ast }}_{it}=\sigma +{\alpha }_1{Feag}_{it}+{\alpha }_2{X}_{j,it}+{\mu }_i+{\epsilon }_{it}\hfill & {{Wpgp}^{\ast }}_{it}>0\hfill \\ 0\hfill & {{Wpgp}^{\ast }}_{it}\le 0\hfill \end{array}\right.$$

(5)

$$Wpgp=\left\{\begin{array}{ll}{{Wpgp}_{it}}^{\ast }=\sigma +{\alpha }_3{Feag}_{i(t-1)}+{\alpha }_4{X}_{j,it}+{\mu }_i+{\epsilon }_{it}\hfill & {{Wpgp}_{it}}^{\ast }>0\hfill \\ 0\hfill & {{Wpgp}_{it}}^{\ast }\le 0\hfill \end{array}\right.$$

(6)

In Formulas (5) and (6) of this study, Wpgp*_it was the dependent variable, representing the water pressure in grain production in region i in t year; Feag_it was the core explanatory variable, meaning the agricultural fiscal expenditure in region i in t year, and Feag_i(t−1) represented the agricultural fiscal expenditure in region i in the t − 1 year. X_j,it were control variables, representing other factors affecting WPGP. σ represented the constant term of the equation. A represented the corresponding coefficients of each equation; μ_i meant the provincial effect, which was difficult to observe in each province. ε_it represented the random disturbance term. Meanwhile, i = 1, 2, …, 31 meant 31 provincial administrative regions in mainland China; t = 2003, 2004, …, 2018 represents 16 years; and j = 1, 2, …, 5, respectively, represented the proportion of wetland, net profit of grain, proportion of nature reserves, irrigation and water conservancy facilities, and soil erosion control level.

Since agricultural fiscal expenditure, grain production technology, and WPGP were all continuous variables, hierarchical regression analysis is suitable for testing moderation effects. The first step was to carry out the regression of agricultural fiscal expenditure and grain production technology to the WPGP (Formula (7)). The second step was the regression of agricultural fiscal expenditure, grain production technology, and agricultural fiscal expenditure × grain production technology to a WPGP (Formula (8)). For the goodness-of-fit (R²) increases of the two regression models and the sign and significance of the interaction term coefficients, the validity of the regulating effect was judged. Suppose the goodness of fit R² rose and the interaction term coefficient was significant. In that case, it it suggests that grain production technology moderates a moderating role in the impact of agricultural fiscal expenditure on the WPGP.

$$Wpgp=\left\{\begin{array}{ll}{{Wpgp}^{\ast }}_{it}=\sigma +{\lambda }_1{Feag}_{it}+{\lambda }_2{Gpte}_{it}+{\lambda }_3{X}_{j,it}+{\varphi }_i+{\epsilon }_{it}\hfill & {{Wpgp}^{\ast }}_{it}>0\hfill \\ 0\hfill & {{Wpgp}^{\ast }}_{it}\le 0\hfill \end{array}\right.$$

(7)

$$Wpgp=\left\{\begin{array}{ll}{{Wpgp}_{it}}^{\ast }=\sigma +{\xi }_1{Feag}_{it}+{\xi }_2{Gpte}_{it}+{\xi }_3{Feag}_{it}\times {Gpte}_{it}+{\xi }_4{X}_{j,it}+{\omega }_i+{\epsilon }_{it}\hfill & {{Wpgp}_{it}}^{\ast }>0\hfill \\ 0\hfill & {{Wpgp}_{it}}^{\ast }\le 0\hfill \end{array}\right.$$

(8)

In Formulas (7) and (8), Gpte_it was the moderating variable, representing the Grain production technology in region i in t year; λ, ξ represented the corresponding coefficients of each equation; φ_i and ω_i represented the provincial effect that was difficult to observe in each province, and the other symbols were interpreted in the same way as Formulas (5) and (6).

2.1.4. Spatial Econometric Model

The global Moran’s I index [34] can be expressed as follows.

$$\mathrm{Moran} I={\displaystyle \frac{{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{k=1}^m{\omega }_{ik}}}\left(Wpg{p}_i-Wpg{p}_a\right)\left(Wpg{p}_i-Wpg{p}_a\right)}{S^2{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{k=1}^m{\omega }_{ik}}}}}$$

(9)

In Formula (9), i represented the variance of the index S², Wpgp_a represented the mean of Wpgp_i, and ω_ik was the spatial distance weight matrix, defined as the reciprocal of the distance between province i and province k. n and m represented the total number of provinces. The global Moran’s I index value range is [−1, 1]. If it is greater than 0, it is positively correlated; if i is less than 0, it is negatively correlated. A value approaching ±1 in absolute terms indicates a higher degree of spatial dependence.

The Spatial Panel Durbin Model [35] was expressed as follows.

$$Wpg{p}_{it}=\alpha +\rho {\displaystyle \sum _{k=1}^n{\omega }_{ik}}Wpg{p}_{it}+\psi x_{l,it}+\theta {\displaystyle \sum _{k=1}^n{\omega }_{ik}}{x}_{l,kt}+{\phi }_i+{\tau }_t+{\epsilon }_{it}$$

(10)

$$Wpg{p}_{it}=\alpha +\zeta Wpg{p}_{i(t-1)}+\rho {\displaystyle \sum _{k=1}^n{\omega }_{ik}}Wpg{p}_{it}+\psi x_{l,it}+\theta {\displaystyle \sum _{k=1}^n{\omega }_{ik}}{x}_{l,kt}+{\phi }_i+{\tau }_t+{\epsilon }_{it}$$

(11)

In Formulas (10) and (11), x_l,it represented the l variable value of region i in t year, x_l,kt represented the l variable value of province k in t year, l = 1, 2, …, 6 referred to the level of government expenditure subsidizing agriculture, the proportion of wetlands, net profit of grain, reservoir capacity, irrigation and water conservancy facilities, and soil erosion control; ρ, as the spatial lag coefficient of WPGP, represented the spatial correlation of WPGP among different regions, and its magnitude reflected the degree and direction of correlation. ψ referred to X’s unknown parameter vector, representing the influence direction and degree of each explanatory variable on the water pressure of regional grain production. θ, as the spatial lag coefficient of the explanatory variable, represented the influence coefficient of explanatory variables in other regions on the WPGP in this region, and its magnitude reflected the degree and direction of influence. φ_i referred to the regional fixed effect; τ_t was the time-fixed effect; ε_it represented the random disturbance term. Additionally, Formula (11) was based on Formula (10) to include the first-order lag of WPGP (ζ represented grain production water pressure time first-order lag term impact on the current grain production water pressure coefficient) to build the Dynamic Spatial Panel Dubin model, which allows for the examination of water pressure on the space and time lag effect. It could accurately measure the spatial spillover effect of agricultural fiscal expenditure on the water pressure of regional grain production [31]. Hence, this study adopted the Dynamic Spatial Panel Dubin model for testing.

