The Effect of Farmers’ Insurance-Adoption Behavior on Input for Beef-Cattle Disease Prevention: Endogenous Switching Regression Model

Abstract

This study selects the Inner Mongolia Autonomous Region (IMAR), among the most crucial beef-cattle farming areas in China, to obtain data from the micro-surveys of 447 beef-cattle farmers. Utilizing an endogenous switching regression (ESR) model, this research empirically investigates the effect of farmers’ beef-cattle insurance enrollment behavior on their input of disease prevention. This study finds that farmers adopting beef-cattle insurance reduce beef-cattle disease-prevention input. Based on counterfactual assumptions, if insured farmers had not adopted insurance, their input in disease prevention would increase by 33.45%. Further research confirms that a decrease in the market purchase price of beef cattle enhances the negative effect of farmers’ insured behavior on input for beef-cattle disease prevention. The heterogeneity analysis leads to two more conclusions. One is that insured farmers have the largest reduction in shed-disinfection input, the smallest reduction in voluntary vaccination input, and an intermediate reduction in deworming input. The other is that the act of adopting insurance reduces disease-prevention input to a greater extent for farmers who are far from the core areas of beef-cattle farming or who have not experienced beef-cattle deaths.

1. Introduction

1.1. Current Challenges in Beef-Cattle Disease Prevention

The adverse impact of animal diseases extends beyond animals and animal husbandry itself [1]. These impacts affect the income and livelihood of farmers [2,3] and threaten human health through diseases arising from food-borne pathogens or direct contact between humans and livestock [4]. Furthermore, animal diseases pose a significant and growing threat to global food systems and can even exacerbate the environmental damage caused by carbon emissions [5,6,7]. As the largest fecal-producing livestock species, beef cattle are among the domestic animals with the greatest individual impact on the environment [8], and beef production has also consistently been identified as the highest emission source of the livestock sector [7]. Therefore, solving the disease problem in the process of beef-cattle farming is essential and will ensure farmers’ income, reduce industrial environmental harm, protect food security, and promote the development of the beef-cattle industry. China, which is the world’s third-largest beef producer, after the United States and Brazil, and the world’s largest beef importer [9,10], has been grappling with the persistent threat of animal diseases but has also perennially been vulnerable to animal diseases [1]. Common beef-cattle diseases in China are foot and mouth disease (FMD) and bovine spongiform encephalopathy, which are Class I animal diseases; brucellosis and bovine nodular skin disease, of Class II; and bovine pasteurellosis and bovine viral diarrhea (BVD), of Class III [11]. For instance, FMD has persistently been classified as a high-risk disease by the World Organization for Animal Health [7,12]. In 2024, China reported three cases of FMD outbreaks. Among them, the infected animals in two cases were all beef cattle. There have also been many cases of beef cattle contracting Class II and Class III diseases, even leading to death, demonstrating that the current beef-cattle industry in China is greatly affected by diseases (Appendix A 

Table A1. Report of major diseases of beef cattle in China.

Month	Class I	Class II			Class III
	FMD	Bovine Nodular Skin Disease	Infectious Bovine Rhinotracheitis	Bovine Tuberculosis	BVD
January	−	1	35	1	884
February	−	−	15	2	709
March	−	−	52	1	1456
April	−	−	25	1	810
May	−	3	19	−	901
June	−	3	9	43	659
July	−	7	11	2	891
August	−	22	56	32	1291
September	7	2	11	30	741
October	2	−	11	5	564
November	10	11	49	31	724
December	−	−	11	34	1205
Total	19	49	304	182	10,835

Note: The data are derived from the “Report on Major Animal Diseases in China” issued by the Bureau of Animal Husbandry and Veterinary Medicine of the Ministry of Agriculture of the People’s Republic of China.

).

By referring to the compulsory immunization schemes implemented in nations like those in Europe [13,14], China has likewise adopted preventive measures for beef-cattle diseases, covering compulsory immunization and proactive disinfection and isolation [15]. Vaccines are among the most cost-effective and sometimes the only means to preventing disease in livestock populations [2]. Therefore, China has implemented a government-subsidized compulsory immunization (CI) policy for three beef-cattle diseases including FMD, brucellosis, and hydatid disease [16]. Nevertheless, implementing beef-cattle immunization also faces many difficulties. First, the government is obliged to expend considerable manpower and financial resources for the administration and implementation of immunization [17]. Second, extant vaccines have encountered difficulties in attaining a high safety standard [3]. Therefore, the effect of the CI policy currently implemented in China has also been limited. This limitation manifests in over-reliance on government-provided free vaccines by livestock farmers, which are only targeted at certain diseases. Thus, many new epidemic diseases, as well as diseases with low treatment costs (e.g., BVD), cannot be prevented in time.

The insurance coverage for livestock also serves as a valuable means for the management of risks associated with livestock farming [18,19]. As early as 2012, China proposed policy propositions such as actively employing insurance and other policy means to enhance the socialized service capability of animal disease prevention, incorporating key animal diseases into the insurance-coverage scope of the livestock industry [20]. Based on the research carried out by American scholars, government-subsidized livestock insurance is a highly favored livestock compensation policy [21]. At present, China mainly implements a policy-based insurance system for beef cattle (Insurance companies commonly undertake the underwriting of losses resulting from beef-cattle deaths attributed to factors like common diseases, infectious diseases, natural disasters, and accidents during the spring. The guarantee duration is typically 1 year, and the insured entities typically encompass cattle with farming capacity, fattening cattle, and calves. The insurance amount coverage is approximately 70–80% of the farming cost (Appendix A 

Table A2. Average cost of free-range beef cattle by region in China.

Year	Cost of Substances and Services per Head (Yuan)	Labor Cost per Head (Yuan)	Total Cost per Head (Yuan)	Percentage of Insurance Claims (%)
2022	12,506.52	1217.52	13,724.04	72.86%
2021	12,620.99	1175.52	13,796.51	72.48%
2020	11,491.02	1147.56	12,638.58	79.12%

Note: The data are derived from the “National Compilation of Cost and Benefit Data for Agricultural Products (2021–2023)”.

), and the government bears 80% of the total premium, with the remaining 20% borne by individual farmers), which is mainly supported by local government subsidies and plays a significant role in supporting the development of the beef-cattle industry. This system has a particular focus on the prevention and control of diseases [22]. Farmers merely need to make a small premium payment to acquire most of their livestock cost assurance, resulting in a high uptake of insurance in China. Most agricultural insurance products in China are provided through government subsidies rather than market mechanisms. Due to the pervasive market failure in the agricultural insurance market [23], exacerbated by systemic risks inherent in agriculture, farmers’ moral hazard, and adverse selection behaviors [24], government financial subsidies for agricultural insurance have become an inevitable outcome [25]. In the operation of agricultural insurance, government financial subsidies have been shown to significantly boost farmers’ willingness to participate in insurance [26]. However, such subsidies may exacerbate adverse selection behaviors among farmers [27], thereby undermining the intended objectives of the insurance schemes [28]. In addition, subsidies may exacerbate rent-seeking behaviors among agricultural insurance companies, such as misappropriating fiscal premium subsidies [29,30], ultimately resulting in the inefficient allocation or the outright waste of subsidy funds [31]. In conclusion, relying solely on government-subsidized beef-cattle insurance is insufficient to achieve the risk management objectives of beef-cattle farmers.

1.2. Research Gap and Objectives

Insurance may incentivize inputs in prevention [32] as disease prevention is typically a prerequisite for claiming insurance compensation [33]. Disease prevention is a proactive risk management strategy implemented prior to the occurrence of risks. Effective disease prevention measures can significantly reduce the likelihood of risk events, thereby minimizing the economic losses that farmers may incur due to beef-cattle infections or even mortality. Risks can be mitigated but not entirely avoided. At this point, insurance serves as a post-risk management tool, enabling farmers who have insured their beef cattle to receive compensation for losses in the event of beef-cattle mortality. It is evident that, theoretically, beef-cattle farmers who concurrently implement disease-prevention measures and adopt insurance can significantly mitigate actual economic losses, effectively achieve their risk management objectives, and consequently enhance their enthusiasm for beef-cattle farming. Nevertheless, the moral hazard in policy-oriented agricultural insurance is invariably complex to circumvent [34]. Therefore, both disease prevention and insurance adoption can be viewed as risk-management measures. Questions then arise: Is the relationship between the two “mutually stimulating”, “mutually substitutive”, or “unrelated”? Or is there any moral risk in the process of subscribing to policy-based beef-cattle insurance? Only through rigorous research can we establish the relationship between these two measures and thereby determine whether beef-cattle farmers have achieved their individual risk-management objectives. This research holds significant implications for enhancing beef-cattle insurance policies, refining the design of insurance products, and boosting farmers’ motivation for disease prevention.

