How Rural Labor Aging Affects Agricultural Productive Service Development: Threshold Effects and the Transmission Mechanism of Inefficient Cultivated Land Scale Management

Abstract

The accelerating aging of China’s rural labor force—characterized by an ‘earlier and more severe’ trajectory in rural areas compared to urban ones—poses a fundamental challenge to agricultural modernization, yet its implications for the development of agricultural productive services remain poorly understood. This study addresses a central question for agricultural sustainability: does rural labor aging act as a ‘catalyst’ or a ‘stumbling block’ for the development of agricultural productive services? To address this question, we employ fixed-effects, panel threshold regression, and mediation models using provincial panel data from 2011 to 2022, systematically investigating the nonlinear effects and transmission mechanisms. A significant single threshold is identified at an aging rate of 8.17%. Below this threshold, labor aging promotes productive service development via a demand-pull effect; above it, the promoting effect weakens substantially due to an adoption-inhibition effect driven by human capital decline. Mechanism analysis reveals a negative mediating role of cultivated land scale management: land concentration induced by labor aging, without corresponding human capital upgrades, is associated with “inefficient scale management” as reflected in the suppression of service development. While our empirical measure captures operational scale rather than efficiency per se, this negative mediation pattern is consistent with the presence of an inefficiency trap. Heterogeneity analysis shows the strongest inhibitory effect in northeast China, while eastern regions with strong factor substitution capacity are less affected. As most provinces had crossed the threshold by 2022, policies should shift from leveraging the dividend period to addressing deep-aging challenges, including breaking the trap of inefficient scale management and adopting region-specific strategies. These findings highlight that achieving agricultural sustainability in an aging society requires shifting from relying on the demographic dividend to proactively overcoming the structural lock-in of deep aging. Our proposed policies could contribute to fostering a more resilient and sustainable agricultural system in China, and may also offer valuable references for other rapidly aging economies with appropriate contextual adaptation.

1. Introduction

The core of agricultural modernization lies in the systematic transformation of agricultural production modes, with a key indicator being the evolution from traditional manual labor to a modern industrial paradigm dominated by specialized and socialized services [1]. Agricultural productive services serve as a bridge linking smallholders with modern agriculture [2], covering the entire chain of farming, planting, management, harvesting, storage, and marketing. Vigorously developing agricultural productive services is not only an intrinsic requirement for extending and upgrading the agricultural value chain, but also a strategic pathway to address resource and environmental constraints and to achieve the organic connection between smallholders and modern agriculture.

During this transformation, China is experiencing profound demographic changes, and the aging of the rural labor force has become an irreversible trend. As young laborers continue to migrate to urban areas, the rural labor supply exhibits the triple characteristics of “declining quantity, aging population, and deteriorating quality”, forming an aging pattern that is “earlier and more severe in rural areas than in urban areas” [3,4]. This pattern not only directly alters the factor endowment of agricultural labor, but also deeply affects the process of agricultural modernization by changing farmers’ behavioral patterns, technology adoption preferences, and factor allocation methods. Rural labor aging has thus become a key realistic constraint on agricultural transformation.

In this context, how does rural labor aging affect the development of agricultural productive services—is it a “catalyst” or a “stumbling block”? On one hand, the substitution effect triggered by labor scarcity “forces” farmers to outsource production tasks, creating rigid demand for services such as mechanization, plant protection, and land trusteeship, thereby promoting the expansion of agricultural productive services. On the other hand, an “adoption-inhibition” effect caused by human capital decline also emerges: due to reduced learning ability, increased risk aversion, and lower trust in green technologies and new models, older farmers tend to choose traditional and basic services, and lack sufficient willingness to adopt high-level services such as intelligent, green, and integrated ones. This may lead service development into a “low-level cycle” or even induce “inefficient scale management”, thereby restraining the upgrading of agricultural productive services. The interplay of these two forces implies that the impact of rural labor aging on productive services may not be simply linear, but may exhibit stage-wise or threshold characteristics as the degree of aging deepens, with the net effect depending on the balance between the two mechanisms.

Therefore, this paper attempts to fill the above research gap by addressing two interrelated questions: First, does the impact of rural labor aging on agricultural productive services exhibit a nonlinear threshold effect, and if so, at what critical point does the nature of this impact change? Second, through what mechanism does rural labor aging affect service development—specifically, does cultivated land scale management serve as a negative mediating channel? To answer these questions, we use provincial panel data from China for the period 2011–2022 and employ fixed-effects models, panel threshold regression models, and mediation models to empirically investigate the nonlinear impact of rural labor aging on agricultural productive services and its transmission mechanism. The marginal contributions of this paper are threefold. First, it identifies and verifies that this impact is not simply linear but exhibits a threshold effect, and uses a critical value (8.17%) to distinguish a “dividend period” from a “challenge period”. Second, it reveals the negative mediating role of cultivated land scale management, deepening the understanding of the complex process through which rural labor aging affects agricultural productive services. Third, by conducting heterogeneity analyses based on geographical location and topographical conditions, it provides targeted decision-making references for building a differentiated and precise agricultural socialized service system in the context of deep aging. By identifying a precise threshold and the mediating role of land scale management, this study contributes to the broader sustainability literature by revealing the conditions under which rural aging can be reconciled with agricultural modernization. Our findings underscore that sustainable agricultural development in the context of an aging population requires deliberate, stage-specific policy interventions that prevent scale expansion from becoming inefficient. This research thus provides actionable insights for achieving the United Nations Sustainable Development Goals (SDGs), particularly Goal 2 (Zero Hunger) and Goal 8 (Decent Work and Economic Growth), in the Chinese context. While the findings may have implications for other rapidly aging agrarian economies, further empirical validation beyond China would be necessary before generalizing these conclusions.

2. Literature Review and Theoretical Framework

2.1. Literature Review

2.1.1. Connotation and Driving Forces of Agricultural Productive Services

(1)

Connotation of agricultural productive services

Existing studies suggest that the essence of agricultural productive services is a socialized division of labor system that provides specialized and market-oriented services for each link of agricultural production. First, digital technologies such as big data and artificial intelligence reshape the service form, making services digital, intelligent, and networked, with the core goal of improving the synergy and resilience of the industrial chain and supply chain [5]. Second, some studies focus on introducing green technologies (e.g., precision fertilization) into the production process through specialized services, aiming to promote low-carbon agricultural development and enhance total factor productivity [6,7]. Third, organizational carriers such as village collectives and cooperatives aggregate the scattered demand of smallholders, with the core objective of reducing transaction costs and facilitating the combination of advanced production factors with traditional agriculture [8,9].

(2)

Driving forces of agricultural productive service development

First, national strategies and policy guidance are the fundamental driving forces. Top-level designs such as the “Digital Village” strategy and the “dual carbon” goals provide clear strategic directions for the development of agricultural productive services [5,7,10]. Second, addressing practical bottlenecks is crucial. Long-standing structural contradictions such as rural labor aging, off-farm migration, land fragmentation, and agricultural non-point source pollution are important reasons that restrict the innovation and diffusion of agricultural productive services [8,11]. Among these, rural labor aging, as a key structural contradiction, both creates service demand and may inhibit service upgrading through human capital constraints—this is the core issue on which this paper focuses. Third, technological progress provides support. The improvement in digital infrastructure and the popularization of smart equipment have created new service formats, improved service efficiency, and reduced service costs. At the same time, the trend of rural land transfer and scale management has also created factor conditions for agricultural productive services [12,13].