2.2. Materials

The original dataset used in this study comprises data from the statistical data of 31 provincial administrative regions in Mainland China from 2003 to 2018, totaling 496 sample data in 16 years. The data were comprised agricultural fiscal expenditure, GDP, per capita GDP, the added value of the primary industry, the added value of the secondary sector, the added value of the tertiary in the sector output of freshwater products, and electricity, obtained from China Statistical Yearbook. The statistical data of agricultural fertilizer application amount, grain sown area, crop sown area, and soil erosion control area came from China Rural Statistical Yearbook. The statistical data of agricultural water consumption, total water, surface water, water supply, total domestic water consumption, average annual precipitation, and water consumption came from China’s Statistical Yearbook. Statistical data on forest coverage rate, the water consumption rate of ecological environment, sewage discharge, and wetland area came from China Environmental Statistics Yearbook. Statistical data of afforestation area and effective irrigation area came from China Agricultural Statistics. The statistical data of total population, the natural growth rate of population, and the number of employed persons in the primary industry came from China Population and Employment Statistical Yearbook. The land area of each province was from China Land and Resources Statistical Yearbook. The urbanization rate and urban water population were from China Urban Statistical Yearbook. Data on key economic variables were adjusted for price indices using 2003 as the base period. Moreover, the grain patent output data source was retrieved from the Chinese literature search website China National Knowledge Infrastructure (CNKI) patent database.

3. Result

3.1. Analysis of Measurement and Forecast Results of Water Pressure in Grain Production

3.1.1. Analysis of Water Pressure Measurement Results for Food Production

The calculation results of this study showing the specific situation of the mean WPGP and the growth rate in each province from 2003 to 2018 are shown in Figure 1. These results revealed significant regional variations. The WPGP was generally higher in northern China, for example, in Heilongjiang, 2.647; in Ningxia, 2.564; in Jilin, 2.175; in Gansu, 2.013; and in Hebei, 1.929, which were all much higher than the national average value of 1.388 for water pressure in grain production. The provinces with faster growth rates were in mostly in the northern regions, for example, in Anhui, 67.60%; in Henan, 49.90%; in Shandong, 43.10%; in Heilongjiang, 35.00%; in Shanxi, 33.20%, etc. Conversely, the WPGP in southern China was generally low, for example, in Shanghai, 0.263; in Chongqing, 0.662; in Zhejiang, 0.710; in Guangdong, 0.811; and in Fujian, 0.869, etc. In terms of the average annual growth rate from 2003 to 2018, the growth rate of most southern regions was relatively slow, such as 4.40% in Fujian, 7.00% in Hunan, and 7.00% in Xizang. Some provinces even exhibited a downward trend, including Guizhou at −14.10%, Guangdong at −5.60%, Zhejiang at −12.30%, Guangxi at −5.60%, etc. These results were similar to those of existing studies. This was due to the large scale of grain production in the northern region, the relatively large demand for water resources in the agricultural sector, and the generally low water resources carrying capacity in the northern region of China, which made the water pressure in grain production in the northern region generally higher. On the contrary, in the southern region, the scale of grain production was small, the water used for grain production was relatively low, and the water resources carrying capacity in southern China was relatively high. Hence, the water pressure in grain production was generally lower.

In addition, the cultivated land in Northeast China and the Huang-Huai-Hai region accounted for 44.26% of the national total, while the total water only accounts for 9.23%. There was a severe mismatch. Furthermore, from 2003 to 2018, the average agricultural fiscal expenditure accounted for 41.19% of the national total, and the proportion of grain planting area was as high as 54.76%; a higher ratio leading to the strong water pressure index of grain production in these two regions accounted for 43.28% of the national total. The results were similar to those of existing studies [36]. Based on the growth rate from 2003 to 2018, Northeast China (Heilongjiang, 35.00%; Liaoning, 32.80%; Jilin, 26.50%; and Inner Mongolia, 22.50%) and Huang-Huai-Hai (Anhui, 67.60%; Henan, 49.90%; Shandong, 43.10%; and Jiangsu, 28.60%) were far higher than the average growth rate of WPGP of 17.04%. On the contrary, due to the strong water resources carrying capacity, the agricultural fiscal expenditure in South China and Southwest China from 2003 to 2018 was only 27.40% of the national level, and the grain planting area in South China and Southwest China was only 21.48% of the national level. Therefore, the water pressure index ratio for grain production in these two regions is only 21.78% of the national level. While the growth rates of South China (4.40% in Fujian, 5.60% in Guangdong, 1.30% in Guangxi, and 2.20% in Hainan) and Southwest China (14.1% in Guizhou, 5.00% in Yunnan, and 7% in Tibet) were far lower than the national average growth rate of WPGP of 17.04%. Therefore, it could be preliminarily seen that there may be a specific connection between agricultural fiscal expenditure and the WPGP. A follow-up study would focus on the relationship between them.

The evolution of water pressure in grain production in China is shown in Figure 2. From 2003 to 2018, most of the provinces are distributed between 0.50 and 2.00, but the distribution curve of the kernel density function gradually transitions from a single-peak pattern to a multi-peak pattern, indicating that water pressure in grain production in China has a tendency of extreme divergence due to changes in regional resource endowments and socioeconomic and policy factors. In addition, the overall peak per year gradually shifts to the right as time advances, with some values even exceeding 2.5, indicating that the level of water pressure in grain production in most provinces in China has been increasing from 2003 to 2018.

