Smart Farming Technology, Scale Economies and Carbon Efficiency: Evidence from Chinese Dairy Farms

Abstract

Carbon emissions from dairy farms have significantly hindered the advancement of sustainable agriculture, and improving carbon efficiency is a key pathway to mitigate these emissions. As a critical technological innovation, smart farming technology exerts a substantial impact on boosting carbon efficiency in dairy farms. Based on field survey data collected from Chinese dairy farms, this study employs an integrated empirical approach, including endogenous switching regression, two-stage least squares, and propensity score matching, to rigorously evaluate the impact of smart farming technology on economies of scale. A mediation analysis is further conducted to examine the interrelationships among smart farming technology, economies of scale, and carbon efficiency, while the moderating role of government regulation is also empirically tested. The findings reveal three key results: (1) Smart farming technology exerts a direct and positive influence on the economies of scale in dairy farms, with this effect becoming more pronounced as farm size increases. (2) Economies of scale serve as a partial mediator in the relationship between smart farming technology and carbon efficiency. This indicates that smart farming technology not only directly enhances carbon efficiency but also does so indirectly by facilitating the expansion of production scale and reducing unit costs. (3) Government regulation positively moderates this mediating pathway. Specifically, through standardizing production practices, offering policy incentives, and guiding the application of technology, government interventions strengthen the ability of smart farming technology to foster economies of scale. These insights underscore the importance of steering dairy farms toward the adoption of smart farming technologies to simultaneously improve scale efficiency and carbon performance, thereby supporting the transition toward low-carbon and sustainable agricultural development. Finally, this study proposes three policy implications: strengthening institutional support for the adoption of smart farming technologies in dairy production systems, significantly enhancing training programs related to these technologies, and systematically guiding dairy farms toward smart technology adoption to simultaneously improve economies of scale and carbon efficiency.

1. Introduction

China, as the foremost contributor to global carbon dioxide (CO₂) emissions, is confronted with the substantial imperative to curtail its carbon emissions [1]. During the seventy-fifth session of the United Nations General Assembly, the Chinese government officially presented China’s objective of reaching the peak of carbon dioxide emissions by the year 2030 and achieving carbon neutrality by 2050 [2]. Reducing carbon emissions stands as a strategic key task for China in achieving high-quality development in the future. The Food and Agriculture Organization (FAO) of the United Nations pointed out that carbon dioxide emissions from cattle farming account for 18% of the world’s total greenhouse gas emissions, among which dairy cows produce more than 150 million tons of methane each year [3]. The carbon emissions (in CO₂ equivalent) from Chinese dairy farms had reached 45.151 million tons by 2020, and it is projected that this figure will rise to 57.315 million tons by 2030 [4]. The upward trend is likely to exacerbate regional greenhouse effect and impose additional pressure on China’s national carbon neutrality goals. Therefore, dairy farms have been recognized as a primary contributor to the overall carbon emissions associated with agricultural activities in China [5]. Chinese dairy farms are currently undergoing a crucial phase of low-carbon advancement, and the endeavor to mitigate carbon emissions poses a significant challenge [6]. With the continuing expansion of dairy farms and the escalating environmental pollution in China, enhancing carbon efficiency has emerged as an imperative approach to attain sustainable development in Chinese dairy farms. The concept of carbon efficiency pertains to the degree to which undesirable outputs may be minimized while maintaining constant levels of inputs and outputs [7,8,9]. This concept holds significance in the context of mitigating carbon emissions and pollution in dairy farms. Enhancing carbon efficiency is a viable approach to optimize the operational effectiveness of dairy farms in China while mitigating carbon emissions and pollution. This strategy aligns to foster low-carbon growth in the dairy farms sector.

Technological innovation serves as the primary mechanism for tackling both economic growth and carbon emission pollution [10,11]. Currently, the profound amalgamation of smart farming technology and agriculture has yielded substantial reductions in agricultural production costs and environmental degradation, while concurrently offering technological assistance to augment scale economies and carbon efficiency [12,13,14,15]. The Ministry of Agriculture and Rural Affairs of the People’s Republic of China has issued the Guiding Opinions on Vigorously Developing Smart Agriculture, attaching great importance to the application of smart farming technologies [16]. For dairy farms, precision feeding, production monitoring, and environmental control represent the three core digital technologies of smart farms. Precision feeding enables accurate nutrient delivery through automated ingredient mixing and rationed dispensing. Production monitoring utilizes sensors and wearable devices to track individual animal growth, health, and reproductive status in real time. Environmental control optimizes livestock living conditions by automatically regulating barn temperature, humidity, ventilation, and lighting [17,18]. With the ongoing integration of smart farming technology in dairy farms, the dairy farms have been able to enhance their scale and professionalism through adjustments and optimizations in breeding practices [19,20]. This has resulted in improved production efficiency and a reduction in non-essential carbon emissions, ultimately contributing to the enhancement of carbon efficiency. The application of smart farming technology in managing dairy farms’ resources can facilitate resource conservation, enabling herdsmen to reinvest and promote the reproduction of resources. This, in turn, enhances the return rate of production factors and establishes the necessary conditions for the expansion of dairy farms. By achieving scale economies, the development of dairy farms can be further incentivized, leading to increased specialization and scale [18]. To achieve greater returns, dairy farmers should allocate additional capital toward the enhancement of modern dairy farming technology. This strategic investment would facilitate the transition from rudimentary operations to a low-carbon and efficient development model, consequently mitigating carbon emissions and augmenting the carbon efficiency of dairy farms. Following the expansion of dairy farms, there has been a corresponding increase in the revenue generated by dairy farming. Consequently, the primary stakeholders in the dairy farming industry are more inclined to allocate additional resources towards the development of this sector. This, in turn, facilitates the enhancement of expertise and proficiency in managing dairy farms, leading to the establishment of high-quality dairy farms. Moreover, this development also results in a reduction in the utilization of carbon-emitting resources within dairy farms, thereby improving their carbon efficiency.

Simultaneously, there has been a notable increase in the extent of dairy farms in China. In 2020, the Opinions on Promoting the High-Quality Development of Animal Husbandry were issued by the General Office of the State Council in China. These opinions emphasize the need to develop livestock and poultry farming on a large scale, taking into account local conditions. The document also guides farms to expand their farming operations and enhance the standardization of farming practices. Furthermore, it sets a target for the proportion of large-scale livestock and poultry farming to exceed 75% by the year 2030 [21]. China’s dairy farming policies have driven a substantial rise in large-scale dairy cattle production, with its proportion reaching 70% in 2022. Despite this growth, the figure remains below the targeted objective. As a result, the future development of China’s dairy sector will continue to prioritize the construction of large-scale farms [22]. Nevertheless, despite the significant expansion of dairy farms in China, the establishment of scale economies has not been achieved. This indicates that Chinese dairy farms continue to encounter challenges such as increased farming expenses and diminished farming profits [23,24]. In the context of the advancement of dairy cattle production, the attainment of scale economies in dairy farms emerges as a crucial concern pertaining to the sustainable development of such feedlots.

