Assessing Income Heterogeneity from Farmer Participation in Sustainable Management of Forest Health Initiatives

Abstract

Farmers’ participation in sustainable forest management plays a significant role in increasing their income and contributing to the comprehensive advancing of the rural revitalization strategy. This study focuses on farmers living near existing national forest health bases in Inner Mongolia. Using the endogenous switching regression model (ESRM), we empirically examine the income effects of farmers’ participation in sustainable forest management through employment and land leasing. The robustness of the model estimation is tested through various methods, including replacing the dependent variable. Furthermore, heterogeneity analysis is conducted using quantile regression. The results show the following: (1) Participation in sustainable forest management through employment (p < 0.001) and land leasing (p < 0.001) significantly increases annual household income by 4.28% and 1.44%, respectively. The income effect for farmers participating through employment is 2.84% higher than for those participating through land leasing. (2) For farmers who did not participate in sustainable forest management, the counterfactual scenario indicates a reduction in annual household income by 5.87% and 2.55%, respectively, highlighting a greater potential income improvement for non-participating farmers if they were to engage in sustainable forest management. (3) Heterogeneity analysis reveals that the income effects of the two participation forms vary across income levels. Employment participation in forest health bases has a more significant impact on low-income (QR_10) farmers, while land leasing participation has a greater impact on high-income (QR_90) farmers.

1. Introduction

The sustainable management of forest resources is a long-term endeavor. To fully leverage the benefits of forest resources, it is essential to comprehensively promote high-quality forestry development and explore multiple models for sustainable forest management [1]. Forest health is rooted in the forest’s ecological environment and aims to promote public health. It utilizes forest resources and corresponding facilities to support forest recreation, vacation, convalescence, sports, and education. By integrating theories and technologies from multiple disciplines, including medicine, forestry, and tourism, the forest health industry provides services that promote physical and mental well-being, making it an emerging industry.

In 2017, forest health was first introduced in China [2], and its development has been emphasized annually up to 2025. In 2019, China issued guidelines to Promote the Development of the Forest Health Industry, outlining strategies for advancing the sector. Active initiatives have been implemented nationwide to foster the deep integration of forest and health resources. By 2022, forest health bases—including full-area forest health counties, cities, and districts—had received nearly 500 million visitors [3] (A forest health base is a specific area that relies on forest resources, takes health as the theme, and integrates various elements such as natural landscapes, ecological environments, leisure tourism, and health preservation. It provides forest health and wellness services for people.).

Recent research indicates that “forest health base” pilot projects in China primarily concentrate in regions with abundant natural resources but relatively underdeveloped economies, such as the southwest, northeast, and central regions [1]. As a western underdeveloped region with some of the country’s largest forested areas and highest forest coverage rates, Inner Mongolia currently hosts 26 national forest health base pilot units. These bases offer unique forest resource advantages and serve as a “green bank”, providing stable income opportunities for forest-area farmers. Consequently, Inner Mongolia has been recognized as one of China’s top ten national forest health base pilot sites.

In recent years, China has continuously promoted ecological industry poverty alleviation, developed the forest health industry, driven county-level economic development, and increased farmers’ income. By the end of 2020, these efforts had cumulatively helped increase the income of more than 20 million impoverished people, with an average annual income increase of about RMB 5500 (“RMB” stands for “Renminbi”, which is the legal tender of China and is widely used in various transaction and payment scenarios within the Chinese mainland) for farmers relying on forest resources (based on the exchange rate of RMB 1 to USD 0.1381 on 9 March 2025, this is about USD 759.55) [4]. However, with the rapid development of the forest health industry, under the premise of optimizing the allocation of production factors, how should farmers participate in forest health bases? What is the income effect of farmers participating in different forms? Understanding these issues will help understand the income effect of farmers participating in forest health bases in different forms, which is of great significance for fully advancing the rural revitalization strategy.

Previous studies on forest health have primarily focused on various aspects, including current development status [5,6,7], model developments [8,9], path development [10,11,12], industrial integration [13,14,15,16,17], evaluation indicators [18], influencing factors [15,19,20], and the spatial distribution of forest resources [21]. However, as the forest industry continues to develop, it has optimized the social, economic, and ecological benefits of the forestry industry, gradually becoming a major direction and key driver of social progress [22]. With the advancement of the rural revitalization strategy, scholars have linked the future development of forest health with rural revitalization. The forest health industry has significant development potential as a crucial component of the health industry under this strategy [23]. It fully leverages the natural environmental advantages of rural areas, stimulates tourism-driven economic growth [24], and contributes to the implementation of the rural revitalization strategy, ultimately increasing employment and income for farmers [25].

Research has explored various factors influencing the development of forest health, proposing measures and strategies to integrate it with rural revitalization [26]. Starting from the issue of farmers’ lack of subjectivity, previous research has confirmed that forest health and wellness development plays a key role in promoting comprehensive rural revitalization [27]. By constructing an index system of influencing factors, researchers identified key elements affecting farmers’ well-being through the forest health industry [28]. Studies have demonstrated that the forest health industry’s development contributes to farmers’ income, but existing research has not sufficiently examined its impact from the perspective of farming households. How does it affect farmers’ income? What are the heterogeneity effects of this impact? These are critical questions that urgently need to be addressed.

However, the method used to analyze the effects of income on farmers is also crucial. Previous studies have primarily employed ordinary least squares (OLS) regression [29], propensity score matching (PSM) [30], the endogenous switching regression model (ESRM) [31], the Heckman two-stage method [32], and difference-in-differences (DID) [33] to assess income effects. However, to account for the heterogeneity issues arising from omitted and unobservable variables, the endogenous switching regression model (ESRM) is particularly effective, as it directly addresses the challenge. Therefore, the ESRM model is more suitable.

Despite the growing research on the rural revitalization strategy and forest health industry, only a few studies have specifically examined the income effects of forest health based on the perspective of farmers. This study aims to fill the gap by empirically analyzing the income effects of farmers’ participation in forest health bases in different forms using the ESRM. The robustness of the results will be tested by varying the dependent variables, and a heterogeneity analysis will be conducted using quantile regression. Additionally, the study proposes policy recommendations to promote the sustainable development of forest health bases and enhance farmers’ income. This research has significant implications for the coordinated advancement of high-quality forest sustainability and rural revitalization strategies.

2. Theoretical Analysis and Research Hypotheses

2.1. Impact Mechanism of Sustainable Management of Forest Health Bases on Farmers’ Income

Farmers’ participation in the sustainable management of forest health bases affects their income by reorganizing production factors, primarily focusing on land and labor resources [34].