Recent empirical evidence further corroborates the role of agricultural productive services in driving agricultural modernization. Using provincial panel data from China spanning 2000 to 2024, Bo [14] finds that agricultural socialized services significantly suppress agricultural carbon emissions, and this effect operates partially through the channel of promoting agricultural modernization. The study also reveals that the emission-reduction effect is particularly pronounced in major grain-producing regions and varies nonlinearly with the stage of economic development. These findings reinforce the view that agricultural productive services, as an institutional arrangement facilitating scale and specialization, not only address practical production bottlenecks but also serve as a strategic instrument for achieving green agricultural transformation—a point closely aligned with the “dual-carbon” policy context discussed earlier.

Furthermore, recent studies have revealed nonlinear patterns in the relationship between agricultural services and productivity. For instance, using county-level data, ref. [15] finds an inverted U-shaped relationship between agricultural socialized services and agricultural green total factor productivity, and shows that land scale management significantly moderates this relationship—as land scale increases, the inflection point of the inverted U-curve shifts leftward and the curve flattens. This implies that blindly pursuing land scale without corresponding technological and managerial capabilities may weaken the positive effects of services, which is highly consistent with the “inefficient scale management” trap proposed in this paper. Another study focusing on Jiangxi Province [16] demonstrates the spatial spillover effects of agricultural productive services, indicating significant regional heterogeneity in service impacts. Together, these findings suggest that the relationship between agricultural services and productivity is nonlinear and spatially heterogeneous, which indirectly supports the possibility that the impact of rural labor aging on services may also exhibit nonlinear patterns depending on multiple conditions.

2.1.2. Impact of Rural Labor Aging on Agricultural Productive Services

(1)

Constraint effect

With increasing age, farmers’ physical strength and energy irreversibly decline, making it difficult for older farmers to undertake high-intensity tasks such as plowing, planting, managing, and harvesting. Labor scarcity and the relative rise in labor prices lead to the substitution of other factors [17]. Therefore, labor aging provides a basic “demand-pull” for the development of agricultural productive services: older farmers, to compensate for their own labor decline, are forced to outsource production tasks, thus creating rigid demand for services such as mechanized operations [18,19]. This path explains the logical starting point for labor aging to act as a “catalyst” for the development of agricultural productive services.

A recent study on wheat production further refines the understanding of this constraint effect. Wang and Zhong [20] find that the relationship between agricultural labor aging and wheat production efficiency is not linear but exhibits an inverted U-shaped pattern: production efficiency peaks when the average age of a household’s agricultural labor force approaches 55 and declines thereafter. Importantly, agricultural mechanization significantly moderates this relationship, alleviating the negative effects of labor aging on production efficiency [20]. This finding directly supports the “substitution effect” logic: mechanization—as a typical productive service—can effectively compensate for the physical decline in older farmers, but only up to a certain point. Beyond that threshold, the cognitive and behavioral constraints of aging become increasingly difficult to offset through mechanization alone.

(2)

Cognitive and preference effect

Based on the life-cycle theory [21], as farmers age, their human capital (learning ability, risk preference) changes, making production decisions more conservative. This manifests in three aspects:

First, risk preference declines significantly. Older farmers exhibit stronger risk aversion in agricultural production decisions, thus tending to adopt traditional varieties and production methods that are technologically mature and market-stable, while being cautious or even resistant to new technologies and service models with uncertain returns (e.g., digital agricultural services, green production services) [22,23].

This mechanism has been directly verified at the micro level. Using farm-level data from corn farmers, Mao et al. [24] find that agricultural labor aging significantly inhibits both the adoption of improved varieties and the number of improved varieties adopted. The study further reveals that aging inhibits adoption behavior through two channels: increasing risk aversion and reducing farm scale [24]. This evidence directly supports the “cognitive and preference effect” logic—older farmers tend to choose lower-risk, traditional options and are reluctant to adopt new technologies that require higher cognitive engagement. At the same time, using provincial panel data from China, Huang et al. [25] find that rural labor aging enhances forestry economic resilience by promoting large-scale forest land management, driving forestry technological innovation, and increasing government capital investment, but it also weakens forestry economic resilience by reducing educational and health human capital.

Second, learning ability and cognitive limitations become prominent. Facing increasingly digital and intelligent modern agricultural productive services (e.g., drone plant protection, smart irrigation), older farmers encounter high cognitive and learning barriers, exhibiting a significant “digital divide” that directly inhibits their ability and willingness to adopt high-level services [26,27].

Third, production decision “path dependence” is reinforced. The experience accumulated over long-term practice becomes the main basis for their decisions, leading them to trust personal empirical judgments rather than professional service recommendations based on scientific data (e.g., adopting soil-testing formula fertilization services). This behavior solidifies traditional production methods and constitutes an “adoption-inhibition” force for the upgrading of productive services.

Recent micro-level evidence from Jiangsu Province [28] further refines this picture. The study finds that overall participation in agricultural productive services significantly improves farmers’ technical efficiency, but the effects are stage-specific: services for land preparation, sowing, and harvesting are associated with higher efficiency, while services for fertilization and pesticide application do not consistently improve efficiency and may reflect potential overuse of chemical inputs. This suggests that the “adoption-inhibition” effect may be particularly acute for certain types of services (e.g., those requiring precise decision-making), whereas basic mechanization services are readily adopted even by older farmers. Such stage-specific heterogeneity reinforces the importance of examining service types separately, as we do in our empirical analysis of different service categories (though our main measure aggregates overall service development).

Furthermore, some studies point out that the “adoption-inhibition” effect is not uniform across all service types. For example, mechanized services such as land preparation, sowing, and harvesting significantly improve the technical efficiency of older farmers, while services requiring precise decision-making (e.g., fertilization and pesticide application) may lead to overuse due to cognitive limitations [28]. This indicates that older farmers’ resistance to services is mainly concentrated on high-order services that require greater cognitive involvement. At the same time, Ding and Zhao [29] find that agricultural productive services have a more significant promoting effect on the adoption of cultivated land protection technologies by farmers with a high degree of labor aging, suggesting that services themselves can become a tool to overcome aging constraints, thereby making the net effect nonlinear.

This nonlinear pattern is further corroborated by a large-scale study on grain production. Hu et al. [30] find that agricultural labor aging generally inhibits both grain production efficiency and grain eco-efficiency across different grain crops, but the impact exhibits crop-specific nonlinear characteristics: a positive U-shaped relationship for maize, an inverted U-shaped relationship for rice, and no significant nonlinear relationship for wheat [30]. This finding provides important evidence that the relationship between aging and agricultural performance is not only non-linear but also varies across different crop systems—a form of heterogeneity that mirrors our regional heterogeneity analysis. It reinforces the importance of examining threshold effects and crop- or region-specific patterns rather than assuming a uniform linear relationship.

