Evaluation of Food Legumes Pest and Disease Control in China: Evidence Using a Provincial-Level Dataset

Abstract

Food legumes play a pivotal role in China’s food security, nutritional health, and green development strategies due to their unique advantages. This paper presents an empirical study on the economic evaluation of scientific research on pest and disease control for food legumes. Using panel data from 31 Chinese provinces from 2008 to 2023, we employ a Double Machine Learning (DML) approach to identify the impact of investment in plant protection research on food legume outputs. The results indicate a steady increase in China’s investment in this field, with an average annual growth rate of 5.19% from 2008 to 2023, and the total investment in 2023 was 2.14 times that of 2008. Investment in plant protection research effectively mitigates output losses and leads to significant production increases. Specifically, a 1% increase in research investment corresponds to a 0.2% increase in food legume output. This effect remains robust across various algorithms, time windows, and control variable settings. Based on these findings, we recommend: (1) increasing financial support and talent acquisition for research on food legume pests and diseases to enhance the stability and sustainability of research investment; (2) strengthening cooperation mechanisms between research institutions and enterprises to leverage their respective strengths and promote the commercialization of research outcomes and regional variety extension; (3) establishing a diversified research investment system that explores a co-construction model guided by the government, involving enterprises, and utilizing public–private partnerships to reconcile the conflict between long research cycles and market demands; (4) fostering a dual-track linkage between regional technological innovation and enterprise product commercialization to improve the efficiency of technology transfer and application; and (5) strengthening R&D in cutting-edge fields like Artificial Intelligence to improve the efficiency and precision of pest and disease control.

1. Introduction

The food legume industry is crucial for global food security and nutrition. In China, “food legumes” refer to a variety of legume crops (excluding soybeans and peanuts) that are primarily harvested for their grains and tender pods, which are used for human consumption or as livestock feed. They are a vital source of plant-based protein. According to the Food and Agriculture Organization (FAO), food legumes are cultivated worldwide [1], with major varieties including mung beans, adzuki beans, common beans (dry), broad beans (green and dry), peas (green and dry), chickpeas, and lentils. China’s annual cultivation area for food legumes is approximately 4 million hectares, producing around 30 million tons. The FAO defines food security as existing “when all people, at all times, have physical, social and economic access to sufficient, safe and nutritious food that meets their dietary needs and food preferences for an active and healthy life.” The global food system confronts two main challenges: ensuring food security and nutrition in a health-conscious manner, and promoting environmental sustainability and the resilience of agricultural systems in the face of climate change. In developing countries, plant-based proteins constitute nearly 80% of dietary protein [2], and food legumes help address both challenges. They enrich the soil with essential nutrients, maintain biodiversity, improve soil structure [3], and contribute to climate change mitigation and adaptation [4]. As nutrient-dense foods, they offer significant health benefits, reducing the risk of certain chronic diseases [5]. For the most impoverished populations with limited access to meat, food legumes can significantly improve dietary quality. A classic example is the combination of rice and beans: rice is rich in methionine but low in lysine, while most beans are rich in lysine but low in methionine. Consumed together, they provide a more complete protein profile [6].

However, escalating global climate change has made pests and diseases a primary threat to food legume production. Strengthening investment in plant protection research is a key measure to enhance the crop’s resilience [4]. China has achieved significant advancements in the science and technology of pest and disease control for food legumes, supporting the industry’s safe, green, and efficient growth. As a major global producer and consumer, China contributes approximately 5.6% to the world’s total output [1] and leads in the production and trade of various legumes, including broad beans, peas, mung beans, adzuki beans, and kidney beans. Over the past two decades, supported by national programs such as the Modern Agricultural Industry Technology System, the National Natural Science Foundation, and the National Key R&D Program, China has significantly improved its pest and disease control capabilities. (1) Key advancements encompass comprehensive investigations and identifications of pests and diseases, screenings for resistant resources, and the development of green control technologies. Numerous diseases, including gray mold on mung beans, were identified for the first time, and the genetic variation in pathogens for major diseases such as broad bean anthracnose was uncovered. This foundational work has facilitated a transition from reactive to targeted control strategies. (2) Establishment of techniques for identifying resistance to diseases such as mung bean wilt and pests like bean weevils has led to the discovery of superior resistant germplasm, the mapping of multiple resistance genes, and the creation of new disease-resistant varieties, filling a critical gap in China’s resources. (3) The integration of comprehensive pest management technologies has shown remarkable increases in output. For example, an integrated technique for combating common bacterial blight and late-stage leaf spot in kidney beans has demonstrated output increases ranging from 11.8% to 122.3% in trial stations and is now widely and effectively applied in North and Northwest China [7]. (4) Active training and technical services have been provided to disseminate knowledge and skills, significantly raising the overall level of pest and disease control.