Previous research has explored the concepts of scale economies and agricultural carbon efficiency [25,26,27]. However, these studies have mostly focused on macro-level analysis. Consequently, there is a scarcity of research examining the micro-level dynamics of scale economies and carbon efficiency specifically in the context of dairy farms. Furthermore, it is worth noting that smart farming technology represents a recent technical advancement that has not yet been extensively explored by researchers within the context of theoretical and empirical studies. Several issues warrant further investigation within the existing research framework, for instance, does smart farming technology effectively produce scale economies in dairy farms, particularly in relation to the management and operations of these agricultural areas? Does smart farming technology lead to enhanced carbon efficiency in dairy farms by leveraging scale economies? This paper examines the theoretical influence mechanism of smart farming technology, scale economy, and carbon efficiency in dairy farms. It also utilizes micro-research data from dairy farms to analyze the impact of smart farming technology on the scale economy, employing the endogenous switching model. Additionally, the mediation effect model is applied to analyze the role of scale economy in mediating the relationship between smart farming technology and carbon efficiency in dairy farms. The subsequent sections are organized as follows. Section 2 presents the theoretical framework and research hypotheses pertaining to the utilization of smart farming technology, scale economies, and carbon efficiency within the context of dairy farms. Section 3 provides a comprehensive account of the data and empirical model utilized in this research article. Section 4 examines and deliberates the empirical findings. Section 5 gives a comprehensive summary of the whole article and presents suggestions based on the findings.

2. Theoretical Basis and Research Hypothesis

Agricultural producers aim to achieve economies of scale but are constrained by internal factors from reaching the optimal production state. The core objective of agricultural producers is to attain economies of scale. To achieve this goal, they continuously expand their operational scale and reduce production costs to improve agricultural production efficiency [28]. However, due to constraints such as limited management capacity and asset stock, producers struggle to maintain production at the optimal level where long-term average costs are minimized. This imposes restrictions on their efforts to maximize profits by expanding production scale [29].

Smart farming technology can overcome the limitations of traditional agriculture, align with the cost characteristics of dairy farms, and support the realization of economies of scale in both agriculture and dairy farms. Integrating smart farming technology into agricultural development can optimize various links in agricultural production, processing, and distribution, thereby addressing the limitations of traditional agriculture—such as low efficiency, high costs, and risks associated with small-scale decentralized operations. Meanwhile, through optimizing models and promoting institutional reforms, it facilitates the efficient flow of industrial factors and the expansion of agricultural scale, ultimately supporting the achievement of economies of scale in the agricultural sector [30,31]. Furthermore, smart farming technology enables the optimal allocation of production factors, reduces both marginal and average production costs, alleviates constraints on production scale expansion, and broadens the boundaries of large-scale production—all of which provide support for the development of economies of scale [32,33]. In the context of dairy farms, smart farming technologies significantly boost carbon emission efficiency by optimizing key farm operations. Digital precision feeding tailors diets to individual cows, which enhances feed conversion rates and directly reduces methane emissions from digestion and manure. Simultaneously, production monitoring technology improves herd health and productivity, meaning more milk is produced per cow for the same level of emissions. Furthermore, environmental control systems efficiently manage energy use in barns for heating, cooling, and ventilation, slashing the overall carbon footprint of the farming operation [34,35]. Therefore, these integrated technologies enable dairy farms to achieve a higher output with lower emissions per unit of milk. Additionally, smart farming technology exhibits the characteristic of “high fixed costs and low marginal costs”. This cost structure incentivizes herders to expand production for profit maximization, and in the long run, the unit cost associated with scaling up dairy farming decreases [36,37]. These technologies can also significantly reduce unit management costs and drive the transition from labor to machinery, contributing to carbon pollution control in dairy farming. Practices have shown that replacing labor with technology can improve output efficiency, optimize dairy production efficiency and economic benefits, thereby laying a foundation for the development of economies of scale in dairy farms [38].

Based on the above analysis, this paper puts forward the following hypothesis:

Hypothesis 1.

Smart farming technology can promote the realization of economies of scale in dairy farms.

Economies of scale can enhance production and carbon efficiency, and the consolidation of dairy farm scales can further reduce costs, promote green development, and improve overall benefits. The formation of economies of scale reduces the consumption of raw materials and energy, improves factor utilization efficiency, lowers energy consumption, carbon emissions, and pollution intensity, and ultimately drives the improvement of production efficiency, carbon efficiency, and profitability [36]. The growth in profits, in turn, encourages producers to increase investment in carbon emission reduction, further enhancing their capacity for carbon emission and pollution management. For dairy farms, the consolidation of production scale can further foster economies of scale, reducing the costs of fixed assets such as smart farming equipment, green storage cellars, and manure storage tanks [39]. This reduction in fixed costs effectively lowers long-term average costs, creating conditions for the establishment of regional specialized production while reducing management costs associated with specialized production. Cost reduction is key to deepening the division of labor in specialized production, which can further improve production efficiency. The saved costs can then be invested in green production technologies, reducing pollution by improving the substitution rate and utilization efficiency of factors such as energy [40]. Ultimately, economies of scale help build a low-carbon and environmentally friendly new model for dairy farming, achieving improvements in both production efficiency and ecological benefits.

Based on the above analysis, this paper puts forward the following hypothesis:

Hypothesis 2.

Economies of scale play a positive mediating role in the impact of smart farming technology on carbon efficiency in dairy farms.

Through the above analysis, it can be found that economies of scale serve as a crucial mediating variable in the transmission chain of smart farming technology empowering the improvement of carbon efficiency in dairy farms. As an important external institutional variable, government regulation can exert a significant moderating effect on the mediating process through which smart farming technology affects the carbon efficiency of dairy farms. Government regulation refers to the government’s direct policy intervention and indirect policy intervention through market mechanisms. Its purpose is to adjust the production behaviors of actors, correct market failures, and ultimately achieve the goals of reducing environmental pollution and promoting economic development [41]. On the one hand, through regulations such as promoting smart breeding technology and providing training for dairy farms, the government can improve dairy farmers’ cognitive ability and technical level regarding smart farming technology. This enhancement enables dairy farmers to achieve better practical results when using smart farming technology for production and operation. The improved application level of smart farming technology can further promote the scientific allocation of various production factors in dairy farms. It also provides basic conditions for dairy farms to expand their production scale, thereby accelerating the formation of economies of scale in dairy farms [17]. On the other hand, government regulation standardizes the order of the market and reduces disorderly competition [42]. This provides a stable development environment for dairy farms, further encouraging them to increase investment in smart farming technology. Such investment lays a technical foundation for the large-scale development of dairy farms. It also enables smart farming technology to play a more effective role in promoting dairy farms to achieve economies of scale, ultimately driving the improvement of carbon efficiency in dairy farms.

Based on the above analysis, this paper puts forward the following hypothesis:

Hypothesis 3.

Government regulation exerts a moderating effect on the mediating process through which the application of smart farming technology influences the carbon efficiency of dairy farms.

3. Methods and Data

3.1. Data Sources

The data presented in this study originates from field research conducted by a research group focusing on dairy farms in Heilongjiang province, which is recognized as a prominent region for dairy farming in China. The Heilongjiang province possesses favorable conditions for dairy farming, characterized by a substantial size of operations, notable regional representation, and robust data accessibility. This study used a combination of stratified and typical sampling to select representative dairy farms from each research area based on breeding scale, breeding technology, and pollution control. The final distribution of the surveyed areas is shown in Figure 1. The research was carried out throughout the period of June 2022 to August 2024. To ensure sample representativeness and data reliability, the researchers conducted in-depth, one-on-one interviews with ranchers, technical directors, milking parlor supervisors, and other relevant farm personnel. These interviews, based on a structured questionnaire, were designed to gain comprehensive insights into dairy farming practices. The research team assigned unique identifiers to the dairy farms under investigation and conducted a comprehensive and extended monitoring process to ensure the rigor and viability of the empirical analysis. A total of 278 valid questionnaires were collected, primarily focusing on the current utilization of smart farming technology in dairy farms, as well as personal and familial information of the herdsmen, and the state of dairy cattle breeding.

