Sustainable Rural Livelihoods and Equity: A Comparative Analysis of Land Transfer and Non-Farm Employment in Sichuan Province, China

Abstract

While agricultural modernization improves productivity, it may worsen rural inequality. Without systematic guidance and institutional rules, it harms inclusive and sustainable rural development. To examine the income distribution effects of two distinct modernization pathways, this study uses an innovative dual-mode framework integrating resource endowment, mechanism, and distribution to compare Land Transfer and Non-farm Employment. Based on a survey of 963 farm households in modern agricultural parks of Sichuan Province, we apply regression, endogeneity correction, mechanism and heterogeneity analysis. The study found that Land Transfer exhibits a significant positive correlation with income growth through economies of scale and labor release effects, yet its benefits primarily flow to local elite groups with superior resource endowments, demonstrating an “elite capture” tendency; Non-farm Employment is closely linked to income growth by raising wage levels, enhancing skill levels, and improving employment stability. Its benefits are more likely to reach ordinary, low-income, and less-educated farmers, reflecting the characteristic of “inclusive growth.” The framework reveals divergent equity outcomes of efficiency-oriented reforms, providing new insights for building fair and sustainable agricultural systems. It also provides micro-level policy references for SDG 10 (reduced inequalities) and SDG 8 (decent work and economic growth).

1. Introduction

1.1. Research Background and Problem Statement

Income inequality poses a global challenge to sustainable development. Within China’s policy context of “rural revitalization” and “common prosperity,” a key issue is whether agricultural modernization can avoid exacerbating rural inequality. This issue is central to assessing the inclusiveness of relevant institutions [1]. As agricultural modernization accelerates, China’s farming system has moved beyond traditional fragmented management. The Modern Agricultural Industrial Park has become a key platform for resource integration. By 2024, there will be 95 star-rated industrial parks in Sichuan Province, China [2]. Modern agricultural parks in the province have the core functions of aggregating elements, integrating industrial chains, and benefiting farmers. They generally adopt a multi-level management model of “management committee + platform company + cooperative + farmers” [3]. Through organizational and institutional innovations, these parks reorganize land and labor allocation. They are widely seen as effective tools for boosting smallholder incomes [4,5]. In practice, however, ordinary farmers often lack sufficient land, labor, or social capital. This limits their ability to share equally in the benefits of development [6,7,8,9]. Therefore, Land Transfer and Non-farm Employment are core pathways through which modern agricultural parks integrate resources. They also serve as key arenas for examining the inclusiveness of agricultural modernization institutions. Their income distribution effects require systematic evaluation [10,11,12,13].

1.2. International Experience and Chinese Practice

An in-depth exploration of this issue must necessarily focus on the specific pathways through which modern agricultural parks achieve resource integration. International experiences provide diverse points of reference. For instance, Thailand in Southeast Asia has promoted its “One Village One Product” initiative to develop rural specialties [14], while Vietnam has relied on cooperatives to integrate smallholders into value chains [15,16]. In Africa, countries such as Ethiopia have attempted to integrate agriculture with tourism to revitalize rural areas [17,18]. Such developing countries generally face common developmental challenges, including fragmented small-scale farming operations and insufficient industrial drive [19,20]. Their success and failure experiences provide crucial comparative benchmarks for understanding the China model. The common lesson from these practices is that, in the absence of systematic national policies and infrastructure support, small farmers’ participation is often limited in depth, making it difficult to achieve inclusive development. In contrast, China’s Modern Agricultural Industrial Park, characterized by “government guidance and market dominance,” see their “guidance” specifically manifested in planning, land consolidation, basic investment, and rule-setting, while the “dominance” grants market operation rights to the operating entities [21,22]. This approach has established a more localized and diversified agricultural ecosystem, providing a unique platform for the deep integration of small-scale farmers.

Although farmer integration models are diverse in form [23,24,25,26], they can be summarized into two core pathways from both theoretical and practical perspectives. The first is the land resource integration model centered on Land Transfer, which achieves intensive production through consolidating land-management rights [27]. The second is the labor resource absorption model oriented toward Non-farm Employment, which provides farmers with Non-farm job opportunities within industrial chains [28]. The two differ in their distributional effects. Land Transfer, by revitalizing land assets and enabling scale operations, theoretically offers multiple benefits to farmers, including rental income, dividends, and Non-farm Employment opportunities [29,30]. This process is often embedded in the existing rural social power structure, with elite groups frequently dominating the transfer process due to their resource and informational advantages. As a result, ordinary smallholders, after receiving fixed rents, may still be excluded from the distribution of value-added benefits, falling into the dilemma of “elite capture” and experiencing increased livelihood vulnerability [31,32].

Non-farm Employment more directly promotes farmer income by providing Non-farm jobs [33]. Evidence from numerous developing countries demonstrates that Non-farm Employment is a crucial pathway for rural poverty reduction, from Indonesia’s promotion of high-value agriculture and rural Non-farm Employment to tribal villages in eastern India where non-agricultural wage income accounts for 63% to 74% of total household income [34,35]. The wage income earned through such employment has a significant poverty-reduction effect for low-income households and contributes to skill accumulation and employment stability [36,37]. In the Chinese context, the development of secondary and tertiary industries within parks is regarded as an important avenue for achieving “inclusive growth,” enabling the benefits of growth to reach ordinary farmers more broadly [38].

1.3. Research Questions and Potential Marginal Contributions

In summary, while both Land Transfer and Non-farm Employment can theoretically enhance household income, their underlying distributional logic and potential social effects may differ substantially. Land Transfer may reinforce resource endowment disparities [39], whereas Non-farm Employment tends to be more inclusive [40]. Distinguishing between these two pathways and their underlying benefit-distribution mechanisms is of great significance for building equitable and sustainable agricultural systems. Therefore, this study focuses on the following key questions: Can Land Transfer and Non-farm Employment significantly increase farmers’ income? How do different modes of participation affect farmers with varying resource endowments? Does the Modern Agricultural Industrial Park policy alleviate or exacerbate income inequality within rural areas?

Departing from previous studies that examine a single engagement mode in isolation, this research develops a dual-mode comparative analytical framework (Figure 1).

Its potential marginal contributions are as follows: First, it highlights the interaction and contrasting distributional logics of Land Transfer and Non-farm Employment within a complex rural governance setting, providing a renewed perspective for assessing institutional inclusiveness. Second, drawing on the phenomena of “elite capture” and “inclusive growth,” the study constructs a three-dimensional analytical system encompassing “group heterogeneity, net distributional effects, and intrinsic mechanisms,” establishing a more refined analytical toolkit. Third, this framework aids in a deeper understanding of the trade-offs and distributional differences in benefits among households with different endowments under the two participation modes. It further sheds light on the politics of distribution during China’s rural transition, providing empirical evidence for achieving equitable and sustainable rural revitalization.