3.1.2. Analysis of Water Pressure Forecast Results for Food Production

From the forecast growth trend of WPGP from 2003 to 2030 (Figure 3), the WPGP in all regions except South China was expected to show an increasing trend. This finding is similar to that of the Zhao et al. [37] study. The Northeast region, Northwest area, and Huang-Huai-Hai region would continue to grow rapidly. This is similar to the Wu et al. [38] study. Due to the relatively large share of arable land and relatively low water endowment in the three regions, coupled with the expansion of regional financial support for agricultural expenditure, the trend of expanding food production would continue to rise. This led to a rapid climb in water pressure for food production. Judging from the forecast results, compared with 2018, the predicted growth rate of water pressure in grain production in 2030 is as follows. The Northwest area would increase by 25.35%, the Northeast region would increase by 20.53%, the Southwest area would increase by 14.01%, the Huang-Huai-Hai area would increase by 13.39%, and the middle and lower reaches of the Yangtze River would increase by 11.64%, while the South China region would decrease by 5.03%. Therefore, we suggested that the water rights trading market gradually be improved in the Northeast region, Northwest area, and Huang-Huai-Hai region. The water price formation mechanism should be promoted to increase the cost of water for grain production. In this way, the water efficiency of grain production would be improved, and the regional water would be saved. Hence, we could relieve the water pressure in grain production.

3.2. Empirical Test Results

3.2.1. Test Results of the Impact of Agricultural Fiscal Expenditure on the Water Pressure in Grain Production

(i)

Agricultural fiscal expenditure has intensified the water pressure of food production.

It could be seen from the estimation results of Model 1 in

Table 3. Test on the relationship between fiscal expenditure for agriculture and water pressure in grain production.

Variables	WPGP
	Model 1	Model 2 (IV)	Model 3
Fiscal expenditure for agriculture (Feag)	0.031 *** (5.838)	0.026 *** (3.664)
Lag_Feag			0.029 *** (4.928)
Grain sown area (Grsa)	0.138 *** (7.940)	0.167 *** (7.264)	0.128 *** (6.958)
The proportion of wetlands (Wetl)	−0.304 (−0.991)	−0.073 (−0.218)	−0.303 (−0.998)
Grain net profit (Mpmg)	0.439 (0.979)	−0.199 (−0.314)	0.170 (0.372)
The proportion of nature reserves (Rese)	0.003 *** (2.692)	0.003 ** (2.347)	0.003 ** (2.474)
Irrigation facilities (Wafa)	0.055 (0.356)	0.057 (0.347)	0.059 (0.380)
Soil erosion control level (Waso)	−0.312 (−1.436)	−0.286 (−1.222)	−0.386 * (−1.786)
F statistics in phase one (F-statistic)		17.42
Wald chi2		41.83 ***
Constant	0.786 *** (5.685)	0.755 *** (6.651)	0.898 *** (6.330)
Observations	496	496	465

Note: ***, **, and * represent significance at 1%, 5%, and 10%, respectively, and the numbers in brackets are z values.

(Based on Formula (5)) that agricultural fiscal expenditure had a significant impact on the WPGP at a 1% level, with a coefficient of 0.031, indicating that agricultural fiscal expenditure had a significant positive impact on the WPGP. Additionally, the rising WPGP meant governments needed to regulate it. To protect the ecological environment, the government might increase agricultural fiscal expenditure in this area and took measures such as building water conservancy facilities or providing fallow grants to control the excessive WPGP. Therefore, there might be reverse causality between WPGP and financial expenditure for agriculture. This study dealt with reverse causality using the instrumental variable. Due to the government’s “general public budget income” and “general public spending” being highly correlated, and that the following transaction expenditure “accounts for the proportion of the” general public spending that was “usually was relatively stable; thus, the” available public budget income “with the endogenous variable” fiscal expenditures for subsidizing agriculture “correlation, and regression model of the current period of the random error term and the pressure of water use in grain production, were not related. Therefore, this study chose “general public budget revenue (Reve)” as an instrumental variable. We first tested the effectiveness of instrumental variables. As shown in Model 2, the estimated F value of 17.42 in the first stage was far higher than the critical value of 10 in the weak instrumental variable test. The p-value of the Wald endogeneity test significantly indicated that “agricultural fiscal expenditure” was an endogenous variable. This showed that “general public budget revenue” was a more appropriate instrumental variable of “agricultural fiscal expenditure”. The two-stage estimation results showed that the regression coefficient of agricultural fiscal expenditure on the WPGP was 0.026, which was significant at the level of 1%, further indicating that agricultural fiscal expenditure would aggravate the WPGP. In addition, considering that farmers’ grain production decisions might be affected by the fiscal expenditure on agriculture in the previous year, that is, it may take some time for the agricultural fiscal expenditure to transmit the impact of regional water pressure on grain production, the first-order lag term of agricultural fiscal expenditure was added for regression. The results were shown in Model 3 (Based on Formula (6)) and passed the significance test at 1% level with a coefficient of 0.029, indicating that there was indeed a certain time lag in the impact of agricultural fiscal expenditure on the water pressure of regional grain production.

(ii)

Robustness test

To further test the robustness of the estimation results, this study divided the samples into northern and southern regions for testing (

Table 4. Robustness testing in the north and south regions between fiscal expenditure for agriculture and the water pressure index of grain production.

Variables	WPGP in Northern China			WPGP in Southern China
	Model 4	Model 5 (IV)	Model 6	Model 7	Model 8 (IV)	Model 9
Fiscal expenditure for agriculture (Feag)	0.040 *** (3.716)	0.036 ** (2.572)		0.030 *** (5.907)	0.015** (2.054)
Lag_Feag			0.033 *** (3.003)			0.028 *** (4.970)
Grain sown area (Grsa)	0.116 *** (4.659)	0.151 *** (4.060)	0.113 *** (4.314)	0.132 *** (4.951)	0.272 *** (6.724)	0.120 *** (5.214)
Proportion of wetlands (Wetl)	0.013 (0.014)	0.048 (0.048)	−0.161 (−0.188)	−0.591 ** (−2.423)	−0.102 (−0.342)	−0.685 *** (−2.842)
Grain net profit (Mpmg)	0.149 (0.146)	−1.274 (−0.833)	−0.633 (−0.597)	0.862 *** (3.106)	−0.022 (−0.036)	0.777 *** (2.980)
Nature reserves proportion (Rese)	0.003 * (1.713)	0.004 * (1.714)	0.003 * (1.704)	0.004 *** (3.133)	0.005 *** (3.355)	0.004 ** (2.540)
Irrigation facilities (Wafa)	−0.160 (−0.658)	0.006 (0.023)	−0.125 (−0.512)	0.479 *** (2.581)	0.639 *** (2.878)	0.569 *** (3.062)
Soil erosion control level (Waso)	−0.519 (−1.201)	−0.853 * (−1.795)	−0.873 ** (−2.037)	−0.158 (−0.696)	0.323 (1.212)	−0.171 (−0.745)
F statistics in phase one (F-statistic)		14.64			15.47
Wald chi2		15,554.38			17,494.08
Prob > chi2		0.000			0.000
Constants	1.362 *** (6.071)	1.408 *** (6.401)	1.590 *** (6.933)	0.250 (1.533)	−0.249 (−1.372)	0.304 ** (1.995)
Observations	240	240	225	256	256	240

Note: ***, **, and * represent significance at 1%, 5%, and 10%, respectively, and the numbers in brackets are z values.