3.2. Variable Selection

(1) Dependent variable: carbon efficiency (CE). Carbon efficiency refers to the measure of how effectively and sustainably resources are utilized in the production of dairy products within these agricultural systems. This study utilizes the scholarly research conducted by various scholars [43] to examine the relationship between dairy farms and their impact on the environment. To assess carbon efficiency, the Undesirable Outputs-SBM model is employed. The total input is considered the independent variable, while the total output and carbon emission pollution are regarded as the dependent variables representing desired and non-desired outputs, respectively. Input variables include forage, labor, fixed assets, water-electricity-fuel, medical epidemic prevention, and other materials; dependent variables cover desired outputs (raw milk yield, by-product income) and non-desired output (carbon emissions from rumen fermentation, manure, energy use).

Scale economies (Scale). The concept of scale economies refers to the phenomenon in which the average cost of production decreases as the quantity of output increases. In a state of consistent technological circumstances, the rise in product output accompanied by a gradual decline in the average cost of producing a single unit of product signifies the presence of scale economies. Conversely, a gradual increase in the average cost of producing a unit of product indicates the diseconomies of scale. This paper utilizes scholarly research [44] to examine the scale economies. This is achieved by considering the average cost and the ratio of the marginal cost of dairy farms, which collectively determine the scale economies index. It is worth noting that the scale economies index greater than 1 indicates the presence of scale economies. Research findings suggest that the cost associated with expanding the production of dairy farms for milk production is lower than the average cost of milk production. This implies the presence of scale economies, whereby the increase in production may be achieved by enlarging the dairy farms. A scale economies index below 1 suggests that the cost of increasing each unit of milk production exceeds the average cost of milk production. This indicates the presence of diseconomies of scale, wherein expanding the production scale of dairy farms results in a further rise in the cost of dairy farms. Consequently, this situation is not favorable for maximizing the benefits derived from dairy farms.

(2) Explanatory variable: smart farming technology (Tech). In this paper, we summarize the smart farming technology equipment including total mixed ration weighing controller, digital feeding wagon, automatic calf weigher, automatic milking controller, milk composition analyzer, milk flow meter, automatic temperature control panel, automatic spraying equipment and other digital equipment [17,18,19]. If one of the smart farming technology devices is used, then smart farming technology has been applied in dairy farms and is assigned a value of 1. Conversely, it is assigned a value of 0.

(3) Control variables: The control variables in this study include the herdsmen’s years of education (Edu), age (Age), village cadres (Cad), farming experience (Exper), risk perception (Risk), breeding training (Train), Cooperative (Coop), urban household (Urban), the proportion of farming income (Income), and the total number of family laborers [45]. A detailed description of these variables can be found in

Table 1. The meaning and descriptive statistics of the main variables.

Variable Name	Variable Symbol	Variable Meaning	Mean	Standard Deviation
Carbon efficiency	CE	The results obtained from the Undesirable Outputs-SBM model	0.6222	0.1130
Scale economies	Scale	The value of the average cost of producing milk divided by the marginal cost of dairy farms	1.2201	0.1143
Smart farming technology	Tech	Application of smart farming technology equipment. Yes = 1, no = 0	0.3669	0.4828
Years of education	Edu	Number of years of education for herdsmen	9.7230	1.2854
Age	Age	The age of the herders	46.6007	7.2288
Village cadres	Cad	Yes = 1, no = 0	0.0683	0.2528
Farming experience	Exper	The number of years of dairy farming	16.6835	9.8318
Risk perception	Risk	Level of concern about the risks of dairy farming (from 1 to 5)	4.5540	0.6264
Breeding training	Train	Yes = 1, no = 0	0.3705	0.4838
Cooperative	Coop	Yes = 1, no = 0	0.4137	0.4934
Urban household	Urban	Yes = 1, no = 0	0.3345	0.4727
Proportion of farming income	Income	The proportion of income from dairy farming in total income	78.3453	21.1483
Number of family laborers	Labor	Number of household members available for dairy farming	3.8705	1.3666

.

(4) Moderating variables: Government regulation (Regula). Government regulation specifically includes restrictive regulation (RR), incentive regulation (IR), and guiding regulation (GR). Drawing on the research of Lin et al. (2018) [46] and combining the actual conditions of dairy farms, this study constructs an evaluation index system for government regulation. First, from a dimensional perspective, it measures the regulatory intensity of restrictive regulation, incentive regulation, and guiding regulation, respectively; then, it calculates the comprehensive value of government regulation as a whole. Restrictive regulation is characterized by indicators such as the completeness of government environmental protection laws and regulations, the intensity of government supervision over environmental pollution, the intensity of government prohibition on pollution discharge from dairy farming, and the severity of government penalties for environmental pollution caused by dairy farming. Incentive regulation is represented by indicators including the intensity of government recognition for green demonstration households in dairy farming, the intensity of government credit support for dairy farming, the degree of government subsidies for green production technologies in dairy farming, and the intensity of government subsidies for the purchase of breeding equipment. Guiding regulation is reflected through indicators such as the intensity of government promotion of dairy farming technologies, the intensity of government provision of dairy farming information, the degree of government publicity and education on environmental protection and governance, and the intensity of government training and guidance for dairy farming. Government regulation is measured based on the actual experience of dairy farmers, with a 1–5 point rating scale (from low to high). Principal component analysis and varimax factor rotation are used to extract three common factors: “Restrictive Regulation”, “Incentive Regulation”, and “Guiding Regulation”. Finally, the comprehensive value of government regulation is calculated based on the score of each factor and its corresponding variance contribution rate. The Kaiser-Meyer-Olkin (KMO) test statistic assesses data suitability for factor analysis by comparing simple and partial correlations, and a value closer to 1 denotes stronger correlations and a more appropriate dataset for the analysis. The KMO value of each item is greater than 0.7, and Bartlett’s test of sphericity statistic reaches the 1% significance level. These results indicate that factor analysis of the data is feasible.

3.3. Descriptive Statistics of the Sample

Among the 278 dairy farmers surveyed, 102 have adopted smart agricultural technologies on their farms, representing 36.69% of the total sample. The remaining 176 farmers have not yet implemented such technologies, accounting for 63.31% of the sample. This suggests that the current adoption of smart agriculture in dairy farming remains relatively limited, indicating considerable potential for its further promotion. Regarding technical training, 55.76% of the dairy farmers reported that they had not participated in any training programs, reflecting a relatively low participation rate in technical training among this group. In terms of household registration type, 215 farmers hold rural household registration, making up 77.34% of the sample, while 63 have urban registration, accounting for 22.66%. This indicates that rural household registration remains the dominant type among dairy farmers in the sample area. Only 19 farmers serve as village cadres, constituting less than one-tenth of the total sample. The vast majority (93.17%) do not hold major village cadre positions.

