Sustainable Agriculture in the Face of Climate Change: Exploring Farmers’ Risk Perception, Low-Carbon Technology Adoption, and Productivity in the Guanzhong Plain of China

Abstract

Agriculture is a significant contributor to global greenhouse gas emissions, and reducing carbon emissions in this sector is essential for mitigating global warming. To achieve China’s targets of carbon peak by 2030 and carbon neutrality by 2060, promoting low-carbon agricultural technology (LCAT) is fundamental. This study examines the impact of farmers’ risk perception on LCAT adoption behavior and its productivity effects with the Ordered Probit regression method, using micro survey data from 531 farmers in Shaanxi Province, China. The results show that farmers with stronger risk perceptions were more likely to adopt LCAT, based on their loss aversion characteristics. Additionally, farmers’ perceptions of yield, market, and climate risks positively influence the adoption of LCAT, with market risk perception having the strongest effect. Adopting LCAT has significant production and spillover effects, improving the output rate of farmers’ operating farmland and neighboring plots by 2.4% and 1.2%, respectively, for each additional measure adopted. This study contributes to the perception and loss aversion literature by examining farmers’ adoption of low-carbon agricultural practices. This study sheds light on the importance of risk perception in the adoption of sustainable agricultural practices and can inform policies aimed at promoting the adoption of LCAT for achieving sustainable agriculture and mitigating climate change, highlighting the crucial role of sustainable environmental management in the agricultural sector.

1. Introduction

Global climate change has led to frequent extreme weather events, such as floods, droughts, and high temperatures, resulting in reduced food production and quality, as well as increased distribution of pests and diseases, increasing the global food security risk [1,2]. In addition to the risks caused by climate change, external shocks such as the COVID-19 epidemic, the Russia–Ukraine conflict, and the energy crisis threaten the global food production and supply system. Global food security tends to deteriorate and resolving the global food crisis remains a primary prerequisite for ensuring human security. Human-induced carbon emissions are the main cause of climate change, and agriculture is a major contributor to global carbon emissions (The Intergovernmental Panel on Climate Change, 2021), accounting for about 13% (Data from United Nations Framework Convention on Climate Change, Climate Action and Support Trends) of total global carbon emissions from the agricultural sector alone [3,4]. In 2018, China’s total agricultural carbon emissions amounted to 801.61 million tons of carbon dioxide, accounting for 6.85% of China’s total carbon emissions. Consequently, the matter of low-carbon agricultural development has garnered significant attention across various sectors in China. The challenge lies in achieving consistent production and supply while simultaneously reducing carbon emissions in agriculture. This task requires joint efforts from national policymakers and participants in agricultural production [5]. As a result, the low-carbon agricultural development model is gradually emerging as a prominent trend in sustainable agricultural practices.

The low-carbon agricultural development model primarily involves transforming unsustainable production methods in agriculture. This includes reducing reliance on industrial products, actively enhancing the amount of carbon sequestered by forest vegetation, crops, and soil, promoting resource recycling, and ultimately achieving maximum output with minimized carbon costs [6,7]. In the traditional agricultural production method, extensive and intensive tillage and irrigation practices are commonly employed. Additionally, a significant amount of pesticides, agricultural films, fertilizers, and other chemical elements are used. These practices have the potential to disrupt the structure of the soil and result in substantial carbon emissions. Various aspects of agricultural production, such as tillage, irrigation, agricultural inputs, and waste disposal, contribute significantly to carbon emissions, making them the primary source of agricultural-related carbon emissions [8]. Meanwhile, however, we cannot ignore the fact that agricultural production itself has the dual effect of carbon emissions and carbon sinks, and forests, grasses, crops and soils are important aspects of carbon sequestration in agriculture [9]. Therefore, LCAT, as an important tool for carbon sequestration and emission reduction, is also the core driver of the low-carbon agricultural production model.

LCAT can effectively reduce agricultural carbon emissions, increase the soil’s ability to store water, and improve air quality, and has been gradually implemented in the form of fallow, reduced tillage, straw mulching, deep land loosening, and comprehensive pest control [10]. Although existing studies have extensively discussed low-carbon agriculture, they have not strictly defined and differentiated the types of LCAT. Wenjing Li et al. [11] investigated the factors influencing the adoption of LCATs among rice farmers, where LCATs included ten technologies such as no-till direct seeding, intermittent irrigation, soil testing and fertilizer application, and straw return. Kai Li et al. [12] discussed the diffusion of LCAT among rice farmers in the Zhejiang province, where LCAT includes four techniques of organic fertilization, stalk shredding and plowing, integrated pest management and slow-release fertilizer. It is observed that the selection for LCAT currently mainly depends on the technology produced and the type of crops grown. Therefore, regarding the study area and the type of crops grown in this paper, as well as the type of LCAT that farmers practically adopt, the following four types of technologies were identified for this study: minimum and no tillage, subsoil loosening, straw returning and integrated pest and weed control. Available literature on farmers’ LCAT adoption behavior has been relatively abundant, mainly in terms of farmers’ endowment [13], technology awareness [14], social networks [15], technology environments [16], and government extensions [17]. However, the focus on farmers’ risk perceptions is still lacking. The individual risk-decision behavior of farmers is mainly determined by their subjective judgment of objective probability. Therefore, exploring the influence of farmers’ risk perceptions on the adoption of this technology is of great theoretical and practical importance.

Although LCAT has unparalleled effects on traditional farming technologies, the widespread dissemination and diffusion of new technologies is a slow process [18]. Farmers are the main adopters of technologies, and their willingness and behavior to adopt each technology directly determine its effectiveness. Most previous studies on farmers’ decision-making behavior have been conducted under the Expected Utility Theory (EUT), which assumes that farmers are ‘perfectly rational’ [19]. However, in reality, decision makers have very limited access to information in the decision-making process and are often influenced by personal subjective factors and the environment they are in [20]. Loss aversion is an important criterion of individual behavior in PT, which refers to behavioral decision makers who show more sensitivity to losses when faced with the same level of gains and losses. Some studies have used this theory to analyze farmers’ decision to adopt agricultural technology; however, research on the effect of loss aversion on behavior is still in its infancy. Farmers are accustomed to coping with year-to-year climate changes. Woods et al. (2017) [21] assessed how farmers perceive climate change and their likelihood of undertaking adaptive actions, indicating that the more concerned they are about climate change, the more they can accommodate it. Farmers’ adoption of climate change responses is determined by both their loss aversion and benefit preferences. He et al. (2019) [22] explored the effect of farmers’ risk preferences and loss aversion on their energy-efficient appliance use behavior in rural China. They find that more loss-averse farmers are more willing to purchase and more likely to use durable energy-efficient appliances. Kibet et al. (2018) [23] conducted a social experiment on French bean farmers living in Kenya and found that the degree of risk aversion had a significant positive effect on compliance with good agriculture practices (GAP). By contrast, loss aversion has a significant negative effect on farmers’ GAP compliance. Ward and Singh (2015) [24] investigated this issue in India and found that farmers with higher loss aversion were more inclined to adopt new rice cultivars with risk resilience under adverse conditions. Based on the available literature, the rationality of explaining farmers’ decision-making behavior from a behavioral economics perspective is confirmed, but few studies have applied it to farmers’ LCAT adoption behavior decisions.