Continuous advances in pest and disease control technology are a vital driver for reducing crop losses and promoting sustainable development. The existing literature widely acknowledges that technological progress in this area is a core pathway to increasing crop outputs and achieving green development [8]. As climate change intensifies the frequency and severity of pest and disease outbreaks, research in this field has become even more critical [9]. Biotic stressors are responsible for approximately 30% of global crop output losses, manifesting as reduced outputs and deteriorated nutritional quality [10]. It is estimated that pests and diseases result in annual global crop losses ranging from 13% to 22%, which equates to billions of dollars in economic costs for major staples such as rice, wheat, corn, and potatoes [11]. This reality necessitates a transition to more efficient control strategies, such as integrated pest management (IPM), which integrates agronomic practices, genetic resistance, and chemical control [11], and improves output stability by optimizing biotic community interactions [12]. However, traditional chemical control methods are encountering challenges such as increasing pathogen and pest resistance, escalating research and development costs for new pesticides, and mounting ecological risks [13]. These issues necessitate technological innovation to achieve a balance between economic viability and ecological sustainability. For food legumes, the effect of technological advances in pest control on output improvement is particularly pronounced. Studies have shown that they possess strong resistance to abiotic stresses, such as drought and heat, but are highly vulnerable to biotic stresses [14]. Mung bean production losses are primarily caused by root-knot nematodes, dry root rot, powdery mildew, Cercospora leaf spot, anthracnose, and various viruses [15], which lead to severe production losses [16,17]. For instance, a 1% increase in rust severity can reduce food legume output by 8.39% [18]. Although existing control systems, such as resistant varieties and biopesticides, have proven effective in laboratory settings, their real-world application is often hindered by delays in technology dissemination [19]. Particularly in smallholder farming scenarios, the nonlinear relationship between control costs and returns, coupled with barriers to technology adoption, constrains the practical efficiency of control measures [20].

Currently, global food security and nutrition remain pressing issues. According to the Food and Agriculture Organization’s (FAO) report “The State of Food Security and Nutrition in the World 2024,” an estimated 713 to 757 million people faced hunger in 2023 [21]. As a vital source of plant-based protein and micronutrients, food legumes play an irreplaceable role in diversifying diets, improving nutrition for impoverished populations, and promoting sustainable agriculture. In China, food legumes are not only a cornerstone of the traditional diet but also a key component in achieving the “Grand Food Vision” and “Healthy China 2030” strategic goals. As climate change intensifies, pests and diseases have emerged as the primary obstacles to the stability and increase in food legume production. Consequently, green and efficient control measures have become a crucial aspect in ensuring output. In recent years, China has steadily increased its investment in this research area, achieving significant progress. However, there is a lack of quantitative evaluation of the benefits from this investment, which hampers the translation of China’s practical experience into global solutions. This study aims to fill that gap by systematically analyzing investment and output data since 2008 to quantify the contribution of plant protection research to food legume production. The findings are intended to provide a scientific basis for optimizing research investments in China and offer valuable experience for global efforts to combat malnutrition and ensure food security.

2. Methods and Data Sources

2.1. Method: The Construction of Double Machine Learning Model

This study employs a Double Machine Learning (DML) model to estimate the impact of investment in plant protection research on food legume output. Compared to traditional linear regression models, DML accommodates complex, nonlinear relationships between control variables and the dependent variable, more accurately reflecting the realities of agricultural production.