Concerning educational attainment, most dairy farmers have only received primary or junior high school education. Those with higher education levels and richer knowledge reserves represent a relatively small proportion. In terms of age distribution, 154 farmers (55.40% of the sample) are between 40 and 50 years old. The proportion of young and middle-aged individuals aged 40 or below who are willing to engage in dairy farming is very small. Overall, dairy farmers in the sample area possess relatively rich breeding experience, with a considerable number having been engaged in dairy farming for 10 to 20 years. In relation to risk perception, more than half of the farmers expressed concern or extreme concern about the risks associated with dairy farming. This reflects a relatively high degree of risk aversion among the sampled dairy farmers. Regarding family size, 128 farmers come from households with three or fewer members, accounting for 46.04% of the sample. Most dairy farming households engage in part-time farming, relying on both agricultural and non-agricultural income to meet household consumption needs.

3.4. Model Construction

Herdsmen, acting as rational economic agents, may exhibit a greater propensity to adopt smart farming technology as a means to enhance production scale and optimize revenue. Consequently, this phenomenon may introduce a self-selection bias, thereby giving rise to endogeneity issues that can compromise the accuracy of model estimation outcomes. Hence, this research study leverages the scholarly works of several researchers [47,48,49,50] and uses the endogenous switching model (ESR) to examine the effects of smart farming technology on scale economies in dairy farms. The endogenous switching model enables the estimation of changes in scale economies of smart farming technology in dairy farms, both in real-world scenarios and hypothetical circumstances. This allows for the identification of the specific impacts generated by smart farming technology in dairy farms.

Furthermore, the use of the ESR model is highly suitable for analyzing binary explanatory variables. Compared to other methods, this endogenous switching regression offers two key advantages. First, it constructs a decision equation to estimate the probability of adopting smart agricultural technologies, separately modeling the selection process and the outcome process. Second, it estimates both the actual economies of scale of dairy farms and those under counterfactual scenarios. This enables an accurate measurement of the average treatment effect (ATT/ATU) of smart agricultural technologies on economies of scale, while avoiding estimation bias caused by self-selection. The estimation process of the endogenous switching model can be divided into two primary steps. The first step involves constructing a decision equation to examine the potential factors that influence the adoption of smart farming technology by herders. The second step involves constructing an outcome equation to estimate the productivity of dairy farms in two separate groups: one group utilizing smart farming technology and another group not utilizing smart farming technology. This allows for an examination of the variations in scale economies for dairy farms under different scenarios of smart farming technology usage. The decision equation pertaining to the utilization of smart farming technologies may be expressed as follows:

$${T_i}^{*}=\Phi \left(Z_i\right)+{\mu }_{\mathrm{i}}$$

(1)

The variable “T_i” denotes the implementation of smart farming technology within dairy farms. A value of T_i = 0 indicates the absence of smart farming technology adoption in dairy farms, while a value of T_i = 1 signifies the presence of smart farming technology use in dairy farms. The variable “Z_i” is used to represent a collection of control factors that have an impact on the utilization of smart farming technology in dairy farms. The symbol Φ(·) is employed to describe a schematic function, while μ_i represents a random perturbation term. To evaluate the influence of smart farming technology on the scale economies, the equation is formulated as follows:

$$Y_i={\beta }_i{X}_i+\delta T_i+{\epsilon }_i$$

(2)

The variable Y_i represents the scale economy, while X_i serves as the control variable in examining the influence of smart farming technology on scale economy. Additionally, ε_i represents the random perturbation term in this context. The impact of unobservable elements on the smart farming technology, T_i, in dairy farms can lead to a correlation between T_i and the residual term, which in turn can introduce bias in the estimation findings. Hence, the endogenous switching model further formulates the resulting equation of smart farming technology in the following manner:

$$\begin{array}{l}{Y}_{1i}={\beta }_1{X}_{1i}+{\rho }_{1u}{\lambda }_{1i}+{\epsilon }_{1i},T=1\\ Y_{0i}={\beta }_0{X}_{0i}+{\rho }_{0u}{\lambda }_{0i}+{\epsilon }_{0i},T=0\end{array}$$

(3)

The variables Y_1i and Y_0i represent the degree of scale economies in dairy farms when employing smart farming technology and when not employing smart farming technology, respectively. The variables X and ε represent the relevant control variables and random disturbances, respectively. Additionally, ρ and λ represent the covariance and inverse Mills ratio, respectively. A study was conducted to determine the anticipated economic benefits of using smart farming technology in dairy farms, both in real-world scenarios and hypothetical settings. By comparing the outcomes of dairy farms with and without smart farming technology, the average treatment impact of utilizing smart farming technology in dairy farms is determined.

The expected value of scale economies for dairy farms using smart farming technology is as follows:

$$E[\left.Y_{\mathrm{i}i}\right|T_i=1]={\beta }_1{X}_{1i}+{\rho }_{1u}{\lambda }_{1i}$$

(4)

The expected value of scale economies for dairy farms not using smart farming technology is as follows:

$$E[\left.Y_{0i}\right|T_i=0]={\beta }_0{X}_{0i}+{\rho }_{0u}{\lambda }_{0i}$$

(5)

The level of scale economies for dairy farms using smart farming technology in the counterfactual scenario is:

$$E[\left.Y_{0i}\right|T_i=1]={\beta }_0{X}_{1i}+{\rho }_{0u}{\lambda }_{1i}$$

(6)

The level of scale economies for dairy farms not using smart farming technology in the counterfactual scenario is as follows:

$$E[\left.Y_{1i}\right|T_i=0]={\beta }_1{X}_{0i}+{\rho }_{1u}{\lambda }_{0i}$$

(7)

The average treatment effect on scale economies for dairy farms using smart farming technology is as follows:

$$AT{T}_i=E[\left.Y_{1i}\right|T_i=1]-E[\left.Y_{0i}\right|T_i=1]$$

(8)

The average treatment effect on scale economies for dairy farms not using smart farming technology is as follows:

$$AT{U}_i=E[\left.Y_{1i}\right|T_i=0]-E[\left.Y_{0i}\right|T_i=0]$$

(9)

4. Results and Discussion

4.1. Correlation Analysis of Variables

The Pearson correlation coefficient test was utilized to examine the correlation between each variable. The outcomes of the test are presented in

Table 2. Correlation coefficient matrix.

	1	2	3	4	5	6	7	8	9	10	11	12
1	1
2	0.0845 ***	1
3	0.1269 ***	0.1355 ***	1
4	−0.0377 ***	−0.0530	0.0851	1
5	0.0322 *	0.0783	0.0542	−0.0172	1
6	0.0570 ***	−0.0812	0.0276	0.0202	−0.0557	1
7	−0.1245 ***	−0.1970	−0.0762	0.0391	0.0665	−0.0046	1
8	0.1801 **	0.0505 **	0.0595	−0.0609	0.0197	−0.0037	−0.0252	1
9	0.0645 *	0.0972 **	0.0417	−0.0885	0.0221	−0.0492	−0.0583	0.0397	1
10	−0.0810	−0.0968 **	−0.0015	0.0238	−0.0375	−0.0276	−0.0086	−0.0654	0.0053	1
11	0.1479	−0.0658	−0.0961	−0.0389	−0.0473	0.0013	−0.0107	0.0005	−0.0193	−0.0512	1
12	−0.1025	0.0121	−0.0164	−0.0249	−0.0786	0.0315	−0.0577	−0.0614	−0.0468	−0.0612	0.0677	1

Note: 1—Scale; 2—Tech; 3—Edu; 4—Age; 5—Cad; 6—Exper; 7—Risk; 8—Train; 9—Coop; 10—Urban; 11—Income; 12—Labor. ***, **, and * are the significant levels at 1%, 5%, and 10%, respectively.