2.1.1. Transformation of Land Use

Traditionally, farmers use their land for crop cultivation or tree planting to generate agricultural or forestry income [35]. However, with the active promotion of forest health base pilot projects and the rapid development of the forest health industry, the construction of forest health bases has been continuously expanding. According to the natural level forest health base requirements, these bases must have a total forest area of no less than 1000 hectares, a forest coverage rate exceeding 50%, and well-developed supporting infrastructure and good transportation conditions. Consequently, establishing forest health bases necessitates the conversion of farmland and forestland from their original uses. Farmers in the surrounding can lease their land to the forest health base and earn rental income.

2.1.2. Reallocation of Labor Resources

Effective labor allocation is a key factor influencing farmers’ household income. Previous studies have shown that forest health bases establish close economic linkages with local village farmers, creating employment opportunities for surplus labor [36]. One of the key drivers of high-quality development in forest health bases is the availability and quality of labor [37]. However, based on the smallholder rationality hypothesis, farmers, as rational economic agents, seek maximum household income. Therefore, how does choosing to work at a forest health base or leasing land to a forest health base affect farmers’ household income? This question requires further discussion, as its impact is closely tied to the farmers’ personal characteristics, household attributes, and regional factors.

2.2. Income Effects of Farmers’ Participation in Sustainable Management of Forest Health Bases

The forest health industry plays an important role in rural economic prosperity, the comprehensive implementation of the rural revitalization strategy, and the creation of vibrant rural communities [38]. Farmers, as the primary drivers of rural economic development, significantly influence the overall economic landscape of the countryside. The development of the forest health industry contributes significantly to increasing farmers’ income [27,28].

Farmers can engage in the sustainable management of forest health bases either by leasing their land to generate rental income or by working as laborers. Research indicates that non-agricultural employment positively affects household income [39]. By leveraging forest health bases, farmers can achieve effective transformation and the rational allocation of production factors, thereby increasing household income. Based on this, the following research hypotheses are proposed:

H1: 

Participation in the sustainable management of forest health bases has a positive impact on increasing farmers’ income.

H1a: 

Employment positively impacts farmers’ income.

H1b: 

Land leasing positively impacts farmers’ income.

2.3. Impact of Different Forms of Farmers’ Participation

According to the Department for International Development (DFID) sustainable livelihood analysis framework, differences in resource endowments, livelihood environments, and strategies lead to varying income effects depending on farmers’ livelihood choices [40]. Additionally, household income levels influence livelihood strategy preferences [41].

For low-income farmers, a limited capacity to diversify land often results in relatively stable but lower income levels [42]. However, rising household expenses, such as living costs, education, healthcare, and asset investments, necessitate adjustments in their livelihood strategies to improve household income. This is particularly relevant for farmers who choose to remain in their local communities rather than migrate for work. The expansion of forest-based industries, such as forest tourism and forest health initiatives, creates additional employment opportunities for these farmers [43]. Transitioning to tertiary industries, which typically offer higher labor productivity, can lead to substantial income growth compared to relying solely on agricultural employment [44].

For instance, participating in the sustainable operation of forest health bases through employment can significantly increase farmers’ income. Field interviews revealed that forest health bases often lease or purchase farmland or forest land from farmers at reasonable rents or through one-time compensation, ensuring the sustainable management of these bases. Diversifying land use, rather than relying on a single mode of land use, can enhance household income [45]. Under similar conditions, low-income farmers are more likely to benefit from alternative land use strategies that yield higher income.

In contrast, high-income farmers, who have greater advantages in livelihood strategy selection, are less dependent on income from employment or land resources [46]. These farmers tend to focus on higher-income livelihood strategies while maintaining stable income from land assets. Based on this theoretical analysis of income groups, the following hypotheses are proposed:

H2: 

The impact of employment varies across income levels, with a greater effect on low-income farmers.

H3: 

The impact of land leasing varies across income levels, with a greater effect on high-income farmers.

Participation in the sustainable management of forest health bases leads to different land use and labor allocation strategies among farmers, resulting in heterogeneous income effects across income levels. Based on this, a theoretical analysis framework for farmers’ participation in the sustainable management of forest health bases is constructed (Figure 1).

3. Data and Model

3.1. Data

The data used in this study were obtained from field surveys conducted between May and September 2024 among farmers living near forest health bases in various leagues and cities of Inner Mongolia. The survey targeted farmers who had not chosen to remain in their hometowns rather than seek work elsewhere or lease their land for purposes other than forest health bases. A combination of stratified and random sampling methods was employed. The data analysis was conducted using Stata version 22.0.

First, based on the number of forest health bases and the level of the sustainable management of these bases in Inner Mongolia (Figure 2), four cities were selected: Hulunbuir, which has the highest forest health bases; Chifeng, the second highest; and Tongliao and Ulanqab, which have the most effective sustainable management of forest health bases. Next, 1–3 national-level forest health demonstration bases in each city were selected according to their approval batches, operational status, and base construction conditions. Surrounding farmers were then randomly selected for in-depth income-related interviews using a combination of household questionnaires and interviews. A total of 506 questionnaires were distributed. After excluding those with contradictory information or missing key variable data, 458 valid questionnaires were obtained, resulting in an effective response rate of 90.51%.

In the surveyed regions, the overall sample participation rate for farmers working in forest health bases was 57.86%, higher than the 31.66% participation rate of farmers who leased their land to forest health bases. Analyzing participation by region, Chifeng had the highest participation rate for farmers working in the sustainable management of forest health bases, at 87.20%, followed by Tongliao (62.50%) and Ulanqab (58.42%). Hulunbuir had the lowest participation rate at only 27.21%, primarily because most forest health bases in Hulunbuir are operated by state-owned forest farms, where former employees comprise the majority of the workforce, resulting in a lower proportion of participation among farmers. Regarding land leasing, participation rates varied significantly across regions: Hulunbuir had the highest participation rate at 41.91%, as farmers in this region generally own larger forestland areas than those in central and western regions, increasing the likelihood of leasing land to forest health bases. In contrast, Chifeng had the lowest participation rate for land leasing, at only 18.40% (

Table 1. Participation of farmers in sustainable management of forest health bases in surveyed regions.