2. Theoretical Analysis and Research Hypotheses

2.1. Theoretical Analysis

2.1.1. Distinguishing Rural Elites from Ordinary Farmers

This study defines “rural elites” based on their typical characteristics in terms of resource control and community influence. If a farmer’s household has members serving as village officials, establishing or operating enterprises, or being large-scale farmers and breeders, any of these identities will classify the household as a rural elite. These three groups were selected mainly because they usually have significant advantages in political resources, economic capital, or production scale and are more likely to occupy advantageous positions in the allocation of resources such as Land Transfer. This simplified operation is necessary for empirical measurability, but this paper also acknowledges that other types of elites that exert influence through implicit resources such as social capital and information networks may play more covert roles in the allocation of park resources, providing an expansion direction for subsequent research. Households meeting none of these criteria are defined as “ordinary farmers.” Comparing benefits between these two groups clarifies the distributional bias.

2.1.2. Theoretical Distinction: “Elite Capture” Versus “Inclusive Growth”

Since the rural elite group may rely on their resource endowment advantages to gain an advantage in the park’s income distribution, thereby excluding the interests of ordinary farmers [41], this study’s core theoretical concern lies in distinguishing between two distribution logics. “Elite Capture” refers to the fact that rural elites, by virtue of their endowment advantages, occupy a dominant position in resource allocation and thereby consolidate or even expand the existing inequality [42]; “Inclusive Growth” emphasizes providing low-barrier participation opportunities to enable growth outcomes, especially income increments, to be more equitably distributed to disadvantaged groups [43]. For instance, large-scale farmers in the park rent out the land of ordinary farmers at low prices and then obtain scale operation benefits, which is a typical example of “Elite Capture”; while the park provides processing jobs without experience requirements for low-educated farmers and offers training and skills development, this reflects “Inclusive Growth”. These two concepts provide ideal types for describing the tendencies of distribution outcomes, but in reality, they may not be mutually exclusive but coexist or interweave in different fields.

2.1.3. Theoretical Framework Construction

Based on these concepts, we build a comparative theoretical framework (Figure 2). Land Transfer requires initial land endowment, capital, and social networks. The entry barrier is high. Non-farm Employment mainly depends on personal labor. The barrier is low [44,45,46]. This determines the different primary beneficiary groups. Land Transfer benefits village cadres, enterprise owners, and large-scale farmers more. Non-farm Employment provides broader participation and income channels for ordinary farmers, low-income households, and less-educated groups [47]. Regarding core mechanisms, Land Transfer promotes income growth through agricultural economies of scale and labor release [48]. Non-farm Employment raises income directly through wages, skill accumulation, and stable Non-farm jobs [49]. However, their distributional effects may differ sharply. Land Transfer may worsen community inequality due to resource disparities, showing “elite capture” features [50,51]. Non-farm Employment helps narrow income gaps, demonstrating stronger “inclusive growth” attributes [52]. This framework delineates logical distinctions between the two path tendencies rather than definitive correspondences. Specific contexts and policies modulate their actual manifestation, and the two effects may coexist; this contrast provides the analytical foundation for subsequent empirical testing in this study.

2.2. Research Hypotheses

The construction of modern agricultural parks provides farmers with two ways to increase their income by reshaping the allocation of the two core elements of land and labor (Figure 3).

Land Transfer generates stable asset-based income for households through rental returns, equity dividends, and value-added benefits from scaled operations, while also releasing surplus labor to engage in other economic activities [53]. Non-farm Employment, on the other hand, directly raises total household income via wage compensation and facilitates sustained income growth through skill training and occupational advancement [54,55]. Both forms of participation effectively overcome the income limitations inherent in traditional smallholder farming, diversify income sources, and demonstrate robust and significant positive effects on raising household income. Accordingly, the following hypothesis is proposed:

Hypothesis 1.

Both Land Transfer and Non-farm Employment within modern agricultural parks can significantly boost farm household income.

While both Land Transfer and Non-farm Employment can contribute to farmers’ income growth, their underlying operational mechanisms may differ. Land Transfer directly activates a core rural production factor. Land Transferees can reduce unit production costs and more easily introduce modern inputs such as machinery and improved seeds by consolidating fragmented plots, thereby capturing returns to scale [56,57]. Land Transferors, on the other hand, are able to reallocate labor to sectors with higher productivity. Through the labor transfer effect, they gain wage income, leading to a structural increase in total household income [58]. Consequently, the distribution of benefits from Land Transfer may be asymmetric: gains from scale operations flow mainly to elite groups with capital advantages, whereas benefits from labor reallocation accrue more to ordinary farmers.

The income growth mechanism of Off-farm employment in agricultural parks focuses more on empowering the local labor market. It not only provides stable wage income that exceeds agricultural earnings but also enhances human capital through enterprise-based training. Moreover, by offering relatively standardized labor contracts, it helps reduce income volatility and strengthens the sustainability of income growth [59]. This pathway places lower demands on initial resource endowments and is thus more inclusive. Accordingly, the following hypotheses are proposed:

Hypothesis 2.

Land Transfer promotes income growth through economies of scale and labor release, primarily benefiting rural elites.

Hypothesis 3.

Off-farm employment promotes income growth by raising wage income, accumulating labor skills, and enhancing employment stability, with particularly empowering effects for ordinary farmers.

Furthermore, the distribution of benefits from the two pathways may exhibit heterogeneity. Resource endowment theory suggests that differences in households’ economic, social, and human capital affect their ability to capture opportunities [60]. Rural elites with higher resource endowments (e.g., village cadres, large-scale farmers), leveraging advantages in capital and information, tend to dominate Land Transfers. They obtain excess profits through scale operations and value-chain extension, creating a “rich-get-richer” cycle [61]. Ordinary smallholders often passively transfer out their land, receiving only fixed rents and struggling to share in the value-added gains [62,63]. This results in a progressive income effect from Land Transfer, which may exacerbate income inequality and reflect a tendency toward “elite capture” [64,65,66].

The income growth mechanism of Non-farm Employment is more inclusive. The theory of inclusive growth emphasizes that the benefits of growth should particularly reach disadvantaged and ordinary groups [67,68]. Off-farm jobs in agricultural parks (e.g., in processing, logistics, or services) require lower initial physical capital and rely more on labor and skills, thereby providing a relatively equitable channel for income growth for low-resource-endowment households [69]. This not only directly delivers wage income higher than farming but also enhances long-term capabilities through skill training [70]. The marginal income effect is especially pronounced for low-income, small-scale, and less-educated farmers, helping mitigate distributional imbalances caused by resource disparities. Accordingly, the following hypotheses are proposed:

Hypothesis 4.

Land Transfer exhibits a tendency toward “elite capture”, which has a stronger effect on high-income and large-crop farmers with better resource endowments or may exacerbate income inequality.

Hypothesis 5.

Non-farm Employment has the characteristic of “inclusive growth”, which has a more significant effect on low-income, small-crop, and low-education farmers and can alleviate income inequality.

3. Study Area and Data Sources

3.1. Study Area

As illustrated in Figure 4, this study selected Sichuan Province, China as an exploratory case study. The selection of this study area was based on three primary considerations. First, as a major agricultural province in western China and a key grain-producing region nationwide, Sichuan is regionally representative in terms of farmer income increase and rural revitalization [71]. Second, boasting a leading number of agricultural parks across the country, Sichuan has intensive practices of Land Transfer and Non-farm Employment, which facilitates the simultaneous observation of elite capture risks and actual effects of inclusive growth [72]. Third, the province features diverse landforms including plains, hills, mountains and plateaus, with remarkable differences in agricultural development conditions, thus bearing typical research value [73].