To conclude, the key question of this study, the relationship between farmers’ risk perceptions and LCAT adoption behavior, is proposed, and possible innovations are attempted in two aspects. First, it is about the innovation of research perspective. Few studies have analyzed farmers’ adoption of LCAT from the perspective of behavioral economics. To fill this gap, our study analyzes the role of the psychological characteristics of farmers’ loss aversion in the influence of risk perception on LCAT adoption behavior from the perspective of farmers’ behavioral economics and PT. Second, it is about the innovation of research content. In addition to analyzing the impact of farmers’ risk perception on their adoption behavior of LCAT, this study further analyzes the effectiveness of the application of LCAT, which determines the sustainability of this technology adoption. Combining the key issues and possible innovations of the study, the specific objectives of this study are proposed, which are as follows: using survey data from 531 farmers in the GuanZhong Plain of Shaanxi Province, China, we analyzed the influence of farmers’ risk perception on their adoption of LCAT, using a Multivariate Ordered Probit model based on their psychological characteristics of loss aversion. We further examined the effect of technology adoption at the crop yield level, which provides a reference for understanding farmers’ technology adoption decisions under uncertainty.

The remainder of this study is organized as follows. Section 2 introduces LCAT, conducts a theoretical analysis of its practice in China, and develops the testable hypotheses. Section 3 presents the materials and methods, focusing on the definition of data and variables and the econometric model. Section 4 reports the empirical results, and Section 5 draws conclusions from the study and summarizes the main findings and policy implications.

2. The Theoretical Basis

2.1. LCAT and Its Application in China

Agricultural carbon emissions include greenhouse gases emitted during the production process of agricultural land and the change in agricultural land use, where “carbon” refers to the carbon standard converted from greenhouse gases (methane, carbon dioxide, etc.), not just carbon dioxide. In 2018, non-CO₂ greenhouse gas emissions such as methane and nitrous oxide accounted for 85.67% of China’s total agricultural carbon emissions, which amounted to 377.45 Mt CO₂ eq and 309.30 Mt CO₂ eq, respectively [25]. Agricultural carbon emissions primarily stem from production activities conducted on agricultural land, forest land, and grassland. These emissions occur through processes such as soil respiration, enteric fermentation of livestock, and manure management [26]. Key agricultural production activities that contribute to these emissions include tillage, irrigation, fertilization, pesticide application, agricultural film usage, agricultural machinery usage, straw management, as well as enteric fermentation of livestock and manure emissions.

Depending on the climatic and topographical characteristics of the region, the LCAT used may also vary, such as the “co-culture of mulberry and fish” in the Pearl River Delta [27] and the “co-culture of rice and fish” in Yangtze River Delta [28]. While the promotion of Low-Carbon Agriculture Technology (LCAT) in Shaanxi Province, a region known for its dry farming and the Loess Plateau, has been underway, several challenges persist. These include a reliance on a single mode of operation, inadequate deep loosening and plowing technology, subpar performance of no-till seeding equipment, and low, inconsistent grain yields. Considering the specific conditions of the study area, our focus was on carbon emissions associated with farming production activities. Through extensive survey interviews, we identified the primary Low-Carbon Agriculture Technologies (LCAT) adopted by farmers in Shaanxi Province, China. These include minimum and no tillage practices, soil subsoiling, straw incorporation, and integrated pest and weed control. The definitions and roles of each technology are detailed in

Table 1. Types of LCAT and their carbon reduction effects.

Type	Definition	Carbon Reduction Effects
Minimum and no tillage	It refers to a type of tillage method in which the crop is sown directly on stubble land without plowing or harrowing before sowing, without tillage during the reproductive period of the crop after sowing, and chemical herbicides are sprayed before and after sowing to eliminate weeds.	Compared with the traditional tillage method, reduced tillage can reduce the damage of the soil layer by farm machinery, which in turn increases soil organic matter and soil carbon sinks [29]. The carbon sink content of the 0–20 cm soil layer under no-till is 30% higher than that of tillage [30].
Subsoil loosening	It is a mechanized land preparation technology that loosens the soil without disturbing the original soil structure using tractor traction loosening equipment.	The deep loosening technology can loosen the soil without affecting the soil layer, effectively avoiding water loss leading to soil carbon exposure, thus achieving the best effect of soil carbon sequestration [31].
Straw returning	It is a method of applying straw (wheat straw, corn straw, and rice straw, etc.), which is not suitable for direct use as feed, directly, or in piles after it has been rotted into the soil.	Straw return technology can not only fully offset the carbon emissions from burning, but also enhance the soil carbon sequestration potential, which has a certain carbon sequestration and emission reduction effect [32].
Integrated pest and weed control	It is used to mitigate or prevent pathogenic microorganisms and pests from harming crops, human, and animals by artificially adopting certain means, which can generally be divided into chemical control using chemical substances such as pesticides and physical control using physical energy such as light or rays, or the construction of barriers.	Oak Ridge National Laboratory estimates that every 1 kg reduction in pesticide application will achieve 4.934 kg of carbon emissions, the technology can achieve a reduction in pesticide use pest control, can effectively reduce the density of pathogens and harmful microorganisms effectively and quickly, can maintain and increase the number of soil microorganisms and their enzyme activity, increase soil quality, and achieve carbon emissions reduction [33].

.

2.2. Theoretical Framework and Research Hypothesis

2.2.1. Prospect Theory

Prospect theory (PT), as a risk decision model, has been widely used in finance [34], insurance [35,36], consumption-savings decisions, and other domains where risk attitudes play a central role. However, its application in the Three Rural Issues (Agriculture, Rural and Farmer) is still very limited. Only a few scholars have applied this theory to farmers’ land transfer behavior [37] and local land supply decisions [38]. Owing to the lack of application of this theory in the field of agricultural production, no scholars have analyzed the decisions of farmers to adopt LCAT by combining this theory.

The core concern of this study is whether the relationship between farmers’ risk perception and LCAT adoption behavior can be explained by the loss aversion feature of PT. This question can be answered in the following three dimensions. First, regarding the research subject, farmers are a typical group with incomplete rationality, owing to their incomplete cognition and poor information in rural areas, which is consistent with the premise of the theory. Second, PT itself is a model of risky decision making that applies psychological research to economics and is suitable for analyzing human judgment and decision analysis under uncertainty. Third, in the research context, widely accepted findings in cognitive psychology posit that decision makers exhibit aversion to losses rather than to potential corresponding gains. Alternative research has also demonstrated that actors exhibit a preference for action when it involves the possibility of improvement [39]. Accordingly, when farmers perceive the risks faced in the agricultural production process, that is, when they realize that certain losses will be incurred, they develop an aversion psychology to maximize their avoidance of these possible losses. Farmers are more inclined to adopt LCAT, which provides both environmental and economic benefits. As mentioned above, it is appropriate to use the loss aversion characteristics of PT to analyze the influence of farmers’ risk perceptions on LCAT adoption behavior.