Region	Sample Size	Forms of Participation in Sustainable Management of Forest Health Bases
		Employment (Participation Rate %)	Non-Employment (Participation Rate %)	Land Leasing (Participation Rate %)	Non-Leasing (Participation Rate %)
Total	458	265 (57.86)	193 (42.14)	145 (31.66)	313 (68.34)
Eastern: Hulunbuir	136	37 (27.21)	99 (72.79)	57 (41.91)	79 (58.09)
Central: Chifeng	125	109 (87.20)	16 (12.80)	23 (18.40)	102 (81.60)
Central: Tongliao	96	60 (62.50)	36 (37.50)	30 (31.25)	66 (68.75)
Western: Ulanqab	101	59 (58.42)	42 (41.58)	35 (34.65)	66 (65.35)

).

3.2. Variable Selection and Statistical Analysis

The variables in this study are categorized into four groups: dependent variables, core explanatory variables, identification variables, and control variables. The definitions, assignments, and descriptive analysis results of each category are shown in

Table 2. Explanation and description of variables.

Variable Type	Variable Name	Description	Mean	Standard Deviation
Dependent Variable	Annual Household Income	Average annual household income over the past 3 years	10.9896	0.0298
Participation in Forest Health Base	Participation in Sustainable Management of Forest Health Base	Participation in employment: Yes = 1; No = 0	0.5786	0.0231
		Participation in land leasing: Yes = 1; No = 0	0.3166	0.0218
Identification Variable	Distance to Forest Health Base	Distance between the household and the forest health base	6.4559	0.5724
Individual Characteristics	Gender	Male = 1; Female = 2	1.4127	0.0230
	Age	25 years and below = 1; 26–35 years = 2; 36–45 years = 3; 46–55 years = 4; Over 55 years = 5	3.7450	0.0492
	Education Level	No schooling = 1; Primary school = 2; Middle school = 3; High school = 4; College and above = 5	3.4301	0.0508
	Personal Health Status	Very healthy = 1; Healthy = 2; Average = 3; Unhealthy = 4; Very unhealthy = 5	2.0459	0.0399
	Marital Status	Married = 1; Unmarried = 2; Divorced = 3	1.1572	0.0226
Household Characteristics	Family Members’ Health Status	Very low = 1; Low = 2; Average = 3; High = 4; Very high = 5	2.1376	0.0473
	Number of Family Members	Actual number of family members	3.5633	0.0542
	Annual Household Expenditure	Average annual household expenditure over the past 3 years	10.0770	0.0278
	Household Fixed Asset Investment	Actual annual investment in fixed assets	8.0716	0.0352
	Per Capita Arable/Forest Land Area	Total arable or forest land area/number of family members	4.5253	0.1763
	Household Social Relations Status	Very weak = 1; Weak = 2; Average = 3; Strong = 4; Very strong = 5	3.3537	0.0339
Type Dummy Variable	Agriculture-oriented	Agriculture-oriented = 1; Others = 0	0.3690	0.0199
	Forestry-dependent	Forestry-dependent = 1; Others = 0	0.3668	0.0232
	Non-agricultural Non-forestry-oriented	Non-agricultural non-forestry-oriented = 1; Others = 0	0.2641	0.0229
Regional Dummy Variable	Hulunbuir	Hulunbuir = 1; Others = 0	0.2969	0.0214
	Tongliao	Tongliao = 1; Others = 0	0.2729	0.0208
	Chifeng	Chifeng = 1; Others = 0	0.2096	0.0190
	Ulanqab	Ulanqab = 1; Others = 0	0.2205	0.0194

Note: The table also shows the mean and standard deviation of each variable.

.

3.2.1. Dependent Variable

The dependent variable in this study is farmers’ average annual household income. The logarithm of the average annual income over the past three years is used as the dependent variable.

3.2.2. Core Explanatory Variables

The core explanatory variables indicate that farmers participate in the sustainable management of forest health bases. Based on the allocation of production factors and the actual condition in the survey area, participation through employment and land leasing are selected as core explanatory variables. If a farmer participates in sustainable management through employment or land leasing, the variable is assigned a value of 1; otherwise, it is assigned a value of 0.

3.2.3. Identification Variable

At least one control variable is not included in the outcome equation to ensure the identifiability of the decision and outcome equations. This identification variable should not directly affect the dependent variable but must directly influence the core explanatory variables. Based on these criteria, the distance between a farmer’s household and the forest health base is selected as the identification variable. For farmers participating in sustainable management through employment and land leasing, proximity to the forest health base is a key consideration. However, distance does not directly impact a household’s annual income, making it a strong exogenous variable [47].

3.2.4. Control Variables

This selection of control variables is based on previous research [48,49] and is adapted to the actual conditions of the survey area. Four categories of factors that may affect farmers’ participation in the sustainable management of forest health bases and income effects are included as control variables: personal characteristics, household characteristics, type characteristics, and regional characteristics. For type characteristics, farmers are classified into three categories based on a comparison of agriculture and forestry income in the survey area: an agriculture-oriented type, forestry-dependent type, and non-agriculture non-forestry-oriented type. Different types of farmers exhibit varying impacts on their annual agricultural income [50]. For regional characteristics, the sample data are divided into four regions: Hulunbuir, Tongliao, Chifeng, and Ulanqab. The farmer type and regional characteristics are treated as dummy variables. The farmer-type dummy variable is used to explore differences in participation and its impact on household annual agricultural income. The regional characteristic dummy variable is used to explore the impact of farmers’ participation in forest health bases across different geographic regions.

The average annual household income (logarithm) for farmers participating through employment is 10.8903, compared to 10.8471 for non-participating households, indicating that participating farmers earn 0.0498 more than non-participating farmers. The average annual household income (logarithm) for farmers participating through land leasing is 10.8796, compared to 10.8471 for non-participating households, showing income differences of 0.0325 in favor of participating farmers. These results suggest a significant income disparity between participating and non-participating farmers, with employment participation leading to a greater household income increase than land leasing participation (

Table 3. Descriptive comparative analysis of annual household income between households participating in sustainable management of forest health base and those not participating.

Variable		Participating Households		Non-Participating Households		Mean Difference
		Mean	Standard Deviation	Mean	Standard Deviation
Employment	Average household annual income	10.8903	0.0328	10.8405	0.0388	0.0498
Land Leasing		10.8796	0.0313	10.8471	0.0416	0.0325

Note: Mean difference is the difference between the mean values of participating households and those of non-participating households.

).