Three main conclusions can be drawn from scholars’ studies on the interrelationship between farmers’ insurance-adoption behavior and risk-prevention input behavior. (I) Some scholars have posited that, if farmers choose to adopt insurance, they will change their risk-prevention input behavior under the effect of moral hazard, considering that they will receive a certain degree of financial compensation after the occurrence of risks. In one sense, moral hazard can manifest directly in reduced risk-prevention inputs. After empirical analysis, Lin and Wang [35] conclude that adopting poultry insurance negatively affects input for disease prevention. Similarly, when farmers choose to adopt crop insurance, they reduce inputs of chemical or organic fertilizers under moral hazard [36,37,38,39] and also reduce the use of pesticides [40,41]. This implies that there is a substitution relationship between agricultural insurance and farmers’ traditional ex ante risk-prevention inputs [42]. However, moral hazard can also be expressed in terms of increasing the “risk-increasing” factor input while decreasing the “risk-reducing” factor input [43,44]. For example, insured farmers may increase the use of antibiotics in pig disease treatment while reducing the use of non-antibiotic drugs [45], or they may increase the use of fertilizers in corn cultivation while reducing the use of organic fertilizers [46]. (II) Some other scholars have found that farmers perceive that being insured enhances their risk resilience and thus choose to increase the input of risk prevention to stabilize or even increase their output. Scholars have concluded that farmers significantly increase pesticide, fertilizer, and film inputs after adopting insurance [47,48,49], implying that the act of adopting insurance has an incentive effect on farmers’ input for risk-prevention behavior. (III) In addition to the above viewpoint, scholars have also concluded that no significant relationship exists between insurance-adoption behavior and other risk-prevention inputs and argued that farmers do not have moral hazard after adopting insurance. When studying the interrelationship between pig or dairy farmers’ insurance-adoption behavior and the input behavior of disease prevention, multiple researchers have found that farmers’ insurance-adoption behavior has no significant effect on their inputs of disinfectants, detergents, vaccines, disease check-ups, and treatments and have argued that no moral-hazard problem arises [50,51]. A substantial body of research examines the interrelationship between farmers’ insurance-adoption behavior and their input in risk-prevention measures. As large livestock, beef cattle exhibit distinct characteristics compared to crop cultivation and other forms of livestock farming in terms of insurance product design and farmers’ risk-management strategies. However, there is a relative paucity of research examining the interrelationship between insurance and other risk-management measures, specifically for beef cattle as the insured asset. Then, in risk management in beef-cattle farming in China, does a reciprocal interrelationship exist between the input for disease prevention and the act of adopting insurance? Are there any other significant factors exerting an influence? Furthermore, the existence of moral hazard among beef-cattle farmers following insurance uptake has not been adequately investigated in China. All these questions require further investigation to answer them.

1.3. Research Contributions and Article Structure

By comparing with the existing studies, we can say that the marginal contributions of this paper are as follows: First, this research bridges the gap in existing research concerning the relationship between adopting insurance and the input for disease prevention in China, which is among the world’s most significant beef-producing and importing nations. Secondly, this study examines the intrinsic relationship between risk-management measures and moral hazard from a risk-management perspective, thereby providing a theoretical foundation for policy formulation and product enhancement. Third, this paper takes pre-event preventive input, which has been less studied separately in previous research, as the object of study and further divides it into voluntary vaccination, deworming, and shed disinfection to explore heterogeneity. Fourth, considering the current situation of the international and Chinese beef-cattle markets, this study introduces the acquisition price of the beef-cattle market as a crucial influencing factor in the mechanism analysis and conducts empirical verification to further clarify the crucial influence of price fluctuations on the behavioral choices of beef-cattle farmers. Fifth, the ESR model is innovatively utilized in insurance-adoption and disease-prevention behavior research. The model considers the influence of both observable and unobservable factors on the dependent variable, and it uses full information maximum-likelihood estimation to better avoid the problem of missing valid information. Counterfactual analysis can also be carried out.

The rest of this paper is organized as follows: Section 2 offers a theoretical analysis and research hypotheses. Section 3 pertains to the study area, data, and methods, simultaneously listing the research variables and clarifying their implications. Section 4 offers an analysis of empirical results. Section 5 contains the discussion. Section 6 presents the conclusions of this study.

2. Theoretical Analysis and Research Hypotheses

Based on the theory of expected utility maximization, livestock farmers make decisions on disease-prevention behavior based on the subsidies provided by the government, the risk protection provided by insurance companies, and the compensation for losses. Farmers’ decision paths are shown in Figure 1.

2.1. Basic Model and Cost Functions

Drawing on Reeling and Horan [52] and Tian et al. [53], we set the disease-prevention input cost function as follows:

$$B(b)=\chi b^2+\alpha b.$$

(1)

In addition, we set the probability function of disease occurrence as follows:

$$p(b)=\xi (1+\mu b).$$

(2)

In Equations (1) and (2), $b$ represents the extent of disease-prevention input by livestock farmers, and $0\le b\le 1$. It can be readily shown that ${\displaystyle \frac{\partial B}{\partial b}}>0$ and ${\displaystyle \frac{{\partial }^{2}B}{\partial b^2}}>0$. Here, $b=0$ means that the livestock farmers have not made any input for disease prevention. At this point, $p(0)=\xi$, so $0\le \xi \mu \le 1$. $b=1$ is the maximum input of the livestock farmers in disease prevention. At this point, $p(1)=\xi +\xi \mu$, so ${\displaystyle \frac{\partial p}{\partial b}}=\xi \mu <0$.

The loss function of beef-cattle disease mortality faced by farmers can be derived using the probability function of disease occurrence and is formulated as follows:

$$L(b)=Mp(b)=M\xi (1+\mu b).$$

(3)

In Equation (3), $M$ represents the farmer’s market purchase price for beef cattle, which, as an exogenous variable, is not controlled by the farmer. In China’s beef-cattle market, market prices are determined by the aggregate supply and demand dynamics, and individual producers lack the ability to influence pricing. Consequently, for individual beef-cattle farmers, the market purchase price for beef cattle is an exogenous variable. $p(b)$ is a monotonically decreasing function of $b$, so $L(b)$ is also a monotonically decreasing function, which means that farmers can reduce beef-cattle deaths by increasing their input for disease prevention.

The utility function of the farmers, typically denoted by $u$ ($u^{\prime }>0$, and $u^{″}<0$), is subsequently calculated. It is also assumed that livestock farmers are risk-averse [54,55,56]. The livestock farmer’s utility-maximizing input decision for disease prevention depends on the lowest $L$ and the smallest $B$. That is, it satisfies the utility function

$$u_{\max }=\min [L(b)+B(b)].$$

(4)

The maximum-value problem of Equation (4) is solved at $b\ge 0$ and must satisfy the first-order necessary condition ${\displaystyle \frac{\partial L}{\partial b}}+{\displaystyle \frac{\partial B}{\partial b}}=0$ and ${\displaystyle \frac{\partial B}{\partial b}}=-{\displaystyle \frac{\partial L}{\partial b}}$. Since $L(b)$ is a monotonically decreasing function, ${\displaystyle \frac{\partial L}{\partial b}}<0$. It can therefore be transformed without affecting its economic significance into ${\displaystyle \frac{\partial B}{\partial b}}=\left|{\displaystyle \frac{\partial L}{\partial b}}\right|$. The equation collapses to

$${\displaystyle \frac{\partial (\chi b^2+\alpha b)}{\partial b}}={\displaystyle \frac{\partial [M\xi (1+\mu b)]}{\partial b}}.$$

(5)

Collating Equation (5) finally gives $b={\displaystyle \frac{M\xi \mu -\alpha }{2\chi }}$. This indicates that a higher $M$ leads to a higher $b$, implying that higher market purchase prices for beef cattle incentivize farmers to increase their input for disease prevention.

2.2. Incorporating Insurance

Next, we conduct a further analysis of the situation of farmers after adopting beef-cattle insurance. Usually, the higher the probability that beef cattle is infected with a disease, the more likely the farmer is to receive an insurance payout (Incorporating moral hazard into the assumptions of the insurance payout function may result in a systematic overestimation of the actual insurance payouts). Assume that the farmer receives the insurance payout function as

$$C_I(b)=C_{I}p=C_I\xi (1+\mu b).$$

(6)

In Equation (6), $C_I$ represents the insurance compensation amount received by the farmers. It shows more clearly the extent of the influence of the insurance-adoption behavior on the input behavior of beef-cattle disease prevention through mathematical and theoretical methods, so we assume that $C_I$ is a continuous variable. However, the compensation for beef-cattle insurance in the actual research site is fixed, so $C_I$ is regarded as a constant in the subsequent study, and $C_I=0$ when no insurance compensation occurs. At this point, the farmer faces a loss of ($L-C_I$) from beef-cattle disease deaths, and the cost of risk prevention to share the disease risk is ($B+C_0$), where $C_0$ is the cost of the insurance purchase (i.e., the premiums borne by the individual farmer).

The optimal decision for disease-prevention input maximization by livestock farmers also depends on the lowest $L$ and the smallest $B$, where the utility function is as follows:

$$U_{\max }=\min [L(b)-C_I(b)+B(b)+C_0].$$

(7)

Collating Equation (7) finally gives $b={\displaystyle \frac{(M-C_I)\xi \mu -\alpha }{2\chi }}$. At this point, the difference between $M$ minus $C_I$ is positively correlated with $b$. Other things being equal, $C_I$ is negatively correlated with $b$ if $M$ is also constant, and, similarly, $M$ is positively correlated with $b$ if $C_I$ is also constant.

2.3. Hypothesis Development

Further discussion is based on the premise that “$C_0$ is a silent cost and $p$ is unpredictable”. If the farmer predicts that beef cattle will not die from the disease and that, at this time, increasing the input for disease prevention will only increase the cost and not increase the sales revenue, then the farmer can reduce the farming cost by decreasing the input for disease prevention. If the farmer predicts the death of a beef cattle from a disease, at this time, the farmer can obtain $C_I$. Increasing disease-prevention input at this point is equivalent to giving up $C_I$ in exchange for a net gain from selling beef cattle, but the extent of the net gain cannot be determined [57]. Then, driven by the interest in obtaining insurance indemnity, livestock farmers may face moral hazard, compelling them to lower the input in disease prevention for the sake of obtaining $C_I$. The following research hypothesis is therefore proposed:

$H_1$: Farmers reduce their input for the prevention of beef-cattle diseases after insuring their cattle; that is, there exists moral hazard after insurance adoption.