To explain loss aversion in PT, three basic elements explain how individuals make decisions under uncertainty [40]. First, reference dependence exists, in which gains or losses are examined relative to a reference point. The second is loss aversion (relative to a reference point), in which the value function of the losses is steeper than that of the gains of equal size. Third, the marginal value of gains or losses decreases with the size of gain or loss. Collectively, these characteristics can be expressed using the Formula (1).

$$V\left(X\right)=\{\begin{array}{c}{\left(X-r\right)}^{\alpha } if X\ge r\\ \beta {\left(r-X\right)}^{\lambda } if X<r\end{array}$$

(1)

where $V\left(X\right)$ is the value function of X, $r$ is the reference point for non-zero values (reference dependence), a < 1 and λ < 1 (diminishing sensitivity), and β > 1 (loss aversion). The above characteristics of PT can be expressed by Figure 1.

In the following analysis, we used farmers’ current year’s grain production status as the reference point [41]. This is because the current agricultural production conditions are a natural focus for farmers, who often make production decisions based on them. Using the current year’s production status as a reference point, farmers assess the risk of food production, anticipate the magnitude of potential losses, and decide on future food production behavior under the influence of loss-averse psychological characteristics.

2.2.2. Loss Aversion Characteristics of Farmers in the Technology Adoption Process

Agricultural risk encompasses unforeseen events resulting from unavoidable circumstances experienced by farmers during their involvement in agricultural production and operations. Risk perception is a subjective judgment made by people regarding the characteristics and severity of a particular agricultural risk [42]. Experienced farmers draw upon their natural environment and production expertise to subjectively assess the risks encountered during the agricultural production process. These assessments serve as the foundation for their decision-making processes. In the process of subjective risk determination, farmers’ loss aversion psychology is an important factor that influences their behavioral decisions. Therefore, we used loss aversion to explain the influence of farmers’ risk perception on their LCAT adoption behavior.

The stronger the risk perception of farmers, the more pronounced their loss aversion characteristics [43]. When farmers face yield declines, price fluctuations, and extreme weather, loss-averse farmers are motivated to minimize losses through various measures, such as reducing soil erosion on agricultural land, improving the agricultural production environment, and increasing agricultural production efficiency [42]. As an environmentally friendly and risk-reducing farming technique, LCAT can precisely compensate for the needs of farmers owing to the loss avoidance psychology. LCAT offers many benefits that traditional farming cannot offer, mainly in terms of reduced workload (labor time), fuel savings, improved soil quality, and better air quality. Therefore, LCATs are more easily accepted and applied by farmers than systematic innovative recycling technologies (agroecological technology), and do not require large prior capital investment and professional technical training.

According to the loss aversion feature of PT, a farmer facing potential losses would be more inclined to adopt a new technology than those with an expected gain of the same size [41]. Based on the survey data, we found that the average number of years of farming experience was more than 15 years. Thus, farmers had considerable experience in farming, and their adoption decision was not only whether to adopt but also the number of LCATs adopted. Farmers facing expected losses can also reduce their losses by increasing their technology adoption level. While it may not be feasible to quantitatively measure the exact expected losses of a farmer and determine their “true losses”, it is possible to conduct scientific interviews with farmers to gauge their level of pessimism regarding the anticipated losses they may face in the presence of risk.

2.2.3. Farmers’ Risk Perception and Technology Adoption Behavior

The complexity and variety of risk factors faced in agricultural production and operation processes leads to the broadness, complexity, and diversity of agricultural risks [44]. In addition to measuring the overall risk perception of farmers, we classified agricultural risks into the following three types: yield risk [45], market risk [46], and climate risk [47], combining existing research results with actual surveys of agricultural production and operation processes. The perceptions of each type of risk and farmers’ LCAT adoption behavior were analyzed separately for the characteristics of farmers’ loss aversion, and hypotheses were formulated.

(1)

Field risk perception and technology adoption behavior

Stable food production is essential to ensure food security. For farmers, yield levels directly reflect the agricultural production results. In general, some of these are caused by uncontrollable natural factors, including meteorological, biological (diseases, insects, and rodents), geological, and environmental factors. Others are due to farmers’ behavior, including the unreasonable use of pesticides and fertilizers, frequent changes in new species, shortage of agricultural labor, and incomplete understanding of new technologies. Since farmers have been cultivating for a longer time and have abundant farming experience, they tend to form yield expectations in advance and, at the same time, will show more sensitivity to yield loss aversion. Therefore, farmers with production experience tend to be willing to pay higher costs to reduce losses from yield risks. This is consistent with Patrick’s [48] study, in which loss-averse individuals were more inclined to select new varieties that could reduce risk. Yield-risk-averse farmers will also be more inclined to choose LCAT with increased yield effectiveness.

(2)

Market risk perception and technology adoption behavior

The factors that cause agricultural market risks are often diverse, and there are large uncertainties such as changes in market conditions, shifts in consumer demand, and changes in economic policies. The causes of agricultural market risk can be analyzed from the following two aspects: on one hand, it can be analyzed from the perspective of individual farmer information asymmetry. As a production subject, farmers’ judgment of risk and expectations of return will have a greater impact on their production behavior. Due to the constraints of the poor cultural quality of the farmer group, the low level of household income, traditional concepts, and short-sightedness can lead to some incompletely rational behavior of farmers in the process of agricultural production [49], which increases the possibility of farmers’ exposure to market risks. On the other hand, the analysis was conducted from the perspective of the geographical conditions of the study area. The complex topographic conditions and severe land fragmentation in rural areas increase the difficulty of transporting agricultural products, which in turn increases the possibility of price fluctuations in agricultural products and the uncertainty of farmers’ expected returns. The prominent manifestation of agricultural market risk is the problem of price fluctuations and the stagnation of agricultural products. The increase in the magnitude and frequency of price fluctuations of agricultural products and the increase in storage costs owing to the stagnation of agricultural products can cause significant losses to agricultural production and operation. To avoid possible losses caused by market risks to the greatest extent possible, farmers will try to ensure food production and improve the quality of agricultural products to satisfy consumers’ needs as much as possible, given that they cannot change the geographical and objective conditions. Low-carbon agriculture, to maintain green and sustainable production in agriculture, is of great importance in reducing agricultural production costs, improving the yield and quality of agricultural products, and enhancing the competitiveness of agricultural markets. Therefore, the greater the perceived market risk, the more willing farmers are to adopt LCAT because of their loss aversion characteristics for market risk.