3.2. Data Sources

The data were obtained from field surveys conducted at representative agricultural parks in Sichuan Province from 1 March 2024 to 31 March 2025. The study employed a purposeful sampling strategy based on terrain heterogeneity, with sampling density aligned with the actual weight and theoretical significance of each terrain type in small-scale farming. To ensure balance, randomly selected farming households were surveyed within each park. Specifically, 71 parks in the hilly areas of eastern Sichuan were selected for saturated surveys as the main sample, where smallholders are the densest and agricultural models the most complex, to fully present the mechanism spectrum of Land Transfer and Non-farm Employment. One park in the alpine and canyon areas of western Sichuan was chosen as an extreme terrain test site to verify the external validity of the main case conclusions in high-altitude environments. Four parks in the plain areas of southern Sichuan were included as a scale operation control group to test the robustness of Non-farm Employment effects in highly commercialized agricultural areas. Simultaneously, regional fixed effects are employed to control for regional heterogeneity, balancing sampling efficiency with scientific rigor. Data collection combines structured questionnaires with semi-structured interviews; the questionnaires cover modules such as household income and expenditure, land-management practices, and employment participation, while interviews verify key information to ensure data authenticity. The sampling process strictly adheres to terrain weighting principles to guarantee sample representativeness. In total, the study covered 76 parks at four administrative levels (county, municipal, provincial and national). At least 15 farmer households were randomly selected from each park for a questionnaire survey, resulting in 963 valid farmer samples with an effective response rate of 84.47%. Among these samples, 68.63% and 44.86% of the farmers participated in Land Transfer and Non-farm Employment, respectively.

4. Research Methodology and Variable Selection

4.1. Research Methodology

This study develops a comprehensive methodological framework to systematically assess the impact of Land Transfer and Non-farm Employment on farmers’ net income (Figure 5).

First, a binary regression model was constructed to confirm that both Land Transfer and Non-farm Employment significantly promote rural income growth. Second, potential endogeneity issues were addressed to ensure the robustness of the core findings. Subsequently, the analysis examined the dual mechanisms of Land Transfer and the heterogeneous effects of three Non-farm Employment pathways on farmers with different resource endowments. Finally, heterogeneity analysis reveals that Land Transfer exhibits a tendency toward “elite capture,” while Non-farm Employment demonstrates characteristics of “inclusive growth.” This multi-angle, step-by-step approach provides solid methodological support for understanding income growth paths and distributional differences.

4.2. Model Construction

4.2.1. Binary Regression Model Construction

To examine the impacts of Land Transfer and Non-farm Employment on the income growth of farm households within parks, a binary regression model is constructed as follows [74]:

Y_ir = α₀ + α₁LT_ir + α₂NE_ir + α₃Z_ir + δ_r + ε_ir

(1)

where Y_ir denotes the net income growth of household i in region r, serving as the dependent variable; the core explanatory variables LT_ir and NE_ir are binary dummy variables, coded as 1 if the household participates in Land Transfer or Non-farm Employment respectively, and 0 otherwise; α₁ and α₂ represent the estimated coefficients of the core explanatory variables, while α₃ is the coefficient vector of control variables; Z_ir denotes the control variables; δ_r stands for the regional fixed effects; and ε_ir is the random error term.

4.2.2. Mediating Effect Model

To explore the impact mechanisms of Land Transfer and Non-farm Employment on the income growth of farm households, the specific model is designed as follows [75]:

M_kir = β₀ + β₁X_ir + β₂Z_ir + δ_r + μ_kir

(2)

where M_kir denotes the mediating variable, with 2 mediating variables corresponding to Land Transfer and 3 to Non-farm Employment, and k represents the serial number of the mediating variable; X_ir is the core explanatory variable, i.e., Land Transfer (LT_ir) or Non-farm Employment (NE_ir); β₁ is the estimated coefficient of the core explanatory variable on the mediating variable, and β₂ is the coefficient vector of control variables; the definitions of other variables are consistent with those in Equation (1); and μ_kir is the random error term.

4.3. Variable Selection

4.3.1. Dependent Variable

Income growth. This was obtained through a questionnaire survey, covering all income sources such as household operating income, wage income, property income and transfer income. The total household income before and after the farmers’ participation in the park was separately calculated. Then, referring to existing studies [76], the extent of income growth achieved by farmers through participating in the modern agricultural park was selected as the explained variable. This indicator directly captures the net economic gain from park involvement and is used to assess the scale of impact that contributing land or labor has on household income. To prevent interference from the initial income level, the growth rate, binary variables, and quantile regression were also added subsequently.

4.3.2. Explanatory Variables

Based on field survey data and combined with the actual process of agricultural production and operation, two modes of household participation in modern agricultural parks were incorporated into the benchmark regression model. Land Transfer is defined as 1 if a farmer engaged in Land Transfer-in or out of the park and 0 otherwise. Non-farm Employment, defined as 1 if a farmer gained wage employment within a park enterprise and 0 otherwise.

4.3.3. Mechanism Variables

To test the transmission mechanisms in Hypotheses 2 and 3, we specify the following mediator variables. For Land Transfer, we draw on Xu et al. [77] and measure economies of scale by “actual operating area after Land Transfer-in.” Following Huang et al. [78] we measure labor release by “proportion of Non-farm or part-time income.” For Non-farm Employment, drawing on Su et al. [79], Nakano et al. [80], and Bhattacharya et al. [81], we measure skill upgrading by “Proportion of farmers in the park who receive skill training,” wage income improvement by “proportion of wage income,” and employment stability by “proportion of formal workers in park enterprises.”

4.3.4. Control Variables

In addition to the two park participation modes, other variables also influence farmers’ income growth. Based on questionnaire design and prior studies [82,83], this study selects control variables across three dimensions (

Table 1. Variable definitions and descriptive statistics.