Following the DML framework proposed by Chernozhukov et al. [22], we decompose the causal identification process into two stages. In the first stage, machine learning methods (e.g., Random Forest, LassoCV) are used to construct conditional expectation functions for the dependent and key explanatory variables while controlling for confounding factors such as production scale, input costs, and environmental variables. In the second stage, an orthogonal score function is constructed from the residuals to isolate the influence of control variables, thereby enabling a robust estimation of the net effect of research investment on output. Based on existing literature [23,24,25], the DML model is specified as follows:

$$Y_{it}={\theta }_{0}RD{I}_{it}+g\left(C{V}_{it}\right)+U_{it}$$

(1)

$$E\left(U_{it}\mid RD{I}_{it},C{V}_{it}\right)=0$$

(2)

where $Y_{it}$ is dependent variable, representing food legume output in the province i of the year t. $RD{I}_{it}$ is the core explanatory variable, namely scientific research investment in plant protection for food legumes. $g\left(C{V}_{it}\right)$ indicates high-dimensional control variable $C{V}_{it}$, which will affect the food legume output through an unknown functional form. $g\left(.\right)$ needs to be estimated for its specific form $\widehat{g}\left(C{V}_{it}\right)$ by machine learning algorithm. ${\theta }_0$ is coefficient. $U_{it}$ is error term.The machine learning algorithm is adopted to estimate Equations (1) and (2) to make ${\theta }_0$ obtain biased estimator ${\widehat{\theta }}_0$, where $n$ is sample size.

$${\widehat{\theta }}_0={\left({\displaystyle \frac{1}{n}}{\displaystyle \sum }_{i\in I,t\in T}R\mathrm{D}{I}_{it}^2\right)}^{-1}{\displaystyle \frac{1}{n}}{\displaystyle \sum }_{i\in I,t\in T}RD{I}_{it}\left(Y_{it+1}-\widehat{g}\left(C{V}_{it}\right)\right)$$

(3)

To ensure that coefficient estimator satisfies unbiasedness under small sample size, the following auxiliary regression equation is constructed:

$$RD{I}_{it}=m\left(C{V}_{it}\right)+V_{it}$$

(4)

$$E\left(V_{it}|C{V}_{it}\right)=0$$

(5)

where $m\left(C{V}_{it}\right)$ is the regression function of core explanatory variable on the high-dimensional control variable, and a machine learning algorithm is involved to estimate its specific form $\widehat{m}\left(C{V}_{it}\right).V_{it}$ represents error term. The specific operation process is as follows: first, using machine learning algorithm to estimate auxiliary regression $\widehat{m}\left(C{V}_{it}\right)$ and taking its residuals $V_{it}=RD{I}_{it}-m\left(C{V}_{it}\right)$, followed by estimating $\widehat{g}\left(C{V}_{it}\right)$, and then performing regression based on ${\widehat{V}}_{it}$ to obtain unbiased estimated coefficient.

$$\begin{array}{r}\hfill {\check{\theta }}_0={\left({\displaystyle \frac{1}{n}}{\displaystyle {\displaystyle \sum }_{i\in I,t\in T}}{\widehat{V}}_{it}RD{I}_{it}\right)}^{-1}{\displaystyle \frac{1}{n}}{\displaystyle {\displaystyle \sum }_{i\in I,t\in T}}{\widehat{V}}_{it}\left(Y_{it+1}-\widehat{g}\left(C{V}_{it}\right)\right)\end{array}$$

(6)

2.2. Variable Selection and Data Description

2.2.1. The Scientific Research Investment in Plant Protection

Investment in plant protection research provides critical technological support for stabilizing and increasing crop outputs. While chemical pesticides played a significant role since the mid-20th century [26], their overuse has led to issues like resistance, environmental risks, and secondary disasters [27]. By the late 20th century, a paradigm shift occurred towards more ecologically friendly and sustainable integrated control strategies [28]. Research focus shifted to pesticide reduction, biological control, and green alternatives. These technologies have longer R&D cycles and higher barriers to adoption, making it difficult for private enterprises to sustain investment based solely on short-term returns [29]. Consequently, public funding has become crucial for advancing cutting-edge research and promoting technology adoption in this field [30].