. There exists a statistically significant positive correlation between smart farming technology and scale economies at a significance level of 10%. This suggests that smart farming technology is associated with scale economies in dairy farms. It is observed that there existed a noteworthy positive correlation between the educational attainment of herders, village cadres, and breeding training with the scale economy. Conversely, a substantial negative association was found between the risk perception and scale economy. As shown in Table 2, a negative correlation exists between the adoption of smart farming technology and urban household registration. One plausible explanation is that dairy farmers with rural household registration often rely on livestock farming as their primary income source, leading to a stronger inclination toward smart agricultural technologies. In contrast, those with urban household registration have access to non-agricultural employment opportunities in cities, resulting in a less urgent need for such technologies in dairy production.

4.2. Multicollinearity Test

To address the problem of multicollinearity among variables, the present study employed the variance inflation factor (VIF) method to conduct a multicollinearity assessment. The results of the test are displayed in

Table 3. Results of multicollinearity test.

Variable	VIF	Variable	VIF
Scale	1.92	Risk	1.37
Tech	2.52	Train	1.59
Edu	3.19	Coop	2.76
Age	1.92	Urban	1.56
Cad	1.31	Income	1.34
Exper	1.23	Labor	1.35

. The regression equation’s expansion factor for each variable is determined to be less than 4, indicating the lack of multicollinearity among the variables.

4.3. The Effect of Smart Farming Technology on Scale Economies in Dairy Farms

This section employs the endogenous switching model to estimate the varying effects of smart farming technology on the scale economies. The instrumental variable for smart farming technology is “the mean value of the degree of use of smart farming technology in other dairy farms.” The estimation results are presented in

Table 4. Estimation results of the endogenous switching model.

Variable	Decision Equation	Outcome Equation
	Smart Farming Technology	Dairy Farms That Have Applied Smart Farming Technology	Dairy Farms That Have Not Applied Smart Farming Technology
Instrumental variables	1.4031 ***
	(0.0617)
Edu	0.0625 ***	0.0471 ***	0.0449 ***
	(0.0188)	(0.0107)	(0.0114)
Age	−0.0009	−0.0049	−0.0050 ***
	(0.0023)	(0.0034)	(0.0016)
Cad	−0.0368	0.1019 ***	0.1059 ***
	(0.0565)	(0.0371)	(0.0383)
Exper	0.0004	0.0010	0.0022 **
	(0.0015)	(0.0009)	(0.0010)
Risk	−0.0136	−0.0412 ***	−0.0429 **
	(0.0238)	(0.0149)	(0.0173)
Train	0.0668 *	0.0257 *	0.0215 *
	(0.0339)	(0.0218)	(0.0231)
Coop	0.2150 ***	0.0625 **	0.0573 **
	(0.0395)	(0.0248)	(0.0277)
Urban	0.0899 **	−0.0082	−0.0143
	(0.0351)	(0.0218)	(0.0247)
Income	0.0006	0.0005	0.0004
	(0.0007)	(0.0005)	(0.0005)
Labor	−0.0491	−0.0100	−0.0082
	(0.0331)	(0.0070)	(0.0073)
Constant	0.8712 ***	0.7683 ***	0.8135 ***
	(0.2685)	(0.1659)	(0.1855)
Correlation coefficient of residuals		0.0096 ***	0.0084 ***
		(0.0011)	(0.0011)

Note: ***, **, and * are the significant levels at 1%, 5%, and 10%, respectively. Clustered standard errors are in parentheses.

. The significance of the residual correlation coefficients at the 1% level suggests that the adoption of smart farming technology is not a random behavior but rather a self-selected behavior. Consequently, the results obtained using ordinary least squares estimation may be biased. To address this bias, this paper employs the endogenous switching model to correct the estimation bias arising from the self-selection of dairy farms.

The decision equation for the use of smart farming technology in dairy farms revealed a statistically significant positive relationship between the amount of education attained by ranchers and the scale economies seen. This impact was found to be significant at a confidence level of 1%. There is a positive correlation between the level of education among herders and their access to contemporary knowledge and information channels in the field of dairy farming. Additionally, higher education levels enable herders to better comprehend the significant potential of smart farming technology in improving efficiency. Consequently, the promotion of smart farming technology in dairy farms is facilitated by the increased education levels of herders. The act of engaging in training has a favorable impact on the utilization of smart farming technologies in dairy farms, with statistical significance observed at the 10% level. Technical training can enhance herders’ digital literacy and agricultural knowledge, thereby providing them with a more comprehensive understanding of the evidence-based application of smart farming technologies. This, in turn, fosters greater adoption and utilization of these technologies within the dairy sector. The results of the analysis demonstrate a statistically significant positive relationship between participation in cooperatives and cost savings in dairy farms. This suggests that cooperatives play a pivotal role in supporting dairy farms through the provision of essential inputs and equipment, thereby facilitating operational expansion and investment in smart farming technologies. The presence of an urban hukou demonstrates a statistically significant positive impact on the utilization of smart farming technology in dairy farms. This finding suggests that in urban regions, the enhanced smart infrastructure facilitates the acquisition of information regarding smart farming technology by herdsmen. Consequently, they are more inclined to leverage this expertise to pursue technological innovation and industrial upgrading in their dairy operations.

The outcome equations demonstrate the assessment of scale economies in dairy farms, comparing those utilizing smart farming technology with those that do not. The findings of this study demonstrate a noteworthy positive correlation between the number of years of education and the adoption of smart farming technology in dairy farms. This correlation is consistent across feedlots, regardless of their adoption of smart farming technology. The results indicate that higher-educated herders are more likely to adopt cost-reduction strategies, which in turn drive the attainment and enhancement of scale economies. The utilization of smart farming technology was shown to have a considerable positive impact on dairy farms, as indicated by the absence of age-related negative effects among herders, with a statistically significant level of 1%. One plausible explanation for this phenomenon is that older herders tend to exhibit greater risk aversion in their investment strategies, leading to a lower adoption rate of modern, efficient practices such as smart farming technology. This constrained technological uptake subsequently impedes the development of scale economies within cattle feedlots. There exists a notable positive correlation between village cadres and the scale economies in dairy farms, regardless of whether smart farming technology is employed or not. This association can be attributed to village cadres’ herder status, which affords them greater familiarity with government policies and modern agricultural practices, enabling them to more effectively leverage resources for operational expansion and the realization of scale economies. The scale economies of dairy farms, both those utilizing smart farming technology and those that do not, are negatively affected by risk perception. The realization of scale economies in dairy farming, driven by factors such as operational size, feeding techniques, and technology, typically requires substantial input investments to increase output. However, risk-averse herdsmen are often susceptible to apprehension regarding the potential hazards of these elevated costs. Consequently, this risk sensitivity can deter the necessary capital commitment, thereby acting as a significant impediment to achieving scale economies. The involvement in cooperative organizations has a notable and favorable impact on the scale economies within dairy farms, regardless of their utilization of smart farming technology. This effect is statistically significant at a confidence level of 10%. This phenomenon can likely be attributed to the advantages that dairy farms gain by participating in cooperative alliances, through which they are able to acquire and implement advanced technologies and modern production equipment at a significantly lower cost. Such a cost-efficient strategy not only helps conserve valuable production inputs but also establishes a solid foundation to support and facilitate the future expansion and scaling of dairy farming operations.