Variable Type	Variable Name	Variable Definition or Assignment	Mean	Std. Dev.
Dependent Variable	Income Growth	Absolute income growth of smallholders after joining parks (10,000 RMB)	1.7726	1.7539
Explanatory Variables	Land Transfer	1 = engaged in Land Transfer-in or out within the park, 0 = otherwise	0.6864	0.4642
	Non-farm Employment	1 = engaged in wage employment within the park, 0 = otherwise	0.4486	0.4976
Mediating Variables	Economies of Scale	Actual operating area of farmland after land inflow (10,000 mu)	0.0053	0.0260
	Labor Release	Proportion of non-agricultural or part-time income in total household income	0.5576	0.2510
	Wage Income	Proportion of wage income in total income	0.2278	0.0220
	Labor Skill Improvement	Proportion of farmers in the park who receive skill training	0.7373	0.3301
	Employment Stability	Proportion of regular employees in enterprises within the park	0.0467	0.0983
Control Variables	Age	Age of the farmer at the time of the survey	55.4533	12.0127
	Communist	Farmer’s political status, 1 = Communist Party member, 0 = Otherwise	0.3313	0.4709
	Gender	Gender of the farmer, 1 = Male, 0 = Female	0.6687	0.4709
	Health	Farmer’s health level, 1 = Poor, 2 = Average, 3 = Good	2.7591	0.4606
	Education	Farmer’s education level, 1 = Junior high school or below, 2 = Junior high school, 3 = Senior high school/vocational secondary school, 4 = Junior college or above	2.0704	0.9760
	Social Capital	1 = Have family members or direct relatives working as public servants, 0 = Otherwise	0.1578	0.3648
	Family Size	Number of people in the farmer’s household	4.3954	1.7133
	Labor Force Ratio	Number of people aged 18–65 in the household/Total number of people in the household	0.6298	0.2919
	Farmland Area	Area of farmland owned by the farmer (10,000 mu)	0.0064	0.0287
	Park Level	County-level = 1, Municipal-level = 2, Provincial-level = 3, National-level = 4	1.9014	0.9197
	Park Area	Planned area of the park (10,000 mu)	2.7030	6.2164
	Village Number	Number of villages in the Park	7.0265	21.3266
	Park Output Value	Annual output value of the park (10,000 RMB)	8.1577	19.2462
	Enterprise Number	Number of enterprises in the park	8.4686	11.9204
	Support Policies	1 = with government policy support, 0 = otherwise	0.6947	0.4608
	Planting as Leading Industry	1 = the park takes planting as the leading industry, 0 = otherwise	0.8442	0.3628
	Breeding as Leading Industry	1 = the park takes breeding as the leading industry, 0 = otherwise	0.1537	0.3608
	Processing as Leading Industry	1 = the park takes the processing industry as the leading industry, 0 = otherwise	0.1246	0.3304

): farmer characteristics (age, gender, political affiliation, health status, and education level); household characteristics (family size, social capital, labor ratio, and farmland area); and park characteristics (park level, planned area, number of villages, output value, number of enterprises, support policies, and dominant industries). To control for potential biases from regional socioeconomic conditions, regional fixed effects are included in the model.

5. Empirical Results

5.1. Binary Regression Results

5.1.1. Binary Regression and Robustness Tests

This study examines the income growth effects of two primary pathways for farmer participation in the park, based on a binary regression analysis of 963 farm household samples. As presented in columns 1 and 2 of

Table 2. Binary regression and robustness test results.

Variables	Income Growth		Rate of Income Growth	Binary Variable	Quantile25	Quantile50	Quantile75
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Land Transfer	0.7465 ***	0.7416 ***	0.1066 ***	0.2300 ***	0.6439 ***	0.8079 ***	0.9983 ***
	(0.1169)	(0.1152)	(0.0204)	(0.0334)	(0.0886)	(0.1104)	(0.0414)
Non-farm Employment	1.0795 ***	1.0770 ***	0.1282 ***	0.2945 ***	0.7535 ***	1.2454 ***	1.0005 ***
	(0.1115)	(0.1106)	(0.0223)	(0.0281)	(0.1047)	(0.1235)	(0.0405)
Constant	0.7760 ***	2.0477 **	0.5064 ***	0.4631 **	0.2894	1.5654 *	3.0014 ***
	(0.0999)	(0.6435)	(0.1384)	(0.1665)	(0.3848)	(0.6856)	(0.6976)
Controls	no	yes	yes	yes	yes	yes	yes
Region fixed effect	yes	yes	yes	yes	yes	yes	yes
R-squared	0.1979	0.2734	0.2199	0.3370
Sample size	963	963	963	963	963	963	963

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

, both Land Transfer and Non-farm Employment show a statistically significant positive impact on farm household income growth at the 1% level, regardless of whether control variables are included. This provides preliminary evidence for the effectiveness of these two models in raising incomes. To verify the robustness of the findings, this paper conducts robustness checks using three alternative approaches (columns 3–7): first, re-estimating the regression by re-measuring the dependent variable as the income growth rate of the sample households (column 3); second, re-specifying the dependent variable as a binary indicator of whether a household achieved income growth (1 = yes, 0 = no) and re-estimating the model (column 4); third, performing quantile regression analyses at the 25th, 50th, and 75th percentiles, corresponding to low, medium, and high levels of income growth, respectively (columns 5–7). The results from these robustness checks are consistent with the conclusions from the main regression analysis, thereby validating Hypothesis 1.

5.1.2. Income Effects Comparison and Distributional Differences

Farm households participating in park-based Non-farm Employment gained about 10,000 RMB on average. This exceeds the 7400 RMB for Land Transfer households. This gap likely stems from different mechanisms and beneficiary structures. Non-farm Employment mainly provides stable wage income. It has low barriers and wide coverage. It particularly boosts income for middle- and low-income groups. Land Transfer income shows sharp polarization. Most farmers only receive fixed rents. High returns concentrate among a few who can integrate resources and expand operations. Thus the overall mean is lower.

Quantile regression further confirms this. The income effect of Land Transfer strengthens steadily as quantiles rise. This shows high-income farmers gain excess returns from Land Transfer-in. Low-income farmers hardly share value-added benefits. The income effect of Non-farm Employment peaks at the 50th quantile. This indicates that middle-segment farmers effectively convert employment opportunities. It remains significant at low quantiles, providing stable income growth for low-income earners. At high quantiles, its marginal effect is limited due to diversified income sources. This creates a more inclusive income growth path.

In summary, Non-farm Employment raises average income gains through universal benefits. Land Transfer intensifies internal differentiation among farmers due to heavy reliance on capital and scale. This difference essentially reflects how the two modes depend differently on farmers’ initial endowments. It also suggests they may have opposite effects on rural income distribution. Therefore, further analysis from a heterogeneity perspective is needed. We should explore income mechanisms and distributional effects of the two modes across different groups.

5.2. Endogeneity Treatment

5.2.1. Instrumental Variable Test

• IV Selection and Justification for Land Transfer

To address endogeneity concerns arising from omitted variables, reverse causality, and measurement bias, this study employs an instrumental variable (IV) approach to estimate the effects of two significant models (

Table 3. Instrumental variable regression results.

Variables	Land Transfer	Land Transfer	Income Growth	Non-Farm Employment	Non-Farm Employment	Income Growth
	(1)	(2)	(3)	(4)	(5)	(6)
Altitude	0.0007 ***
	(0.0001)
Farmland Transfer Support Policy		0.2566 ***
		(0.0284)
Land Transfer			2.3437 ***
			(0.4012)
Park Location				0.0051 **
				(0.0019)
Employment Support Policy					0.3850 ***
					(0.0360)
Non-farm Employment						2.6605 ***
						(0.3412)
Constant	0.4932 **	0.6885 ***	0.9138	0.5963 *	0.5133 **	0.7782
	(0.1892)	(0.1768)	(0.7502)	(0.2355)	(0.1891)	(0.7042)
Controls	yes	yes	yes	yes	yes	yes
Region fixed effect	yes	yes	yes	yes	yes	yes
R-squared	0.2268	0.2598	0.1304	0.1032	0.2014	0.0909
Sample size	963	963	963	963	963	963
Kleibergen-Paap rk LM statistic	95.281 (0.0000)			101.280 (0.0000)
Kleibergen-Paap Wald rk F statistic	57.024			59.230
Stock-Yogo weak ID test critical values:10%	19.93			19.93
Hansen J statistic	0.735 (0.3913)			0.077 (0.7814)
Endogeneity test	17.815 (0.0000)			29.199 (0.0000)

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

).