2.1.3. Summary of the Literature

Despite the richness of the above literature, two critical gaps remain unresolved. The above literature review shows that there are two opposite behavioral effects of rural labor aging on agricultural productive services, which co-exist and interact in reality. However, most studies either list the two effects in parallel or focus on only one of them. They fail to systematically explain, within a unified analytical framework, how these two effects dynamically evolve as aging deepens, nor have they tested whether the net effect exhibits threshold characteristics; moreover, they have not revealed the potential negative mediating role of cultivated land scale management in this process. In other words, under what conditions does labor aging mainly manifest as “demand-pull,” and at what stage does “adoption-inhibition” become dominant, so that the overall impact shows non-linearity? This question has not been fully answered.

To fill the above research gap, this paper constructs an analytical framework that incorporates a non-linear threshold and a behavioral transmission mechanism, aiming to dissect the complex process through which rural labor aging affects the servitization level of agriculture.

2.2. Theoretical Framework and Research Hypotheses

These two mechanisms constitute the theoretical backbone of our analysis and are revisited across different sections to demonstrate their implications at various stages of the empirical investigation.

Based on the dual-mechanism framework of “demand-pull” and “adoption-inhibition”, the impact of rural labor aging on the development of agricultural productive services is the result of the dynamic interplay of two opposing forces.

On one hand, labor aging, through physical decline-induced labor scarcity, generates rigid demand for basic productive services such as mechanized operations and plant protection, “forcing” the expansion of service supply. Thus, in the early stage it mainly shows a promoting effect. On the other hand, as aging deepens, the decline in cognitive ability and risk preference of older farmers, along with higher technology learning costs and low trust in emerging services, make farmers more inclined to choose traditional basic services and lack sufficient adoption of high-quality services, thereby weakening the promoting effect. The relative strength of these two effects may change with the degree of aging itself, resulting in a non-linear overall impact.

From a transmission perspective, the reduction in labor quantity and physical capacity prompts some farmers to exit agricultural production or transfer out their land, promoting the concentration of cultivated land among new agricultural operators and forming scale management. However, when scale management remains at the superficial level of “land factor concentration” without corresponding technological innovation, management upgrading, and service coordination, it may fall into the trap of “inefficient scale management”. Conceptually, ‘inefficient scale management’ refers to a situation in which operational scale expands (i.e., more land is cultivated per laborer) without proportional improvements in human capital, managerial capacity, or technological adoption. In other words, it is not scale expansion per se that is problematic, but the quality deficit accompanying that expansion. Since this inefficiency is inherently a latent construct, it cannot be observed directly; however, the expansion of scale under conditions of aging—where human capital is declining—serves as a proxy signal for the likely presence of inefficiency. Specifically, if land concentration occurs primarily because younger, more educated farmers exit agriculture while older farmers remain and expand their cultivated area, the resulting scale expansion is more likely to be ‘inefficient’—extensive rather than intensive—relying on traditional practices rather than modern services. Thus, while our empirical measure (cultivated land area per laborer) captures the scale dimension, its mediating role in the aging–APS relationship is interpreted as evidence for the inefficiency trap because aging-induced scale expansion, as well as absent human capital upgrading, tends to suppress service adoption. It is important to emphasize that we are inferring inefficiency from the observed negative mediation pattern, not measuring it directly.

On one hand, simple land scale expansion may homogenize service demand and inhibit the diversification and upgrading of service supply. On the other hand, scale management dominated by older farmers, constrained by human capital, finds it difficult to effectively access and adopt high-quality productive services. Therefore, cultivated land scale management does not act as a positive transmission channel for the development of agricultural productive services, but rather plays a negative mediating role by weakening the positive impact of labor aging through “inefficient scale management”.

Based on the above analysis, this paper proposes the following two research hypotheses:

H1. 

The impact of rural labor aging on the development of agricultural productive services exhibits a non-linear threshold effect.

H2. 

Cultivated land scale management plays a negative mediating role in the process through which rural labor aging affects the development of agricultural productive services.

3. Materials and Methods

3.1. Model Specification

3.1.1. Benchmark Regression Model

To examine the baseline impact of rural labor aging on the development of agricultural productive services, we specify the following two-way fixed-effects model:

$${\mathrm{A}\mathrm{P}\mathrm{S}}_{\mathrm{i}\mathrm{t}\mathrm{ }}={\alpha }_1+{\alpha }_2\times {\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}+{\alpha }_3\times {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}\mathrm{ }}+{\mu }_{\mathrm{i}}+{\lambda }_{\mathrm{t}}+{\epsilon }_{it}$$

(1)

where ${\mathrm{A}\mathrm{P}\mathrm{S}}_{\mathrm{i}\mathrm{t}}$ denotes the development level of agricultural productive services in province i in year t; ${\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}$ is the core explanatory variable representing rural labor aging. ${\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}}$ is a vector of control variables including the proportion of disaster-affected areas (disaster_ratio), human capital level (human_capital), effective irrigated area (ln_irrigation), urbanization rate (urbanization), and land transfer rate (land_transfer), ${\mu }_{\mathrm{i}}$,${\lambda }_{\mathrm{t}}$ are province- and year-fixed effects, respectively, capturing time-invariant provincial characteristics and common time trends; ${\epsilon }_{it}$ is the idiosyncratic error term. The coefficient ${\alpha }_2$ is our main focus.

3.1.2. Mediation Test Model

According to the theoretical analysis, we hypothesize that rural labor aging affects agricultural productive services through the channel of cultivated land scale management. To test this mediating mechanism, we follow the classic mediation analysis procedure proposed by Baron and Kenny [31] and construct the following model:

$$\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}\_\mathrm{o}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}={\beta }_1+{\beta }_2{\times \mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}+{\beta }_3\times {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}\mathrm{ }}+{\gamma }_{\mathrm{i}}+{\theta }_{\mathrm{t}}+{\nu }_{\mathrm{i}\mathrm{t}}$$

(2)

where $\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}\_\mathrm{o}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}$ represents the degree of cultivated land scale management. The definitions of ${\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}$ and ${\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}}$ are identical to those in Equation (1); ${\gamma }_{\mathrm{i}}$, ${\theta }_{\mathrm{t}}$ are province- and year-fixed effects, respectively.

3.1.3. Panel Threshold Regression Model

To systematically test for possible nonlinear effects of aging on agricultural productive service development, we employ the panel threshold regression model developed by Hansen [32]. The determination of the number of thresholds follows a general-to-specific principle: we first test the null hypothesis of a single threshold; if rejected, we proceed to test for double thresholds, and so on, until the null hypothesis for a higher-order threshold cannot be rejected. This stepwise procedure ensures the scientific validity of the model specification and the robustness of the conclusions.