If $M$ declines, the net proceeds received by farmers from the sale of beef cattle will be close to or even lower than the amount of their insurance compensation, at which time the utility-maximizing farmers will choose to reduce the input of disease prevention to achieve the purpose of cost savings or even to obtain the insurance compensation, which is in line with the above theoretical derivation results. It can be concluded that a lower $M$ will raise the real value of $C_I$, while too high an insurance amount will trigger moral hazard (i.e., farmers are more likely to reduce their input in beef-cattle disease prevention because they are insured) [58]. As such, the following research hypothesis is proposed:

$H_2$: A decline in the market purchase price of beef cattle enhances the negative effect of insurance-adoption behavior on input for beef-cattle disease prevention.

3. Study Area, Data, and Methods

3.1. Study Area

The research area of this paper is the IMAR of China, as shown in Figure 2. In terms of international standing, both the number of beef cattle reared and the output of beef in China occupy the third position globally [59]. Moreover, as one of the five major pastoral areas in China, the IMAR possesses a marked advantage in the grassland ecological environment and occupies a leading position in the beef-cattle production level [60]. By conducting a meticulous comparison of the total beef output, the quantity of beef cattle slaughtered, and the end-of-year beef-cattle inventory in the top-ranked regions of China from 2013 to 2022, we find that the national ranking and the national proportion of the total beef output, the number of cattle slaughtered, and the end-of-year cattle inventory in the IMAR consistently rise (Appendix A 

Table A3. The total output of beef in the top 10 regions of China from 2013 to 2022 (unit: million tons).

Year	2022	2021	2020	2019	2018	2017	2016	2015	2014	2013
China	718.3	697.5	672.4	667.3	644.1	634.6	716.8	700.1	689.2	673.2
Hebei	58.1	55.8	55.6	57.2	56.5	55.6	54.3	53.2	52.4	52.3
Inner Mongolia	71.9	68.7	66.3	63.8	61.4	59.5	55.6	52.9	54.5	51.8
Liaoning	32.3	31.5	31.0	29.6	27.5	25.1	41.6	40.3	42.8	43.2
Jilin	44.3	40.8	38.7	41.9	40.7	38.0	47.1	46.6	46.0	45.0
Heilongjiang	52.7	50.7	48.3	45.5	42.6	43.9	42.5	41.6	40.6	39.7
Shandong	60.4	61.3	59.7	73.3	76.4	75.9	67.0	67.9	66.6	67.9
Henan	36.7	35.5	36.7	36.2	34.8	35.0	83.0	82.6	82.1	80.6
Sichuan	38.6	36.9	37.0	36.4	34.5	33.3	36.9	35.4	33.4	31.1
Yunnan	43.6	42.0	40.9	39.0	36.0	35.8	35.2	34.3	33.6	31.8
Xinjiang	49.4	48.5	44.0	44.5	42.0	43.0	42.5	40.4	39.2	37.8
Proportion of Inner Mongolia (%)	10.0	9.8	9.8	9.6	9.5	9.4	7.8	7.6	7.9	7.7
Ranking of Inner Mongolia	1	1	1	2	2	2	3	4	3	4

Note: The data are sourced from the “China Rural Statistical Yearbook (2014–2023)”.

,

Table A4. The quantity of beef cattle slaughtered in the top 10 regions of China from 2013 to 2022 (unit: 10,000 heads).

Year	2022	2021	2020	2019	2018	2017	2016	2015	2014	2013
China	4839.9	4707.4	4565.5	4533.9	4397.5	4340.3	5110.0	5003.4	4929.2	4828.2
Hebei	353.2	339.9	335.2	349.1	345.6	340.5	331.9	325.4	320.6	325.3
Inner Mongolia	428.8	410.3	397.0	383.3	375.1	363.2	339.7	326.4	336.8	320.2
Liaoning	203.5	198.7	195.8	188.1	175.1	159.9	272.3	266.3	283.3	287.2
Jilin	262.3	242.4	238.7	258.7	249.6	233.6	306.4	303.2	299.6	297.0
Heilongjiang	311.4	299.7	289.4	281.0	270.2	281.5	274.3	269.7	263.6	256.2
Shandong	275.6	280.0	275.7	345.9	363.4	361.6	445.5	447.5	440.8	443.4
Henan	243.5	235.9	241.2	238.4	231.2	233.0	550.2	548.6	546.0	535.5
Sichuan	306.0	293.1	296.4	291.7	276.2	267.3	305.2	295.5	278.7	264.7
Yunnan	360.1	345.2	335.9	326.4	309.1	307.8	300.4	292.8	287.3	275.7
Xinjiang	292.6	289.2	266.3	270.9	253.5	259.3	258.1	247.3	239.4	230.3
Proportion of Inner Mongolia (%)	8.9	8.7	8.7	8.5	8.5	8.4	6.6	6.5	6.8	6.6
Ranking of Inner Mongolia	1	1	1	1	1	1	3	3	3	4

Note: The data are sourced from the “China Rural Statistical Yearbook (2014–2023)”.

and

Table A5. The end-of-year beef-cattle inventory in the top-ranked regions of China from 2013 to 2022 (unit: 10,000 heads).

Year	2022	2021	2020	2019	2018	2017	2016	2015	2014	2013
China	8454.1	8004.4	7685.1	6998.0	6618.4	6617.9	7441.0	7372.9	7040.9	6838.6
Inner Mongolia	658.9	585.9	538.3	499.8	489.8	526.5	444.8	423.2	388.3	369.9
Jilin	377.8	323.6	270.5	316.2	309.4	322.0	400.4	420.8	401.8	408.6
Henan	322.8	289.3	270.0	257.3	231.1	230.5	620.8	650.4	626.6	610.1
Sichuan	583.6	524.8	547.8	502.3	476.2	494.8	552.8	561.8	529.4	487.9
Yunnan	835.0	824.5	810.4	777.5	755.8	747.7	721.8	688.2	681.3	658.9
Tibet	555.5	544.7	531.3	525.2	498.4	470.0	466.6	471.3	467.5	467.5
Gansu	495.5	480.1	450.3	427.1	410.5	394.1	416.4	420.1	423.8	402.1
Xinjiang	632.3	628.7	634.8	475.5	4920.	521.9	457.9	429.6	427.1	423.6
Proportion of Inner Mongolia (%)	7.8	7.3	7.0	7.1	7.4	8.0	6.0	5.7	5.5	5.4
Ranking of Inner Mongolia	2	3	4	4	4	2	6	6	8	8

Note: The data are sourced from the “China Rural Statistical Yearbook (2014–2023)”.

). Furthermore, given the statistics, among the 12 cities of the IMAR, the top two in terms of the year-end inventory of live cattle, the quantity of beef cattle slaughtered, and the total output of beef are Tongliao City and Chifeng City (Appendix A 

Table A6. Beef-cattle farming in various municipalities of the IMAR, China.

League (or City)	Beef-Cattle Output (10,000 Head/Year)	Total Beef Production (Tons/Year)	Year-End Stock of Beef Cattle (10,000 Head/Year)
Tongliao City	105.09 (Ranked first)	178,669 (Ranked first)	237.2 (Ranked first)
Chifeng City	80.37 (Ranked second)	132,886 (Ranked second)	139.01 (Ranked second)
Xilingol League	71.3	119,909	121.03
Hulunbuir City	60.26	94,590	103.35
Hinggan League	29.44	51,556	76.32
Baotou City	21.25	37,139	15.69
Hohhot City	17.59	31,009	38.57
Ordos City	15.64	26,035	33.37
Ulanqab City	13.06	22,003	23.63
Bayannur City	11.84	20,039	25.6
Alxa League	2.74	4282	5.67
Wuhai City	0.23	589	0.91
Total	428.81	718,706	820.35
Proportion of Tongliao City (%)	0.25	0.25	0.29
Proportion of Chifeng City (%)	0.19	0.18	0.17

Note: The data originate from the Inner Mongolia Statistical Yearbook 2023.

). Consequently, the selection of this surveyed area for the pertinent research on beef cattle is typical and representative, fulfilling the analytical demands of this paper.

3.2. Data Collection

The data used in this study come from the subject group’s July–August 2023 and January 2024 national project surveys. The survey locations include eight flag counties in Tongliao City and Chifeng City, with the sampling principle of “two townships per flag county, two villages per township, and 10–20 farming households per village”. A total of 484 questionnaires were collected in the survey. The outliers were removed (this paper determined outliers based on the farming scale and eliminated individual samples whose farming scale is approximately 20 times the average), and, finally, 447 valid samples were selected for this study, resulting in a questionnaire adoption rate of 92.36%. The survey design’s primary principle is collecting critical information, including objective indicators and subjective perceptions, from farmers. Objective indicators encompass the personal characteristics and the farming characteristics of farmers, disease prevention measures, insurance-adoption status, etc., for instance, age, farming year, farming quantity, vaccination coverage rate, etc. Subjective perceptions encompass risk perception, risk disposition, insurance awareness, etc. To ensure the effective transmission of information, members of the investigation team conducted one-on-one interviews to collect questionnaire data.

3.3. Econometric Methods

The following equations are constructed for the impact of farmers’ insurance-adoption behavior on input for beef-cattle disease prevention:

$$\ln Epi{P}_i=\beta X_i+\delta IN{S}_i+{\mu }_i.$$

(8)

In Equation (8), $\ln Epi{P}_i$ denotes the log-treated disease-prevention input, $X_i$ denotes control variables such as personal characteristics and farming characteristics, $IN{S}_i$ is a binary choice variable for whether farmer $i$ takes out beef-cattle insurance, $\beta$ and $\delta$ are the parameters to be estimated, and ${\mu }_i$ is the random interference term.