For Land Transfer, drawing on existing research [84,85], we select the average altitude of the township where the park is located and whether the park has Land Transfer support policies as instruments. Altitude is relevant because high-altitude areas in Sichuan are mostly mountains and plateaus with scattered plots and difficult cultivation, making farmers more willing to transfer land, while low-altitude plains and hilly areas have the opposite situation (Column 1). Supportive policies such as subsidies and standardized contracts significantly reduce transfer costs and enhance farmers’ willingness to engage in Land Transfer (Column 2). In terms of exogeneity, altitude, as a natural geographical endowment, is not influenced by farmers’ income or operational decisions, nor does it directly affect income levels. Farmland Transfer policies are formulated by the government based on regional development planning aimed at optimizing agricultural layout rather than targeting individual household income, thus satisfying the exclusion restriction.

2.

IV Selection and Justification for Non-farm Employment

For Non-farm Employment, with reference to existing studies [86], the distance from the park to the city center and whether the park implements employment support policies are selected as instrumental variables. The spatial distance between parks and city centers significantly influences farmers’ employment choices. While nearby parks attract farmers to urban jobs due to the urban pull effect, distant parks, limited by high commuting costs and unfamiliar environments, cause farmers to prioritize local park employment (Column 4). Employment support policies in parks (e.g., labor subsidies, vocational training, and entrepreneurship incentives) directly stimulate enterprises to expand employment scale and improve job supply, thereby significantly enhancing parks’ capacity to absorb employment (Column 5). The relative position between the park and the city center is an objective geographical attribute, not affected by farmers’ employment decisions or income levels. Employment support policies are formulated by the government based on regional economic plans, focusing on macroeconomic industrial layout and employment growth, not directly intervening in farmers’ income, thus meeting the exogeneity requirement. Therefore, both variables satisfy the relevance and exogeneity conditions required for valid instruments.

3.

IV Tests and Second-Stage Regression Results

Endogeneity tests show the following results: the Kleibergen-Paap rk LM statistic rejects the null hypothesis of underidentification at the 1% significance level (p < 0.01). The Kleibergen-Paap rk Wald F statistic exceeds the 10% critical value of the Stock-Yogo weak identification test, indicating that the instruments are not weak. The Hansen J statistic yields p > 0.1, supporting the exogeneity of the instruments. Additionally, the endogeneity test of the endogenous regressors returns p < 0.01, confirming the necessity of the IV approach. The results of the second stage regression (columns 3 and 6) indicate that both participation methods in the two parks have contributed to the increase in farmers’ income, thereby further verifying Hypothesis 1.

5.2.2. Propensity Score Matching Test

• PSM Design and Tests

To mitigate sample selection bias, this study further adopts the Propensity Score Matching (PSM) method, designating farm households participating in Land Transfer or Non-farm Employment as the treatment group and non-participants as the control group. The matching variables include all regression control variables and county-level dummy variables. We first perform kernel matching for the two participation modes, followed by the common support test and balancing test. For the common support test, the kernel density plots in Figure 6 and Figure 7 show a higher degree of overlap in the kernel density curves after matching, which satisfies the common support assumption and attests to the reliability of the matching results. In terms of the balancing test, the standardized bias plot in Figure 8 reveals that the standardized bias of nearly all variables is reduced post-matching, verifying the validity of the matching and confirming that the balancing test is passed.

2.

Matching Results and Robustness Verification

To enhance result robustness, we conduct nearest-neighbor matching and radius matching again. After matching, the Land Transfer sample sizes are 334, 899, and 908 under nearest-neighbor, radius, and kernel matching respectively. The Non-farm Employment sample sizes are 454, 903, and 957 under the three methods. We input the successfully matched samples into the benchmark regression model (

Table 4. PSM regression results.

Income Growth
Matching Method	One-to-One Near Neighbor	Radius	Kernel	One-to-One Near Neighbor	Radius	Kernel
Variables	(1)	(2)	(3)	(4)	(5)	(6)
Land Transfer	0.7011 ***	0.7017 ***	0.7250 ***
	(0.1715)	(0.1146)	(0.1147)
Non-farm Employment				0.9483 ***	1.0518 ***	1.0576 ***
				(0.1528)	(0.1111)	(0.1113)
Constant	2.0572	1.9826 **	1.9907 **	2.2842 *	2.2900 ***	2.2314 ***
	(1.1012)	(0.6355)	(0.6325)	(1.0378)	(0.6898)	(0.6664)
Controls	yes	yes	yes	yes	yes	yes
Region fixed effect	yes	yes	yes	yes	yes	yes
R-squared	0.2853	0.2579	0.2594	0.1992	0.2457	0.2718
Sample size	334	899	908	454	903	957

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

). Results show that both modes positively affect park farmers’ income at the 1% significance level across all three matching methods. Endogeneity test results are basically consistent with benchmark regressions. This verifies Hypothesis 1.

5.3. Mechanism Analysis

5.3.1. Mechanism Analysis of Land Transfer

• Mechanism Tests: Economies of Scale and Labor Release Paths

Based on Hypothesis 2, Land Transfer affects farm income through economies of scale and labor release. According to the “Panel A: Mechanism Test” in Figure 9 (detailed results in Appendix A,

Table A3. Mechanism test results for Land Transfer.

Variables	Economies of Scale	Labor Release	Economies of Scale		Labor Release
			Ordinary Farmers	Rural Elites	Ordinary Farmers	Rural Elites
	(1)	(2)	(3)	(4)	(5)	(6)
Land Transfer-in	0.0033 **		0.0006	0.0182 *
	(0.0011)		(0.0004)	(0.0076)
Land Transfer-out		0.0823 ***			0.0975 ***	0.0303
		(0.0146)			(0.0166)	(0.0327)
Constant	0.0032	0.3966 ***	−0.0020	0.0686	0.3903 ***	−0.0706
	(0.0058)	(0.0761)	(0.0022)	(0.0507)	(0.0797)	(0.2821)
Controls	yes	yes	yes	yes	yes	yes
Region fixed effect	yes	yes	yes	yes	yes	yes
R-squared	0.7937	0.4310	0.8710	0.7642	0.4140	0.5021
Sample size	963	963	811	152	811	152

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

), the 95% confidence intervals for both “Economies of Scale” and “Labor Release” exclude zero and lie entirely to the right of zero. This indicates that both mechanisms are statistically significant. The results confirm the theoretical hypothesis. Land Transfer boosts production efficiency by integrating land resources and expanding the operation scale. It also frees farmers from land constraints, allowing them to shift to Non-farm Employment or more efficient activities. This generates dual positive effects.

2.