3.3. Model

The most common methods used for analyzing income effects include instrumental variables (IV), propensity score matching (PSM), and the endogenous switching model (ESM). These methods aim to address sample selection bias and endogeneity issues. However, each method has its limitations; the IV method does not consider the heterogeneity among income effects, and while the PSM method can address the heterogeneity problem of observable variables, it cannot solve the heterogeneity problem of unobservable variables. The ESM can overcome the impact of both observable and unobservable variables on sample selection bias. Additionally, it allows for the estimation of both decision equation and “counterfactual” equation for income levels, providing a more intuitive assessment of how participation in the sustainable management of forest health bases influences farmers’ income [51]. Given these advantages, this study employee the ESRM, proposed by Lokshin et al. [52], to estimate the income effects of farmers’ participation in forest health bases.

Based on this, a farmer income model is constructed:

$${\mathrm{Y}}_{\mathrm{i}}=\mathsf{\alpha }{\mathrm{X}}_{\mathrm{i}}+\mathsf{\eta }{\mathrm{P}}_{\mathrm{i}\mathrm{j}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(1)

In Equation (1), ${\mathrm{Y}}_{\mathrm{i}}$ represents the annual household income of farmers, ${\mathrm{X}}_{\mathrm{i}}$ represents the personal characteristics, household characteristics, type characteristics, and regional characteristics that influence farmers’ participation in the sustainable management of forest health bases, and ${\mathrm{P}}_{\mathrm{i}\mathrm{j}}$ represents whether farmer i participates in the sustainable management of forest health bases with decision behavior j (j = 0–1). If a farmer participates in the sustainable management of forest health bases, ${\mathrm{P}}_{\mathrm{i}\mathrm{j}}=1$; if not, ${\mathrm{P}}_{\mathrm{i}\mathrm{j}}$ = 0. $\mathsf{\alpha }$ is the parameter to be estimated, describing the impact of all control variables on income effects, and $\mathsf{\eta }$ is the parameter to be estimated, describing the income effect of farmers’ participation in the sustainable management of forest health based on their annual household income. ${\epsilon }_i$ is the random disturbance term.

According to the estimation logic of the ESRM, the income effect regression model is constructed in two steps:

Step 1: Use full information maximum likelihood estimation (FIML) to perform a regression analysis on the decision behavior of farmers participating in the sustainable management of forest health bases and construct the decision equation and outcome equation:

$${\mathrm{P}}_{\mathrm{i}}={\mathsf{\gamma }\mathrm{Z}}_{\mathrm{i}\mathrm{j}}+{\mathsf{\mu }}_{\mathrm{i}}$$

(2)

$${\mathrm{Y}}_{\mathrm{i}0}={\mathsf{\beta }}_{\mathrm{i}0}{\mathrm{X}}_{\mathrm{i}\mathrm{j}}+{\mathsf{\epsilon }}_{\mathrm{i}0}{\mathrm{P}}_{\mathrm{i}\mathrm{j}}=0$$

(3)

$${\mathrm{Y}}_{\mathrm{i}\mathrm{n}}={\mathsf{\beta }}_{\mathrm{i}1}{\mathrm{X}}_{\mathrm{i}\mathrm{j}}+{\mathsf{\epsilon }}_{\mathrm{i}1}{\mathrm{P}}_{\mathrm{i}\mathrm{j}}=1$$

(4)

In Equation (2), ${\mathrm{Z}}_{\mathrm{i}\mathrm{j}}$ is a vector of personal, household, type, and regional characteristics that influence farmers’ participation in the sustainable management of forest health bases. $\mathsf{\gamma }$ is the parameter vector to be estimated, and ${\mathsf{\mu }}_{\mathrm{i}}$ is the error term in the farmers’ participation decision equation.

In Equations (3) and (4), ${\mathrm{Y}}_{\mathrm{i}0}$ represents the annual household income of farmers who do not participate in the sustainable management of forest health bases, and ${\mathrm{Y}}_{\mathrm{i}1}$ represents the annual household income of farmers who participate in the sustainable management of forest health bases. ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}$ represents the personal characteristics, household characteristics, type characteristics, and regional characteristics that influence farmers’ participation in the sustainable management of forest health bases. ${\mathsf{\beta }}_{\mathrm{i}0}$ and ${\mathsf{\beta }}_{\mathrm{i}1}$ are the parameters to be estimated, and ${\epsilon }_{i0}$ and ${\epsilon }_{i1}$ are the error terms in the outcome equation.

Step 2: To solve the estimation bias problem caused by the introduction of unobservable factors, use the inverse Mills ratio and ${\lambda }_{i1}$ calculated in the first step, and their corresponding covariances ${\mathsf{\sigma }}_{\mathsf{\mu }1}=\mathrm{c}\mathrm{o}\mathrm{v}\left({\mathsf{\mu }}_{\mathrm{i}},{\mathsf{\epsilon }}_{\mathrm{i}1}\right)$ and ${\mathsf{\sigma }}_{\mathsf{\mu }0}=\mathrm{c}\mathrm{o}\mathrm{v}\left({\mathsf{\mu }}_{\mathrm{i}},{\mathsf{\epsilon }}_{\mathrm{i}0}\right)$, and substitute them into the above Equations (3) and (4) to obtain the new equations:

$${\mathrm{Y}}_{\mathrm{i}0}={\mathsf{\beta }}_{\mathrm{i}0}{\mathrm{X}}_{\mathrm{i}}+{\mathsf{\sigma }}_{\mathsf{\mu }0}{\mathsf{\lambda }}_{\mathrm{i}0}+{\mathsf{\zeta }}_{\mathrm{i}0}{\mathrm{P}}_{\mathrm{i}\mathrm{j}}=0$$

(5)

$${\mathrm{Y}}_{\mathrm{i}1}={\mathsf{\beta }}_{\mathrm{i}1}{\mathrm{X}}_{\mathrm{i}}+{\mathsf{\sigma }}_{\mathsf{\mu }1}{\mathsf{\lambda }}_{\mathrm{i}1}+{\mathsf{\zeta }}_{\mathrm{i}1}{\mathrm{P}}_{\mathrm{i}\mathrm{j}}=1$$

(6)