). The estimation results for both the northern and southern regions were similar to those of the national sample. It was confirming that the baseline regression results were relatively robust through the sample test from South to North China.

In addition, samples were divided into major and non-major grain-producing areas for testing (

Table 5. Robustness testing of Major and Non-Major Grain-production areas between fiscal expenditure for agriculture and the water pressure index of grain production.

Variables	WPGP in Major Grain-Producing Areas			WPGP in Non-Major Grain-Producing Areas
	Model 10	Model 11	Model 12	Model 13	Model 14	Model 15
Fiscal expenditure for agriculture (Feag)	0.035 *** (4.303)	0.049 *** (4.909)		0.029 *** (3.463)	0.014 (1.110)
Lag_Feag			0.033 *** (3.718)			0.026 *** (2.835)
Grain sown area (Grsa)	0.119 *** (4.586)	0.111 *** (3.296)	0.113 *** (4.012)	0.262 *** (4.556)	0.332 *** (5.001)	0.241 *** (4.190)
Proportion of wetlands (Wetl)	−0.401 (−0.461)	−1.159 (−1.213)	−0.775 (−0.867)	−0.175 (−0.556)	0.085 (0.243)	−0.145 (−0.480)
Grain net profit (Mpmg)	1.565 (1.098)	−0.294 (−0.158)	0.584 (0.356)	0.216 (0.437)	−0.383 (−0.581)	0.035 (0.072)
Nature reserves proportion (Rese)	0.005 ** (2.310)	0.008 *** (3.233)	0.005 ** (2.065)	0.001 (1.094)	0.000 (0.184)	0.001 (1.017)
Irrigation facilities (Wafa)	0.426 (0.867)	0.652 (1.165)	0.563 (1.095)	0.089 (0.553)	0.180 (1.035)	0.081 (0.518)
Soil erosion control level (Waso)	0.060 (0.123)	−0.416 (−0.769)	−0.008 (−0.017)	−0.554 ** (−2.258)	−0.396 (−1.455)	−0.626 *** (−2.623)
F-statistic		12.72			10.53
Wald chi2		13,739.33			16,512.81
Prob > chi2		0.000			0.000
Constants	0.412 * (1.819)	0.611 *** (2.856)	0.566 ** (2.295)	0.748 *** (3.624)	0.673 *** (3.986)	0.850 *** (4.137)
Observations	208	208	195	288	288	270

Note: ***, **, and * represent significance at 1%, 5%, and 10%, respectively, and the numbers in brackets are z values.

). The estimation results for the sample of major grain producing areas were similar to those of the national sample; the results were similar to the [39] research. Meanwhile, the estimation results for non-major grain producing areas were different. We took “general public budget revenue” as the instrumental variable (Model 14). The two-stage estimation results showed that the regression coefficient of agricultural fiscal expenditure on the WPGP was 0.014, which was statistically insignificant. This indicates that fiscal agricultural-related expenses in non-major grain-producing areas would not aggravate the WPGP. The regression result was shown in Model 15 by adding the first-order lag term of agricultural fiscal expenditure. The regression coefficient of agriculture-related fiscal costs on the WPGP was 0.026, which was significant at the 10% level, indicating that there was a specific time lag of agricultural fiscal expenditure on the WPGP in non-major grain-producing areas. Still, the impact was weak compared with that in major grain-production regions.

3.2.2. Moderating Effect Test

The estimation results are shown for Model 16 to Model 18 in

Table 6. The moderating effect test of grain production technology.

Variables	WPGP
	Model 16	Model 17	Model 18
Fiscal expenditure for agriculture (Feag)	0.031 *** (5.838)	0.027 *** (4.440)	0.031 *** (5.009)
Grain production technology (Gpte)		0.004 * (1.734)	0.021 *** (3.109)
Interactive items (Feag × Gpte)			−0.002 *** (−2.681)
Grain sown area (Grsa)	0.138 *** (7.940)	0.139 *** (7.996)	0.138 *** (8.006)
Proportion of wetlands (Wetl)	−0.304 (−0.991)	−0.354 (−1.152)	−0.361 (−1.181)
Nature reserves proportion (Rese)	0.439 (0.979)	0.504 (1.124)	0.480 (1.072)
Grain net profit (Mpmg)	0.003 *** (2.692)	0.003 *** (2.778)	0.003 *** (2.953)
Irrigation facilities (Wafa)	0.055 (0.356)	0.021 (0.134)	−0.017 (−0.114)
Soil erosion control level (Waso)	−0.312 (−1.436)	−0.256 (−1.170)	−0.367 * (−1.661)
Constants (Cons)	0.786 *** (5.685)	0.786 *** (5.690)	0.807 *** (5.833)
Observations	496	496	496

Note: *** and * represent significance at 1% and 10%, respectively, and the numbers in brackets are z values.

(Based on Formulas (7) and (8)). The results for Models 16 and 17 showed that agricultural fiscal expenditure had a significant positive impact on the WPGP. The regression coefficient of the interaction term Feag × Gpte on the WPGP was −0.002, which passed the significance test at the level of 1% (Model 18). These results indicated that the technical environment of grain production played a negative moderating role in the impact of agricultural fiscal expenditure on the WPGP. In better grain production technology, the aggravating effect of agricultural fiscal expenditure on the WPGP is effectively restrained. It could be seen that under the incentive effect of agriculture-related fiscal expenses, regional planting of more grain could be induced to increase the WPGP.