Group Heterogeneity: Elite Dominance and Ordinary Farmer Constraints

We further analyze group heterogeneity under these two mechanisms. As illustrated in the “Panel B: Economies of Scale” of Figure 9, ordinary farmers’ confidence interval crosses zero, showing no significant effect. Rural elites’ 95% confidence interval excludes zero, showing a significantly positive effect. This indicates that scale economies only significantly benefit rural elites. These groups possess sufficient capital, technology, and market resources. They can integrate land more efficiently to achieve scale-based value addition. Ordinary farmers struggle to gain from this path.

In “Panel B: Labor Release”, ordinary farmers’ 95% confidence interval excludes zero, showing a significantly positive effect. Rural elites’ confidence interval crosses zero, showing no significant effect. This means ordinary farmers mainly rely on Land Transfer-out to release family labor and obtain Off-farm income. Rural elites have diversified income structures. Their labor is not highly bound to land. Thus Land Transfer has weak marginal effects on labor release for them. Hypothesis 2 is verified. They also imply that future policies should emphasize the inclusiveness of Land Transfer and adopt differentiated measures to promote more balanced income distribution.

5.3.2. Mechanism Analysis of Non-Farm Employment

• Mechanism Tests: Triple Optimization of Wages, Skills, and Stability

Consistent with Hypothesis 3, Non-farm Employment optimizes farm income structure and employment quality through three mechanisms: wage income growth, labor skill accumulation, and employment stability enhancement. In Figure 10 (detailed results in Appendix A,

Table A4. Mechanism test results for Non-farm Employment.

Variables	Wage Income	Labor Skill Improvement	Employment Stability	Wage Income		Labor Skill Improvement		Employment Stability
				Ordinary Farmers	Rural Elites	Ordinary Farmers	Rural Elites	Ordinary Farmers	Rural Elites
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Non-farm Employment	0.0038 **	0.0444 **	0.0299 ***	0.0048 **	−0.0019	0.0399 *	0.0460	0.0163 ***	0.0784 ***
	(0.0014)	(0.0143)	(0.0057)	(0.0016)	(0.0035)	(0.0159)	(0.0257)	(0.0044)	(0.0230)
Constant	0.2315 ***	0.5928 ***	0.0001	0.2311 ***	0.2025 ***	0.5300 ***	1.3672 ***	0.0068	0.3363
	(0.0071)	(0.0881)	(0.0319)	(0.0075)	(0.0225)	(0.0904)	(0.3359)	(0.0229)	(0.2634)
Controls	yes	yes	yes	yes	yes	yes	yes	yes	yes
Region fixed effect	yes	yes	yes	yes	yes	yes	yes	yes	yes
R-squared	0.1306	0.6588	0.2800	0.1702	0.0369	0.6887	0.5327	0.3021	0.2692
Sample size	963	963	963	811	152	811	152	811	152

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

), the “Panel A: Mechanism Test” shows that the effect coefficients for “Wage Income,” “Labor Skill Improvement,” and “Employment Stability” all lie to the right of zero. Their 95% confidence intervals all exclude zero. This indicates all three mechanisms are statistically significant. Non-farm Employment clearly promotes all three aspects. The results confirm that Non-farm Employment comprehensively optimizes farmers’ income structure and employment quality by improving income composition, accumulating human capital, and reducing employment risks.

2.

Group Heterogeneity: Universal Benefits for Ordinary Farmers and Limited Marginal Gains for Elites

We further analyze group heterogeneity under these three mechanisms. As shown in Figure 10, in the “Panel B: Wage Income, Labor Skill Improvement”, ordinary farmers’ 95% confidence intervals exclude zero, showing significantly positive effects. Rural elites’ confidence intervals cover zero, showing no significant effects. This indicates that wage growth and skill enhancement from Non-farm Employment only significantly benefit ordinary farmers. These groups have weaker initial income and skill bases. They can quickly improve core income capacity and market skills through Non-farm Employment. Rural elites have solid skill foundations and diversified income sources. The marginal returns from this path are limited for them.

In “Panel B: Employment Stability”, both groups’ 95% confidence intervals exclude zero, showing significantly positive effects. However, the elite group benefits more significantly from employment opportunities. This disparity suggests potential stratification within the Non-farm Employment sector. Leveraging their educational advantages and social capital, elites are more likely to secure long-term, stable positions such as managerial or technical roles, whereas ordinary farmers predominantly occupy seasonal or temporary jobs. Although these positions offer greater income stability compared to farming, they still expose workers to higher employment volatility risks. This finding indicates that the ‘inclusiveness’ of Non-farm Employment primarily manifests in the broad accessibility of job opportunities rather than in equal employment quality. Policy design should focus on optimizing job structures to prevent Non-farm employment positions from becoming new arenas for elite capture. Yet ordinary farmers still significantly gain employment protection. This reflects the basic universal nature of this path. Overall, although all three transmission mechanisms of Non-farm Employment are significant overall, their effects are more pronounced for ordinary farmers. Hypothesis 2 is verified.

This provides empirical evidence for improving park employment policies and promoting inclusive growth. Future policies can continue strengthening wage growth and skills training support for ordinary farmers. Meanwhile, employment stability should be guaranteed for all farmer types. This will advance inclusive rural transformation.

5.4. Heterogeneity Analysis

5.4.1. Framework and Grouping Design

Based on earlier hypotheses and mechanism tests, we systematically examine elite capture effects in Land Transfer and inclusive features of Non-farm Employment. Besides classifying farmers into ordinary and elite groups, we build a heterogeneity identification system using three dimensions of resource endowment: economic capital, physical capital, and human capital (

Table 5. The heterogeneity analysis of the regression results.

Variables	Farmer Type		Initial Household Income Level			Farmland Area		Education Level
	Ordinary Farmers	Rural Elites	Low	Middle	High	Small	Large	Low	High
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Land Transfer	0.6027 ***	1.2604 *	0.5963 *	0.5942 **	0.7168 ***	0.5436 **	0.9058 ***	0.6597 ***	0.5665 *
	(0.1157)	(0.5000)	(0.2400)	(0.2117)	(0.1881)	(0.1759)	(0.1634)	(0.1293)	(0.2538)
Non-farm Employment	1.1474 ***	0.5489	0.9706 ***	0.9622 ***	0.8598 ***	0.9601 ***	0.9898 ***	1.2392 ***	0.5867 *
	(0.1133)	(0.3421)	(0.2477)	(0.1882)	(0.1725)	(0.1857)	(0.1517)	(0.1240)	(0.2457)
Constant	1.8874 **	11.2453 ***	1.3107	2.6576 *	3.1759	3.1351 **	1.5566	2.1657 **	4.1945 *
	(0.6282)	(2.9245)	(1.0648)	(1.1336)	(1.6764)	(1.1133)	(0.8269)	(0.7094)	(1.7685)
Controls	yes	yes	yes	yes	yes	yes	yes	yes	yes
Region fixed effect	yes	yes	yes	yes	yes	yes	yes	yes	yes
R-squared	0.2594	0.1550	0.2473	0.2987	0.2975	0.2284	0.2913	0.2962	0.2719
Sample size	811	152	321	328	314	355	608	701	262

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

).