Here, ${\mathsf{\lambda }}_{\mathrm{i}0}$ and ${\mathsf{\lambda }}_{\mathrm{i}1}$, respectively, control the selective bias problem caused by unobservable variables in the case of farmers not participating in or participating in the sustainable management of forest health bases and ${\mathsf{\zeta }}_{\mathrm{i}1}$ and ${\mathsf{\zeta }}_{\mathrm{i}0}$ are the zero-mean hypothesis. Using full information maximum likelihood estimation (FIML), the equations are jointly estimated to obtain the joint model:

$${\mathrm{Y}}_{\mathrm{i}}\left\{\begin{array}{c}{\mathrm{Y}}_{\mathrm{i}0}={\mathsf{\beta }}_{\mathrm{i}0}{\mathrm{X}}_{\mathrm{i}}+{\mathsf{\sigma }}_{\mathsf{\mu }0}{\mathsf{\lambda }}_{\mathrm{i}0}+{\mathsf{\zeta }}_{\mathrm{i}0} {\mathrm{P}}_{\mathrm{i}\mathrm{j}}=0\\ {\mathrm{Y}}_{\mathrm{i}1}={\mathsf{\beta }}_{\mathrm{i}1}{\mathrm{X}}_{\mathrm{i}}+{\mathsf{\sigma }}_{\mathsf{\mu }1}{\mathsf{\lambda }}_{\mathrm{i}1}+{\mathsf{\zeta }}_{\mathrm{i}1} {\mathrm{P}}_{\mathrm{i}\mathrm{j}}=1\\ {\mathrm{P}}_{\mathrm{i}}={\mathsf{\gamma }\mathrm{Z}}_{\mathrm{i}\mathrm{j}}+{\mathsf{\mu }}_{\mathrm{i}}\end{array}\right.$$

(7)

Finally, the average treatment effect is calculated. The average treatment effect (ATT) of farmers participating in the sustainable management of forest health bases and the average treatment effect (ATU) of farmers not participating are calculated as follows:

$$\mathrm{A}\mathrm{T}\mathrm{T}=\mathsf{{\rm E}}\left[\left.{\mathrm{Y}}_{\mathrm{i}1}\right|{\mathrm{P}}_{\mathrm{i}}=1\right]-\mathsf{{\rm E}}\left[\left.{\mathrm{Y}}_{\mathrm{i}0}\right|{\mathrm{P}}_{\mathrm{i}\mathrm{j}}=1\right]={\mathrm{X}}_{\mathrm{i}1}\left({\mathsf{\beta }}_{\mathrm{i}1}-{\mathsf{\beta }}_{\mathrm{i}0}\right)+{\mathsf{\lambda }}_{\mathrm{i}1}\left({\mathsf{\sigma }}_{\mathsf{\mu }1}-{\mathsf{\sigma }}_{\mathsf{\mu }0}\right)$$

(8)

$$\mathrm{A}\mathrm{T}\mathrm{U}=\mathsf{{\rm E}}\left[\left.{\mathrm{Y}}_{\mathrm{i}0}\right|{\mathrm{P}}_{\mathrm{i}}=0\right]-\mathsf{{\rm E}}\left[\left.{\mathrm{Y}}_{\mathrm{i}1}\right|{\mathrm{P}}_{\mathrm{i}\mathrm{j}}=0\right]={\mathrm{X}}_{\mathrm{i}0}\left({\mathsf{\beta }}_{\mathrm{i}0}-{\mathsf{\beta }}_{\mathrm{i}1}\right)+{\mathsf{\lambda }}_{\mathrm{i}0}\left({\mathsf{\sigma }}_{\mathsf{\mu }0}-{\mathsf{\sigma }}_{\mathsf{\mu }1}\right)$$

(9)

4. Empirical Analysis

The estimation results of the ESRM for the impact of farmers’ participation in the sustainable management of forest health based on their income are shown in Section 4.3. Model 1 uses employment in forest health bases in the decision equation as the dependent variable and annual household income as the outcome equation’s the dependent variable. Model 2 uses land leasing participation in the sustainable management of forest health bases as the decision equation’s dependent variable and annual household income as the outcome equation’s dependent variable. The results indicate that both Model 1 and Model 2 reject the null hypothesis at the 1% significance level, as confirmed by the Wald test. The log-likelihood values for Models 1 (−364.5145) and 2 (−412.4019), indicate good model fit. In Model 1, lns0, lns1, r0, and r1 are significantly different from zero at the 1% significance level, while in Model 2, lns0, lns1, and r1 are significantly different from zero at the 1% significance level. These findings suggest that unobservable variables affect farmers’ participation in the sustainable management of forest health bases and their annual household income, necessitating estimation for selection bias. The endogeneity test results also indicate the presence of endogeneity in variables affecting annual household income. Therefore, using the ESRM for estimation is appropriate.

4.1. Factors Influencing Farmers’ Participation in Sustainable Management of Forest Health Bases

The estimation results of the decision equations in Models 1 and 2, as shown in

Table 4. ESRM estimation results of the impact of farmers’ participation in sustainable management of forest health bases on annual household income.