The single-threshold model is specified as

$${\mathrm{A}\mathrm{P}\mathrm{S}}_{\mathrm{i}\mathrm{t}\mathrm{ }}={\mathrm{Ƴ}}_0+{\mathrm{Ƴ}}_1\times {\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}\times \mathrm{I}\left({\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}\le \eta \right)+{\mathrm{Ƴ}}_2\times {\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}\times \mathrm{I}\left({\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}>\eta \right)+{\mathrm{Ƴ}}_3\times {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}\mathrm{ }}+{\gamma }_{\mathrm{i}}+{\theta }_{\mathrm{t}}+{\nu }_{\mathrm{i}\mathrm{t}}$$

(3)

where η is the threshold value, and I(⋅) is an indicator function that equals 1 if the condition in parentheses is satisfied and 0 otherwise. This specification allows the marginal effect of the core explanatory variable ${\mathrm{a}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}}_{\mathrm{i}\mathrm{t}}$ on the dependent variable ${\mathrm{A}\mathrm{P}\mathrm{S}}_{\mathrm{i}\mathrm{t}}$ to differ across regimes. ${\mathrm{Ƴ}}_1$ applies when the aging level is at or below the threshold, ${\mathrm{Ƴ}}_2$ applies when it exceeds the threshold.

3.2. Variable Definitions and Data Sources

3.2.1. Dependent Variable

Agricultural productive services (APS): Agricultural productive services cover the entire chain of agricultural production and represent a highly specialized form of social services. Following Dai et al. [7], we use the ratio of the output value of agriculture, forestry, animal husbandry, and fishery services (i.e., the value of service activities provided to agricultural production) to the gross output value of agriculture, forestry, animal husbandry, and fishery (i.e., the total value of all agricultural products and services produced) as the proxy variable for agricultural productive services. This ratio reflects the relative importance of service activities within the overall agricultural economy. The numerator captures the monetary value of specialized service inputs (e.g., mechanization, plant protection, technical advisory), while the denominator represents the total value of agricultural production. A higher ratio indicates a greater degree of servitization in agricultural production.

3.2.2. Independent Variable

Rural labor aging (aging): The age structure of the labor force is a key factor influencing agricultural production decisions. Referring to Ma [18], we measure the degree of rural aging by the proportion of the rural population aged 65 years and above to the total rural population, thereby reflecting the aging trend of the rural labor structure.

3.2.3. Control Variables

To account for potential confounding factors, we introduce the following control variables:

Human capital (human_capital): Measured by the ratio of the number of enrolled students in higher education institutions to the year-end permanent resident population in each province.

Disaster area ratio (disaster_ratio): Measured by the ratio of the area affected by natural disasters to the total sown area of crops.

Land transfer rate (land_transfer): Measured by the ratio of the total area of household contracted cultivated land transferred to the area of household contracted cultivated land under management.

Effective irrigated area (ln_irrigation): Natural logarithm of the effective irrigated area.

Urbanization rate (urbanization): Urban population as a share of total population.

3.2.4. Mediating Variable

Cultivated land scale operation (Scale_operation): Cultivated land scale operation (Scale_operation). Cultivated land scale management is generally considered to be conducive to improving agricultural production efficiency. Using farm-level data from China, Zhang et al. [33] find that renting in land increases agricultural labor productivity by about 43%, with the expansion of operational scale being the most important mediating pathway.

Therefore, following Li et al. [34], we use the cultivated land area per laborer as a measure of the degree of scale operation. The formula is as follows: cultivated land area per laborer = total cultivated land area/number of persons employed in the primary industry. This variable is used to test the mediating role in the relationship between rural labor aging and agricultural productive services.

3.2.5. Data Sources

This study uses provincial-level panel data from 2011 to 2022, covering 30 provinces (excluding Tibet, Hong Kong, Macao, and Taiwan due to data availability. Tibet is excluded due to severe data gaps in key variables such as rural labor age structure and agricultural service output value; Hong Kong, Macao, and Taiwan are excluded because their statistical systems and agricultural economic structures are not comparable with mainland provinces). Data preprocessing includes the following: (1) linear interpolation was applied to fill a small number of missing values in certain years to maintain time-series continuity; (2) to avoid the potential influence of extreme values, all continuous variables were winsorized at the 1st and 99th percentiles. The data are obtained from the China Rural Statistical Yearbook, China Population and Employment Statistics Yearbook, Rural Statistical Yearbook, China Agricultural Mechanization Statistical Yearbook, and provincial annual data from the National Bureau of Statistics of China.

The data used in this study are publicly available from the above sources. The processed dataset and computer codes are available from the corresponding author upon reasonable request.

3.3. Descriptive Statistics

Table 1. Descriptive statistics of variables.

Variable	Symbol	Mean	Std. Dev.	Min	Max
Agricultural productive services	APS	0.042	0.019	0.012	0.106
Rural labor aging	aging	0.133	0.045	0.05	0.275
Rural labor aging	Old_dependency	0.197	0.077	0.07	0.469
Effective irrigated area (ln)	Ln_irrigation	7.283	1.057	4.693	8.805
Urbanization	urbanization	0.601	0.121	0.35	0.896
Human capital	human capital	0.021	0.006	0.008	0.044
Land transfer rate	land_transfer	0.332	0.69	0.049	0.873
Disaster area ratio	disaster ratio	0.165	0.223	0.004	1.666
Cultivated land scale operation	Scale_operation	6.866	5.399	1.530	34.495

The old dependency ratio is not used in the main models but is reported for potential robustness checks.

presents the descriptive statistics of the main variables, including the mean, standard deviation, minimum, and maximum values.

4. Empirical Results Analysis

4.1. Benchmark Regression Results

Table 2. Benchmark regression results.

	Agricultural Productive Services
	(1)	(2)
aging	0.036	0.061
	(0.034)	(0.080)
Control variables	NO	YES
Provincial FE	YES	YES
Year FE	YES	YES
Constant	0.005	0.017
	(0.006)	(0.048)
N	360	360
R2	0.244	0.466

Note: Cluster-robust standard errors are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01. The same applies to the following tables.

reports the benchmark regression results of the impact of rural labor aging on agricultural productive services. To control for unobserved individual and time factors, all models employ two-way fixed effects.

As shown in Table 2, regardless of whether control variables are included, the coefficient of rural labor aging is positive but not statistically significant. This indicates that, within the sample period of this study, there is no reliable statistical evidence that a deepening of labor aging directly affects the development of agricultural productive services—in other words, there is no simple linear relationship between rural labor aging and agricultural productive services. This result is consistent with our hypothesis H1: labor aging simultaneously triggers two opposite forces—“demand-pull” and “adoption-inhibition”—and their net effect may cancel each other out or exhibit a more complex nonlinear pattern. This provides a crucial logical starting point for further exploring the threshold effect and transmission mechanism behind it.

4.2. Endogeneity Tests

Although the benchmark regression controls for a series of variables and two-way fixed effects, the model may still suffer from endogeneity bias due to omitted variables, measurement errors, or reverse causality. To mitigate possible endogeneity (especially reverse causality), we employ two complementary instrumental variable (IV) approaches.

4.2.1. Lagged Aging as an Instrumental Variable

First, we re-estimate the model using an IV approach, taking the second and third lags of the aging indicator as instruments.

Table 3. Instrumental variable estimation results (lagged aging).