Farmers’ insurance-adoption behavior and disease-prevention input behavior can be affected by a complex mix of observable and unobservable variables, giving rise to self-selection and endogeneity problems in the sample. Therefore, in this paper, the ESR model has a strong advantage in solving the above problems [61]. By employing this method, we aim to determine whether the effect of farmers’ insurance-adoption behavior on input for beef-cattle disease prevention is causal and whether it can produce an unbiased estimate of the impact effect. The model is estimated to have two stages, as follows.

The first stage estimates the factors affecting whether farmers adopt beef-cattle insurance and constructs the decision equation as follows:

$$IN{S}_i=\alpha Z_i+\gamma Co{g}_i+{\epsilon }_i,IN{S}_i=\left\{\begin{array}{cc}1\hfill & IN{S}_i>0\hfill \\ 0\hfill & IN{S}_i\le 0\hfill \end{array}\right..$$

(9)

In Equation (9), $IN{S}_i=1$ indicates that farmer $i$ has taken out beef-cattle insurance, while $IN{S}_i=0$ means that farmer $i$ has not taken out such insurance and $Z_i$ is the types of factors that influence whether a farmer is insured. To ensure that the model is identifiable, at least one of the variables in the decision equation must not appear in the influence effect equation. Therefore, insurance perception (i.e., $Co{g}_i$) is included as an identifying variable in the decision equation. $\alpha$ and $\gamma$ are the parameters to be estimated, and ${\epsilon }_i$ is the random interference term.

In the second stage, the effects of farmers’ insured and uninsured behaviors on beef-cattle disease-prevention input were estimated separately, and the impact effect equations are constructed as follows:

$$\ln Epi{P}_{i1}={\beta }_1{X}_{i1}+{\mu }_{i1},\mathrm{if}IN{S}_i=1.$$

(10)

$$\ln Epi{P}_{i0}={\beta }_0{X}_{i0}+{\mu }_{i0},\mathrm{if}IN{S}_i=0.$$

(11)

In Equations (10) and (11), $\ln Epi{P}_{i1}$ and $\ln Epi{P}_{i0}$ denote the disease-prevention input for the insured and uninsured groups, respectively; $X_{i1}$ and $X_{i0}$ are control variables. When unobservable factors affect both $IN{S}_i$ and $\ln Epi{P}_i$, a correlation exists between the residuals of the decision equation and the influence effect equation, so the inverse Mills ratio calculated based on the decision equation for insured behavior was introduced into the influence effect equation to obtain

$$\ln Epi{P}_{i1}={\beta }_1{X}_{i1}+{\sigma }_{\mu 1}{\lambda }_{i1}+{\mu }_{i1},\mathrm{if}IN{S}_i=1.$$

(12)

$$\ln Epi{P}_{i0}={\beta }_0{X}_{i0}+{\sigma }_{\mu 0}{\lambda }_{i0}+{\mu }_{i0},\mathrm{if}IN{S}_i=0.$$

(13)

In Equations (12) and (13), ${\lambda }_{i1}$ and ${\lambda }_{i0}$ represent the choice of insurance coverage due to unobservable variables. ${\sigma }_{\mu 1}$ and ${\sigma }_{\mu 0}$ denote the covariance of the error terms of the decision equation and the impact effect equation, which, if statistically significant, indicate a need for a solution to the problem of “simultaneous decision-making” and “self-selection” [62]. The estimates obtained at this point will be unbiased and consistent.

After estimating the correlation coefficients from the ESR model, we can also calculate the two sets of average treatment effects of insurance-adoption behavior on the input of disease-prevention factors, including the average treatment effect for the treated group (hereafter referred to as ATT) and the average treatment effect for the untreated group (hereafter referred to as ATU). However, the ATU includes the effect of samples that are not affected by being insured, and their results are of little significance [63,64]. The most crucial parameter to estimate is the ATT [65], so, in this paper, only the ATT was estimated to measure the effect of farmers’ insurance-adoption behavior on the input of beef-cattle disease prevention.

The conditional expectation for beef-cattle disease-prevention input in the insured group is as follows:

$$E[\ln Epi{P}_{i1}|IN{S}_i=1]={\beta }_1{X}_{i1}+{\sigma }_{\mu 1}{\lambda }_{i1}.$$

(14)

If the insured group does not adopt insurance, the conditional expectation of its beef-cattle disease-prevention input is as follows:

$$E[\ln Epi{P}_{i0}|IN{S}_i=1]={\beta }_0{X}_{i1}+{\sigma }_{\mu 0}{\lambda }_{i1}.$$

(15)

Then, the actual insured group is regarded as the treatment group, and the difference between Equations (14) and (15) is taken as the average treatment effect of the treatment group’s input in disease prevention, denoted as follows:

$$ATT=E[\ln Epi{P}_{i1}|IN{S}_i=1]-E[\ln Epi{P}_{i0}|IN{S}_i=1]=X_{i1}({\beta }_1-{\beta }_0)+{\lambda }_{i1}({\sigma }_{\mu 1}-{\sigma }_{\mu 0})$$

(16)

3.4. Variable Design

The dependent variable is disease-prevention input. Comprehensively considering multiple factors such as cost and technological level, we established the main disease-prevention measures commonly employed in beef-cattle farming in China at the present stage as follows: The first is voluntary vaccination. In China, the predominant policy implemented for beef cattle is compulsory immunization against specific diseases for prevention purposes, which is offered free of charge to farmers by the government. Therefore, this paper investigates the inputs made by beef-cattle farmers in voluntary vaccination beyond compulsory immunization, encompassing the inputs in domestic or imported vaccines that are voluntarily administered. The second is deworming. Beef-cattle parasites, both internal and external to the animal, are eliminated via subcutaneous injection and medicinal bath cleansing. This constitutes a disease-prevention modality that demands the contemplation of factors such as seasonality and the farming environment and is voluntarily implemented and self-financed by beef-cattle farmers. The third is shed disinfection, which refers to the expenses borne by beef-cattle farmers for cleaning the environment of the cattle shed (i.e., cleaning and disinfecting the cattle shed and its surrounding environment with caustic soda, quicklime, and bleaching powder). Accordingly, the disease-prevention inputs presented in this paper encompass the sum of the expenditures for voluntary vaccination, deworming, and shed disinfection, with a logarithmic treatment.

The treatment variable is the insurance-adoption behavior of beef-cattle farmers. Measured by whether the farmer is insured for beef cattle, we set the variable to 1 when insured for beef cattle and 0 when uninsured.

For the control variables, the following selections were made: Past research has indicated that personal and farming characteristics influence insured behavior and other risk-prevention behaviors [35,48,66]. Disease prevention, among the green livestock-farming technologies, is vulnerable to the impact of other green livestock-farming technologies (The green breeding technologies defined in this paper include breed improvement technology, seeding and breeding combination technology, composting technology, substrate technology and harmless treatment technology, among others [67,68]). Consequently, we assessed this variable by the quantity of green livestock-farming technologies adopted. Risk characteristics such as the beef-cattle mortality rate, risk perception, and risk disposition influence farmers’ insurance-adoption behavior and their input in disease prevention [45,69]. Furthermore, they include the characteristics of disease prevention, summarized based on the theory of agricultural risk management and variables such as the fluctuations in the purchase price of beef cattle in the market (In China, three principal pathways exist for livestock farmers to sell beef cattle: selling to third parties who come for acquisition at the doorstep, selling in the live animal trading market, and selling directly to slaughterhouses. In terms of distance, farmers usually choose to sell locally. The first two are categorized as Category I, both involving a third party; that is, the third party acquires live cattle and then uniformly channels them into the next-level market for sale. Moreover, the selling price of this method is calculated based on each cattle and is negotiable. The third one is categorized as Category II. The livestock farmers, acting as the producers, sell directly to slaughterhouses. The selling price is calculated per kilogram, and slaughterhouses stipulate a unified purchase price that is nonnegotiable. Given the actual circumstances of the survey area, the “price fluctuations” investigated in this paper pertains to Category I.), as deduced from the previous theoretical exposition.

The identifying variable is insurance awareness. To address the problem of endogeneity bias due to possible omitted variables and bidirectional causality, we used farmers’ knowledge of beef-cattle insurance policies and terms as the identifying variable. Usually, the better one understands the policies related to agricultural insurance, the more one is aware of its positive effects, and the more one is inclined to purchase agricultural insurance [70]. Of the insured farmers in the sample, 66.7% choose “very much understanding” and “more understanding”, while 18.6% choose “less understanding” and “very little understanding”. That is, in the samples with a higher degree of understanding, the proportion of insured farmers is greater, demonstrating that the insurance awareness of farmers is closely related to whether they choose to adopt insurance. By contrast, insurance awareness does not directly affect farmers’ disease-prevention input (i.e., it is exogenous to the dependent variable).

Accordingly, this study identified the dependent variable, treatment variable, and identifying variable and selected 16 control variables, which were classified into six characteristic types. The specific variable definitions and assignments are shown in

Table 1. Variable definitions.