Variables	Model 1 (Employment, n = 458)			Model 2 (Land Leasing, n = 458)
	Decision Equation	Result Equation		Decision Equation	Result Equation
		Participation	Non-Participation		Participation	Non-Participation
Gender	0.4924 *** (0.1673)	−0.2112 *** (0.0489)	0.0948 (0.0717)	0.0279 (0.0672)	−0.0148 (0.0653)	−0.0818 (0.0553)
Age	−0.1715 * (0.0942)	0.0596 * (0.0285)	−0.1046 ** (0.0424)	0.0586 (0.1319)	0.0235 (0.0371)	0.0185 (0.0326)
Education Level	−0.4782 *** (0.0917)	0.1381 *** (0.0310)	−0.0250 (0.0426)	0.0175 (0.0189)	0.0379 (0.0370)	0.0541 * (0.0310)
Personal Health Status	−0.2175 ** (0.1097)	0.1383 *** (0.0477)	−0.0471 (0.0526)	−0.1586 (0.1249)	0.0294 (0.0409)	−0.0127 (0.0378)
Marital Status	−0.0067 (0.1753)	0.0126 (0.0330)	0.0354 (0.0819)	−0.0681 (0.0709)	0.0655 (0.0686)	−0.0175 (0.0584)
Family Members’ Health Status	−0.2284 *** (0.0742)	0.0005 (0.0216)	−0.2287 *** (0.0518)	−0.0015 (0.0586)	−0.0231 (0.0264)	−0.0269 (0.0288)
Number of Family Members	0.6081 ** (0.2749)	0.1837 ** (0.0747)	0.1016 (0.1263)	0.0264 (0.2005)	0.0042 (0.0977)	0.0010 (0.0865)
Annual Household Expenditure	0.3617 ** (0.1578)	0.3406 *** (0.0499)	0.6632 *** (0.0739)	0.0388 *** (0.0836)	0.0949 (0.0772)	0.3699 *** (0.0553)
Household Fixed Asset Investment	0.1869 * (0.1111)	0.0511 * (0.0340)	0.2346 *** (0.0487)	0.3621 (0.1359)	0.1025 ** (0.0432)	0.0535 (0.0382)
Per Capita Arable/Forest Land Area	0.08504 *** (0.0231)	0.0308 *** (0.0075)	0.0297 *** (0.0113)	0.0872 (0.0873)	0.0092 (0.0072)	0.0114 (0.0087)
Household Social Relations Status	−0.0112 (0.1174)	−0.0090 (0.0340)	−0.2079 *** (0.0555)	−0.0197 (0.0918)	−0.0864 (0.0443)	−0.0659 * (0.0399)
Agriculture-oriented	−1.305 *** (0.2804)	0.0803 (0.0674)	0.1239 (0.1445)	0.2130 (0.1993)	0.1061 (0.0935)	0.3055 *** (0.0842)
Forestry-dependent	−1.6304 *** (0.2931)	0.2517 *** (0.0888)	0.0058 (0.1529)	−0.6043 *** (0.1973)	−0.0693 (0.1041)	0.3499 *** (0.0984)
Hulunbuir	−1.0090 *** (0.2635)	0.0358 (0.0973)	−0.2305 ** (0.1123)	−0.3044 (0.2058)	0.1613 * (0.0988)	−0.1488 (0.0996)
Chifeng	0.3800 (0.2497)	−0.2399 *** (0.0684)	0.1505 (0.1310)	−0.3424 *** (0.2283)	−0.1734 (0.1140)	−0.0689 (0.0882)
Ulanqab	−0.7915 *** (0.2559)	−0.0609 (0.0767)	−0.5335 *** (0.1163)	0.2130 (0.1993)	−0.0114 (0.0948)	−0.1628 *** (0.0921)
Distance to Forest Health Base	0.2256 *** (0.0694)	-	-	0.0913 *** (0.0357)	-	-
Constant	−2.1658 (2.0724)	6.4199 *** (0.6380)	4.1558 *** (0.8804)	−4.352 *** (1.6312)	9.3166 *** (1.0072)	6.8004 *** (0.7018)
lns1	-	−0.8251 *** (0.0805)	-	-	−0.6942 *** (0.0451)	-
r1	-	0.8187 *** (0.2800)	-	-	2.7632 *** (0.2781)	-
lns0	-	-	−1.0490 *** (0.0519)	-		−1.1413 (0.1016)
r0	-	-	−0.4792 *** (0.1863)	-		0.3027 *** (0.4100)
Log likelihood	−364.5146			−397.1396
Wald chi2	319.9700 ***			28.1700 ***

Note: * represents significant at the 10% level, ** represents significant at the 5% level, and *** represents significant at the 1% level. The non-agriculture non-forestry-oriented type is the reference group for the farmer-type dummy variable, and Tongliao City is the reference group for the regional characteristic dummy variable. lns0 and lns1 are the square root of the residual variance of the decision equation, and the result equation, r0 and r1 are the residual correlation coefficient. Robust standard errors are in parentheses.

, indicate that the factors influencing participation in the sustainable management of forests differ based on the participation form. In Model 1, the distance from the farmer’s household to the forest health base significantly positively impacts the decision to participate through employment at the 1% significance level. Among personal characteristics, gender, age, education level, and personal health status significantly negatively impact the decision to participate through employment. Among household characteristics, the health status of household members and per capita arable land area have a significant negative impact at the 1% significance level. In contrast, the number of household members, annual household expenditure, and fixed asset investment have a significant positive impact. The results for type characteristics and regional characteristics indicate that different types of farmers and regions have varying impacts on participation behavior, consistent with the findings of previous studies [45]. In Model 2, the distance from the farmer’s household to the forest health base significantly positively impacts the decision to participate through land leasing at the 1% significance level. Annual household expenditure has a significant positive impact. Among type characteristics, the forestry-dependent type has a significantly greater positive impact on the decision to participate through land leasing, consistent with the actual conditions in the survey area. Among regional characteristics, the impact on participation behavior varies significantly by region, with Tongliao having the most significant positive impact.

4.2. Factors Influencing Annual Household Income

From the estimation results of the outcome equations in Models 1 and 2, shown in Table 4, the factors influencing annual household income differ significantly between participating and non-participating farmers in the sustainable management of forest health bases. Among personal characteristics, gender significantly impacts the income of farmers who participate through employment but has no significant effect on the income of non-participating farmers or those who participate through land leasing. Age has a significant impact on the income of both participating and non-participating farmers through employment. Education level significantly affects the income of farmers who participate through employment and land leasing, while personal health status has a significant positive impact on the income of farmers who participate through employment.

Among household characteristics, the health status of household members and annual household expenditure have a significant negative impact on the income of non-participating farmers. The number of household members and per capita arable/forest land area significantly impact the income of farmers who participate through employment, while fixed asset investment has a positive impact on the income of participating farmers.

Among farmer types, the impact on annual household income varies significantly, with forestry-dependent farmers having significant positive impact on the income of farmers who participate through employment, which is consistent with conditions observed in the survey area. Among regional characteristics, the impact on annual household income varies significantly by region. For farmers participating in the sustainable management of forest health bases, the impact on income is most significant in Chifeng and Hulunbuir, compared to Tongliao as the reference group. In contrast, for non-participating farmers, the impact on income is most significant in Ulanqab.

4.3. Average Treatment Effect of Farmers’ Participation in Sustainable Management of Forest Health Bases on Annual Household Income

The average treatment effects (ATT) for farmers participating through employment and land leasing are 0.5177 and 0.1967, respectively, both significant at the 1% level. This indicates that employment and land leasing participation significantly increase annual household income. In terms of the rate of change, the income effect for farmers participating through employment is 2.84 percentage points higher than for those participating through land leasing (

Table 5. Average treatment effect of participation in sustainable management of forest health bases on annual household income.

		Engagement		Treatment Effect		Change Rate
		Participation	Non-Participation	ATT	ATU
Employment	Participating Households	11.6030	11.0853	0.5177 ***	-	4.28%
	Non-participating Households	11.6033	10.8564	-	0.7469 ***	5.87%
Land Leasing	Participating Households	11.2599	11.0632	0.1967 ***	-	1.44%
	Non-participating Households	11.4852	10.6483	-	0.8369 ***	2.55%

Note: *** represents significant at the 1% level. The rate of change refers to the income change brought about by participation in the sustainable management of forest health bases compared to non-participating farmers.

).