(3)

Climate risk perception and technology adoption behavior

What distinguishes agriculture from other industries is that the main activities of agriculture are carried out in the open air, which means that the production and operation activities of agriculture depend more directly and closely on natural conditions and are most vulnerable to the influence of the natural environment. Located between the Loess Plateau and the Qinling Mountains, the Guanzhong Plain has a special geographical location and an unstable ecological environment, which makes it more vulnerable to climate change, such as droughts, floods, frosts, hailstorms, high winds, and other extreme weather. Weather changes will lead to higher production and maintenance costs for farmers and even affect their ‘subsistence’ production methods. Consequently, climate change can expose farmers to a high risk of loss. Owing to farmers’ loss aversion characteristics, they adopt technical measures with improvement possibilities to minimize their perceived climate risk losses [50].

To summarize, farmers face risks in the agricultural production process due to yield fluctuations, price volatility, and climate change. However, these risks also present opportunities for improvement, particularly in terms of farmers’ aversion to losses. These opportunities include introducing new crops and varieties, motivating farmers to adopt advanced technologies, and ultimately achieving the objectives of stabilizing production expectations and increasing crop yields. Following the process of theoretical analysis we constructed the analytical framework of this paper, as shown in Figure 2. The following hypotheses are proposed in this study:

Hypothesis 1:

The stronger the risk perception of the farmer, the more inclined he/she is to adopt LCAT.

Hypothesis 2:

Farmers with a perception of yield, market, and climate risks are more willing to adopt LCAT than those who do not perceive risks.

3. Methodology

3.1. Study Area and Data Description

For this study, data were gathered through a field survey of wheat growers in the Guanzhong Plain of Shaanxi Province, China. The survey was carried out by the Grain Production Status Research Group between July and August 2020. A combination of stratified and random sampling methods was employed to ensure representative data collection. The Guanzhong Plain in China is bounded to the north by the Loess Plateau in northern Shaanxi and to the south by the basin in southern Shaanxi, as shown in Figure 3. It is an influential grain production base in China and the province’s largest grain production area. The cropping pattern in this region is mainly wheat–corn rotation, and wheat production accounts for more than 80% of the province, which is a highly intensive grain production area. Based on a comprehensive consideration of grain production levels, regional economic levels, and research feasibility, the group selected Jing Yang and Xing Ping counties in Xianyang City and Pucheng and Fuping counties in Weinan City as research areas for investigation. The specific research process was as follows: 2–3 townships were randomly selected in each district (county and city), 3–5 administrative villages were selected in each township, and finally, 20–30 farming households were randomly selected in each village to represent the overall characteristics of the region, while 550 questionnaires were distributed, 544 questionnaires were returned, and 531 valid questionnaires were obtained after eliminating invalid samples with missing key information and inconsistencies. The validity rate was 97.61%. The farmers’ questionnaire survey was conducted mainly through face-to-face interviews, and the questionnaire contained information on the wheat growers’ household characteristics, wheat growing situation, cultivation risks, and village characteristics. By utilizing various scientific data collection techniques, including pre-surveys, questionnaire modifications, and formal surveys, we conducted an initial verification of the measurement questionnaire’s internal consistency and validity.

The basic description of the sample is described as follows: the sample proportions of Xianyang City and Weinan City were 34.84% and 65.16%, respectively, with Xingping City and Jingyang County accounting for 77.17% and 22.83% of the sample in Xianyang City, and Fuping County and Pucheng County accounting for 57.30% and 42.70% of the sample in Weinan City, respectively. Considering the individual characteristics of the sample, the average age was 61 years, and the proportions of education levels below junior high school, junior high school, and above junior high school were 48.59%, 37.85%, and 13.56%, respectively. Looking at the characteristics of the sample households, the mean number of household laborers engaged in agriculture was approximately two, the mean annual household income was 93,500 yuan, the average percentage of non-agricultural income was 81.6%, and the percentage of purchasing agricultural insurance was only 9.4%. From the characteristics of farming land, the mean value of practical operation of farmland by households was 5.40 hectares, the mean value of land parcels owned was 2.35, the proportion of land rights was 86.1%, and the proportion of farmers who had experienced shocks in the process of food production in the last three years was 35.2%.

3.2. Description of Model Variables

3.2.1. Dependent Variable

The dependent variable in this study was ‘farmers’ adoption of LCAT. According to the connotation of LCAT and the analysis of the actual research situation, LCAT in the study area mainly consists of the following four types of measures: minimum and no tillage, subsoil loosening, straw returning, integrated pest, and weed control. According to the survey, the proportion of farmers adopting LCAT was high, which was related to the ecological vulnerability characteristics of the region. Therefore, the binary method of ‘adoption or not’ is not comprehensive, and this study further investigates farmers’ degree of adoption of LCAT [51]. According to the number of items adopted by farmers for the four technologies, values were assigned from 0 to 4.

3.2.2. Core Independent Variable

The core explanatory variables of this study are farmers’ risk perceptions, which include farmers’ overall risk perceptions and specific risk perceptions, and the specific risk perceptions include yield risk, market risk, and climate risk perceptions. Subsequent evaluation of farmers’ risk perceptions can better reflect their degree of loss aversion [52]. Farmers’ overall risk perception was measured by the following question: ‘How much risk do you think you face in the process of crop production?’ Respondents’ responses were assigned a value of 1, 2, 3, 4, and 5 from ‘no risk’ to ‘great risk’; the higher the value of this variable, the more pronounced the farmers’ loss aversion characteristics. Farmers’ specific risk perception was measured by the following questions: ‘Do you think you are exposed to yield risk in the process of crop production?’, ‘Do you think you are exposed to market risk in the process of crop production?’, ‘Do you think you are exposed to climate risk in the process of crop production?’. When farmers perceive themselves to be exposed to yield, market, and climate risks, they exhibit loss aversion characteristics.

3.2.3. Control Variable

Based on the existing studies and combined with the research reality, 20 other factors influencing farmers’ adoption of LCAT were selected as control variables in this study: household head characteristics (age, gender, and education level), household production and operation characteristics (annual household income, number of agricultural laborers, proportion of non-farm income, distance from township government, area of land operated, degree of fragmentation of farmland, land titling, social networks, agricultural training, online learning, intergenerational transmission, agricultural surface pollution status, whether they experienced shocks, and whether they purchased insurance), and village characteristics (government subsidies, demonstration areas, and large grain-producing counties). The definitions, assignments, and descriptive statistics of each variable are listed in

Table 2. Definition of variables and descriptive statistics.