3.2.3. Spatial Spillover Effect Test

Before the Spatial Dubin model analysis, we first calculated the global Moran’s I index (Based on Formula (9)) of WPGP in 31 provincial administrative regions in Mainland China from 2003 to 2018 using the “inverse distance weight”. The results showed that Moran’s I index shares passed the 1% significance test each year (

Table 7. WPGP’s global spatial Moran index for 2003–2018.

Year	Value	Year	Value	Year	Value	Year	Value
2003	0.388 *** (4.584)	2007	0.371 *** (4.404)	2011	0.373 *** (4.416)	2015	0.387 *** (4.578)
2004	0.358 *** (4.263)	2008	0.375 *** (4.443)	2012	0.348 *** (4.151)	2016	0.332 *** (3.998)
2005	0.322 *** (3.862)	2009	0.355 *** (4.222)	2013	0.273 *** (3.33)	2017	0.311 *** (3.749)
2006	0.375 *** (4.439)	2010	0.316 *** (3.801)	2014	0.39 *** (4.500)	2018	0.316 *** (3.806)

Note: *** represent significance at 1%, and the numbers in brackets are z values.

). These results indicated that the WPGP in 31 provincial administrative regions in Mainland China had a solid positive global spatial correlation and a significant spatial agglomeration phenomenon from 2003 to 2018.

To enhance understanding, we plotted the scatter plot of the spatial distribution of WPGP for 2003, 2008, 2013, and 2018 (Figure 4). In the figure, most of the provinces’ spatially weighted WPGP values are located in the first and third quadrants in the sample time span. This indicates that provinces with higher WPGP are clustered together, while areas with lower WPGP are adjacent to each other. The results are similar to those of existing studies [40]. Therefore, to address potential estimation biases arising from the omission of geographic spatial distribution factors, it is essential to incorporate spatial effects into the spatial econometric framework.

It could be seen from Model 19 in

Table 8. Regression results of Spatial Panel Dubin model.

Variables and Tests	Dependent Variable: WPGP
	Model 19
ρ	0.395 *** (8.604)
Lag_Wpgp	0.861 *** (38.857)
Fiscal expenditure for agriculture (Feag)	0.039 *** (5.468)	ω_Feag	0.118 *** (6.213)
Grain sown area (Grsa)	0.026 *** (5.729)	ω_Grsa	−0.023 * (−1.810)
Proportion of wetlands (Wetl)	−0.219 ** (−2.371)	ω_Wetl	3.086 *** (14.518)
Grain net profit (Mpmg)	1.644 *** (19.494)	ω_Mpmg	7.965 *** (25.737)
Nature reserves proportion (Rese)	−0.017 *** (−3.730)	ω_Rese	0.185 *** (13.585)
Irrigation facilities (Wafa)	0.381 *** (5.823)	ω_Wafa	1.433 *** (8.136)
Soil erosion control level (Waso)	1.401 *** (15.181)	ω_Waso	6.547 *** (24.696)
AIC	−329.311
BIC	−258.896
Observations	465

Note: ***, **, and * represent significance at 1%, 5%, and 10%, respectively, and the numbers in brackets are z values.

(Based on Formulas (10) and (11)) that the influence coefficient ρ of spatial lag term on the WPGP in the dynamic Spatial Panel Dubin model measured by inverse distance weight was significant at the 1% level, and the coefficient was 0.395. This indicated that the WPGP exhibits a strong spatial dependence across regions, proving that grain production’s water pressure had a specific spatial dependence. The WPGP had a similar level among regions with equal distances, and there was a particular risk of the spillover effect of WPGP. Our results were identical to the Cowan [41] research, as the high fluidity of water resources and their classification as public goods are characterized by rivalry and non-excludability. In regions experiencing intense water stress from grain production, excessive exploitation of groundwater and surface water is likely, which can encroach upon water supplies in adjacent areas. Such behavior can cause the water pressure associated with grain cultivation to spill over into neighboring regions. The coefficient of the first-order lag term for WPGP was statistically significant at the 1% level, with an estimated value of 0.861. This confirms that past WPGP levels exert a notable and positive influence on current WPGP. Furthermore, the inclusion of the lag term helps to capture unobserved dynamic effects and inertia beyond spatial structure alone, thereby enhancing the accuracy of the estimated ρ coefficient. Based on the results presented in Model 19, we also observed that agricultural fiscal expenditure within the region had a coefficient of 0.039, significant at the 1% level. Similarly, fiscal spending in surrounding regions showed a coefficient of 0.118, which also met the 1% significance threshold, suggesting notable spatial spillover effects.

Since the Dynamic Panel Spatial Dubin model was not linear regression, its estimated coefficients cannot directly reflect the marginal effect of agricultural fiscal expenditure on the WPGP. The partial differential method showed that the estimated coefficient must be decomposed into direct, indirect, and total effects (

Table 9. Decomposition results of spatial spillover effect.

Variables	Direct Effect	Indirect Effect	Total Effect
Fiscal expenditure for agriculture (Feag)	0.049 *** (6.153)	0.207 *** (5.136)	0.256 *** (5.625)
Grain sown area (Grsa)	0.025 *** (5.172)	−0.019 (−0.953)	0.006 (0.246)
Proportion of wetlands (Wetl)	0.021 (0.225)	4.756 *** (8.590)	4.778 *** (8.040)
Grain net profit (Mpmg)	2.299 *** (17.825)	13.608 *** (12.176)	15.906 *** (12.970)
Nature reserves proportion (Rese)	−0.003 (-0.558)	0.283 *** (8.424)	0.281 *** (7.731)
Irrigation facilities (Wafa)	0.503 *** (7.840)	2.481 *** (8.125)	2.984 *** (9.293)
Soil erosion control level (Waso)	1.946 *** (15.797)	11.253 *** (10.544)	13.199 *** (11.412)

Note: *** represent significance at 1%, and the numbers in brackets are z values.

).

From the decomposition result of the direct effect, we observe that the immediate impact of agricultural fiscal expenditure on the WPGP was significant at the 1% level, and the coefficient was 0.049. These showed that after considering the feedback effect (that was, the agricultural fiscal expenditure in the region affected the WPGP in the region by affecting the WPGP in the neighboring region and then, in turn, affecting the WPGP in the region), the WPGP in the region would increase by 0.049 units for every increase in agricultural fiscal expenditure in the region. We observe that the estimated coefficient after considering the spatial effect was higher than that of Model 2 (0.026), which meant that the model, without considering the spatial effect, underestimated the impact of agricultural fiscal expenditure on the WPGP in this region to some extent.