First, we divide the sample into low-, middle-, and high-income groups by initial household income. This captures participation ability differences under varying wealth accumulation. Second, we split farmers into high and low groups by median farmland area. This reflects divergence in land element control rights. Finally, we divide groups by education level using high school as the cutoff. This identifies stratification in information access and skill accumulation [87]. This grouping design operationalizes farmer resource endowment differences into observable group characteristics. It provides clear analytical boundaries for revealing “elite bias” versus “universal benefits” distribution logics under the two modes.

5.4.2. Heterogeneity Results

• Multi-dimensional Heterogeneity

Heterogeneity results are shown in Figure 11. By farmer identity, the income growth coefficient for Land Transfer is much higher for rural elites (1.2604*) than ordinary farmers (0.6027***). Non-farm Employment only shows strong significant income effects for ordinary farmers (1.1474***). It has no significant effect on rural elites. This directly confirms that Land Transfer benefits skew toward elites with better resource endowments. It shows “elite capture” characteristics. Non-farm Employment precisely covers ordinary farmers. This matches the core logic of inclusive growth.

By initial household income, the Land Transfer income coefficient expands as income level rises (low-income group 0.5963* → high-income group 0.7168***). This shows high-income groups (close to elites) gain more from Land Transfer. The Non-farm Employment coefficient narrows as income rises (low-income group 0.9706*** → high-income group 0.8598***). This means low-income farmers are the main beneficiaries of Non-farm Employment. It further strengthens the pro-poor nature of Non-farm Employment.

By farmland area, Land Transfer income effects are significantly higher for large-area farmers (0.9058***) than for small-area farmers (0.5436**). Large-area farmers can better achieve scaled operations through Land Transfer. This is essentially a benefit concentration among resource-advantaged groups (elite-like). Non-farm Employment significantly boosts income for both small- and large-area farmers. The coefficient gap is small. This shows ordinary farmers with small land areas can also stably benefit from Non-farm Employment.

By education level, the Non-farm Employment coefficient is much higher for low-education farmers (1.2392***) than high-education groups (0.5867*). This shows ordinary farmers with low education rely more on Non-farm Employment to raise income. In Land Transfer, the higher coefficient for low-education farmers may stem from their lower income base. They depend on passive Transfer-out for short-term rents. This seems to show significant income growth. High-education elites have lower coefficients but gain high added value through management rights control. This reinforces elite capture through resource control.

2.

Divergent Distribution Effects

These multi-dimensional heterogeneity results jointly show that Land Transfer benefits continuously concentrate among high-income, large-farmland, high-resource elites. This is essentially “elite capture” under “management rights consolidation.” High-income farmers gain higher value-added returns from Land Transfer through stronger resource integration capacity. Large-farmland farmers further amplify the benefits through scaled operations. High-education elites gain high added value on the value chain through land-management rights control. Ordinary farmers often can only passively transfer out of land and obtain short-term rents. They are marginalized not only in economic returns but also lose long-term opportunities for independent development based on land. This verifies Hypothesis 4.

In sharp contrast, Non-farm Employment precisely covers low-income, small-farmland, low-education ordinary farmers. It shows clear “inclusive growth” characteristics. Low-income groups obtain the most prominent income growth effects through Non-farm Employment. Small-farmland farmers can stably benefit without relying on land scale. Low-education farmers can break skill bottlenecks and raise income levels through Non-farm Employment. This effect exactly compensates for the elite bias of Land Transfer. It provides ordinary farmers an income growth path bypassing resource endowment constraints. This verifies Hypothesis 5. This also signals potential fairness risks in modern agricultural parks. Without policy intervention, the elite bias of Land Transfer will solidify urban-rural value chain polarization. Measures such as strengthening ordinary farmers’ bargaining power and promoting benefit sharing are needed to achieve inclusive income growth.

5.4.3. Quantitative Validation of Income Inequality Indicators

To further confirm the different effects of Land Transfer and Non-farm Employment on income inequality, we follow existing studies [88,89]. We use household per capita income as the baseline. We calculate changes in the Gini coefficient and Theil index before and after participation. Results are shown in

Table 6. Changes in Gini coefficient and Theil index before and after participation in Land Transfer and Non-farm Employment.

Group	Sample Size	Gini Coefficient		Change Rate	Theil Index		Change Rate	Conclusion
		Before	After		Before	After
Participated in neither	182	0.3473	0.3984	0.1472	0.2006	0.2724	0.3580	Widened income gap
Land Transfer only	349	0.2196	0.2547	0.1597	0.0790	0.1085	0.3737	Widened income gap
Non-farm Employment only	120	0.3387	0.2717	−0.1977	0.2109	0.1323	−0.3725	Reduced income gap
Participated in both	312	0.1901	0.1889	−0.0062	0.0625	0.0644	0.0297	No significant change in income gap

The change rate is calculated as (after − before)/before.

. For farmers only participating in Land Transfer, both indicators rise significantly. The Gini coefficient increases by 15.97%. The Theil index increases by 37.37%. This shows Land Transfer intensifies income differentiation. For farmers only participating in Non-farm Employment, both indicators drop significantly. The Gini coefficient decreases by 19.77%. The Theil index decreases by 37.25%. Income gaps narrow markedly. For farmers participating in neither activity, income gaps continue expanding. For farmers participating in both activities, income distribution remains relatively stable. These clear quantitative changes directly confirm the essential distributional differences between Land Transfer and Non-farm Employment. They further verify Hypotheses 4 and 5. They provide strong evidence for understanding distribution issues in agricultural transformation.

6. Discussion

6.1. Research Framework Innovation and Core Marginal Contributions

This study builds a comprehensive dual-mode comparative framework. It innovatively introduces “elite capture” and “inclusive growth” into the evaluation of distributional effects in modern agricultural parks. This represents a breakthrough in inclusive rural transformation research. Our method offers three advantages over existing literature [90,91,92]. First, it moves beyond the isolated examination of single farmer-linkage modes. It systematically reveals the mechanism differences between the two income growth paths. Second, through multi-dimensional heterogeneity analysis, it first quantitatively verifies the progressive distribution characteristics of Land Transfer and the inclusive poverty reduction effects of Non-farm Employment. Third, it discovers a fundamental tension between land factor marketization and labor factor empowerment. Land Transfer benefits skew toward rural elites and easily intensify rural class differentiation [93,94,95]. Non-farm Employment benefits more ordinary farmers and helps to narrow the income gap [96,97]. This finding provides new analytical perspectives for constructing equitable and sustainable agricultural systems. It has important academic value and theoretical significance.