	Variable	Description	MEAN	SD	MIN	MAX
Dependent variable	Conversation tillage technology adoption	No adoption = 0, One tech = 1, Two techs = 2, Three techs = 3, Four techs = 4	2.759	0.994	0	4
Core independent variable	General risk perception	How much risk do you think you face in the process of crop production? No risk = 1, Not too much = 2, Average = 3, Quite a lot = 4, Very much = 5	3.556	1.229	1.000	5.000
	Yield risk perception	Do you think you are exposed to yield risk in the process of crop production? No = 0, Yes = 1	0.872	0.335	0.000	1.000
	Market risk perception	Do you think you are exposed to market risk in the process of crop production? No = 0, Yes = 1	0.608	0.489	0.000	1.000
	Climate risk perception	Do you think you are exposed to climate risk in the process of crop production? No = 0, Yes = 1	0.716	0.452	0.000	1.000
Production effect	Farming land productivity	How do you think LCATs have affected the productivity of your own farming operation? Significantly lower = 1, Slightly lower = 2, Basically the same = 3, Slightly higher = 4, Significantly higher = 5	3.738	0.725	2.000	5.000
Spillover effect	Adjacent farmland productivity	How do you think LCATs affect the productivity of adjacent plots? Significantly lower = 1, Slightly lower = 2, Basically the same = 3, Slightly higher = 4, Significantly higher = 5	3.290	0.915	1.000	5.000
Control variable	Household head characteristics
	Age	Actual age of household head (years)	60.829	9.585	24.000	86.000
	Gender	Female = 0, Male = 1	0.940	0.238	0.000	1.000
	Education level	Illiterate = 1, Elementary school = 2, Middle school = 3, High school/junior high school = 4, College and above = 5	2.565	0.876	1.000	5.000
	Family characteristics
	Income	Annual total household income (million yuan)	9.35	5.888	0.33	37.016
	Agricultural labor	Number of laborers actually engaged in agricultural production by households	1.896	0.634	0.000	4.000
	Non-farm income	Share of non-farm income in total household income (%)	0.816	0.203	0.000	0.997
	Distance from township government	km	2.546	1.704	0.000	20.000
	Land	Current operating land area (hm2)	5.396	4.089	1.000	64.000
	Degree of cultivated land fragmentation	Number of agricultural parcels	2.350	1.120	1.000	7.000
	Land titling	Whether or not the family land is titled? No = 0, Yes = 1	0.861	0.347	0.000	1.000
	Social network	Share of family name in village (%)	0.470	0.330	0.010	1.000
	Training	Have you participated in training on LCAT? No = 0, Yes = 1	0.036	0.186	0.000	1.000
	E-learning	Have you ever learnt about agricultural technologies or acquired knowledge through your cell phone? No = 0, Yes = 1	0.358	0.480	0.000	1.000
	Intergenerational transmission	Do you have a willingness to pass on generational responsibility for food production and management? No = 0, Yes = 1	0.073	0.261	0.000	1.000
	Agricultural surface pollution	Do you think agricultural surface source pollution in your area is serious at present? No pollution = 1, Relatively minor = 2, Average = 3, Relatively serious = 4, Very serious = 5	1.900	0.943	1.000	5.000
	Experiencing shocks	In the last three years, has your household’s food production process suffered a shock? No = 0, Yes = 1	0.352	0.478	0.000	1.000
	Agricultural insurance	Have you purchased agricultural insurance? No = 0, Yes = 1	0.094	0.292	0.000	1.000
	Village characteristics
	Government subsidies	No = 0, Yes = 1	0.038	0.191	0.000	1.000
	Demonstration area	No = 0, Yes = 1	0.032	0.176	0.000	1.000
	Major food-producing counties	No = 0, Yes = 1	0.369	0.483	0.000	1.000

.

3.3. Model Specification

Let T denote the type of LCAT adopted by farmers to examine the effect of their risk perceptions on the adoption of LCAT. According to the number of technology adoptions, this study assigned values from low to high (0–4) to no adoption, one adoption, two adoptions, three adoptions, and four adoptions, respectively. Because the dependent variable was an ordered categorical variable, a Multivariate Ordered Probit model [53] was used for the empirical estimation. The general form of the model is as follows.

$$T={\beta }_0+{\beta }_{1}P+{{\displaystyle \sum }}_{n=1}{\beta }_{2n}C{V}_n+\epsilon$$

(2)

where T is the dependent variable, the degree of adoption of LCAT by farmers, P is the core explanatory variable, the risk perception of farmers, representing the overall risk perception and separate risk perception of farmers, the separate risk perception includes yield risk perception, market risk perception, and climate risk perception, $C{V}_n$ denotes the vector of control variables, including the variables of household head characteristics, family characteristics, and village characteristics, ${\beta }_0$, ${\beta }_1,$ and ${\beta }_{2n},$ are the coefficients to be estimated, and ε is the residual term, which follows a normal distribution and has a variance of ${\sigma }^2$, that is $\epsilon ~N\left(0, {\sigma }^2\right)$. Simultaneously, we also perform regression analysis using OLS models to ensure the robustness of the ordered Probit model regression results.

4. Result and Discussion

4.1. Farmers’ Risk Perception and Adoption Behavior of LCAT

Our study initially focused on assessing the overall risk perceptions experienced by farmers during grain production. Subsequently, we investigated the influence of farmers’ perceptions regarding yield, market, and climate risks on their grain production process. Among them, 87.2%, 60.8%, and 71.6% of farmers considered that they faced yield risk, market risk, and climate risk, respectively. In addition, this study measured the extent to which farmers simultaneously faced risks, and the results are shown in

Table 3. Statistics on the intensity of risk faced by farmers.

Simultaneous Risk Exposure	N	%
0	51	9.6
1	74	13.94
2	126	23.73
3	280	52.73

. The results showed that the percentage of farmers who considered facing three risks at the same time was the highest (52.73%), and the percentage of farmers who considered no risk of food production in the process of food production was only 9.6%.

Based on the survey, we summarized the specific reasons for farmers’ perceptions of agricultural risks (

Table 4. Statistical analysis of specific reasons for the generation of farmers’ risk perceptions.

Risk Type/Reason	N	%
Yield risk perception	463	87.19
Pests and diseases	333	71.92
Irrigation conditions	113	24.41
Not proficient in farming techniques	35	7.56
Others (high cost, land repossessed at any time, etc.)	254	54.86
Market risk perception	323	60.83
High price fluctuations	90	27.86
Significant changes in supply and demand	77	23.84
Difficulties in transportation and circulation	22	6.81
Others (insecure fixed investment, rice supply outside the province, etc.)	241	74.61
Climate risk perception	380	71.56
More extreme weather	237	62.37
Increased costs due to weather changes	176	46.32
Inadequate agricultural insurance	98	25.79
Others	9	2.37

). The measurement of specific causes of risk generation validates and distinguishes the different types of risks faced by farmers, further increasing the reliability of the measurement of the core explanatory variables. For yield risk prediction, farmers’ judgments are based on the pest and disease situation, agricultural irrigation conditions, and the mastery of agricultural technology faced during the previous year’s grain production [45]. For market risk prediction, farmers’ judgments are based on the previous year’s market price fluctuations, transportation and circulation, and grain supply in neighboring areas [46]. For climate risk prediction, farmers’ judgments are based on the number of extreme weather events, such as droughts and floods, inadequate agricultural insurance systems, and increased costs due to weather changes [47].

Table 5. Adoption of LCAT by farmers.