From the decomposition result of the indirect effect, we find that the spillover impact of agricultural fiscal expenditure on the WPGP was significantly positive. The coefficient was 0.207, indicating that for every 1 unit increase in fiscal agricultural-related expenses in this region, the WPGP in the neighboring region would increase by 0.207 units.

From the decomposition result of the total effect, we observe that the overall impact of agricultural fiscal expenditure on the WPGP was significantly positive. The coefficient was 0.255; that was, every 1 unit increase of agricultural fiscal expenditure in this region would increase the WPGP in the region and adjacent regions by 0.255 units. These results indicated that the increase in agricultural fiscal expenditure in this region would increase the water pressure in grain production in this region and increase the water pressure in grain production in neighboring regions. The results were similar to those of existing studies [31].

4. Discussion

Since 2004, the Chinese government has gradually established four agricultural subsidies. After the three subsidy reforms launched in 2016, the agriculture-related fiscal subsidies were divided into protection subsidies and agricultural machinery purchase subsidies. The Chinese government hoped that financial and agricultural subsidies could improve farmers’ enthusiasm for production, guarantee an adequate supply of food and other agricultural products, and improve international competitiveness [42]. However, they ignored the distortive effects of such subsidies on the allocation of grain production factors [43], which led to the increasingly severe mismatch between grain production factors [44]. Major water resources projects such as large-scale water conservancy facilities construction, river and lake network repairs, and South-to-North Water Diversion projects focused on utilizing water resources rather than protection. Therefore, it had caused tremendous pressure on water in the region. The details are as follows: Firstly, most agricultural fiscal expenditure was used to directly or indirectly subsidize grain production [45]. The subsidies directly provided to agriculture mainly include arable land fertility protection subsidies, target price subsidies, straw returning subsidies, and agricultural machinery purchase protection subsidies. In addition, there were also infrastructure construction and scientific research investment that indirectly affected food production. The number of subsidies for food and agriculture mainly depended on cultivated land, grain output, and the purchase of agricultural machinery and tools. Secondly, both China’s cultivated land and subsidy objects that boosted land fertility protection were concentrated in the north. Thus, the north’s distortion of grain production factors allocation was severe. In addition, the extent of continuous cultivation of grain in northern China was excellent, and the purchase of agricultural machinery was correspondingly large. Thus, the investment in agricultural machinery subsidies was mainly concentrated in northern China. Simultaneously, grain production depended on arable land. Therefore, target price subsidies and straw returning subsidies focused on northern China. Consequently, the focus of China’s grain production gradually shifted to the north, which was relatively rich in arable land but low in water, exacerbating the already critical water shortages in northern China [46].

4.1. Agricultural Fiscal Expenditure Would Aggravate the Regional Water Pressure in Grain Production

Based on the baseline estimation results, it is evident that agricultural fiscal expenditure exacerbates regional WPGP. This finding implies that agriculture-related fiscal expenditures significantly influenced the structural allocation of grain production factors, resulting in spatial mismatches between grain cultivation and water availability. Specifically, a substantial portion of China’s agricultural fiscal expenditure was directed toward subsidizing grain production, both directly and indirectly [43].

From a theoretical perspective, these fiscal interventions functioned as economic incentives that altered farmers’ production behavior by shifting their marginal cost–benefit calculations. Direct subsidies encouraged scale expansion by increasing the expected return of grain production, thereby enlarging the cultivated area. Additionally, these incentives lowered the opportunity cost of remaining in agriculture compared to engaging in non-agricultural employment, especially for large-scale or part-time farmers, effectively anchoring labor and land in grain cultivation.

Indirectly, agricultural fiscal expenditure influenced the production environment through investment in public goods such as agricultural infrastructure and technological advancement. These expenditures improved production efficiency—irrigation facilities enhanced water accessibility, post-harvest infrastructure reduced storage losses, and R\&D investments promoted yield improvements through advanced crop varieties and optimized cultivation techniques. According to production economics theory, improvements in total factor productivity reduce the unit cost of production [47], which can lead to an output expansion effect. In water-scarce regions, this expansion intensifies water consumption per hectare, thereby aggravating the WPGP [48].

However, the structure of these expenditures reveals a critical shortcoming in policy design. Fiscal spending has disproportionately focused on enhancing water extraction and delivery—such as building reservoirs, canals, and ditches—rather than investing in water-saving technologies or promoting demand-side management. This reflects a utilitarian approach that prioritizes short-term yield maximization over long-term sustainability, deviating from the principles of integrated water resource management.

In conclusion, the current production-oriented configuration of agricultural fiscal support has unintentionally amplified water pressure in grain-producing regions, largely due to its bias toward expanding production capacity rather than improving resource-use efficiency. Given that water resources function as public goods with high externalities, the failure to internalize water scarcity costs in fiscal allocation decisions has contributed to regional water stress. Therefore, future fiscal policy design should integrate environmental externalities and adopt a resource-conservation-oriented framework. This includes rebalancing fiscal spending toward water-saving irrigation technologies, enhancing the efficiency of water allocation, and incentivizing sustainable agricultural practices through ecological compensation and differentiated subsidy schemes.

This study fills a crucial research gap by empirically linking agricultural fiscal expenditure to regional water pressure in grain production [49], which has been insufficiently explored in prior research that mostly focused on production output or income effects [50]. Our findings support and extend earlier theoretical and empirical work on the unintended environmental consequences of agricultural subsidies [51], highlighting the need to incorporate water resource constraints into subsidy policy design. By combining TOPSIS analysis with economic theory and spatial data, this article advances understanding of how fiscal policies can exacerbate resource pressures, thus providing a basis for more sustainable agricultural fiscal reforms.

4.2. Agricultural Fiscal Expenditure in Major Grain-Producing Areas Substantially Affected the Aggravation of Water Pressure in Grain Production

In the sub-sample regression test, it was observed that agricultural fiscal expenditure in non-major grain-producing areas did not significantly exacerbate regional water pressure in grain production (WPGP). This suggests that financial subsidies in these regions, while intended to stimulate grain production, did not translate into substantial increases in grain planting or associated water use. This outcome can be theoretically explained through a resource allocation and behavioral response framework grounded in production economics and opportunity cost theory.