Variable		Definition
Dependent variable	Disease-prevention input	The sum of the costs for voluntary vaccination, deworming, and shed disinfection (yuan)
Treatment variable	Insurance-adoption behavior	Whether to adopt beef-cattle insurance: Yes = 1; No = 0
Control variables
Personal characteristics	Sex	Sex of respondents: Male = 1; Female = 0
	Age	Actual age of respondents (years)
	Educational level	Respondents’ actual years of education (years)
Farming characteristics	Farming year	Respondents’ actual years of farming (years)
	Farming scale	Maximum farming quantity of this year (head)
	Farm size	Actual size of the beef-cattle farm (square meters)
	Net income from farming	Net income of farming in the current year after deducting the cost (years)
	Green farming technology	Number of adoptions of green farming technology (pieces)
Risk characteristics	Mortality rate	Proportion of the number of beef-cattle deaths to the maximum farming quantity in the current year (%)
	Risk perception	Whether due attention is given to the potential risks such as epidemic diseases and natural disasters that may arise during the farming process: Yes = 1; No = 0
	Risk disposition	Extent of risk appetite: Risk appetite = 1; Risk neutrality = 2; Risk avoidance = 3
Disease-prevention characteristics	Disease cognitive skill	Number of diseases to which beef cattle are usually prone that the respondents have already known (pieces)
	Unit frequency of disease prevention	Ratio of the total frequency of disease prevention this year to the maximum quantity of farming (times)
	Distance	Whether the distance from the livestock sheds to the local animal husbandry and veterinary station exceeds the average distance from the village where the farmer is located to the station: Yes = 1; No = 0
External characteristic	Price fluctuations	Alteration in the acquisition price of the beef-cattle market in the present year when compared with that of the previous year: Lower = 1; Not lowered = 0
Regional characteristic	Regional variable	Tongliao city = 1; Chifeng city = 0
Identifying variable	Insurance awareness	Respondents’ knowledge of beef-cattle insurance policies and provisions: very little understanding = 1; less understanding = 2; fair = 3; more understanding = 4; very much understanding = 5

.

The descriptive statistics of the sample are presented in

Table 2. Descriptive statistics of the sample.

Variable	Full Sample	Insured Group n = 195	Uninsured Group n = 252
Disease-prevention input a	1294.899 (1772.630)	1379.708 (1905.035)	1229.274 (1663.870)
Insurance-adoption behavior	0.436 (0.496)	1.000 (0.000)	0.000 (0.000)
Sex	0.975 (0.155)	0.985 (0.123)	0.968 (0.176)
Age	48.767 (9.574)	48.041 (9.388)	49.329 (9.697)
Educational level	8.389 (2.807)	8.672 (2.740)	8.171 (2.844)
Farming year	26.539 (11.647)	25.697 (11.085)	27.190 (12.046)
Farming scale	41.354 (41.051)	39.275 (32.481)	42.963 (46.616)
Farm size	50.855 (51.447)	43.369 (42.287)	56.647 (56.953)
Net income from farming b	47,291.960 (208,654.900)	35,439.65 (176,482.600)	56,463.380 (230,420.600)
Green farming technology	3.192 (0.763)	3.441 (0.674)	3.000 (0.773)
Mortality rate	3.070 (5.941)	3.605 (6.325)	2.657 (5.605)
Risk perception	0.955 (0.207)	0.969 (0.173)	0.944 (0.230)
Risk disposition	2.228 (0.902)	2.200 (0.906)	2.250 (0.900)
Disease cognitive skill	6.197 (3.275)	6.256 (3.244)	6.151 (3.304)
Unit frequency of disease prevention	0.510 (0.528)	0.549 (0.584)	0.479 (0.479)
Distance	0.351 (0.478)	0.385 (0.488)	0.325 (0.469)
Price fluctuations	0.512 (0.500)	0.554 (0.498)	0.480 (0.501)
Regional variable	0.582 (0.494)	0.723 (0.449)	0.472 (0.500)
Insurance awareness	2.960 (1.342)	3.641 (1.241)	2.432 (1.170)

Note: ^a and ^b are absolute values in the descriptive statistics, and both use logarithmic values in the model fitting below. Standard errors in parentheses.

. From the sample composition, out of 447 households, 195 adopted insurance, representing 43.62% of the total sample. This suggests that, while the coverage rate for beef-cattle insurance is relatively high, there remains potential for further improvement. From the perspective of personal characteristics, most insured farmers are male, exhibit a relatively younger age profile, and possess a higher level of education. From the perspective of farming characteristics, insured farmers exhibit shorter farming years, smaller farming scales, reduced farm sizes, and lower net incomes than uninsured farmers. However, they demonstrate a higher adoption rate of green farming technologies. From the perspective of risk characteristics, insured farmers experience a higher beef-cattle mortality rate and exhibit stronger risk perception yet are more inclined to engage in risk-taking behavior. From the perspective of disease-prevention characteristics, insured farmers demonstrate higher disease cognitive skills and a greater frequency of disease-prevention measures per unit. The distance between their livestock sheds and the local animal husbandry and veterinary station is significantly greater. Furthermore, a higher proportion of insured farmers are located in Tongliao City and experience more frequent price declines.

4. Results

This paper performs empirical analysis utilizing Stata 17 software. Before model fitting, the correlations between the dependent variable and other variables have been examined to obtain an initial understanding of the relationships among these variables (Appendix B,

Table A7. Correlation Test.

Variable	Disease-Prevention Input
Insurance-adoption behavior	0100 **
Sex	0.152 ***
Age	−0.152 ***
Educational level	0.116 **
Farming year	−0.085 *
Farming scale	0.259 ***
Farm size	0.247 ***
Net income from farming	0.230 ***
Green farming technology	0.080 *
Mortality rate	0.098 **
Risk perception	0.109 **
Risk disposition	0.053
Disease cognitive skill	0.135 ***
Unit frequency of disease prevention	−0.088 *
Distance	0.103 **
Price fluctuations	−0.127 ***
Regional variable	0.145 ***

Note: * p < 10%, ** p < 5%, *** p < 1%.

). Simultaneously, it is imperative to examine potential multicollinearity issues to prevent estimation inaccuracies that may arise from high correlations among the independent variables. This study employs the variance inflation factor (VIF) to assess the extent of multicollinearity within the model. The results of the decision equation and the impact effect equation indicate that the VIF values are 1.61 and 1.58, respectively, both well below the threshold of 5 (Appendix B,

Table A8. The multicollinearity test for decision equation.

Variable	VIF	1/VIF
Sex	1.05	0.95
Age	2.22	0.45
Educational level	1.14	0.88
Farming year	2.12	0.47
Farming scale	4.15	0.24
Farm size	3.44	0.29
Net income from farming	1.34	0.74
Green farming technology	1.12	0.89
Mortality rate	1.05	0.95
Risk perception	1.06	0.94
Risk disposition	1.04	0.96
Disease cognitive skill	1.12	0.89
Unit frequency of disease prevention	1.33	0.75
Price fluctuations	1.08	0.93
Regional variable	1.31	0.76
Insurance awareness	1.19	0.84
Mean VIF	1.61	−

and

Table A9. The multicollinearity test for the impact effect equation.

Variable	VIF	1/VIF
Insurance	1.26	0.80
Sex	1.05	0.95
Age	2.22	0.45
Educational level	1.14	0.88
Farming year	2.12	0.47
Farming scale	4.06	0.25
Farm size	3.44	0.29
Net income from farming	1.38	0.73
Green farming technology	1.18	0.85
Mortality rate	1.07	0.94
Risk perception	1.06	0.94
Risk disposition	1.04	0.96
Disease cognitive skill	1.11	0.90
Unit frequency of disease prevention	1.33	0.75
Distance	1.05	0.95
Price fluctuations	1.09	0.92
Regional variable	1.29	0.77
Mean VIF	1.58	−

). These findings indicate the absence of multicollinearity among the variables. Based on these findings, this study will proceed with ESR-model estimation and treatment effect estimation, and robustness checks will be conducted on the resulting estimates. This study will further investigate the impact of price fluctuations on the estimation results of the main model and conduct heterogeneity analysis based on different types of disease prevention inputs and varying resource endowments among farmers.

4.1. ESR-Model Estimation Results

The estimated results of the ESR model for the impact of farmers’ insurance-adoption behavior on disease-prevention input are shown in

Table 3. ESR model estimation results.

Variable	Decision Equations (Insured Behavior)	Impact Effect Equation (Disease-Prevention Input)
		Insured Group	Uninsured Group
Sex	0.824 ** (0.392)	0.434 (0.483)	0.872 (0.538)
Age	0.015 (0.010)	−0.007 (0.010)	−0.008 (0.014)
Educational level	0.044 * (0.024)	0.008 (0.023)	0.064 * (0.034)
Farming year	−0.009 (0.008)	−0.004 (0.008)	−0.012 (0.011)
Farming scale	−0.007 ** (0.003)	0.009 *** (0.004)	0.007 * (0.004)
Farm size	−0.003 (0.002)	0.005 ** (0.003)	−0.005 (0.003)
Net income from farming	0.271 *** (0.051)	0.116 * (0.062)	0.220 *** (0.066)
Green farming technology	0.470 *** (0.089)	−0.079 (0.115)	0.421 *** (0.124)
Mortality rate	3.690 *** (1.090)	1.520 (1.051)	4.506 *** (1.616)
Risk perception	0.405 (0.306)	0.219 (0.351)	0.850 ** (0.416)
Risk disposition	−0.024 (0.068)	−0.015 (0.066)	0.146 (0.100)
Disease cognitive skill	0.004 (0.020)	−0.002 (0.019)	0.070 ** (0.028)
Unit frequency of disease prevention	0.092 (0.148)	0.150 (0.119)	0.147 (0.210)
Distance	−	0.172 (0.124)	−0.016 (0.149)
Price fluctuations	−0.038 (0.127)	−0.112 (0.122)	−0.215 (0.185)
Regional variable	0.590 *** (0.141)	0.166 (0.170)	0.998 *** (0.201)
Insurance awareness	0.313 *** (0.048)	−	−
Constant	−7.552 *** (0.971)	4.613 *** (1.403)	0.898 (1.176)
Observations	447	195	252
$\ln {\sigma }_{i1}$		−0.221 *** (0.052)
${\rho }_{i1}$		0.064 (0.312)
$\ln {\sigma }_{i0}$			0.475 *** (0.055)
${\rho }_{i0}$			2.265 *** (0.424)
Wald test	99.480 ***
Log likelihood	−844.953
LR test	66.420 ***

Note: * p < 10%, ** p < 5%, *** p < 1%. Standard errors in parentheses.