Table 5. Average treatment effect estimation results.

	Applying Smart Farming Technology	Not Applying Smart Farming Technology	ATT	ATU
Dairy farms that have applied smart farming technology	1.3042	1.1696	0.1345 ***
	(0.0117)	(0.0124)	(0.0168)
Dairy farms that have not applied smart farming technology	1.3098	1.1603		0.1495 ***
	(0.0138)	(0.0125)		(0.0143)

Note: *** is the significant levels at 1%. Clustered standard errors are in parentheses.

presents the projected mean treatment impact of smart farming technology on scale economies in dairy farms. The empirical findings indicate a statistically significant positive average treatment impact of smart farming technology on the scale economies in dairy farms, with a confidence level of 1%. The estimated value for the scale economies in dairy farms utilizing smart farming technology is 1.3042. Under the counterfactual hypothesis, it can be shown that the use of smart farming technology in dairy farms significantly impacts their scale economies, resulting in a decrease to a value of 1.1696 when smart farming technology is not utilized. Moreover, there would be a substantial decrease in the extent of scale economies. dairy farms that do not employ modern technologies. The projected scale economies for dairy farms that do not utilize smart farming technologies is 1.3098. Under the counterfactual hypothesis, it is posited that the adoption of smart farming technology by dairy farms that are currently not utilizing such technology would result in a notable rise in scale economies, namely to a value of 1.1603. Furthermore, it is anticipated that this increase in scale economies will occur at a quick pace. The ATT and ATU values, namely 0.1345 and 0.1495, respectively, suggest that the utilization of smart farming technologies in dairy farms yields more pronounced scale economies. The utilization of smart farming technology in dairy farms, specifically, can lead to a more pronounced development of scale economies. Hypothesis 1 has been confirmed.

4.4. Heterogeneity Analysis

This paper categorizes dairy farms into three distinct scales: small-scale, medium-scale, and large-scale. It investigates the varying effects of smart farming technology on scale economies within these different farming scales. Specifically, small-scale dairy farms are defined as those with a farming scale of fewer than 50 cows, medium-scale dairy farms encompass those with a farming scale of more than 50 cows but less than 500 cows, and large-scale dairy farms include those with a farming scale of over 500 cows. The results of the estimation are presented in

Table 6. Effect of smart farming technology on scale economies for dairy farms of different scales.

Variable	Small Scale	Medium Scale	Large Scale
	Scale	Scale	Scale
Tech	0.0695 **	0.1599 ***	0.2432 ***
	(0.0307)	(0.0238)	(0.0300)
Edu	0.0242	0.0359 ***	−0.0131
	(0.0147)	(0.0101)	(0.0146)
Age	−0.0056 ***	−0.0012	−0.0059 ***
	(0.0016)	(0.0014)	(0.0020)
Cad	0.0614**	0.0405 *	0.3271 ***
	(0.0256)	(0.0229)	(0.0640)
Exper	0.0006	0.0031 ***	0.0047
	(0.0008)	(0.0007)	(0.0033)
Risk	0.0021	−0.0084	−0.0491 *
	(0.0170)	(0.0101)	(0.0259)
Train	0.0416 *	0.0302 **	−0.0316
	(0.0229)	(0.0142)	(0.0713)
Coop	0.0260	0.1002 ***	0.2228 ***
	(0.0383)	(0.0328)	(0.0402)
Urban	0.0179	0.0048	−0.0275
	(0.0231)	(0.0163)	(0.0272)
Income	0.0004	−0.0002	−0.0010
	(0.0005)	(0.0004)	(0.0011)
Labor	0.0179 ***	0.0008	−0.0002
	(0.0059)	(0.0051)	(0.0217)
Constant	0.3763	0.2903 **	0.5646 ***
	(0.2638)	(0.1455)	(0.1752)
R2	0.491	0.620	0.707
Adj. R2	0.438	0.592	0.675

Note: ***, **, and * are the significant levels at 1%, 5%, and 10%, respectively. Clustered standard errors are in parentheses.

. The utilization of smart farming technology in dairy farms of varying sizes, namely small-scale, medium-scale, and large-scale, has a notably positive impact on scale economies. This finding further reinforces the beneficial role of smart farming technology in facilitating scale economies. Additionally, the coefficients representing the influence of smart farming technology in the three categories of dairy farms are 0.0695, 0.1599, and 0.2432, respectively. These coefficients suggest that as the size of dairy farms increases, the scale economies effect facilitated by smart farming technology becomes more pronounced. The coefficients of the effect of smart farming technology on dairy farms were 0.0695, 0.1599, and 0.2432, respectively. A potential explanation for this phenomenon is that smart farming technology can rapidly modify the original factor inputs and production methods employed in larger-scale dairy farms. Consequently, this can lead to a reduction in the average cost associated with operating such large-scale dairy farms, thereby facilitating their expansion and enabling the attainment of scale economies. Conversely, smaller-scale dairy farms may encounter higher costs when attempting to fully leverage smart farming technology due to their limited size. Additionally, the investment required for smart farming technology is comparatively lower for smaller-scale dairy farms. The utilization of smart farming technology in dairy farms incurs higher costs for small-scale operations. This is primarily due to the smaller size of these feedlots and the longer time required to recoup the investment in technology. Consequently, the ability to rapidly expand the scale of small-scale dairy farms within a short timeframe is challenging. As a result, the impact of scale economies, resulting from the adoption of smart farming technology, is comparatively weaker in small-scale dairy farms when compared to their larger counterparts.

4.5. The Effect of Smart Farming Technology on the Carbon Efficiency of Dairy Farms Mediated by Scale Economies

Table 7. The result of smart farming technology on the carbon efficiency of dairy farms mediated by scale economies.

Variable	Regression (1)	Regression (2)	Regression (3)
	CE	Scale	CE
Tech	0.1324 ***	0.1380 ***	0.0701 ***
	(0.0117)	(0.0143)	(0.0113)
Scale			0.4517 ***
			(0.0417)
Edu	0.0258 ***	0.0283 ***	0.0130 ***
	(0.0055)	(0.0068)	(0.0048)
Age	0.0004	0.0065 ***	−0.0005
	(0.0007)	(0.0009)	(0.0007)
Cad	0.0761 ***	0.0957 ***	0.0329 **
	(0.0181)	(0.0221)	(0.0156)
Exper	0.0008 *	0.0016 ***	0.0015 ***
	(0.0005)	(0.0006)	(0.0004)
Risk	−0.0322 ***	−0.0329 ***	−0.0174 ***
	(0.0075)	(0.0092)	(0.0064)
Train	0.0248 **	0.0252 *	0.0362 ***
	(0.0107)	(0.0131)	(0.0090)
Coop	0.0334 **	0.0996 ***	0.0116
	(0.0130)	(0.0159)	(0.0116)
Urban	0.0048	−0.0152	0.0117
	(0.0109)	(0.0133)	(0.0091)
Income	−0.0004 *	0.0003	−0.0005 ***
	(0.0002)	(0.0003)	(0.0002)
Labor	0.0052	0.0120 ***	0.0106 ***
	(0.0034)	(0.0042)	(0.0029)
Constant	0.3510 ***	0.8396 ***	−0.0283
	(0.0835)	(0.1024)	(0.0780)
Observations	278	278	278
R2	0.646	0.480	0.754
Adj. R2	0.631	0.459	0.743