For non-participating farmers, the average treatment effects (ATU) are 0.7469 and 0.8369, respectively. Compared to the counterfactual scenario, non-participating farmers’ annual household income is reduced by 5.87% and 2.55%, respectively. This suggests that participation in the sustainable management of forest health bases through employment and land leasing can significantly increase annual household income, with the income effect being more pronounced for non-participating farmers after participation. This suggests that participation in the sustainable management of forest health bases through employment and land leasing increases the stability of household income for surrounding farmers. At the same time, with the rapid development of the forest health industry, the demand for labor and land in forest health bases has expanded, providing more opportunities for farmers to engage in employment and land leasing. This validates Hypotheses H1a and H1b.

4.4. Robustness Check

To further test the estimation results, robustness checks were conducted by replacing estimation methods and the dependent variable.

4.4.1. Replacing Estimation Methods

Table 6. Comparison of model estimation results using different estimation methods.

Method		ESRM	PSM	OLS
Employment	ATT	0.5177 ***	0.1955 ***	-
	ATU	0.7469 ***	0.0058 ***	-
	Coefficient	-	-	0.2641 ***
Land Leasing	ATT	0.1967 ***	0.5584 ***	-
	ATU	0.8369 ***	0.5402 ***	-
	Coefficient	-	-	0.5552 ***

Note: *** represents significant at the 1% level. PSM uses local linear regression matching.

presents the results of the ESRM, PSM, and OLS regression methods for model estimation. The average treatment effects estimated by PSM and the endogenous switching regression model (ESRM) show that the T-tests are significant at the 1% level. Both estimation methods indicate that participation in the sustainable management of forest health bases increases farmers’ annual household income. However, for farmers participating through employment, the ESRM estimation coefficient is significantly higher than the PSM estimation coefficient, whereas the average treatment effect (ATT) for farmers participating through land leasing is lower in the ESRM estimation compared to the PSM estimation. In contrast, in the OLS regression, participation through employment and land leasing has significant positive impacts on farmers’ annual household income. This further demonstrates that the income effect of participation in the sustainable management of forest health bases through employment and land leasing is significant. Therefore, the robustness of the income effect of farmers’ participation in forest health bases is confirmed under various estimation methods. However, the ESRM was chosen to address the endogeneity and selection bias issues for more effective estimation.

4.4.2. Replacing the Dependent Variable

Following previous research on replacing dependent variables [37,43], this study substitutes agricultural income with per mu household income and the dependent variable and re-estimates the ESRM. The results (

Table 7. Average treatment effect of participation in sustainable management of forest health bases on per mu household income.

		Engagement		Treatment Effect		Change Rate
		Participation	Non-Participation	ATT	ATU
Employment	Participating Households	9.8456	8.7234	1.1221 ***	-	2.55%
	Non-participating Households	9.6935	8.6135	-	1.0800 ***	2.82%
Land Leasing	Participating Households	9.6638	7.5726	2.0911 ***	-	3.46%
	Non-participating Households	9.6642	8.2352	-	1.4290 ***	2.08%

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

) show that the average treatment effects of participation in the sustainable management of forest health based on per mu household income through employment and land leasing are significant at the 1% level, confirming the robustness of the empirical results.

4.5. Heterogeneity Analysis

The effects of different participation forms, farmer types, and regions on farmers’ annual household income differ significantly. To further analyze the varying effect of participation forms on farmers’ annual household income, and to compare the heterogeneity in the impact of different farmer types and regions, quantile regression (QR), as proposed by Koenker and Bassett [53], was used. This method was used to examine the heterogeneity of income effects across participation forms, farmer types, and regional characteristics, providing results for the 10th, 25th, 50th, 75th, and 90th percentiles of farmers’ annual household income (

Table 8. QR results of factors affecting farmers’ annual household income.

		QR_10	QR_25	QR_50	QR_75	QR_90
Participation Behavior	Employment	0.4170 *** (0.0724)	0.3070 *** (0.0554)	0.2110 ** (0.0583)	0.1300 ** (0.0625)	0.0293 (0.0733)
	Land Leasing	0.4380 *** (0.0955)	0.3680 *** (0.0598)	0.4530 *** (0.0474)	0.5880 *** (0.0274)	0.6340 *** (0.0322)
Farmer Types	Agriculture-oriented Type	0.1570 (0.1530)	0.0961 (0.1170)	0.0770 (0.1230)	-0.0779 (0.1320)	-0.2410 (0.1550)
	Forestry-dependent Type	0.3250 ** (0.1560)	0.2400 ** (0.1200)	0.1240 (0.1260)	-0.0708 (0.1350)	-0.2640 * (0.1580)
	Non-agriculture Non-forestry-oriented Type	−0.0274 (0.1200)	−0.1110 (0.0920)	−0.0825 (0.0967)	−0.1720 * (0.1040)	−0.2380 * (0.1220)
Regional Characteristics	Hulunbuir	−0.0798 (0.1030)	−0.1070 (0.0789)	−0.1100 (0.0829)	−0.1290 (0.0888)	−0.0364 (0.1040)
	Chifeng	−0.1330 (0.0912)	−0.1390 ** (0.0699)	−0.2190 *** (0.0735)	−0.2750 *** (0.0787)	−0.2260 ** (0.0923)
	Ulanqab	−0.0798 (0.0942)	−0.1510 ** (0.0722)	−0.1220 (0.0759)	−0.1290 (0.0813)	−0.0427 (0.0954)

Note: * represents significant at the 10% level, ** represents significant at the 5% level, and *** represents significant at the 1% level.

).

4.5.1. Participation Through Employment

Employment participation has a significant positive impact on the annual household income at all levels from low to middle income (QR_10, QR_25, QR_50, and QR_75) (QR-10, QR-25, QR-50, QR-75, and QR-90 represent the results of the 10th/25th/50th/75th/90th quantile regression in Stata’s quantile regression analysis), with the highest effect at the low-income level (QR_10). However, the impact gradually decreases as income rises. The impact of participating through land leasing follows a “decline first, then rapidly increase” trend, with the highest impact observed at the high-income level (QR_90). This indicates heterogeneity in the income effects of the two participation forms, meaning that the influence of employment and land leasing participation varies across income levels. Additionally, significant differences exist at each income level: employment participation has a greater effect on low-income (QR_10) farmers, while land leasing participation has the highest effect on high-income (QR_90) farmers. Thus, Hypotheses H2 and H3 are thus validated.