Public funding constitutes the primary source of support for scientific research on pest and disease control for food legumes in China. Based on survey data, funding in this field since 2008 has predominantly come from special-purpose funds allocated through science and technology programs administered by national and local governments. At the national level, key funding sources include the National Edible Bean Agricultural Industry Technology System, the National Key Research and Development Program, and the projects of National Natural Science Foundation (NSFC), etc. Provincial-level dedicated science and technology projects for the pest and disease control of food legumes have been established in Jiangsu, Qinghai, Chongqing, Shanxi, Guizhou, Yunnan, and Hubei provinces. Concurrently, provincial-level industrial system projects for food legumes have been implemented in Shaanxi, Hebei, Jiangsu, and Shanxi. Furthermore, a limited number of prefecture-level cities—such as Tangshan in Hebei, Linxia and Dingxi in Gansu, and Chifeng in Inner Mongolia—have also initiated local science and technology projects. Additionally, the Institute of Crop Sciences of the Chinese Academy of Agricultural Sciences (CAAS) has provided continuous support for food legume research since 2013 through its Innovation Engineering Program and basic scientific research funds. The funding analysis spans from 2008 to 2023 and includes 31 provinces, autonomous regions, and municipalities on the Chinese mainland, excluding Hong Kong, Macao, and Taiwan.

2.2.2. Data Sources for Food Legume Output and Extreme Climate

The dependent variable is food legume output, and the core explanatory variable is investment in plant protection research. We control for other important variables that affect output, which are categorized into agricultural production status and external climate shocks. Production variables include planting area and various inputs, while climate shocks include extreme temperatures and precipitation. This comprehensive set of controls allows for a clearer estimation of the net contribution of research investment. The theoretical framework is depicted in Figure 1.

Controlling for agricultural production conditions and climatic factors, this study employs a DML approach to analyze the contribution of plant protection research investment to food legume output, based on panel data from 31 Chinese provinces from 2008 to 2023. Regarding data sources, food legume output data were sourced from the National Bureau of Statistics (NBS) of China and the Food and Agriculture Organization of the United Nations (FAO). The NBS provides planting area and output data for mung beans and adzuki beans. Data sourced from the FAO correspond to its official variable names for several categories, which are as follows: common beans (beans, dry), peas (peas, green & peas, dry), chickpeas (chick peas), lentils (lentils), broad beans (broad beans, green & broad beans, dry), and cowpeas (cow peas, dry). Extreme climate data were sourced from Guo et al. [31]. Other data were primarily sourced from the NBS platform and public yearbooks such as the China Statistical Yearbook. Data for overall crop mechanization, fertilizer use, pesticide use, and plastic film use from the NBS were scaled according to the proportion of the food legume sowing area. For missing values in certain years or regions, this study employed conventional methods such as mean and interpolation for imputation. This study employed interpolation to impute missing values for certain years or regions. During this process, the variable crop disaster affected area generated a small number of negative values because some observations had very low initial disaster rates. Since disaster-affected area cannot be negative in practice, these negative interpolated values were uniformly replaced with 0 to ensure data rationality and interpretability.

2.2.3. The Descriptive Statistics of Overall Sample

The basic information of the variables used in this study is shown in

Table 1. Variable Description and Descriptive Statistics.