Agricultural Productive Services
aging
L.aging	0.193 *
	(0.102)
Control variables	YES
Provincial FE	YES
Year FE	YES
N	270
Kleibergen–Paap rk Wald F	10.247
Hansen J p	0.743
Kleibergen–Paap rk LM χ2(2)	26.919
Cragg–Donald F	22.577
Stock–Yogo 10% critical value	19.93
Hansen J	0.107

Note: L.aging denotes the instrument constructed from the second and third lags of aging. * p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are in parentheses.

reports the IV (2SLS) estimation results based on these lagged terms. This choice is justified by the following considerations: the current agricultural industrial structure cannot affect the past age structure of the population, satisfying the exogeneity condition; meanwhile, population age structure is persistent, and past aging levels are highly correlated with current levels, satisfying the relevance condition.

We acknowledge that lagged values of aging may be correlated with long-term provincial characteristics such as economic development trajectories, urbanization patterns, or historical agricultural policies, which could in principle violate the exclusion restriction if these characteristics also affect current APS independently of their effect through aging. To address this concern, we include a rich set of time-varying controls and two-way fixed effects to absorb time-invariant provincial heterogeneity. The identifying assumption, therefore, is that conditional on these controls and fixed effects, the lagged aging variables affect APS only through current aging. This is plausible because the main channel through which past age structure affects current agricultural service demand is via its persistence in determining current labor force composition, rather than through direct historical influences on service markets. Nevertheless, we supplement the lagged-instrument approach with a Bartik-type instrument that relies on a different source of exogenous variation, and the consistency of results across these two strategies strengthens our confidence in the main findings.

As shown in Table 3, the identification tests for the lagged instruments (Kleibergen–Paap rk LM p = 0.000, F = 10.247, Hansen J p > 0.1) indicate that the model is properly identified and the instruments are valid. After controlling for endogeneity, the positive effect of labor aging on agricultural productive services becomes more significant in magnitude and significance level.

However, it should be noted that the Kleibergen–Paap rk Wald F statistic (10.247) is below the Stock–Yogo 10% critical value (19.93), suggesting a potential weak instrument problem. This limitation is addressed by the second IV strategy (Bartik-type instrument) reported in Section 4.2.2.

4.2.2. Bartik-Type Interaction Instrumental Variable

The Bartik-type instrument is constructed as follows. Let ${aging}_{i,2010}$ denote the rural labor aging rate in province i in the base year 2010, and let $\overline{{ aging}_t}$ denote the national-level rural labor aging rate in year t (2011–2022). The instrument for province i in year t is defined as

$${IV}_{it}={aging}_{i,2010}\times \overline{{aging}_t}$$

(4)

The intuition behind this construction is that the national aging trend ($\overline{{aging}_t})$ captures time-varying, economy-wide demographic shifts that are exogenous to any single province, while the base-year provincial aging rate (${aging}_{i,2010})$ captures cross-sectional variation that is predetermined and unaffected by subsequent agricultural structural changes. Their product generates exogenous cross-provincial variation in the predicted exposure to the national aging trend.

Relying only on lagged aging has limitations: lagged values may partly reflect long-term regional economic structural differences, weakening exogeneity. To strengthen identification, we construct an interaction instrumental variable following the Bartik [35] shift-share approach. Specifically, we interact the provincial rural labor aging level in 2010 (the base year) with the national-level rural labor aging trend from 2011 to 2022. Their product captures the heterogeneous impact of the national aging trend across provinces, representing an exogenous, cross-provincial differential shock. On one hand, the base-period demographic structure is predetermined and unaffected by changes in agricultural industrial structure during the sample period (2011–2022), satisfying exogeneity; on the other hand, the national aging trend exhibits common time-series variation across provinces, ensuring sufficient correlation with local economic structural changes.

The exclusion restriction requires that the instrument affects agricultural productive services only through its effect on provincial labor aging, and not through any other channel. We argue that this condition is plausibly satisfied because the base-year aging rate is historically determined and is unlikely to be correlated with subsequent province-specific agricultural policy shocks or service market developments, except through its persistence in current aging levels. National demographic trends are driven by long-term fertility and migration patterns that are exogenous to agricultural service markets. The Hansen J tests reported in

Table 4. Instrumental variable estimation results (Bartik-type interactive instrument).

	First Stage	Second Stage
	First Stage	Agricultural Productive Services
aging		0.178 ***
		(0.048)
Iv_it	9.567 ***
	(0.904)
Control variables	YES	YES
Provincial FE	YES	YES
Year FE	YES	YES
Kleibergen–Paap LM	0.000	0.000
Cragg–Donald F	225.222	-
Kleibergen–Paap F	111.881	-
Weak instrument test (10% critical value)	16.38	-
F	111.88 ***	16.38
N	360	360
Centered R2	-	0.439

Note: IV_it is the Bartik-type interaction instrument. F-statistic in the first stage is 111.88, far exceeding the 10% critical value of 16.38, indicating no weak instrument problem. * p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are in parentheses.

(p > 0.1) do not reject the validity of the overidentifying restrictions, providing empirical support for the exclusion restriction.

The second-stage results further confirm the robustness of the benchmark regression: the coefficient of aging is 0.178, significant at the 1% level. This result corroborates the conclusion from the lagged instrument approach, confirming that after controlling for endogeneity, rural labor aging has a significant positive effect on agricultural productive services.

4.2.3. Summary of Endogeneity Tests

A natural question arises: why do the benchmark fixed-effects estimates (Table 2) show an insignificant coefficient for aging, whereas the IV estimates (Table 3 and Table 4) reveal a significant positive effect? This discrepancy is not a contradiction but may be partly attributable to attenuation bias due to endogeneity in the baseline FE specification. Possible sources of this bias include measurement errors in the aging variable, omitted time-varying confounding factors (e.g., unobserved regional policy shocks or market conditions), or reverse causality from agricultural service development back to labor migration patterns. In such cases, the fixed-effects estimator may be biased toward zero. The IV approach, by isolating the exogenous component of rural aging through shift-share variation (Bartik instrument) and lagged demographic structure, helps to mitigate these potential biases and provides a more reliable estimate of the causal parameter of interest. This is a well-documented phenomenon in the empirical literature: when the treatment variable is measured with error or subject to reverse causation, IV estimates tend to be larger in magnitude than FE estimates once the instruments are valid [36]. Given that our diagnostic tests (Kleibergen–Paap LM, Hansen J, and Cragg–Donald F) support the validity and relevance of our instruments, we interpret the IV estimates as the preferred causal evidence. The insignificant FE results, in turn, reflect the net balance of two opposing forces—demand-pull and adoption-inhibition—whose effects cancel out in a naive linear specification. This is precisely why a threshold model is needed to uncover the nonlinear structure underlying these two mechanisms.

4.3. Robustness Checks

We employ two robustness checks: replacing the core independent variable and excluding the four municipalities directly under the central government.

4.3.1. Alternative Variable Measurement

Regardless of whether control variables are included, the coefficient of the old dependency ratio (Old_dependency) on agricultural productive services is not statistically significant, which is consistent with the benchmark regression. This suggests that the underlying factors may exhibit offsetting or nonlinear characteristics—the “demand-pull” effect (aging forcing service substitution) and the “adoption-inhibition” effect (care burden crowding out investment in service upgrading) may counteract each other. This result also indicates that the deepening of rural labor aging is not a direct linear driver of agricultural productive services; instead, a nonlinear relationship may exist (

Table 5. Robustness check—replacing the independent variable.