. The Wald test for modeling goodness-of-fit and the likelihood ratio (LR) test for the independence of the two-stage equations both reject the original hypothesis at the 1% level. Both $\ln {\sigma }_{i1}$ and $\ln {\sigma }_{i0}$ are significantly non-zero at the 1% level. This suggests that unobservable factors affect both farmers’ insurance-adoption decisions and the input for disease prevention. Therefore, it is necessary to construct an ESR model to correct for selectivity bias to avoid biased coefficient estimates.

Column 2 of Table 3 reports the estimation results of the various factors influencing the insurance-adoption behavior of farmers.

(I) Among the personal characteristics, the sex and educational level of the farmers are positive at the 5% and 10% significance levels, respectively. This reveals that male farmers or those with higher levels of education have a higher probability of choosing to be insured [71,72]. (II) Among the farming characteristics, the farming scale is negative at a significance level of 5%, suggesting that the larger the scale of the beef-cattle farm, the lower the probability that the farmers will adopt insurance. The net income from farming is positive at the 1% level of significance, indicating that the higher the net income, the higher the probability that the farmers will adopt insurance [73]. Green farming technology is positive at a significance level of 1%, suggesting that the higher the number of green technology technologies adopted by farmers, the higher the probability of their choosing to adopt insurance. (III) Among the risk characteristics, the mortality rate is positive at a significance level of 1%, suggesting that the higher the mortality rate of beef cattle, the greater the probability that farmers will opt for insurance. (IV) The regional variable is positive at a significance level of 1%, suggesting that Tongliao City, as a core beef-cattle farming area and a nationally representative pilot site for policy-based beef-cattle insurance, has a more complete insurance function, and farmers have a higher probability of choosing to adopt insurance. (V) Insurance awareness, as an identifying variable, is positive at a significance level of 1%; that is, the more farmers understand the policies and terms of beef-cattle insurance, the higher the probability of choosing to be insured [71,72,74,75,76].

Columns 3 and 4 of Table 3 present the estimated results of the influence effect of disease-prevention input for farmers in the insured group and the uninsured group, respectively.

(I) Among the personal characteristics, educational level positively affects the uninsured group at the 10% level of significance but does not have a significant effect on the insured group. (II) Among the farming characteristics, the farming scale exerts a positive influence on the insured group and the uninsured group at significance levels of 1% and 10%, respectively. Farm size positively affects the insured group at the 5% significance level but has no significant effect on the uninsured group. Net income from farming exerts a positive influence on the insured group and the uninsured group at the significance levels of 10% and 1%, respectively. Green farming technology positively affects the uninsured group at the 1% significance level but has no significant effect on the insured group. (III) Among the risk characteristics, mortality rate and risk perception positively influence the uninsured group at the significance levels of 1% and 5%, respectively, but have no significant effect on the insured group. (IV) Among the disease-prevention characteristics, disease cognitive skill positively influences the uninsured group at a significance level of 5%, but the influence on the insured group is not significant. (V) Regional variable positively influences the uninsured group at a 1% significance level, as Tongliao City has implemented policies such as beef-cattle breed improvement more effectively, leading to higher farming costs compared with other areas. Consequently, uninsured farmers will increase their input in disease prevention.

4.2. Results of Treatment Effect Estimates

By employing the ESR model to control for selection bias, the treatment effect can be estimated with greater accuracy. Estimates of the mean treatment effect of farmers’ insured behavior affecting input for beef-cattle disease prevention are shown in

Table 4. Estimated treatment effects of farmers’ insurance-adoption behavior affecting prevention input of beef-cattle diseases.

	Insured Beef Cattle	Uninsured Beef Cattle (Counterfactual)	ATT	t
Average treatment effect	6.682 (0.045)	8.917 (0.048)	−2.235 *** (0.066)	−33.973

Note: *** p < 1%. Standard errors in parentheses.

, and the mean treatment effect for insured farmers was negative at the 1% level of significance. Under the counterfactual assumption that insured farmers do not adopt insurance, the input of beef-cattle disease prevention increases by 2.235, or a rate of 33.45% (2.235/6.682). It can be seen that farmers choosing to be insured significantly reduce the input of beef-cattle disease prevention, while the policyholder reduces the preventive measures of the insured subject matter after insuring, which is a typical manifestation of moral hazard [24]. Thus, hypothesis $H_1$ is verified.

4.3. Robustness Check

This study employs robustness checks by replacing both the dependent variable and the estimation methods to verify the reliability of the analysis results.

Robustness check 1 is the replacement of the dependent variable. We refer to the study conducted by Li and Ma [77], replacing the dependent variable with a unit of beef-cattle disease-prevention input (the estimated results of the ESR model based on unit disease-prevention input are presented in

Table A10. ESR model estimation results for the replacement dependent variable.

Variable	Decision Equations (Insured Behavior)	Impact Effect Equation (Disease-Prevention Input)
		Insured Group	Uninsured Group
Sex	0.661 (0.490)	0.260 (0.450)	0.409 (0.411)
Age	0.012 (0.010)	−0.003 (0.009)	−0.005 (0.011)
Educational level	0.052 ** (0.026)	0.003 (0.021)	0.028 (0.026)
Farming year	−0.008 (0.008)	−0.003 (0.007)	−0.004 (0.008)
Farming scale	−0.004 (0.004)	−0.008 ** (0.003)	0.002 (0.003)
Farm size	−0.004 * (0.003)	0.006 ** (0.002)	−0.001 (0.002)
Net income from farming	0.247 *** (0.059)	0.034 (0.053)	0.040 (0.051)
Green farming technology	0.492 *** (0.097)	−0.086 (0.100)	0.186 * (0.099)
Mortality rate	3.517 *** (1.180)	1.942 ** (0.943)	2.508 ** (1.264)
Risk perception	0.309 (0.342)	0.186 (0.330)	0.422 (0.318)
Risk disposition	−0.060 (0.076)	−0.044 (0.062)	0.024 (0.078)
Disease cognitive skill	−0.019 (0.022)	−0.005 (0.018)	0.048 ** (0.022)
Unit frequency of disease prevention	0.121 (0.152)	0.480 *** (0.113)	0.891 *** (0.164)
Distance	−	0.198 * (0.117)	−0.039 (0.139)
Price fluctuations	0.007 (0.140)	−0.030 (0.116)	−0.128 (0.145)
Regional variable	0.584 *** (0.158)	0.129 (0.151)	0.538 *** (0.167)
Insurance awareness	0.432 *** (0.057)	−	−
Constant	−7.302 *** (1.091)	2.743 ** (1.097)	0.903 (0.907)
Observations	447	195	252
$\ln {\sigma }_{i1}$		−0.279 *** (0.051)
${\rho }_{i1}$		−0.044 (0.230)
$\ln {\sigma }_{i0}$			0.138 ** (0.064)
${\rho }_{i0}$			0.883 *** (0.218)
Wald test	68.100 ***
Log likelihood	−803.538
LR test	11.870 ***

Note: * p < 10%, ** p < 5%, *** p < 1%. Standard errors in parentheses.

of Appendix B). Both the Wald test and the LR test reject the original hypothesis at the 1% level, indicating the need to correct for sample selection bias caused by unobservable variables. The estimated results are shown in

Table 5. Robustness test results for the replacement dependent variable.

	Insured Beef Cattle	Uninsured Beef Cattle (Counterfactual)	ATT	t
Average treatment effect	3.317 (0.030)	4.453 (0.041)	−1.137 *** (0.050)	−22.581

Note: *** p < 1%. Standard errors in parentheses.

, and the ATT for the insured farmers is negative at the 1% significance level. That is, under the counterfactual assumption that insured farmers do not adopt insurance, their input in disease prevention increases by 1.137, or a rate of 34.28% (1.137/3.317).

Robustness check 2 is the replacement of the estimation methods. We use the treatment effect model (TEM), which addresses estimation bias due to the self-selection problem [78], and joint-equation maximum likelihood estimation (MLE), which addresses the endogeneity problem due to associative bias [46]. Columns 2 and 3 of

Table 6. Estimated results of the replacement of the estimation methods.

	Two-Stage TEM		Joint-Equation MLE
	Input Equations for Disease Prevention	Insured Behavior Equation	Input Equations for Disease Prevention	Insured Behavior Equation
Insurance-adoption behavior	−0.605 ** (0.309)	−	−1.898 *** (0.153)	−
Control variable	Controlled	Controlled	Controlled	Controlled
Insurance awareness	−	0.446 *** (0.058)	−	0.197 *** (0.038)
Constant	2.224 *** (0.822)	−6.583 *** (1.107)	0.912 (0.936)	−6.866 *** (0.941)
Wald test	172.520 ***		230.360 ***
Log likelihood	−221.748		−875.478
LR test	168.890 ***		82.010 ***

Note: ** p < 5%, *** p < 1%. Standard errors in parentheses.

report the results of the two-stage estimation of the TEM, and columns 4 and 5 report the results of the MLE of the joint equations. The Wald test and LR test for both estimation methods reject the original hypothesis at the 1% level, further proving that endogeneity exists concerning whether farmers adopt beef-cattle insurance. The regression coefficient obtained from the TEM estimation of the effect of insured behavior on disease-prevention input in beef cattle is −0.605, which is significant at the 5% level, and the regression coefficient obtained from the MLE is −1.898, which is significant at the 1% level, both of which indicate that farmers’ adopting insurance for beef cattle significantly reduces disease-prevention input. These results are consistent with the estimation results of the ESR model, validating the robustness of the previous estimation results.