Variable Type	Variable Name	Variable Description	Mean Value	Standard Deviation	Sample Size
Dependent variable	Food legume output	Food legume output (10 thousand tons)	95.936	136.742	496
Independent variable	Scientific research investment of plant protection	Accumulated scientific research investment of plant protection (10 thousand yuan)	381.197	465.735	496
Control variable	Sown area	Sown area of food legume crops (ha.)	61,522.9	75,378.42	496
	Disaster affected rate	Ratio of crop disaster affected area to sown area	0.157	0.14	496
	total power of Agricultural machinery	Total power of agricultural machinery (kilowatt)	314,403.5	314,379	496
	Fertilizer application	Net usage of agricultural chemical fertilizers (ton)	18,083.15	20,504.59	496
	Pesticide application	Pesticide application (ton)	480.335	521.893	496
	Film plastic usage	usage of agricultural plastic film (ton)	822.972	1097.59	496
	Extreme low temperature days	Days with extreme low temperature	42.495	17.637	496
	Extreme high temperature days	Days with extreme high temperature	64.004	14.987	496
	Extreme precipitation days	Days with extreme precipitation	43.06	22.791	496
	Extreme drought days	Days with extreme drought	35.141	14.439	496

. To mitigate potential estimation bias from data distribution, all variables except the “disaster-affected rate” were log-transformed to address heteroscedasticity and reduce the influence of outliers and scale differences.

2.3. Baseline Regression

This study designates food legume output as the dependent variable and employs DML model to identify the impact of plant protection research investment on food legume output. In this study, the regression analysis is implemented using the ddml command in Stata 17, with Random Forest specified as the machine learning estimator. For robustness checks, alternative algorithms are employed in subsequent analyses for comparison. The estimation results, presented in

Table 2. Baseline Regression.

Variable Name	Food Legume Output	Food Legume Output	Food Legume Output
	Random Forest	Random Forest	Random Forest
Scientific research investment in plant protection	0.186 ***	0.192 ***	0.200 ***
	(0.044)	(0.048)	(0.049)
First-order term of control variable	YES	YES	YES
Second-order term of control variable	NO	NO	YES
Provincial fixed effects	YES	YES	YES
Year fixed effects	NO	YES	YES
Observation	496	496	496

Note: *** indicates significance at the 0.01 confidence level, and the values in parentheses are robust standard errors.

, demonstrate a significant and positive effect. In column (1), which includes first-order control variables and province fixed effects, the coefficient is significant. In columns (2) and (3), we sequentially add year fixed effects and second-order control variables, with the results remaining robustly significant. The final specification in column (3), which controls for two-way fixed effects and high-dimensional control variables, shows that investment in plant protection research has a significant promotional effect on food legume output at the 1% confidence level, with a coefficient of 0.200. This indicates that for every 1% increase in research investment, food legume output increases by an average of 0.2%. This effect highlights the crucial role of research investment in reducing output losses and achieving stable production increases by enhancing pest and disease monitoring, adopting green control technologies, and breeding resistant varieties. The output effect of this research can be understood as the continuous recovery of losses that would otherwise be caused by pests and diseases.

2.4. Robustness Test

2.4.1. Winsorization and Adjustment of Control Variables

To mitigate the potential influence of extreme values and outliers on our estimates, we performed winsorization on all continuous variables at the 1% and 3% levels. As shown in

Table 3. Winsorization and Adjustment of Control Variables.

Variable Name	Food Legume Output	Food Legume Output	Food Legume Output
	1% Winsorization	3% Winsorization	Adjusting Control Variables
Scientific research investment in plant protection	0.165 ***	0.169 ***	0.204 ***
	(0.035)	(0.035)	(0.048)
First-order term of control variable	YES	YES	YES
Second-order term of control variable	YES	YES	YES
Provincial fixed effects	YES	YES	YES
Year fixed effects	YES	YES	YES
Observation	496	496	496

Note: *** indicates significance at the 0.01 confidence level, and the values in parentheses are robust standard errors.

, after re-estimating the model with the winsorized data, the regression coefficient for research investment remains positive and significant at the 1% confidence level. This consistency with the baseline results suggests that our findings are not unduly influenced by outliers. In the process of handling missing values, the crop disaster affected rate provided by the National Bureau of Statistics exhibits a substantially higher proportion of missing observations (around 20%) compared with other variables, which show only sporadic missing values in a few years or regions. To further examine whether such data treatment might undermine the robustness of our estimates, we exclude the Disaster affected rate variable in the robustness checks and re-estimate the models with the remaining control variables. The results remain consistent with the baseline regressions, indicating that our findings are robust.