	Agricultural Productive Services
	(1)	(2)
Old_dependency	0.070	0.059
	(0.053)	(0.049)
Control variables	NO	YES
Provincial FE	YES	YES
Year FE	YES	YES
Constant	0.025 ***	0.016
	(0.007)	(0.049)
N	360	360
R2	0.206	0.481

* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are in parentheses.

).

By replacing the core independent variable, this study confirms that, whether measured by population age structure or economic burden structure, the linear impact of labor aging on agricultural productive services is not statistically significant, which is consistent with the benchmark finding. This enhances the credibility of our theoretical framework and empirical results.

4.3.2. Excluding the Four Municipalities

To further verify whether the benchmark results are driven by specific observations, we exclude the four municipalities directly under the central government—Beijing, Shanghai, Tianjin, and Chongqing—from the sample. As provincial-level administrative units, these municipalities have high urbanization rates and relatively small agricultural shares, and their agricultural economic structures differ from those of ordinary provinces, potentially influencing the overall estimation results.

Table 6. Robustness check—excluding the four municipalities.

	Agricultural Productive Services
	(1)	(2)
aging	0.142	0.109
	(0.117)	(0.112)
Control variables	NO	YES
Provincial FE	YES	YES
Year FE	YES	YES
Constant	0.023 **	0.037
	(0.011)	(0.066)
N	312	312
R2	0.4034	0.456

* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are in parentheses.

shows that after excluding the four municipalities, the estimated coefficient of the core explanatory variable (aging) remains positive, small in magnitude, and statistically insignificant, which is largely consistent with the benchmark regression. This indicates that our main findings are not driven by a few special observations and are generally applicable.

4.4. Heterogeneity Analysis

To explore regional heterogeneity in the impact of rural labor aging on agricultural productive services, we divide the sample into four regions—eastern, central, western, and northeastern—following the classification of the National Bureau of Statistics of China, taking into account the distinct economic and demographic characteristics of the northeastern region. We also analyze separately based on topographical conditions (plain vs. non-plain areas).

4.4.1. Geographical Location

Table 7. Heterogeneity analysis by geographical location.

	Agricultural Productive Services
	Eastern	Central	Western	Northeastern
Aging	0.083	−0.343	0.204	−0.127 ***
	(0.070)	(0.277)	(0.173)	(0.001)
Control variables	YES	YES	YES	YES
Provincial FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
N	120	72	132	36
R2	0.7694	0.8268	0.3007	0.9645

* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are in parentheses.

reports the regional regression results. The impact of rural labor aging on agricultural productive services varies across regions: a significant negative effect is found in the northeast, a negative but insignificant effect in the central region, and positive but insignificant effects in the eastern and western regions. These differences reflect that the role of labor aging is moderated by regional factor endowments and development stages, which together determine the balance between the “demand-pull” and “adoption-inhibition” forces.

Specifically, the northeastern region exhibits a significant negative effect (coefficient = −0.127 ***). This provides strong evidence for the “inefficient scale management” trap proposed in this paper. Although the northeast has a nationwide leading foundation for land scale management, it has long suffered from severe out-migration and deep aging. As a result, the capacity of scale-management operators has not simultaneously upgraded, and human capital constraints are particularly prominent. Hence, the “adoption-inhibition” effect far outweighs the “demand-pull” effect, and the large scale of land actually reinforces reliance on traditional, low-level practices, suppressing the overall development of agricultural productive services.

The central region shows a negative but statistically insignificant effect (coefficient = −0.343). As a major grain-producing area, the central region has a relatively homogeneous agricultural production pattern. Labor aging may exacerbate this “structural lock-in”, making it difficult for service demand to diversify toward higher value-added activities. Meanwhile, factor substitution capacity is weaker than in the east but not as extreme as in the northeast, so the two forces largely offset each other, resulting in statistical insignificance.

The eastern region exhibits a positive but insignificant effect (coefficient = 0.083). With a well-developed socialized service system and abundant capital, the eastern region has the strongest factor substitution capacity, enabling it to rapidly fill the gaps created by labor aging and effectively satisfy the “demand-pull” effect. In addition, higher education levels mitigate the “adoption-inhibition” effect to some extent. Consequently, the direct impact of labor aging on agricultural productive services is buffered by these advantageous conditions, appearing statistically insignificant.

The western region also shows a positive but insignificant effect (coefficient = 0.204). Agricultural productive service development in the west is still at an early stage. The “demand-pull” effect of labor aging generates demand for basic services, driving positive growth. However, constraints in technology, capital, and human resources limit the magnitude and speed of this growth, so the overall effect is not significant.

In summary, the regional differences indicate that the direction and significance of the impact of rural labor aging on agricultural productive services depend on regional factor substitution capacity and development stage, exhibiting a distinctly nonlinear pattern.

4.4.2. Topographical Conditions

Table 8. Heterogeneity analysis by topographical conditions.

	Agricultural Productive Services
	Plain Areas	Non-Plain Areas
Aging	−0.068	0.087
	(0.073)	(0.143)
Control variables	YES	YES
Provincial FE	YES	YES
Year FE	YES	YES
N	192	168
R2	0.643	0.251

reports that the impact of aging on agricultural productive services differs in direction between plain and non-plain areas.

In plain areas, the effect is negative but insignificant (coefficient = −0.068). Flat terrain creates natural conditions for large-scale mechanization to substitute for labor. However, it also makes the region more prone to the “inefficient scale management” trap. When labor aging intensifies, farmers can simply outsource mechanized tasks as a passive, low-threshold substitution. Yet, lacking pressure to upgrade and constrained by human capital in management, such scale operation often remains at an extensive level, generating insufficient demand for precise, intelligent, and new-type services. Thus, overall, it exerts a slight inhibitory effect on the development of agricultural productive services.

In non-plain areas (hills and mountains), the effect is positive but insignificant (coefficient = 0.087). Large-scale machinery substitution is hindered, which forces two possibilities: first, farmers must seek more refined and diversified services (e.g., small-scale machinery services, specialized plant protection teams) to solve production difficulties; second, local agriculture may shift toward specialty industries (e.g., fruit trees, traditional Chinese medicine) that rely more on local knowledge and experience, creating demand for new services such as marketing and brand building. Therefore, despite many unfavorable conditions, the “demand-pull” effect of labor aging in non-plain areas may stimulate differentiated development momentum for agricultural productive services.

In summary, geographical endowment differences between plain and non-plain areas moderate the intensity and direction of the impact of labor aging on agricultural productive services.

4.4.3. Summary

The above analyses reveal pronounced regional differentiation in the impact of rural labor aging on agricultural productive services. The root cause lies in differences in regional factor substitution capacity, which determine the ultimate transmission direction of the aging shock. In regions with strong factor substitution capacity (e.g., the eastern region), modern factors effectively buffer the negative impact of aging, making the net effect on agricultural productive services near neutral. In regions where factor substitution is physically constrained (e.g., non-plain areas), the “forced-substitution” mechanism may instead foster more resilient and locally distinctive service formats, exhibiting a potentially positive effect.