4.4. Further Analysis: Purchase Price Movements in the Beef-Cattle Market

According to the theoretical analysis in “Section 2”, the beef-cattle market purchase price is a crucial exogenous variable influencing the decision making of farmers on input for beef-cattle disease prevention. When this price changes, to test whether it changes the extent to which farmers’ insurance-adoption behavior affects input for beef-cattle disease prevention, we conduct a separate treatment-effect estimation for the sample where “price fluctuations = 1” and compare it with the treatment effect estimates for the full sample comparison.

As shown in row 2 of

Table 7. Further impacts of lower purchase prices in the beef-cattle market.

	Sample Type	Sample Size	Insured Beef Cattle	Uninsured Beef Cattle (Counterfactual)	ATT	t
Average treatment effect	Sample of “price fluctuations = 1”	108	6.556 (0.067)	8.905 (0.075)	−2.349 *** (0.101)	−23.344
	Full sample	195	6.682 (0.045)	8.917 (0.048)	−2.235 *** (0.066)	−33.973

Note: *** p < 1%. Standard errors in parentheses.

, the ATT for insured farmers is negative at the 1% level of significance in the case of a lower market purchase price for beef cattle compared with the previous year. In the counterfactual assumption of not taking insurance, the input in beef-cattle disease prevention by the insured group increases by 2.349, or a rate of 35.83% (2.349/6.556), which is higher than the increase rate of all sampled farmers (33.45%). This suggests that, when a decline in the market purchase price of beef cattle occurs, insured farmers reduce their input in beef-cattle disease prevention to a greater extent. In other words, the descent in the market purchase price of beef cattle has augmented the negative impact of farmers’ insurance-adoption behavior on their disease-prevention input, thus verifying hypothesis $H_2$.

4.5. Heterogeneity Analysis

The first heterogeneity analysis focuses on different types of disease-prevention inputs. As previously stated, the main approaches that farmers in the surveyed area use to conduct disease prevention for beef cattle include voluntary vaccination, deworming, and shed disinfection. As shown in

Table 8. Differences in the impact of farmers’ insurance-adoption behavior on different disease-prevention inputs.

Different Inputs for Disease Prevention	Insured Beef Cattle	Uninsured Beef Cattle (Counterfactual)	ATT	t	Rate
Voluntary vaccination	1.664 (0.020)	1.718 (0.074)	−0.054 (0.076)	−0.701	−3.28%
Deworming	6.267 (0.051)	8.466 (0.051)	−2.199 *** (0.072)	−30.358	−35.09%
Shed disinfection	4.703 (0.047)	6.866 (0.055)	−2.162 *** (0.072)	−30.018	−45.97%

Note: *** p < 1%. Standard errors in parentheses.

, under the counterfactual assumption that insured farmers choose not to be insured, the three inputs increase by 0.054, 2.199, and 2.162, in that order. Among these, the ATTs of deworming and shed disinfection are both negative at the 1% significant level, while the ATT of voluntary vaccination is not significant. It is demonstrated that the insurance-adoption behavior of farmers significantly reduces the inputs of deworming and shed disinfection, with reduction rates of 35.09% and 45.97%, respectively.

The second heterogeneity analysis involves different beef-cattle farming regions. The samples are divided into two groups, using Tongliao City as the core area of beef-cattle farming and Chifeng City as the nearby core area. The second row of

Table 9. Differences in the impact of farmers’ insurance-adoption behavior on the input behavior of disease prevention across resource endowments.

Resource Endowment of Different Farmers		Sample Size of Insured Group	Insured Beef Cattle	Uninsured Beef Cattle (Counter -Factual)	ATT	t	Rate
Farming area	Core area	141	6.749 (0.053)	7.269 (0.055)	−0.520 *** (0.076)	−6.814	7.70%
	Nearby core area	54	6.508 (0.115)	4.129 (0.131)	2.379 *** (0.174)	13.643	−36.56%
Beef-cattle deaths	Deaths	86	6.938 (0.073)	8.812 (0.068)	−1.874 *** (0.100)	−18.709	−27.01%
	No deaths	109	6.478 (0.061)	8.674 (0.094)	−2.196 *** (0.112)	−19.633	−33.90%

Note: *** p < 1%. Standard errors in parentheses.

reports that the ATT in the core area of beef-cattle farming is negative at the 1% significance level. Under the counterfactual assumption that insured farmers do not adopt insurance, the disease-prevention input of farmers in the core area increases by 0.520 (with an increase rate of 7.70%). It can be seen that farmers in this area reduce their input in beef-cattle disease prevention after insurance adoption. The third row of Table 9 reports that the ATT in the nearby core area of beef-cattle farming is positive at the 1% significance level. Under the counterfactual assumption that insured farmers do not adopt insurance, the disease-prevention input of farmers in the nearby core area decreases by 2.379 (with a decline rate of 36.56%). It can be seen that farmers in this area increase their input in beef-cattle disease prevention after insurance adoption.

The third heterogeneity analysis focuses on different situations surrounding beef-cattle deaths. Samples with a mortality rate greater than zero are classified as an occurrence of beef-cattle deaths in the current year, while samples with a mortality rate of 0 are classified as a nonoccurrence of beef-cattle deaths in the current year. The samples are subsequently divided into two groups based on this criterion. Rows 4 and 5 of Table 9 report the analysis results, showing that the ATTs for both the deaths and the no-deaths groups are negative at the 1% significance level, indicating that, regardless of whether the beef cattle have died, the insured farmers significantly reduce their input in preventing diseases after insurance adoption. Under the counterfactual assumption that insured farmers do not adopt insurance, the input of disease prevention for farmers whose beef cattle have died increases by 1.874 (with an increase rate of 27.01%), and that for farmers whose beef cattle have not died increases by 2.196 (with an increase rate of 33.90%). It is further demonstrated that, among farmers whose beef cattle have not died, the decrease in beef-cattle disease-prevention input is greater after adopting insurance.

5. Discussion

5.1. Main Findings

Diseases constitute a crucial element exerting an influence on the evolution of the beef-cattle industry. The implementation of disease-prevention measures such as voluntary vaccination, deworming, and shed disinfection can effectively reduce the probability of risks. Beef-cattle insurance adoption can offer loss compensation to farmers after the occurrence of risk. Existing research has contended that there are diverse types of relationships among different management measures before and after the occurrence of a risk. Through theoretical analysis and empirical tests in this paper, we conclude that there exists a substitutive relationship between the input in disease prevention and the insurance-adoption behavior with beef cattle as the risk-protected objects. That is, beef-cattle farmers significantly reduce disease-prevention input after choosing to adopt insurance. The research outcomes can enrich the diversity of disease-prevention measures for beef cattle in China and enhance their penetration rate while expanding the scope of beef-cattle insurance coverage. Thus, our research holds significance in guaranteeing the development of the beef-cattle industry.

This paper concludes that beef-cattle farmers reduce their input in disease prevention after adopting insurance, thereby validating the moral-hazard issue of insured farmers reducing their input in pre-event risk management. The hypothesis $H_1$ proposed in this paper has been empirically validated. This conclusion is the same as that reached by Lin and Wang [35]. However, some scholars have reached other distinct conclusions, contending that the moral-hazard issue is not the principal impediment to the development of livestock insurance in China [50,51]. This also suggests that a more in-depth comprehension and mastery of the relationship between beef-cattle insurance and disease prevention in China requires the consideration of more influencing factors. Consequently, we conduct a further analysis of the diverse effects that different control variables exert on the conclusion.

Based on the ESR model, we conduct research in two stages. In the first stage, the factors influencing whether farmers adopt beef-cattle insurance are estimated, and the following analysis is conducted based on the crucial conclusions: (I) Farming scale has a negative influence on the insurance-adoption choice of farmers. This suggests that, as the farming scale increases, farmers possess greater risk-bearing capacity and adopt more advanced and diversified disease-prevention measures, thereby reducing the likelihood of opting for insurance coverage. (II) Green farming technology has a positive influence on the insurance-adoption choice of farmers. In recent years, the rapid development of farming technology in China has resulted in gains and increased efficiency for farmers, leading them to actively embrace green farming techniques. Nevertheless, new technologies also introduce novel risk factors [79]. Farmers are often risk-averse [54,55,56], so they tend to purchase insurance to share risks to enhance their ability to withstand risks. (III) The mortality rate of beef cattle also exerts a positive influence on the insurance-adoption choice of farmers. As a means of risk management, insurance has the main functions of risk sharing and loss compensation. The higher the mortality rate of beef cattle, the greater the actual economic loss that farmers suffer. Therefore, they are more inclined to obtain loss compensation through insurance.

The second stage estimates the impact factors of farmers’ insurance participation and nonparticipation on the input in cattle disease prevention and concludes accordingly. (I) Educational level only exerts a positive influence on the uninsured group, suggesting that, in the case of not choosing insurance as an effective means for the risk sharing and loss compensation of diseases, the higher the educational level of farmers, the higher the probability of choosing to directly increase the input in disease prevention. (II) Farming scale exerts a positive influence on both the insured and uninsured groups. However, the significance level is higher for the insured group. This indicates that farmers with a larger farming scale attach greater importance to adopting disease-prevention measures [80]. They will not neglect disease prevention due to having insurance, which also implies that the probability of moral hazard occurring is relatively lower. (III) Net income from farming exerts a positive influence on both the insured group and the uninsured group. Commonly, the higher net income from farming, the more financial capacity and willingness farmers have to conduct disease prevention by increasing input [81,82]. Nevertheless, the degree of influence of this variable on the insured group is lower than that on the uninsured group. This is because, after adopting insurance, regardless of the income level of farmers, they can obtain insurance compensation after risks have manifested. Hence, the willingness of insured farmers to increase input in disease prevention is weaker than that of uninsured farmers. (IV) Green farming technology exerts a positive effect only on the uninsured group. This is because green farming technology belongs to the “risk-increasing” input [83], while the input for disease prevention belongs to the “risk-reducing” input [79]. Farmers who have not adopted insurance, a post-event loss-compensation method, have to bear the risk by themselves. Therefore, they are more prone to increasing the disease-prevention input. (V) Both the mortality rate and risk perception exert a positive influence merely on the uninsured group, suggesting that, when farmers are aware that raising beef cattle may lead to cattle deaths due to various risks, uninsured farmers can only reduce the possible economic losses by directly increasing input in disease prevention. This is also the reason why the disease cognitive skill only exerts a positive influence on the uninsured group.