2.4.2. Lag Effects

To address potential endogeneity that may arise from reverse causality (i.e., provinces with higher outputs may invest more in research), we estimated the model using lagged values of research investment. This method helps to isolate the causal effect of past investment on current output. The results, presented in

Table 4. Lag Regression.

Variable Name	Food Legume Output	Food Legume Output
	First Lag Period	Second Lag Period
L.Scientific research investment in plant protection	0.156 ***
	(0.046)
LL.Scientific research investment in plant protection		0.109 ***
		(0.042)
First-order term of control variable	YES	YES
Second-order term of control variable	YES	YES
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
Observation	465	434

Note: *** indicates significance at the 0.01 confidence level, and the values in parentheses are robust standard errors.

, indicate that investment lagged by one and two periods still has a significant positive impact on food legume output, confirming the robustness of our baseline findings. Furthermore, the coefficients exhibit a decaying effect over time, decreasing from 0.156 for a one-year lag to 0.109 for a two-year lag. This suggests that the benefits of investing in research are most impactful in the near term and diminish over time, highlighting the necessity of sustained and continuous investment to maintain output stability against persistent pest and disease pressures.

2.4.3. Replacing Algorithm Model

To further verify the robustness of our conclusions, we replaced the Random Forest algorithm in the baseline model with three different machine learning methods—Lasso, Gradient Boosting, and Neural Networks—and re-ran the DML estimation. As shown in

Table 5. Replacing Algorithm Model.

Variable Name	Food Legume Output	Food Legume Output	Food Legume Output
	Lasso	Gradient Boost	Neural Networks
Scientific research investment in plant protection	0.104 *	0.191 ***	0.090 ***
	(0.053)	(0.047)	(0.025)
First-order term of control variable	YES	YES	YES
Second-order term of control variable	YES	YES	YES
Provincial fixed effects	YES	YES	YES
Year fixed effects	YES	YES	YES
Observation	496	496	496

Note: * and *** indicate significance at the 0.1 and 0.01 confidence levels, respectively, and the values in parentheses are robust standard errors.

, although the estimated coefficients vary across the different algorithm settings, investment in plant protection research consistently shows a significant positive effect on food legume output, which is in line with the baseline regression results. By comparison, the Random Forest algorithm was retained as the primary algorithm for this study due to its superior compatibility with high-dimensional control variables and its better adaptability to the actual production conditions of food legumes.

2.4.4. Alternative Time Windows

To account for potential structural changes over time, we divided the sample into two sub-periods: 2008–2015 and 2016–2023. Separate regressions were conducted for each period. The results in

Table 6. Alternative Time Windows.

Variable Name	Food Legume Output	Food Legume Output
	2008–2015	2016–2023
Scientific research investment in plant protection	0.215 ***	0.312 *
	(0.065)	(0.169)
First-order term of control variable	YES	YES
Second-order term of control variable	YES	YES
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
Observation	248	248

Note: * and *** indicate significance at the 0.1 and 0.01 confidence levels, respectively, and the values in parentheses are robust standard errors.

show that research investment had a significant positive impact in both periods, confirming the stability of our findings. Notably, the coefficient is larger in the second period (0.312) compared to the first (0.215), suggesting a cumulative effect. This indicates that long-term, sustained investment in plant protection research outputs increasing returns over time as technological knowledge accumulates and disseminates.

2.4.5. Changing Control Variable

The goal of scientific research on plant protection for food legumes is to enhance pest and disease control efficiency and to reduce pesticide residues and environmental pollution through scientific strategies such as precision pesticide application, pesticide substitution, green control, biological control, and resistant breeding. Therefore, in our preceding analysis, we controlled for the “physical input” of pesticides so that the model would better reflect the realization of efficient and green control based on scientific pest management strategies. The focus is on whether, while controlling for pesticide dosage, plant protection research for food legumes can make control more efficient and thereby increase output. This reflects the guiding principle of “enhancing efficacy without increasing volume” for pesticides in actual research work.