Technology Types	N	%	Adoption Intensity	N	%
Minimum and no tillage	95	17.89	1	98	18.46
Subsoil loosening	141	26.55	2	281	52.92
Straw returning	465	87.57	3	127	23.92
Integrated pest and weed control	428	80.60	4	22	4.14

shows the farmers’ adoption of LCAT. The percentages of sample farmers adopting minimum and no-tillage, subsoil loosening, straw return, and integrated pest and weed control technologies were 17.89%, 26.55%, 87.57%, and 80.60%, respectively. The adoption rates of straw returning technology and pest control technology were much higher than those of the other two types of technology among the sample farmers. Less than 30% of the sample farmers adopted three or more LCAT, indicating that there is still room for improvement in the degree of adoption of LCAT by farmers. In terms of farmers’ adoption decisions, most farmers adopted LCAT, so for this region, the definition of farmers’ LCAT adoption behavior as a binary variable ignores the examination of the degree of adoption [54].

4.2. Impact of Risk Perception on Farmers’ LCAT Adoption

4.2.1. Benchmark Return

This study used Stata Statistical Software 16.0 (Stata Corp., College Station, TX, USA) for the empirical analysis. Considering that there may be multicollinearity among multiple variables, we carried out a multicollinearity test for each independent variable. The results show that the variance inflation factor (VIF) is 2.06 at the maximum and the mean value is 1.21, which is much lower than 3.0, thus indicating that the degree of correlation collinearity between the variables is within a reasonable range, and there is no apparent multicollinearity.

As the dependent variables were ordered response variables between 1 and 5, the ordered probit model was chosen to perform a maximum likelihood estimation (MLE) of the equations to obtain consistent estimates of the regression coefficients.

Table 6. Effect of farmers’ risk perception on LCAT adoption: benchmark return.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	OLS	OP	OLS	OP	OLS	OP	OLS	OP
General risk perception	0.137 ***	0.159 ***
	(0.039)	(0.045)
Yield risk perception			0.313 **	0.371 **
			(0.150)	(0.171)
Market risk perception					0.392 ***	0.456 ***
					(0.092)	(0.107)
Climate risk perception							0.325 ***	0.372 ***
							(0.095)	(0.108)
Control variables	√	√	√	√	√	√	√	√
_cons	1.930 ***		2.065 ***		2.071 ***		2.076 ***
	(0.673)		(0.675)		(0.675)		(0.668)
R2	0.1401		0.1239		0.1468		0.1340
Pseudo R2		0.0537		0.0472		0.0566		0.0511
N	531	531	531	531	531	531	531	531

Notes: ***, ** and * show significance level at 1%, 5% and 10%. Values in parentheses are robust standard errors. Complete table with control variables can be found in Supplementary Table S1.

reports the overall risk perceptions faced by farmers and the effect of each risk perception (yield, market, and climate) on farmers’ LCAT adoption behavior. The results of the ordered probit model regressions are reported in columns (2), (4), (6), and (8) of Table 6. The regression results show that the coefficients of farmers’ perceived yield risk, market risk, climate risk, and overall risk perception were significantly positive at levels greater than 5%. This indicates that the greater the perceived risk of farmers, the more inclined they are to adopt LCAT, which tentatively confirms Hypothesis 1.This is consistent with Martey et al.’s [55] findings that COVID-19 perceptions of farm household in Guinea, Sudan and Transition zones of Ghana influenced the probability and intensity of adoption of sustainable agricultural practices (SPAs), with COVID-19 shocks promoting the adoption intensity of SPAs.

The Ordered Probit model in Table 6 reports the partial regression coefficients of the explanatory variables that lack real economic implications. For ease of interpretation,

Table 7. Effect of farmers’ general risk perception on LCAT adoption: marginal effect.

	(1)	(2)	(3)	(4)	(5)
	No Adoption	1 Adoption	2 Adoptions	3 Adoptions	4 Adoptions
General risk perception	−0.026 ***	−0.030 ***	0.015 ***	0.026 ***	0.015 ***
	(0.008)	(0.008)	(0.005)	(0.007)	(0.005)
Age	0.001	0.001	−0.000	−0.001	−0.000
	(0.001)	(0.001)	(0.000)	(0.001)	(0.000)
Gender	−0.014	−0.016	0.008	0.014	0.008
	(0.031)	(0.035)	(0.018)	(0.031)	(0.018)
Education level	−0.011	−0.013	0.007	0.011	0.006
	(0.010)	(0.011)	(0.006)	(0.009)	(0.006)
Income	0.012	0.013	−0.007	−0.011	−0.007
	(0.014)	(0.016)	(0.008)	(0.014)	(0.008)
Agricultural labor	−0.020	−0.023	0.012	0.020	0.011
	(0.013)	(0.015)	(0.008)	(0.013)	(0.007)
Non-farm income	−0.063	−0.071	0.037	0.062	0.036
	(0.058)	(0.066)	(0.034)	(0.057)	(0.033)
Distance from township government	0.005	0.006	−0.003	−0.005	−0.003
	(0.004)	(0.005)	(0.002)	(0.004)	(0.002)
Land	−0.003	−0.003	0.002	0.003	0.002
	(0.003)	(0.003)	(0.002)	(0.003)	(0.002)
Degree of cultivated land fragmentation	0.012	0.014	−0.007	−0.012	−0.007
	(0.008)	(0.009)	(0.005)	(0.008)	(0.005)
Land titling	−0.062 ***	−0.071 ***	0.036 ***	0.061 ***	0.035 ***
	(0.023)	(0.025)	(0.013)	(0.023)	(0.013)
Social network	0.019	0.021	−0.011	−0.019	−0.011
	(0.024)	(0.027)	(0.014)	(0.024)	(0.014)
Training	0.006	0.007	−0.003	−0.006	−0.003
	(0.045)	(0.051)	(0.026)	(0.044)	(0.026)
E-learning	−0.042 **	−0.048 **	0.025 **	0.042 **	0.024 **
	(0.017)	(0.019)	(0.011)	(0.017)	(0.010)
Intergenerational transmission	−0.027	−0.031	0.016	0.026	0.015
	(0.034)	(0.039)	(0.020)	(0.034)	(0.020)
Agricultural surface pollution	−0.020 **	−0.023 **	0.012 **	0.020 **	0.011 **
	(0.008)	(0.009)	(0.005)	(0.008)	(0.005)
Experiencing shock	0.023	0.026	−0.013	−0.023	−0.013
	(0.017)	(0.020)	(0.010)	(0.017)	(0.010)
Agricultural insurance	−0.009	−0.010	0.005	0.009	0.005
	(0.030)	(0.034)	(0.017)	(0.029)	(0.017)
Government subsidies	0.043	0.049	−0.025	−0.043	−0.025
	(0.056)	(0.064)	(0.033)	(0.055)	(0.032)
Demonstration area	0.008	0.010	−0.005	−0.008	−0.005
	(0.049)	(0.055)	(0.028)	(0.048)	(0.028)
Major grain-producing counties	−0.090 ***	−0.102 ***	0.053 ***	0.089 ***	0.051 ***
	(0.019)	(0.018)	(0.013)	(0.016)	(0.013)
N	531	531	531	531	531

Notes: ***, ** and * show significance level at 1%, 5% and 10%. Values in parentheses are robust standard errors.

reports the marginal effects of farmers’ overall risk perceptions on the adoption of LCAT, whereas

Table 8. Effect of yield, market, and climate risk perception on LCAT adoption: marginal effect.