Note: ***, **, and * are the significant levels at 1%, 5%, and 10%, respectively. Clustered standard errors are in parentheses.

displays the projected outcomes pertaining to the influence of smart farming technologies and scale economies on the carbon efficiency of dairy farms. In regression (2), the coefficient of smart farming technology on dairy farms scale economy is estimated to be 0.1380, indicating a statistically significant relationship at the 1% level. This finding suggests that the utilization of smart farming technology in dairy farms can successfully leverage its scale effect and facilitate the development of scale economy. This reaffirms the validity of the association between smart farming technology and the influence on the dairy farms scale economy as established in the previous study. This empirical finding is highly consistent with theoretical expectations, indicating that smart farming technologies effectively overcome internal constraints such as management capacity and asset stock by optimizing the allocation of production factors and reducing marginal production costs. This enables dairy farms to fully leverage economies of scale, achieve continuous expansion of production scale and persistent reduction in unit costs, ultimately facilitating the successful realization of scale economies.

The regression coefficients for smart farming technology and scale economy in regression (3) are 0.0701 and 0.4517, respectively. Both values are statistically significant at the 1% level. This suggests that the utilization of smart farming technology in dairy farms enhances carbon efficiency through the augmentation of scale economies, hence establishing a mediating relationship between smart farming technology and carbon efficiency in dairy farms. These findings further validate the transmission pathway outlined in the theoretical mechanism: smart livestock technologies not only directly enhance carbon efficiency but, more importantly, indirectly amplify this improvement by establishing economies of scale as a mediating channel. Specifically, the attainment of scale economies prompts farms to reduce the intensity of material and energy consumption, thereby increasing factor utilization efficiency. Concurrently, by leveraging diluted fixed costs and deepened specialized division of labor, it creates favorable conditions for the innovation and application of green production technologies. The verification of Hypothesis 2 was successfully conducted.

4.6. Analysis on the Moderating Effect of Government Regulation

Based on theoretical analysis, this study further examines the moderating effect of government regulation on the mediating process through which smart farming technology application affects the carbon efficiency of dairy farms. Specifically, following the logic of the stepwise test method, it verifies whether the moderating effect of government regulation—on the impact of smart farming technology application on economies of scale—is significant. The estimation results are presented in

Table 8. The results of the test of the moderating role of government regulation.

Variable	Regression (4)	Regression (5)	Regression (6)	Regression (7)	Regression (8)
	Carbon Efficiency	Scale Economies	Scale Economies	Scale Economies	Scale Economies
Tech	0.0701 ***	0.0091	−0.0011	0.0935 ***	0.0992 ***
	(0.0113)	(0.0224)	(0.0227)	(0.0171)	(0.0195)
Regula		0.0224 *
		(0.0120)
Tech × Regula		0.1389 ***
		(0.0227)
RR			0.0140
			(0.0102)
Tech × RR			0.1286 ***
			(0.0196)
IR				0.0148
				(0.0105)
Tech × IR				0.0519 ***
				(0.0160)
GR					0.0059
					(0.0105)
Tech × GR					0.0450 **
					(0.0181)
Scale	0.4517 ***
	(0.0417)
Constant	−0.0283	0.9998 ***	0.8370 ***	0.9406 ***	0.9211 ***
	(0.0780)	(0.0964)	(0.0931)	(0.1017)	(0.1062)
Control variables	Control	Control	Control	Control	Control
Observations	278	278	278	278	278
R2	0.7544	0.5696	0.5776	0.5220	0.4984
Adj. R2	0.7432	0.5484	0.5569	0.4984	0.4737

Note: ***, **, and * are the significant levels at 1%, 5%, and 10%, respectively. Clustered standard errors are in parentheses.

. Regression (4) shows the overall regression results with the inclusion of smart farming technology application, economies of scale, and control variables. Regression (5) presents the regression results after adding the interaction term between smart farming technology application and government regulation. Regressions (6) to (8), respectively, report the regression results with the inclusion of interaction terms: smart farming technology application and restrictive regulation, smart farming technology application and incentive regulation, and smart farming technology application and guiding regulation.

In Regression (4), the regression coefficient of economies of scale is 0.4517, which is significant at the 1% level. In Regression (5), the regression coefficient of the interaction term between smart farming technology application and government regulation is 0.1389, also significant at the 1% level. The consistent direction of these regression coefficients indicates that government regulation can positively moderate the effect of smart farming technology application on the economies of scale of dairy farms. The implementation of government regulation encourages dairy farms to further expand the scope and depth of smart farming technology application. This enhances the technology substitution effect of smart farming, thereby allowing dairy farms to lower production costs, achieve optimal scale, and ultimately realize economies of scale.

Government regulation consists of three dimensions: restrictive regulation, incentive regulation, and guiding regulation. On the basis of analyzing the moderating effect of overall government regulation on the mediating process, this study further explores the moderating effects of these three specific dimensions, respectively. Regression (6) shows the results after adding the interaction term between smart farming technology application and restrictive regulation. The interaction term is significantly positive at the 1% level, and its regression coefficient direction is consistent with that of economies of scale in Regression (4). This finding suggests that restrictive regulation acts as a positive moderator, enhancing the effect of smart farming technology on economies of scale. It does so by compelling dairy farms to significantly improve their factor allocation efficiency, which in turn creates the necessary conditions for large-scale operations and facilitates the realization of scale economies.

The estimation results for the moderating effect of incentive regulation on the mediating process are shown in Regression (7). The interaction term between smart farming technology application and incentive regulation in Regression (7) is also significantly positive, with a consistent coefficient direction. This indicates that incentive-based regulation positively moderates the impact of smart farming technology on economies of scale. By bolstering herders’ production and operational confidence, such policies encourage the proactive adoption of smart technologies, thereby facilitating the achievement of scale effects.

The estimation results for the moderating effect of guiding regulation on the mediating process are presented in Regression (8). The regression coefficient of economies of scale in Regression (4) is significantly positive, and the interaction term between smart farming technology application and guiding regulation in Regression (8) is also significantly positive, with a consistent coefficient direction. This shows that guiding regulation positively moderates the positive impact of smart farming technology application on economies of scale. The government’s implementation of guiding regulation helps dairy farms optimize the application methods of smart farming technology, avoid obstacles in smart farming technology application, improve the practical effects of smart farming technology use, and create a more solid foundation for dairy farmers to further expand their breeding. Hypothesis 3 is verified.

4.7. Robustness Tests

4.7.1. Two-Stage Least Squares and Shrinking Tail Treatment

This study employs a two-stage least squares (2SLS) approach to mitigate potential endogeneity in the relationship between smart farming technology and dairy farm economies of scale. We instrument for smart farming technology using “the mean value of the degree of use of smart farming technology in other dairy farms.” This instrument is theoretically relevant and, satisfying the exclusion restriction, helps isolate exogenous variation. Therefore, when dealing with endogenous binary regressors, the 2SLS estimator provides more reliable results than standard ordinary least squares (OLS). The regression analysis was conducted using the two-stage least squares method, and the obtained regression results are presented in

Table 9. The result of two-stage least squares and shrinking tail treatment.