In this section, we relax the control on the “physical input” of pesticides to further enhance the robustness of the baseline regression. Column 1 shows the impact while controlling for pesticide input, and column 2 shows the impact without controlling for pesticide input. The baseline regression results are considered relatively robust; regardless of whether pesticide input is considered, the scientific research investment in plant protection for food legumes has a significant impact on food legume output. In terms of coefficient differences, the small change between the two groups of estimated coefficients indicates that the scientific research of plant protection for food legumes does not rely on “increasing pesticide dosage” to achieve extensive output increase. Instead, it enhances the efficiency of pest and disease management through means such as technical upgrading, precision control, and green alternatives, reflecting the scientific control principle of “enhancing efficacy without increasing volume” (

Table 7. Changing Control Variable.

Variable Name	Food Legume Output	Food Legume Output
	Controlling Pesticide Usage	Not Controlling Pesticide Usage
Scientific research investment in plant protection	0.200 ***	0.198 ***
	(0.049)	(0.050)
First-order term of control variable	YES	YES
Second-order term of control variable	YES	YES
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
Observation	496	496

Note: *** indicates significance at the 0.01 confidence level, and the values in parentheses are robust standard errors.

).

3. Discussion

3.1. Investment in Food Legume Plant Protection Research Shows a Significant Upward Trend

China’s investment in plant protection research for food legumes has steadily increased from 2008 to 2023, with an average annual growth rate of 5.19%. The total investment in 2023 was 2.14 times that of 2008. The trend can be characterized by two phases: a rapid growth phase from 2008 to 2015, with an average annual investment of 13.17 million yuan and a growth rate of 9.97%, followed by a period of slower growth from 2016 to 2023, with an average annual investment of 17.60 million yuan and a growth rate of 0.24%. Investment levels vary significantly among provinces, with Gansu, Hebei, Jiangsu, Yunnan, and Shanxi being the top five recipients.

3.2. Contribution of Pest and Disease Control Measures to Mitigating Output Loss

First, diagnostic accuracy for food legume diseases has been improved through methods such as field surveys, molecular detection of pathogens, and technical training. This has reduced the use of indiscriminate control tactics, thereby safeguarding production safety and green quality. Second, a series of resistant crop varieties have been bred. These include new mung bean varieties resistant to bean weevils, Cercospora leaf spot, powdery mildew, and viral diseases; pea varieties resistant to powdery mildew and Fusarium wilt; and broad bean varieties resistant to red spot disease and viral infections. The adoption of these varieties has led to significant reductions in pesticide use and production costs, protecting both the environment and food safety. Third, green pest control technologies have been developed and integrated, combining enhanced field management with the release of natural enemies and the application of biopesticides or high-efficiency, low-toxicity chemical agents. For instance, a green control technology targeting the cowpea pod borer on mung beans exhibited an average efficacy of 87.34% and recovered 43.87% of yield losses in large-scale applications across areas such as Nanyang in Henan, and Wuhan, Ezhou, and Xiangyang in Hubei. Concurrently, research in the field has identified safe and effective herbicides for specific legume-weed combinations, clarified herbicide resistance characteristics and phytotoxicity symptoms, and developed mitigation techniques. This has led to the integration of green weed control protocols for irrigated regions in Northwest China.