	(1)	(2)	(3)	(4)	(5)
	No Adoption	1 Adoption	2 Adoptions	3 Adoptions	4 Adoptions
	Yield risk perception: Marginal effect
Yield risk perception	−0.062 **	−0.071 **	0.036 **	0.061 **	0.035 **
	(0.030)	(0.032)	(0.018)	(0.028)	(0.017)
Control variables	Yes	Yes	Yes	Yes	Yes
N	531	531	531	531	531
	Market risk perception: Marginal effect
Market risk perception	−0.075 ***	−0.086 ***	0.044 ***	0.074 ***	0.043 ***
	(0.019)	(0.020)	(0.013)	(0.017)	(0.012)
Control variables	Yes	Yes	Yes	Yes	Yes
N	531	531	531	531	531
	Climate risk perception: Marginal effect
Climate risk perception	−0.062 ***	−0.070 ***	0.036 ***	0.061 ***	0.035 ***
	(0.019)	(0.020)	(0.012)	(0.018)	(0.011)
Control variables	Yes	Yes	Yes	Yes	Yes
N	531	531	531	531	531

Notes: ***, ** and * show significance level at 1%, 5% and 10%. Values in parentheses are robust standard errors.

reports the marginal effects of yield, market, and climate risk perceptions on farmers’ adoption of LCAT. In terms of the specific results of the marginal effects, for each level of increase in farmers’ overall risk perception, the probability of adopting four LCATs increased by 1.5%, the probability of adopting three LCATs increased by 2.6%, the probability of adopting two LCATs increased by 1.5%, the probability of adopting one LCAT decreased by 3%, and the probability of not adopting LCAT decreased by 2.6%. It was evident that the greater the perceived risk, the more inclined farmers were to adopt LCAT [52]. Similarly, the results of the marginal effects of the risks in Table 8 show that the greater the perceived yield risk, the more farmers tend to adopt LCAT, the greater the perceived market risk, the more farmers tend to adopt LCAT, and the greater the perceived climate risk, the more farmers tend to adopt LCAT. Farmers’ perceptions of the yield, market, and climate risks had the highest marginal impact on the adoption of the three technologies. Farmers were the most sensitive to perceptions of market risks in the adoption of LCAT. The findings align with the research conducted by Albrecht et al. [56], which also observed that acceptance of digital contact-tracing applications (DCTAs) varied based on different risk perceptions. Specifically, individuals who perceived DCTA usage as associated with high data-security risks demonstrated low acceptance, while those perceiving lower risks exhibited high acceptance.

With respect to the control variables, the marginal regression results in Table 7 reveal significant findings. The variable ‘Land titling’ displays a positive coefficient, indicating that farmers are more likely to embrace LCAT once rural land ownership has been clarified [57]. Specifically, the probability of adopting all three LCAT practices increases by 6.1% in comparison to farmers who have not completed the land titling process. This is because of the long-term consideration that farmers are more willing to adopt environmentally friendly technologies to improve and enhance the soil and ecological environment in the future after enjoying land ownership and use. The ‘E-learning’ variable was significant and had a positive coefficient, indicating that farmers who actively learned about agriculture via cell phones were more likely to adopt LCAT [58], with a 4.2% increase in the probability of three adoptions compared to farmers who did not learn online. Farmers who learn online tend to have a stronger learning ability, better understanding, and application of LCAT, and a clearer understanding of the risk-averse function of these technologies, which makes them more likely to use LCAT. The coefficient of the variable ‘Agricultural surface pollution’ was positive and significant, indicating that the more serious the agricultural surface pollution in the farmer’s area, the more inclined the farmer was to adopt LCAT. This may be explained by the fact that the more serious the agricultural surface pollution in the farmer’s area, the more negative the impact on the farmer and, therefore, the stronger the pro-environmental willingness [59]. The significant and positive coefficient of the variable “Major grain-producing counties” in the village characteristics indicates that farmers located in large grain-producing counties were more likely to adopt LCAT, which was related to the production characteristics of the villages and the local government’s policies to promote and support the technology.

4.2.2. Robustness Test

To further test the robustness of the results of the above empirical analysis, we performed robustness tests on the baseline regression results by changing the model form (Ordered Probit model replaced with Ordinary Least Squares model) and core explanatory variables (adjusting the measure of risk perception). First, both Ordinary Least Squares (OLS) methods were used for estimation, as shown in Table 3. Columns (1), (3), (5), and (7) of Table 3 report the OLS estimation results. The OLS regression results were consistent with the Ordered Probit regression results, where the coefficients of farmers’ perceived yield risk, market risk, climate risk, and overall risk perception were significantly positive, at a level greater than 5%. Second, we replaced the core explanatory variable of the overall risk perception by summing the number of farmers subject to all three risk perceptions simultaneously.

Table 9. Robustness tests: replacement of core independent variable.

	(1)	(2)	Marginal Effect
	OP	OLS	No Adoption	1 Adoption	2 Adoptions	3 Adoptions	4 Adoptions
General risk perception	0.222 ***	0.189 ***	−0.036 ***	−0.042 ***	0.021 ***	0.036 ***	0.021 ***
	(0.056)	(0.048)	(0.010)	(0.010)	(0.006)	(0.009)	(0.006)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2		0.1479
Pseudo R2	0.0572
N	531	531	531	531	531	531	531

Notes: ***, ** and * show significance level at 1%, 5% and 10%. Values in parentheses are robust standard errors.

reports the regression results for both the ordered probit and OLS methods, as well as the marginal effects of the replaced explanatory variables, which are consistent with the regression results in Table 3 and Table 4, both in terms of the significance of the variables and the sign of the coefficients. The results of both robustness tests supported the positive effect of farmers’ risk perception on the adoption of LCAT, thus validating research Hypotheses 1 and 2.

4.3. Production and Spillover Effects

We set two outcome variables, farming land productivity and adjacent farmland productivity, as shown in Table 3, to further investigate the adoption efficiency of LCAT by farmers and discuss the production and spillover effects. Whether farmers continue to adopt technologies or increase their adoption level is influenced not only by their risk perceptions but also by whether they could bring positive benefits [60]. As a soil improvement technology, the ability of LCAT to increase productivity while improving soil quality is key to investigating farmers’ adoption of LCAT [61]. Therefore, this study selects farmers’ subjective judgment of benefits as the outcome variable to further explore whether farmers’ adoption of LCAT increases yield and efficiency, and then determines whether farmers are motivated to continue or increase their adoption.