Firstly, farmers’ production decisions in non-major grain-producing regions are fundamentally governed by the principle of profit maximization under resource constraints. Specifically, land, labor, and capital—the three core factors of production—impose strict limitations on farmers’ ability to expand grain cultivation. When the marginal returns from grain production are low and cannot offset the total (explicit and implicit) costs, farmers are rationally disincentivized from expanding grain planting, regardless of subsidy availability. This is particularly relevant in non-major grain-producing regions, where agricultural production is characterized by low efficiency, poor economies of scale, and relatively underdeveloped infrastructure. As a result, even when subsidies are provided, they may not be sufficient to shift the farmers’ production function or alter their optimal crop allocation behavior.

The empirical results (Models 13–15 in Table 5) support this view by showing that agricultural fiscal expenditure has no statistically significant impact on grain production or WPGP in non-major grain-producing areas. This aligns with economic behavior models suggesting that when the net gain from increased production is negligible or negative, financial incentives fail to induce supply-side responses.

Secondly, the limited impact of fiscal expenditure is further reinforced by the high opportunity costs associated with land, labor, and capital in these regions. Land may be more profitably allocated to higher-value or water-efficient crops; labor, particularly younger and more mobile workers, tends to migrate to non-agricultural sectors with higher wage potential; and capital investment is often constrained or redirected to more stable or lucrative ventures [52]. The subsidies provided may only partially compensate for direct input costs (e.g., seeds, fertilizers, fuel), but they do not adequately cover opportunity costs, especially in regions where off-farm employment offers superior returns. Therefore, the marginal utility of subsidies remains low [53].

From a mechanism perspective, the fiscal subsidies in non-major grain-producing regions fail to trigger a chain reaction that would lead to increased production and water use. Unlike in major grain-producing regions—where agricultural infrastructure, market access, and policy targeting may enable fiscal inputs to effectively translate into output expansion—in non-major regions, systemic constraints hinder this transmission channel. The lack of scale economies, weak technological adoption, poor irrigation infrastructure, and low market integration further suppress the responsiveness of grain production to policy stimuli.

In summary, the negligible effect of agricultural fiscal expenditure on WPGP in non-major grain-producing areas can be explained by a constrained optimization framework: farmers act rationally under high opportunity costs and low marginal returns, resulting in limited production adjustments in response to subsidies. This highlights the importance of regional heterogeneity in policy responsiveness and calls for a differentiated approach to agricultural fiscal policy that considers local production structures, economic incentives, and resource constraints.

Moreover, this study fills a significant gap in the existing literature by empirically demonstrating the differential impact of agricultural fiscal expenditure on water pressure between major and non-major grain-producing regions. While prior research has largely focused on the aggregate effects of subsidies on agricultural output or income [54], few studies have distinguished the heterogeneous regional responses, especially in terms of environmental pressures such as water resource use. Our findings corroborate theoretical models of production behavior under resource constraints [55] and extend the understanding of subsidy effectiveness by integrating spatial and economic dimensions. This contribution provides a nuanced basis for designing region-specific fiscal policies that better align agricultural production incentives with sustainable resource management goals.

4.3. The Role of Technical Environment in Mitigating the Impact of Agricultural Fiscal Expenditure on Water Pressure in Grain Production

When discussing the impact of agricultural fiscal expenditure on regional water pressure in grain production (WPGP), it is essential to incorporate a theoretical perspective that emphasizes the moderation effect of the technical environment. Specifically, technical efficiency in agricultural production acts as a moderating variable in the relationship between fiscal expenditure and WPGP, shaping how production incentives translate into resource consumption. This mechanism can be framed within the context of production theory and environmental efficiency, which posits that technological progress improves input–output ratios and reduces negative externalities.

Firstly, improvements in technical efficiency under high-tech agricultural environments enhance the marginal productivity of key inputs, particularly land and water. Through the adoption of advanced technologies—such as precision agriculture and smart irrigation systems—farmers can optimize the input mix, achieving higher grain yields per unit of water use. These technologies enable more precise control over irrigation and fertilization, improving water productivity and reducing input waste. For example, smart irrigation systems adjust water application in real time based on soil moisture and crop needs, thereby mitigating the inefficiencies of conventional irrigation practices. As a result, the increase in grain production spurred by agricultural fiscal expenditure does not lead to a proportionate rise in water consumption, thereby weakening its impact on WPGP [56].

Secondly, the role of irrigation infrastructure and water allocation efficiency serves as a structural mechanism through which technical environments moderate fiscal impacts. Technologies like drip and sprinkler irrigation improve water use efficiency by minimizing evaporation and seepage losses. Moreover, large-scale projects—such as the “South-to-North Water Diversion” initiative—represent institutional and engineering solutions that reallocate water resources more efficiently across regions, especially from surplus to deficit areas. These interventions not only address the supply-side constraints of water availability but also decouple grain output growth from regional water stress, thereby reducing the marginal water pressure associated with fiscal-driven production expansion.

This theoretical mechanism also explains the observed regional heterogeneity. In water-scarce northern regions, where the natural resource constraint is binding, agricultural fiscal expenditure tends to exert greater upward pressure on WPGP due to limited buffering capacity and lower baseline water-use efficiency. Conversely, in southern regions with relatively abundant water resources and higher technological adoption, the same level of fiscal stimulus has a more muted effect on WPGP. Therefore, regional differences in the technical environment result in asymmetrical transmission of fiscal policy effects on water pressure.

4.4. The Spatial Spillover Effect of Agricultural Fiscal Expenditure on Water Pressure in Grain Production

In the spatial econometric estimation results, the indirect effect decomposition reveals a notable phenomenon: As agricultural fiscal expenditure increases within a region, the WPGP in neighboring regions also tends to rise. This spatial spillover is driven by several interrelated mechanisms.

First, fiscal subsidies based on cultivated land area, grain output, and agricultural machinery incentivize farmers to expand grain production. This effect is especially prominent in water-scarce yet arable land-rich regions like the Northeast and Huang-Huai-Hai regions, which receive a significant share of national agricultural fiscal expenditure [57]. The intensified grain production not only elevates local water use but also generates external pressure on surrounding areas. Our spatial econometric analysis confirms this, showing that a 1% rise in agricultural fiscal expenditure leads to a 0.395% increase in WPGP in neighboring regions (ρ = 0.395, p < 0.01).