4.5.2. Farmer Types

The impact is significant for forestry-dependent farmers at the low-income levels (QR_10 and QR_25) but not at high-income levels (QR_50 and QR_75). For non-agriculture non-forestry-oriented farmers, the impact is significant at high-income levels (QR_75 and QR_90), but not at other income levels. Thus, different farmer types have varying impacts across income quantiles.

4.5.3. Regional Characteristics

Regional factors significantly influence farmers’ annual household income. The eastern region has no significant impact, whereas the central region has a significant effect. The western region significantly affects low-income farmers (QR_10) but not other income levels. This finding aligns with previous research indicating heterogeneity in the effects of regional factors on income effects [54].

5. Discussions and Suggestions

5.1. Discussions

Compared to previous studies [5,6,10,11], this research expands the scope of analysis and provides new insights into optimizing production factor allocation in the forest health industry. Rather than solely examining the income effects of traditional agricultural industries, we emphasize the synergistic effects of non-agricultural employment and land transfer, considering the forest health industry’s dual “ecological–economic” attributes. Our primary focus is how sustainable operations in forest health bases impact farmers’ income, particularly whether these effects vary across different farmer types based on employment and land leasing.

To analyze income heterogeneity, we applied the ESRM to uncover participation mechanisms in the ecological economy domain. Furthermore, this study extends theoretical interpretations by integrating labor transfer theory and resource endowment theory to explain why employment participation benefits low-income farmers more, while land transfer has a stronger impact on high-income farmers.

This study focuses on micro-level data, examining how farmers’ participation in forest health bases affects income. However, for future studies, it is necessary to consider macro-level aspects such as geographical characteristics to understand the income effects of farmers participating in forest health bases. The current study relies on cross-sectional data and cannot capture dynamic effects, such as whether the impact of long-term participation on income is nonlinear. Future research should incorporate panel data for a longitudinal analysis.

We primarily consider land use forms and producer labor allocation as the main production factors, selecting land leasing and employment as the two forms of participation for farmers to examine their income effects. However, other factors should be considered in the future. This study mainly focuses on improving farmers’ objective well-being. However, the high-quality ecological environment, convenient transportation infrastructure, and good living conditions provided by forest health bases also significantly impact the subjective well-being of surrounding farmers. Therefore, a more comprehensive and in-depth study of farmers’ objective and subjective well-being surrounding forest health bases is of greater and longer-term significance.

The findings suggest that farmers’ participation in the sustainable management of forest health based on employment and land leasing can improve farmers’ income to varying degrees. However, with the rapid development of the forest health industry, more and more farmers participate in sustainable management. Will increased participation lead to competition among farmers and a reduction in income? In the context of competition, future discussion can be guided by the crowding effect theory to assess whether excessive farmer participation may result in diminishing marginal returns. Additionally, further research should explore potential mitigation strategies, such as policy interventions, cooperative models, and differentiated participation mechanisms. This question warrants further study and discussion.

This study examines how farmers’ participation in sustainable forest health industry operations (employment and land leasing) affects their income. Future research could further explore the direct income channels, such as wage increases from employment and land transfer effects through rental income and labor reallocation (e.g., transition to non-agricultural employment). Additionally, the indirect pathways should be investigated, particularly whether improvements in infrastructure (e.g., roads and healthcare) driven by the forest health industry lower living costs and indirectly enhance farmers’ income.

The study controls individual and household characteristics and regional and type-specific fixed effects. However, potential moderating variables, such as social capital (e.g., farmers’ interpersonal networks) and ecological awareness (e.g., an understanding of the forest health industry), may influence the model’s explanatory power. Future studies could incorporate these factors to better capture their effects on farmers’ income.

While this research focuses on the ecological economy, future studies should consider aligning with the Sustainable Development Goals (SDGs), particularly SDG 1 (poverty reduction) and SDG 15 (land ecosystem protection). Ensuring sustainable development in the forest health industry requires balancing economic benefits and ecological protection, mitigating risks such as over-commercialization and excessive tourism that may strain ecosystems. Additionally, social equity must be considered—high-income farmers may strengthen their resource advantages through land transfers, potentially widening income disparities. Policy interventions, such as benefit-sharing mechanisms, should be explored to prevent economic inequality.

5.2. Suggestions

Based on the previous analysis, participation in the sustainable management of forest health bases, either through employment or land leasing, significantly impacts farmers’ income. Moreover, non-participating farmers would experience a greater increase in income if they were to engage in the sustainable management of forest health bases. Compared to participating farmers, this is also a key reason why non-participating farmers have lower incomes.

Based on the findings, the following policy recommendations are proposed: Accelerate the construction and operation of the sustainable management of forest health bases. Conduct regular assessments of approved national-level forest health pilot bases to ensure the sustainable and efficient development of forest health bases. This will achieve “mutual promotion and win-win” between forest health bases and surrounding farmers, particularly for low-income groups, as it is an important way to increase their income. Enhance efforts to encourage surrounding farmers to participate in the sustainable management of forest health bases. Actively engage surrounding farmers in the sustainable management of forest health bases, particularly high-income groups, and encourage their proactive involvement in various forms of participation. Since the factors affecting farmers’ income from the sustainable management of forest health bases vary by region, policies should be tailored to local conditions to develop effective strategies for the forest health industry and increase farmers’ income through multiple channels.

6. Conclusions and Prospects

This study focused on farmers who participated in the sustainable management of forest health bases through employment and land leasing and employed the ESRM to empirically test the income effects of participation in forest health bases through these forms. The robustness of the model estimation was tested using various estimation methods and by changing the dependent variables. Additionally, quantile regression was used to analyze the heterogeneity further. Participation in the sustainable management of forest health bases through employment and land leasing increases annual household income by 4.28% and 1.44%, respectively. The income effect for farmers participating through employment is 2.84 percentage points higher than those participating through land leasing. For farmers who did not participate in the sustainable management of forest health bases, their annual household income decreased by 5.87% and 2.55%, respectively, compared to the counterfactual scenario. The income effect is more pronounced for previously non-participating farmers after they engage in sustainable management. It was found that the income effects of the two forms of participation vary across different income levels. Participation through employment has a more significant impact on low-income (QR_10) farmers, while participation through land leasing has a greater impact on high-income (QR_90) farmers.

However, this study has limitations. The sample may not be fully representative of the broader research population. With the ongoing development of the forest health industry, changes in farmer participation patterns and the sustainability of their involvement may affect the generalizability of the findings. Future research should examine evolving participation forms and the sustainability of farmer involvement to provide a more in-depth and comprehensive analysis of the long-term impact of farmers’ income. This will improve the accuracy and depth of analysis, offering stronger evidence for policy decisions.