The results of the OLS and Ordered Probit regressions of the effect of farmers’ adoption of LCAT on the productivity of homestead farmlands are reported in columns (1) and (2) of

Table 10. Production effect and spillover effect of LCAT adoption.

	Production Effect: Farming Land Productivity
	(1)	(2)	Marginal Effect
	OLS	OP	Significantly Lower	Slightly Lower	Basically Unchanged	Slightly Higher	Significantly Higher
LCAT adoption	0.074 **	0.121 **	-	−0.008 **	−0.034 **	0.018 **	0.024 **
	(0.032)	(0.051)	-	(0.004)	(0.014)	(0.008)	(0.010)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.1196
Pseudo R2		0.0589
N	531	531	531	531	531	531	531
	Spillover Effect: Adjacent Farmland Productivity
	(3)	(4)	Marginal Effect
	OLS	OP	Significantly Lower	Slightly Lower	Basically Unchanged	Slightly Higher	Significantly Higher
LCAT adoption	0.096 **	0.123 ***	−0.011 **	−0.020 ***	−0.016 **	0.035 ***	0.012 **
	(0.039)	(0.047)	(0.004)	(0.008)	(0.006)	(0.013)	(0.005)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.0877
Pseudo R2		0.0351
N	531	531	531	531	531	531	531

Notes: ***, ** and * show significance level at 1%, 5% and 10%. Values in parentheses are robust standard errors.

, and the regression results are consistent for both methods. The coefficient of farmers’ adoption of LCAT was significantly positive at the 5% level, indicating that farmers’ technology adoption can significantly increase the productivity of farmlands operated by farmers. The results of the marginal effects showed that the probability of significantly increasing the output rate of farmers’ own-operated farmland increased by 2.4% for each increase in farmers’ adoption of LCAT, confirming that the farmers’ adoption of LCAT has certain production effects.

Columns (3) and (4) of Table 10 report the results of the OLS and Ordered Orobit regressions, respectively, on the effect of farmers’ adoption of LCAT on the productivity of adjacent plots. The regression results for both methods were consistent. The coefficient of farmer LCAT adoption was significantly positive at a level greater than 5%, indicating that farmer LCAT adoption can significantly increase the output rate of adjacent farmland plots. From the results of the marginal effects, the probability of significantly increasing the output rate of neighboring farmlands increased by 1.2% for each additional LCAT adopted by farmers, which was less than the output rate of self-managed farmland; however, it further indicated that the adoption of LCAT has a production spillover effect.

5. Conclusions, Policy Implications

5.1. Conclusions

This study utilized micro-survey data from 531 wheat-corn farmers in Shaanxi Province to investigate the adoption of the following four types of Low-Carbon Agriculture Technology (LCAT): minimum and no tillage, subsoil loosening, straw returning, and integrated pest and weed control. Additionally, the study empirically analyzed the impact of farmers’ risk perception on their adoption of LCAT, considering the perspective of loss aversion. Overall, the conclusions drawn in this paper confirmed the content of the theoretical hypotheses, as follows: first, farmers’ risk perceptions have a significant effect on their adoption of LCAT. The stronger the overall risk perception of farmers, the more inclined they are to adopt LCAT. The probability of adopting two, three, and four types of LCAT increased by 1.5%, 2.6%, and 1.5%, respectively, and the probability of adopting one and no adoption decreased by 3% and 2.6%, respectively, with each one-level increase in the farmers’ overall risk perception. Second, farmers who perceived these risks were more likely to adopt LCAT compared to those who did not perceive yield, market, and climate risks. Among them, farmers with perceptions of yield, market, and climate risks had the highest marginal impact on the three adoptions, and their adoption behavior was most sensitive to the perception of market risks. Third, farmers’ adoption of LCAT had significant production and spillover effects. For each increase in the adoption of LCAT by farmers, the probability that the output rate of farmland operated by farmers and the output rate of farmland in neighboring plots will increase significantly by 2.4% and 1.2%, respectively.

5.2. Policy Implication

Based on the above conclusion and field research, the following policy recommendations are proposed for reference. First, the government should strengthen publicity and education on agricultural production risks, and scientifically guide farmers to improve their risk perception. Farmers’ risk perceptions regarding agricultural production vary because of their different living environments, production experiences, and knowledge levels. Hence, it is crucial for the government to enhance the dissemination of information regarding agricultural production risks. This will enable farmers to acquire a clear and comprehensive understanding of the risks involved in agricultural production and operations. Consequently, it will elevate their level of risk perception and stimulate the demand for the adoption of environmentally friendly farming techniques, including LCAT. Second, local governments should use the loss aversion characteristics of farmers to reasonably and effectively guide farmers to carry out carbon reduction production methods. Local governments should emphasize the risk-reducing function of this technology when promoting LCAT to strengthen farmers’ knowledge of this technology. Third, since the adoption of LCAT by farmers has a significant effect on productivity increases, the local government should increase technology promotion efforts. To further increase the adoption rate of Low-Carbon Agriculture Technology (LCAT), it is recommended to continue offering financial support to farmers for technology adoption via government subsidies. Additionally, breaking down technical barriers for farmers can be achieved through technical training programs and technology demonstrations. These measures will contribute to an increased uptake of LCAT among farmers. Fourth, in addition to providing scientific guidance for farmers to improve their technology adoption rate, it is necessary to further accelerate the improvement of the agricultural risk management system. We should continue to improve the quality of all agricultural insurance services, strengthen talent, and business training for agricultural risk management, and insist that technological innovation is an important supporting force for agricultural risk management.

This study attempts to explain the decision-making behavior of farmers from the perspective of behavioral economics, and explains the preliminary conclusion that the stronger the risk perception of farmers, the stronger their motivation to adopt LCAT from the theoretical level. This reflects the current situation of LCAT promotion in the study area and provides certain practical guidance, which helps to promote the development of low-carbon agriculture from the perspective of stimulating the intrinsic motivation of farmers. Simultaneously, the following shortcomings of this study are worthy of continued research. First, in terms of research content, we only considered the impact of farmers’ risk sensitivity on the enthusiasm of technology adoption, but neglected the farmers’ own awareness of low-carbon agriculture. As the name implies, the higher the farmers’ awareness of low-carbon agriculture technology, the easier it is to promote the popularity of the technology. Given that this study did not scientifically measure farmers’ perceptions of low-carbon agriculture during the sample survey, it could not be included in the research framework of this paper. Second, in terms of research data, the micro-data in this study were derived from a field survey of farmers in 2020. This study was conducted during the COVID-19 outbreak, which greatly contributed to the difficulty of the research and affected the number of samples collected. At the same time, due to data limitations, we cannot derive precise values of farmers’ expected benefits, and using only their degree of optimism or pessimism as a proxy variable for true losses is likely to amplify farmers’ negative expectations, and despite the many question items we used to conduct the interviews, we cannot rule out that farmers will amplify their risk perceptions, and thus overestimate their positive impact on technology adoption.