3.3. High Marginal Returns on Investment in Plant Protection Research

The economic contribution of investment in plant protection research is characterized by high marginal returns. According to our model, a 1% increase in investment corresponds to a 0.2% increase in food legume output. The mechanism of action primarily enhances pest and disease control capabilities through monitoring and early warning for pests and diseases, green control, and breeding resistant varieties. This reduces food legume output losses and achieves stable and increased production. The output effect of plant protection research for food legumes can be understood as the continuous recovery of losses caused by pests and diseases. For instance, based on the sample mean estimation, with every 10,000 yuan increase in research investment, the recovered output loss can reach approximately 503.34 tons, equating to an economic benefit of 4.4797 million yuan (at 8900 yuan per ton, or approximately 1251 USD per ton; a ratio equivalent to an approximate return of 4.48 million USD on a 10,000 USD investment). Although this figure may seem high, it does not imply that the research itself has such a direct output conversion. In fact, this high return reflects the multiplier effect of research investment. Through the interactive effects between multiple technical links and production factors—such as the construction of green control systems, resistant breeding, and enhanced precision of application—plant protection research for food legumes indirectly promotes the reduction in pest and disease losses and the improvement of production efficiency, thereby achieving a systemic improvement in output. Therefore, this rate of return can be understood as a long-term, systemic net effect rather than a one-time input-output ratio. The phased regression results (Table 6) indicate that the coefficients for the two periods are 0.215 and 0.312, respectively. During the 2008–2015 period, research investment already had a significant promotional effect on output; in the 2016–2023 period, the output-increasing effect of research investment was further enhanced, with the magnitude of the output increase being approximately 1.45 times that of the earlier period. The effect of plant protection research investment is not immediate but emerges gradually over time; the cumulative technical effects from the 2008–2015 period were further released after 2016. This further verifies the long-term nature of plant protection research, indicating that only through continuous investment can more significant output-increasing effects be achieved in the future.

4. Conclusions and Recommendations

Our study finds that investment in plant protection research has a significant and robust positive impact on food legume output in China. This conclusion holds across a comprehensive series of robustness checks, which confirm that the effect is driven by technological progress that leads to “increased efficiency without increased quantity” of pesticide use. Based on these findings, we propose the following recommendations to advance food legume research and development:

4.1. Strengthen Long-Term Investment in Plant Protection Research

Our study confirms that the benefits of plant protection research are long-term and cumulative, with significant marginal returns. This indicates that such research is a critical driver for increasing production capacity and achieving green output growth. Therefore, policymakers should increase sustained investment in this area. It is also vital to improve mechanisms for technology transfer and promote the widespread adoption of efficient, precise, and green pest control technologies, especially for specialty crops like food legumes.

4.2. Foster Collaboration Between Research Institutions and Enterprises

While enterprises are poised to become the primary innovators in the food legume seed industry, the core innovative capacity currently resides within public research institutions [32]. Policies should be designed to incentivize collaboration between seed companies and academic institutions. Such partnerships can create regional innovation advantages, enhance the promotion of local varieties, and foster agglomeration effects in regional breeding programs.

4.3. Establish a Diversified Funding Mechanism

To reconcile the long-term nature of research with short-term market demands, a diversified funding mechanism is needed. This mechanism should integrate government, financial, and corporate investment. The government should provide stable support for foundational, long-cycle research, such as innovation in germplasm resources and eco-friendly cultivation technologies. In parallel, enterprises can focus on developing and promoting market-oriented technologies on shorter timelines, creating a positive feedback loop between research and market needs. Furthermore, public–private partnerships (PPPs) can be used to attract social capital, ensuring sustained funding for public-interest research while enabling firms to achieve commercial returns.

4.4. Promote a Virtuous Cycle Between Regional Innovation and Commercialization

A dual-track approach can effectively advance the food legume industry. First, the government should fund regional research to address local challenges, such as developing location-specific cultivation techniques and breeding new varieties. This fulfills the government’s role in serving the public interest. Second, enterprises can then commercialize these public-funded technologies, developing innovative products that meet consumer demand. This synergy accelerates the market entry of research outcomes, driving both continuous technological innovation and industrial value creation.

4.5. Enhance the Application of Artificial Intelligence in Plant Protection

Accelerating the integration of AI in food legume plant protection can overcome key challenges, such as over-reliance on manual experience and delayed monitoring. Key priorities should include improving the accuracy of pest and disease monitoring and early-warning systems, applying AI to green pest control, pesticide screening, and the breeding of disease-resistant varieties, developing intelligent equipment for automated inspection and targeted pesticide application, and building efficient, coordinated systems for intelligent decision-making and field operations. These advancements will comprehensively enhance the economic and ecological returns on plant protection investments.