This spillover is further reinforced by the public good nature of water resources and water conservancy infrastructure. Since water systems transcend administrative boundaries, overuse in one region can deplete availability in adjacent ones. Agricultural fiscal expenditure often funds shared infrastructure, which, while improving local productivity, may unintentionally reduce downstream water access [58]. The consistently significant Moran’s I index (p < 0.01) for WPGP from 2003 to 2018 reflects this spatial interdependence.

Additionally, agricultural fiscal expenditure attracts capital inputs—such as irrigation systems, machinery, and technology—that amplify productivity and influence production behaviors across regions. For example, fiscal support for mechanized farming not only stimulates local yield but also encourages adoption in neighboring regions through demonstration effects [59]. This leads to expanded cultivation and increased water demand beyond the focal area [60]. Our findings demonstrate that fiscal spending positively affects WPGP through both direct and indirect effects (indirect effect coefficient = 0.207, p < 0.01).

Internationally, similar concerns have prompted reforms that integrate environmental conditions into agricultural support schemes. For instance, the European Union’s Common Agricultural Policy (CAP) includes cross-compliance rules linking subsidies to environmental standards, aiming to mitigate externalities such as water overuse [61]. In contrast, China’s fiscal support still largely emphasizes output and efficiency without stringent ecological constraints. Moreover, while existing international studies often focus on nutrient runoff or carbon impacts [62], few address the spatial water implications of fiscal policy. Our study, therefore, contributes new evidence to this discourse by highlighting how fiscal expenditure may exacerbate transregional water stress—an issue especially critical in hydrologically connected and administratively fragmented regions like China.

5. Conclusions and Policy Implications

5.1. Conclusions

This study systematically investigates the impact of agricultural fiscal expenditure on water pressure in grain production (WPGP) across regions in China, uncovering both direct and spatial spillover effects, while also highlighting the moderating role of the technical environment. The main findings are as follows:

First, agricultural fiscal expenditure significantly aggravates regional WPGP. While such expenditures stimulate grain output by incentivizing land expansion, labor retention, and infrastructure investment, they also unintentionally exacerbate mismatches between grain production and water availability. The production-oriented design of current fiscal policies emphasizes yield growth over resource-use efficiency, resulting in unsustainable water consumption—particularly in water-scarce areas.

Second, the aggravating effect of fiscal expenditure on WPGP is particularly pronounced in major grain-producing areas. In contrast, in non-major grain-producing regions, agricultural fiscal expenditure does not significantly affect WPGP, mainly due to structural limitations, low production incentives, and high opportunity costs. This regional divergence highlights the importance of accounting for heterogeneity in resource endowments and production constraints when designing fiscal support policies.

Third, technical efficiency serves as a crucial moderating factor in mitigating the adverse effects of agricultural fiscal expenditure on WPGP. The adoption of precision agriculture, smart irrigation, and efficient water allocation mechanisms can decouple yield growth from water consumption. Regions with stronger technical environments are more resilient to the water pressures induced by fiscal stimuli, underscoring the need to integrate technological upgrading into fiscal policy frameworks.

Fourth, agricultural fiscal expenditure exhibits significant spatial spillover effects, increasing WPGP not only within a region but also in adjacent areas. This effect is driven by interregional water flows, shared infrastructure, and the diffusion of production incentives. These findings suggest that WPGP is a spatially interconnected phenomenon, requiring coordinated governance and integrated fiscal planning that transcends administrative boundaries.

5.2. Policy Implications

Precise formulation of policies could play a key role in mitigating WPGP. The government should adjust agricultural fiscal expenditure in a classified and orderly manner to alleviate the spatial mismatch between grain production and water resources.

Firstly, a resource-saving orientation should be embedded in fiscal support policies. The linkage between agricultural fiscal expenditure and water constraints must be strengthened by considering regional WPGP differences in subsidy design. A water-saving compensation system should be introduced to guide grain production towards more sustainable water use, particularly in water-scarce regions such as the North China Plain, where subsidies could be linked to water-efficient practices or low water-demand crop varieties.

Secondly, interregional coordination of fiscal transfers should be enhanced. Fiscal policies should incentivize grain production in water-abundant but underutilized regions (e.g., parts of the Yangtze River Basin), while scaling back subsidies in high-output but water-deficient areas. This spatial reallocation could be supported through targeted transfer payments or ecological compensation mechanisms, thereby improving the match between grain production patterns and water resource distribution.

Thirdly, to improve the effectiveness of fiscal investment in water infrastructure, the government should prioritize the construction of water metering facilities, promote the development of water rights and pricing markets, and establish pricing mechanisms that reflect the true scarcity and opportunity cost of water. A water-saving incentive system should be implemented, especially in regions where the spillover effects of water use are high, to mitigate the inefficiencies caused by the public goods nature of water.

Lastly, the government should strengthen investment in water-saving technologies and infrastructure. This includes increasing R&D funding for water-efficient grain production technologies within the agricultural fiscal budget, supporting region-specific innovation adapted to local water endowments, and improving the conversion rate of scientific achievements. In parallel, greater fiscal input should be directed toward modernizing water conservancy infrastructure—especially water-saving irrigation systems—and expanding the coverage of high-standard cultivated land and facility agriculture. These efforts would maximize the benefit of limited water resources, reduce per-unit water demand for grain output, and ultimately relieve water pressure in grain-producing regions.

Moreover, a more comprehensive sustainable policy design should consider alternative approaches beyond fiscal expenditure. Water pricing mechanisms should be further developed to reflect the true opportunity cost of water, providing economic incentives for conservation. Establishing clear water efficiency standards can drive the adoption of best practices and technologies, encouraging farmers to optimize water use. Additionally, policies aimed at modifying farmer behavior—through education, extension services, and participatory water management—can effectively complement fiscal tools. Such integrative approaches, widely recognized in international water governance frameworks, offer promising pathways to reduce WPGP by aligning economic incentives, regulatory measures, and social engagement. Incorporating these strategies can help address the complex challenges of water scarcity and promote long-term sustainability in grain production.