5.2. Findings of the Heterogeneity Analysis

Rao and Zhang [45] discovered that the downturn in the prices of hog and pork in the market could aggravate moral hazard. In this paper, a separate counterfactual hypothesis analysis is carried out on the sample with “price fluctuations = 1” and compared with the full sample, resulting in the same conclusion: the decline in the purchase price in the beef-cattle market augments the moral hazard of farmers, which is manifested as a significant strengthening of the negative effect of insurance-adoption behavior on disease-prevention input. The hypothesis $H_2$ proposed in this paper has been empirically validated. The study findings show that livestock industries such as beef cattle and pigs not only face risks of death from diseases, natural disasters, and so on, but also market risks from price fluctuations [4,84,85]. As relevant studies have shown, owing to factors such as the substantial import of beef and the continuous escalation of feed prices [86], the Chinese beef-cattle industry is currently exposed to severe influences of price risks [87]. This implies that, in the future development of beef-cattle insurance, it is essential not only to provide risk protection against the mortality of beef cattle but also to incorporate market risks, such as price fluctuations, which cause economic losses to farmers into the scope of coverage.

This article also conducts a heterogeneity analysis on three specific disease-prevention inputs. First, the CI policy is implemented effectively in the surveyed area, with 94.20% of the sampled farmers receiving government-provided free vaccines. The number of farmers who are fully voluntarily vaccinated or voluntarily vaccinated after receiving free vaccines is relatively limited, and they typically have a relatively high awareness of the significance of immunity. Consequently, voluntary vaccination input is not significant after being insured. Second, although deworming has consistently been a disease-prevention modality that is highly recognized and prevalently adopted by livestock farmers, it requires farmers to bear all the expenses. In addition, in the surveyed regions, the majority of farmers fail to achieve satisfactory outcomes owing to the lack of scientific deworming practices. Consequently, after adopting insurance, farmers comprehensively deliberate on the efficacy of deworming and the cost input, leading to a relatively significant reduction in the input in deworming. Third, the reduction in shed-disinfection input by farmers after insurance adoption is the greatest because this input is merely an indirect and auxiliary way of disease prevention, which requires a lot of labor and does not show much improvement in the short term. In addition, the heterogeneity analysis of resource endowments once again discloses that the farming area and the mortality of beef cattle influence the behavior of disease-prevention input by farmers after their insurance selection. The changes in disease-prevention input by farmers in the core area and the nearby core area following insurance adoption exhibit opposite trends. From a sample perspective, the low insurance adoption among farmers in the nearby core area is evident, with only 54 out of 187 farmers subscribing to beef-cattle insurance. In reality, compared with Tongliao City, Chifeng City lacks preferential policies for beef-cattle farming and insurance. Consequently, the influence of beef-cattle insurance in Chifeng City on farmers’ farming practices, including disease-prevention measures, is relatively limited.

5.3. Policy Implications

The research conclusions of this article verify a series of practical impacts. The conclusion that farmers exhibit moral hazard following insurance-adoption confirms that insurance companies face considerable payout pressures, which may impede the expansion and development of beef-cattle insurance. The finding that farmers reduce their input in disease prevention after insuring their beef cattle confirms that the government faces a substantial burden in animal disease control, which poses a potential risk to the sustainable development of the beef-cattle industry. In light of these findings, it is imperative to implement appropriate measures to address the challenges.

First, the coverage of policy-based beef-cattle insurance should be expanded. The development of swine and dairy cow insurance in China has been relatively advanced, largely due to the premium subsidies provided by the central government. In contrast, insurance for beef cattle in China is subsidized exclusively by local governments. Hence, it is imperative to prioritize the exploration of including beef cattle within the central government’s premium subsidy program. The primary objective is to alleviate the financial burden on local governments, reinforcing beef-cattle insurance’s policy-oriented nature and accelerating its broader implementation. Simultaneously, it is essential to progressively incorporate all beef-cattle farmers into the protection system. In this process, efforts should be made to expand the number of farming entities participating in insurance and increase the number of beef cattle insured by each farmer. The achievement of this objective remains contingent upon the effective implementation of financial subsidies provided by the central government.

Second, it is imperative to continuously refine the design of beef-cattle insurance products. To mitigate the moral hazard arising after farmers adopt insurance, it is advisable to implement differentiated premium rates for beef-cattle insurance policies. Based on the conclusions of this study, whether farmers alter their original disease-prevention input behaviors after adopting insurance is influenced by a multitude of factors, including farm scale, geographic region, beef-cattle mortality rates, perceptions of disease risk, etc. Therefore, by implementing differentiated pricing based on these factors, it is possible to effectively balance farming costs and compensation for losses, reducing the likelihood of moral hazard among farming entities. In the specific implementation process, it is essential to rigorously conduct verification and oversight at each stage of underwriting, policy review, and claims settlement. This stringent control aims to minimize instances of insuring unhealthy beef cattle and claims for non-covered liabilities. Furthermore, implementing no-claim discounts can incentivize farmers to safeguard the insured subject matter actively.

Thirdly, enhancing farmers’ proactive engagement in disease prevention efforts is essential. It is imperative to actively implement the CI policy for beef cattle and promptly adjust the scope of compulsory immunization. In addition to addressing identified infectious diseases, it is essential to consider incorporating diseases with high mortality rates or significant treatment costs into the CI policy. To enhance the frequency of disinfection and cleaning practices among farmers, adopting a dual approach combining subsidies and penalties is imperative. The government could centrally procure and distribute disinfection and cleaning supplies, conduct regular inspections and oversight, and impose appropriate sanctions on non-compliant farmers. Furthermore, it is essential to actively promote the socialization of veterinary services, enhance the sense of professional responsibility among veterinarians, ensure the effective implementation of disease-prevention measures, and provide convenient, high-quality, and cost-effective veterinary services to farming entities.

5.4. Limitations and Future Research

The sample size of this study is relatively limited and confined to the IMAR of China, which may constrain the generalizability of the research findings. It is worth contemplating that China’s vast territory encompasses diverse beef-cattle farming regions, each facing distinct natural conditions such as geographical environments, climatic factors, and varied cultural contexts influenced by multiple ethnic groups. Consequently, this diversity results in significant variations in risk profiles, farming practices, disease prevention methods, and the adoption of beef-cattle insurance. This implies that, to truly realize the stable development of China’s beef-cattle industry and effectively exert the positive role of insurance during its development process, it is necessary to consider other beef-cattle farming regions in China.

It is important to acknowledge that potential biases may exist in the data acquisition process during the research phase or in the selection and measurement of control variables when constructing models. Therefore, in subsequent research, it is crucial to continuously refine the design of survey questions and enhance interview techniques during fieldwork. This approach will facilitate the acquisition of more comprehensive and precise data, thereby supporting the rational selection and specification of variables. Furthermore, the study area will be expanded in future research, and the sample size will be increased. It is also imperative to consider the specific circumstances of other beef-cattle farming regions in China and the heterogeneity among local beef-cattle farmers. Therefore, it is essential to introduce new variables in subsequent research to further enhance and refine the research model, including distance from the farming center, weather conditions, governance factors, etc.

6. Conclusions

The development of China’s beef-cattle industry is significantly impacted by disease outbreaks. As critical tools for managing disease risks, beef-cattle insurance and disease-prevention measures hold substantial implications. Investigating the interrelationship between these two elements is of paramount importance. Based on the micro-survey data of 447 farmers in the IMAR of China, this paper utilizes the ESR model and constructs counterfactual hypotheses to analyze the effect of the insurance-adoption behavior of beef-cattle farmers on their input behavior in disease prevention. The main conclusions are as follows: First, farmers’ insurance-adoption behavior significantly reduces beef-cattle disease-prevention input (i.e., moral hazard arises when farmers adopt beef-cattle insurance). Specifically, under the counterfactual assumption that insured farmers do not adopt insurance, the input in the prevention of beef-cattle diseases increases by 33.45%, and this conclusion remains robust after replacing the dependent variable and altering the estimation method. Second, the decline in the market purchase price of beef cattle enhances the negative effect of farmers’ insured behavior on input for beef-cattle disease prevention. Comparisons with the full sample suggest that insured farmers reduce their input for beef-cattle disease prevention to a greater extent when the market purchase price of beef cattle decreases compared with the previous year. Third, heterogeneity analyses show that insured farmers significantly reduce their deworming and shed-disinfection inputs, with the reduction rates being 35.09% and 45.97%, respectively. Through a heterogeneity analysis of farmers with different resource endowments, it was found that disease prevention input among farmers in the core area decreased significantly by 7.70% after being insured, whereas farmers in the nearby core area experienced a significant increase of 36.56%. The reduction in disease prevention input for insured farmers who did not experience beef-cattle deaths is 33.90%, compared to a reduction of 27.01% for insured farmers who experienced beef-cattle deaths.