Therefore, policy design to address the aging challenge must go beyond a “one-size-fits-all” approach and move toward precise diagnosis of the core constraints of different regions, proposing differentiated solutions accordingly.

4.5. Threshold Effect

The insignificant benchmark regression results, combined with the theoretical interplay between the “demand-pull” and “adoption-inhibition” forces, suggest that the impact of rural labor aging is not constant but may exhibit a critical point or “turning point”. When the degree of aging is low, the “demand-pull” effect may dominate; when aging deepens beyond a certain level, the negative “adoption-inhibition” effect may become prominent.

To test this hypothesis, we employ a panel threshold regression model. As shown in

Table 9. Test of threshold effects.

Threshold Variable	Threshold Value	95% Confidence Interval	Sum of Squared Residuals	MSE	F-Statistic	p-Value
aging	0.0817	(0.0781, 0.0825)	0.0137	0.0000	34.31	0.0367

, the single-threshold model is statistically significant at the 5% level, while the double-threshold model is not. This indicates that the impact of rural labor aging on agricultural productive services is characterized by a single threshold. Therefore, we adopt the single-threshold model for subsequent analysis.

To visualize this nonlinear structure, we split the sample into a “low aging group” (aging rate ≤ 8.17%) and a “high aging group” (aging rate > 8.17%) and plot a scatter diagram.

As shown in Figure 1, the slopes differ markedly between the two groups. In the low aging group (green dots), the level of agricultural productive services rises significantly with aging, and the fitted line has a relatively steep slope. In the high aging group (red dots), although the level of agricultural productive services also increases with aging, the fitted line is almost flat, indicating that the promoting effect of aging on productive services is substantially weakened.

Table 10. Threshold regression estimation results.

Variable	Coefficient	Std. Error
Aging (≤0.0817)	0.2112 ***	0.0385
Aging (>0.0817)	0.0676 **	0.0275
Control variables	YES
Constant	−0.009	0.0387
R2	0.5145
Log-likelihood	1320.871
N	360
Threshold value	0.0817

* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are in parentheses.

reports the single-threshold regression results. Before the threshold (aging ≤ 8.17%), the coefficient of aging is 0.2112 and significant at the 1% level, indicating that the “demand-pull” effect dominates at this stage. The substitution pressure caused by labor scarcity “forces” farmers to purchase services, acting as a strong catalyst for agricultural productive services. After crossing the threshold (aging > 8.17%), the promoting coefficient declines to 0.0676 (significant at the 5% level). This suggests that in the deep aging stage, the “adoption-inhibition” effect arising from human capital decline becomes prominent, severely weakening and offsetting the earlier “demand-pull” effect. The decline in learning ability, risk preference, and management capacity among older farmers becomes a bottleneck constraining the development of agricultural productive services.

From an economic perspective, the threshold value of 8.17% can be interpreted as a “turning point” in the impact of rural labor aging on agricultural productive services. When the aging level is low, labor scarcity raises labor costs, and older farmers adopt more technology and mechanized equipment to reduce dependence on human labor, promoting the transformation of agriculture toward servitization (a “dividend period”). However, when the aging level exceeds 8.17%, the disadvantages of declining physical strength, knowledge, and learning ability gradually become apparent. The innovation responsiveness of the agricultural sector weakens, and the negative effects of human capital decline gradually offset the substitution effect, leading to a diminishing marginal promoting effect of aging on agricultural productive services.

What does the 8.17% threshold mean in economic terms? In China‘s rural context, an aging rate of 8.17% represents the point at which the marginal benefit of labor substitution through productive services equals the marginal cost imposed by human capital decline. Below this level, the agricultural sector still has a sufficiently large pool of younger, more educated farmers who can effectively operate modern machinery and adopt new service models; the labor shortage created by out-migration generates strong demand for services without being offset by cognitive or behavioral barriers. Above this threshold, however, the proportion of older farmers exceeds a critical mass; the average learning ability, risk tolerance, and adaptive capacity of the agricultural workforce fall below the level required for effective adoption of high-order services. The 8.17% threshold is the point at which the two opposing forces are in balance, and beyond which the adoption-inhibition effect increasingly dominates. This interpretation is consistent with the behavioral economics literature, which suggests that cognitive decline and risk aversion intensify nonlinearly with age, particularly after the mid-50s [21,24].

This result suggests that, in the context of a continuously aging rural population, governments should adopt differentiated policies at different stages: in the low-aging stage, strengthen mechanization and technology diffusion; in the high-aging stage, implement agricultural technology extension and labor substitution policies to prevent structural upgrading from stagnating.

However, as shown in Figure 2, by 2022, the rural aging level in the vast majority of provinces in our sample had already exceeded this threshold. This indicates that for most rural areas, the “dividend” of agricultural productive services from aging-induced labor scarcity has ended. Greater attention should be paid to addressing the challenges posed by “deep aging”. At this stage, negative effects such as human capital decline and reduced capacity to absorb new technologies dominate. Therefore, the policy focus should shift from the optimistic expectation of “using aging to force upgrading” to a strategy of “breaking the structural lock-in of deep aging”.

What specific micro-mechanisms cause the substantial decline in the promoting effect after crossing the threshold? Attributing it solely to “human capital decline” is insufficient. This paper hypothesizes that it is closely related to changes in land management practices. In the context of deep labor aging, land transfer is accelerating, but without the entry of high-quality management entities, the trap of “inefficient scale management” may emerge. Therefore, the next section will conduct an in-depth analysis of this mechanism.

To assess the stability of the threshold, we re-estimated the threshold model using the old dependency ratio (population aged 65 and above divided by the working-age population) as an alternative measure of rural aging. As reported in

Table 11. Threshold robustness check: alternative aging indicator.

Specification	Threshold	Single-Threshold p-Value	Double-Threshold p-Value
Baseline (aging rate, %)	0.0817	0.0367	0.3700
Alternative (old dependency ratio)	0.1062	0.0180	0.2190

, the single-threshold model remains statistically significant (p = 0.018, bootstrap = 1000 replications), while the double-threshold model does not (p = 0.219). The estimated threshold value under this alternative specification is approximately 0.1062, which is economically comparable to the baseline threshold of 8.17% when converted to the aging rate metric. Critically, the direction of the marginal effect is unchanged: the promoting effect of aging on service development weakens after crossing the threshold. These results confirm that the single-threshold pattern is not an artifact of the specific aging indicator used, but rather a robust empirical regularity.

4.6. Mechanism Test

Table 12. Mediation test—cultivated land scale operation.

Variable	(1) APS	(2) Scale_Operation	(3) APS
aging	0.178 ***	8.468	0.068
	(0.048)	(11.733)	(0.082)
Scale_operation			−0.01 **
			(0.003)
Control variables	YES	YES	YES
Provincial FE	YES	YES	YES
Year FE	YES	YES	YES
N	360	360	360
R2	0.467	0.450	0.481

* p < 0.1, ** p < 0.05, *** p < 0.01.

and

Table 13. Bootstrap mediation effect test results.