6.2. Potential Interaction Between Land Transfer and Non-Farm Employment

This study treats Land Transfer and Non-farm Employment as independent income growth paths. The aim is to clearly identify their net effects. Yet in reality, the two modes may have complex complementary or substitutive relationships [98,99]. From the perspective of complementarity, surplus labor released by Land Transfer may supply workers for park-based Non-farm Employment. After rural elites profit from Land Transfer, if they invest in agricultural product processing industries, they may create more Non-farm jobs [100,101]. This forms a positive cycle of “Land Transfer—industrial upgrading—employment expansion.” From an alternative perspective, for elite farmers with sufficient capital and operational capabilities, the economies of scale from land transfer may prompt them to forgo Non-farm Employment opportunities, establishing a substitution relationship [102]. If elites monopolize Land Transfer benefits without investing in local industries or allocate resources only for self-interest, they may squeeze ordinary farmers’ employment space [103,104]. For ordinary farmers who cannot secure stable non-agricultural jobs, they may face livelihood crises due to land loss [105], in which case the two approaches exhibit forced substitution rather than positive complementarity. In this study, the Gini coefficient of the 312 farming households participating in both types of activities showed no significant change, suggesting that the distribution effects of the two pathways may offset each other within the group. However, future systematic testing using an interaction effect model is still required to examine whether the two pathways exhibit synergistic, substitutive, or crowding-out effects.

6.3. The Potential Risks of Non-Farm Employment

The current non-agricultural employment participation rate in the park stands at only 44.86%, indicating a potential gap in job supply. If farmers cannot secure stable non-agricultural employment after transferring their land rights, they will face the livelihood dilemma of “having no land to cultivate and no work to do.” Long-term non-agricultural employment suffers from low-wage traps and inadequate social security coverage, while seasonal employment leads to fluctuating income levels and stagnant skill development, exacerbating “employment poverty” [106]. Should an economic crisis disrupt the job market, the intergenerational loss of agricultural skills leaves farmers unable to return to farming while lacking alternative income sources, exposing them to systemic risks [107]. These risks suggest that the “inclusiveness” of non-agricultural employment primarily manifests in entry barriers rather than equal quality standards. Policies should expand employment opportunities while optimizing job structures, enhancing social security integration, and establishing crisis safeguards to prevent non-agricultural positions from becoming new arenas for elite capture.

6.4. Methodological Advancement and Targeted Policy Implications

Methodologically, this study advances regional agricultural economic research from average effects to distributional effects [108,109]. It does so through systematic mechanism testing and group heterogeneity decomposition. This progressive analysis not only strengthens the “efficiency-equity” trade-off theoretical framework but also provides policymakers with operable dual-mode synergy allocation schemes. While encouraging Land Transfer to boost efficiency, Non-farm Employment must be strengthened to ensure fairness [110,111]. The study also reveals the moderating roles of park grade, geographic location, and policy support. This highlights the importance of context-specific institutions.

6.5. Research Extensions and Future Directions

This study provides clear and targeted directions for subsequent research on sustainable agricultural systems and inclusive rural transformation. Future research can deepen in three directions. First, collect panel data. Track long-term livelihood trajectories and intergenerational income mobility of farmers after different park participation modes. Second, expand the definition of rural elites. Include hidden resources (social capital, information networks, psychological cognition) in the analytical framework. This more comprehensively depicts diverse mechanisms of “elite capture” and their deep impacts on community power structures. Third, this study treats Land Transfer and Non-farm Employment as independent pathways and fails to adequately examine their interaction effects in the model; future research could enhance the analysis by employing structural equation modeling or similar approaches. Fourth, conduct cross-regional comparative studies. Explore how regional heterogeneity moderates the effects of the two park participation modes. Deepening these research directions will help build a fairer, more sustainable agricultural modernization practice path. This path will truly benefit vast numbers of smallholders.

7. Conclusions and Policy Recommendations

7.1. Research Conclusions

This study evaluates the income effects and distribution logic of two park participation modes using survey data from 963 households across 76 modern agricultural parks in Sichuan Province. The core findings are the following:

First, both Land Transfer and Non-farm Employment significantly boost farmers ‘income, but their effects differ in nature. Non-farm Employment yields a stronger average income increase, particularly pronounced among low- and middle-income farmers, demonstrating a more inclusive benefit pattern; in contrast, the income enhancement from Land Transfer grows with farmers’ income quintile, with benefits concentrating primarily among high-income households.

Second, the mechanisms differ, leading to different group adaptability. Land Transfer relies on two pathways: “economies of scale” and “labor force release.” However, the scale benefits are more likely to be captured by rural elites with resource advantages, while ordinary farmers primarily benefit by transferring their land to release labor capacity. Non-farm Employment, on the other hand, delivers empowerment through three pathways: wage increases, skill acquisition, and job stability. Its core mechanisms better align with the needs of ordinary farmers while offering relatively limited benefits to the elite group.

Third, income distribution exhibits heterogeneous tendencies. Land Transfer exhibits a “capture by elites” tendency, with the benefits more concentrated among rural elites and high-income and large-crop farmers. This reflects the potential inequality risks brought about by the “concentration of the right to operate”. Non-farm Employment, on the other hand, demonstrates a stronger “inclusive growth” tendency, which can more effectively benefit low-income, small-crop, and low-education-level ordinary farmers, providing them with an income-generating channel that is not dependent on initial endowments and helping to alleviate income inequality. These two distribution logics are not absolutely mutually exclusive in practice but rather represent two dominant path dependencies and risk tendencies.

7.2. Policy Recommendations

Based on research findings, this study proposes a “trinity” institutional synergy scheme. It aims to achieve a dynamic balance between efficiency and fairness.

(1) Land System Level: Promote the “guaranteed base income plus dividend by shares” model. Mandate that agricultural park projects set minimum shareholding ratios for ordinary farmers. This ensures they share land value-added benefits and prevents elite monopolies. The experience of “Land Share Cooperatives” can be adopted, where farmers contribute their contracted land-management rights as shares to receive guaranteed dividends [112]. A minimum dividend ratio should be established, with joint supervision of fund flows conducted by the village committee and the platform company. This model not only reduces implementation costs but also qualifies for special fiscal subsidies.

(2) Employment Promotion Level: In areas with concentrated Land Transfer, build labor-intensive processing or service industries. Embed “work-based training” skill-upgrading programs. This converts released labor into local Non-farm Employment. It forms a positive cycle of “transfer promotes employment.” For remote farmers without park coverage, efforts should be made to move the “county employment service center” down to the townships and integrate rural public welfare positions with cross-regional labor matching. Drawing on Zhejiang’s “Mountain-Sea Cooperation” model [113], a targeted job placement mechanism from developed regions should be established to ensure these households can share the benefits of non-agricultural employment.

(3) Linkage Mechanism Level: Establish one-stop service platforms for “land exit—employment resettlement—social security continuation.” This reduces livelihood vulnerability during the farmer transition. Simultaneously, proactive contingency plans for adverse scenarios must be established. Prior to entering the industrial park, enterprises shall specify a minimum year-round employment ratio and implement a cross-enterprise seasonal job allocation mechanism. Incorporate flexible employment personnel into urban and rural residents’ pension insurance systems; activate emergency mechanisms during economic crises, allowing unemployed rural households to redeem their operational rights at original prices. Implement a supplementary “work-for-relief” emergency employment channel to mitigate the risk of “loss of land and income.”

This systematic design, through the coordinated efforts of land revenue sharing, employment expansion, empowerment, and risk guarantee, can not only leverage the scale efficiency of Land Transfer but also rely on non-agricultural employment to ensure fairness. It provides institutional support for building a fair and sustainable agricultural system path.