Variable	First-Stage Least Square		Second-Stage Least Square		Shrinking Tail Treatment
	Coefficient	Standard Deviation	Coefficient	Standard Deviation	Coefficient	Standard Deviation
Tech			0.1562 ***	0.0144	0.1380 ***	0.0143
Instrumental variable	1.4084 ***	0.0333
Control variables	Control		Control		Control
F/Wald χ2	1224.23		464.66		22.33
R2	0.8427		0.4769		0.480
Adj R2	0.8362				0.459

Note: *** is the significant level at 1%.

. The regression coefficient for smart farming technology in the second-stage estimation results of the two-stage least squares method is 0.1562. This coefficient is found to be statistically significant at the 1% level, providing further evidence that smart farming technology has a positive impact on scale economies. These results also suggest that the findings of the previous study are robust.

Furthermore, this paper adheres to the principle of avoiding data censorship by trimming the dataset at the 1% and 99% quantiles. This trimming process aims to mitigate the influence of outliers on the estimation outcomes. Subsequently, the trimmed data samples are subjected to regression analysis to assess the resilience of the estimation results. Table 8 displays the estimation outcomes after the implementation of the shrinkage treatment. The findings reveal a statistically significant positive relationship between smart farming technology and scale economies in dairy farms at a significance level of 1%. This suggests that the utilization of smart farming technology in dairy farms can effectively facilitate the achievement of scale economies. Furthermore, the observed impact of smart farming technology aligns with the findings presented in the preceding section, thereby reinforcing the robustness of the estimation results.

4.7.2. Propensity Score Matching Method Test

This study utilizes the propensity score matching technique to examine the resilience of the influence of smart farming technology on the economy of dairy farms at a large scale. Propensity score matching is employed to mitigate observable selection bias in the adoption of smart farming technologies. This method matches each dairy farm that adopts smart farming technology with a non-adopting farm based on their propensity scores. By constructing comparable treatment and control groups with similar observable characteristics, propensity score matching ensures the reliability of estimates for the binary explanatory variable. The outcomes of the propensity score matching analysis are presented in

Table 10. The result of the propensity score matching method.

Matching Method	Variable	Processing Group	Control Group	ATT	Standard Error	T
Nearest neighbor matching	Scale	1.3049	1.1652	0.1398 ***	0.0140	10.00
Caliper matching	Scale	1.2625	1.1805	0.0820 ***	0.0284	2.89
Radius matching	Scale	1.2498	1.1730	0.0768 ***	0.0269	2.85
Kernel matching	Scale	1.3118	1.1661	0.1457 ***	0.0131	11.09

Note: *** is the significant level at 1%.

. The four matching approaches yield consistent findings on the coefficient and direction of the impact of smart farming technology on the economy of dairy farms. These results align with the empirical findings reported in a prior study. The collective findings from the propensity score matching method reveal that the treatment group and control group demonstrate that smart farming technology has a significant impact on improving the scale economy of dairy farms. This reaffirms the reliability of the previous study’s empirical results and confirms the validity of Hypothesis 1.

4.7.3. Bootstrap Method Test

To assess the strength and reliability of the mediating effect of scale economies, this study used the Bootstrap technique to examine the function of scale economies in mediating the influence of smart farming technology on the carbon efficiency of dairy farms. The regression outcomes are presented in

Table 11. The result of the bootstrap method.

Variable	Type of Effect	Coefficient	Standard Deviation	95% Confidence Interval
Scale	Direct	0.4517 ***	0.0644	0.3254	0.5779
	Indirect	0.1315 ***	0.0339	0.0649	0.1980

Note: *** is the significant level at 1%.

. The coefficient for the direct effect of smart farming technology on the carbon efficiency of dairy farms does not include zero within the 95% confidence interval. Similarly, the coefficient for the indirect effect of scale economy also does not include zero within the 95% confidence interval. These findings suggest that the direct impact of smart farming technology on the carbon efficiency of dairy farms is statistically significant, as is the mediating effect of scale economy. In summary, the findings derived from the Bootstrap mediation effect test method exhibit a high degree of resemblance to the outcomes obtained through the stepwise regression method discussed earlier. This further confirms the mediating role of scale economy in the impact of smart farming technology on the carbon efficiency of dairy farms. Hypothesis 2 has been confirmed yet again.

5. Conclusions and Policy Implications

5.1. Conclusions

Based on field survey data collected from dairy farms in Heilongjiang Province between June 2022 and August 2024, this study employs an endogenous switching regression model and a mediation effects analysis to systematically examine how smart farming technologies and economies of scale jointly influence the carbon efficiency of dairy farms. The main findings can be summarized as follows. First, smart farming technologies exhibit a direct and statistically significant effect on the scale economies of dairy farms. To address potential endogeneity concerns, both two-stage least squares and propensity score matching methods were applied. Even after controlling for selection bias using propensity score matching, the positive effect of smart farming technology adoption on farm economic performance remains robust. Second, the economic impacts of smart farming technologies are heterogeneous across farm types, with these effects becoming more pronounced as the operational scale of dairy farms increases. Third, scale economies play a mediating effect in the relationship between technology adoption and carbon efficiency. Specifically, smart farming technologies indirectly improve carbon efficiency by enabling production scale expansion and lowering marginal costs in dairy operations. Finally, Government regulation, particularly in its restrictive, incentive, and guiding dimensions, significantly reinforces the positive impact of smart farming technology on economies of scale. This enhanced scale effect facilitates the improvement of carbon efficiency in dairy farms.

5.2. Policy Implications

Based on the findings presented, this study proposes a series of strategic recommendations. First, it is essential to strengthen institutional support for the adoption of smart farming technologies in dairy production systems. We recommend that agricultural authorities form specialized research and development units focused on advancing agricultural smart technologies, with particular attention to applications within the dairy sector. Such initiatives should aim to increase the relevance and accessibility of smart technologies for dairy farmers. In parallel, policymakers should design a differentiated incentive structure that reflects the varying economic and environmental impacts of technology adoption across small, medium, and large-scale dairy farms. This tailored approach will help accelerate the deployment of newly developed smart technologies in a manner that aligns with the operational realities of individual farms.

Second, training programs related to smart farming technologies must be significantly improved. The effectiveness of these technologies is highly dependent on the quality of technical instruction and user comprehension. Therefore, government agencies and industry partners should collaborate to enhance training curricula, raise awareness among dairy producers regarding the advantages of smart technologies, and support their phased integration into cattle feeding and herd management practices. Efforts should also be made to encourage the informed and efficient application of these tools by farm operators, thereby improving real-world effectiveness and return on investment of technology implementation.

Third, dairy farms should be systematically guided toward adopting smart farming technologies to simultaneously improve economies of scale and carbon efficiency. Given the current challenges related to limited operational scale and suboptimal carbon performance in China’s dairy sector, the government should encourage farm expansion and intensification facilitated by smart technologies. This strategy can help transition dairy production away from resource-intensive practices toward a more streamlined, low-carbon, and economically efficient model, ultimately elevating both productivity and environmental sustainability across the industry.