Effect Type	Coefficient	Std. Err.	p-Value	95% Confidence Interval
Indirect effect	−0.0072	0.0038	0.058	[−0.015, −0.001]
Direct effect	0.069	0.082	0.408	[−0.099, 0.236]

present the results of the mechanism test. The specific analysis is as follows:

Column (1) shows that rural labor aging has a significant positive total effect on agricultural productive services (coefficient = 0.178, p < 0.01). Column (2) shows that aging has a positive but statistically insignificant effect on cultivated land scale operation (coefficient = 8.468, standard error = 11.733). Column (3) shows that after including the mediator variable, cultivated land scale operation has a significant negative effect on agricultural productive services (coefficient = −0.01, p < 0.05), while the direct effect of aging becomes insignificant (coefficient = 0.068, p > 0.1). The Bootstrap test further confirms that the indirect effect is −0.0072, with a 95% confidence interval of [−0.015, −0.001] (not including zero) and a p-value of 0.058 (marginally significant). This indicates that rural labor aging promotes the expansion of cultivated land scale operation, which in turn inhibits the development of agricultural productive services.

At first glance, the insignificant coefficient of aging on scale operation in column (2) of Table 12 may appear to challenge the presence of a mediation effect. However, this pattern is a well-recognized phenomenon in the mediation literature, referred to as inconsistent mediation or a suppression effect [37,38]. In such cases, the direct effect (aging → APS) and the indirect effect (aging → scale operation → APS) have opposite signs, and their magnitudes are such that the total effect is significant while the X→M path may be insignificant in isolation. This occurs when the mediator acts as a ‘suppressor’ that masks or counteracts part of the direct effect. In our context, labor aging directly promotes APS through the demand-pull channel (positive direct effect), but simultaneously pushes land toward scale operation, which in turn inhibits APS (negative indirect effect). These two opposing forces operate in parallel; the net total effect is positive, but the indirect pathway is negative and significant when tested via bootstrap. The bootstrap method—which does not require the X→M path to be significant in the first step—is now widely recommended as a more robust approach to mediation testing because it directly tests the product of coefficients (a × b) and does not rely on the sequential significance of individual paths [37,38,39]. Therefore, our finding of a significant indirect effect (−0.0072, 95% CI excluding zero) provides valid evidence for the negative mediating role of scale management, even though the X→M coefficient alone does not reach conventional significance levels.

While the bootstrap results provide statistical support for the indirect effect, it should be acknowledged that the direct pathway from aging to scale operation (column 2 of Table 12) is statistically weak. This implies that, from a substantive perspective, the mediating role of cultivated land scale management should be interpreted with some caution. The indirect effect is significant, but its magnitude is modest (−0.0072), and the marginal significance of the bootstrap estimate (p = 0.058) suggests that the mediation mechanism, while present, may not be the dominant channel through which aging affects agricultural productive services. Rather, it appears to be one of several mechanisms operating in parallel with the direct demand-pull effect. Future research with more disaggregated data and more direct measures of management efficiency would help to clarify the relative importance of this transmission channel.

Therefore, hypothesis H2 is supported. The above results demonstrate that “cultivated land scale operation” plays a significant negative mediating role in the process through which rural labor aging affects agricultural productive services.

Through the above analysis, it is clear that the impact of rural labor aging on agricultural productive services is not a simple linear relationship. On one hand, the deepening of labor aging increases farmers’ demand for agricultural machinery services, thereby promoting the development of agricultural productive services. On the other hand, older farmers experience declines in learning ability, capacity to adopt new technologies, and ability to cope with market risks. As a result, the land scale operation induced by labor scarcity lacks the corresponding human capital and knowledge structure updates. It is difficult to achieve the transition from “land scale expansion” to “upgrading of agricultural productive services”. Instead, it easily falls into the stage of “inefficient scale management” characterized by extensive management and technological stagnation (i.e., simple concentration of land factors without coordinated upgrading of technology, management, and services), thereby inhibiting the development of agricultural productive services.

Therefore, to unleash the potential of agricultural transformation in the context of rural labor aging, policy efforts should not blindly pursue land scale. Instead, they must focus on breaking the trap of “inefficient scale management” and ensure that land flows to entities capable of achieving the coordinated advancement of land scale operation and service upgrading.

5. Conclusions and Policy Implications

5.1. Conclusions

Based on provincial panel data from China for the period 2011–2022, this paper empirically examines the impact of rural labor aging on the development of agricultural productive services. The main conclusions are as follows:

First, the impact of rural labor aging on the development of agricultural productive services represents both an “opportunity” and a “challenge”, and is not a simple linear relationship. After controlling for endogeneity, the total effect of labor aging is significantly positive, indicating a macro trend in which labor scarcity “forces” service demand.

Second, this impact exhibits a significant single-threshold characteristic. When the rural aging rate is below 8.17%, rural labor aging acts as a “catalyst” for the development of agricultural productive services. However, after crossing this threshold, the promoting effect diminishes. By 2022, the vast majority of provinces in China had already entered the “challenge period”, and the “dividend period” has largely ended. The policy focus needs to shift from leveraging the opportunity to addressing the challenge.

Third, this study provides indirect evidence suggesting that “inefficient scale management” may serve as a key negative mediating mechanism leading to the attenuation of the promoting effect. The mechanism test shows that the increase in “cultivated land area per laborer” induced by labor aging actually inhibits the development of agricultural productive services. Simple land concentration, if not matched by human capital and management capacity, will fall into the trap of “inefficient scale management” characterized by “scale without services”, becoming an obstacle to agricultural productive services.

Fourth, the impact of rural labor aging exhibits significant regional heterogeneity, rooted in differences in “factor substitution capacity”. The northeastern region, deeply trapped in “inefficient scale management”, shows a significant inhibitory effect. The eastern region, with its strong capital and technology substitution capacity, shows an insignificant effect. These findings provide empirical evidence for region-specific policy implementation.

These findings provide actionable insights for policy-makers to align agricultural adaptation to population aging with broader sustainable development goals.

5.2. Policy Implications

Based on the above analysis, this paper argues that the traditional “one-size-fits-all” approach should be abandoned in favor of a multi-level, targeted, and context-specific policy mix that varies by region and over time. The specific recommendations are as follows:

(1)

Reconstruct the incentive mechanism of land scale policy to promote the coordinated evolution of land scale and efficiency.

Current agricultural subsidy and credit support policies should be optimized to build a comprehensive policy support system linked to the productive service capacity, human capital level, and production and operation performance of business entities. Policy preferences should be given to new types of business entities such as those actively adopting modern agricultural technologies and agricultural cooperatives, guiding land factors to concentrate among entities with strong operational capacity and service interface capabilities, thereby breaking the institutional lock-in of “inefficient scale management”.

(2)

Implement stage-adapted development strategies for agricultural productive services based on threshold characteristics.

Leveraging digital technology and agricultural equipment innovation, efforts should focus on cultivating and promoting agricultural equipment and digital service platforms that are adapted to aging labor forces, lowering operational barriers. User-friendly and practical digital agricultural production tools and full-entrustment service models should be developed to reduce the cognitive and learning costs for aging business entities in adopting modern productive services, effectively mitigating the “adoption-inhibition” effect caused by human capital decline, and achieving the precise matching of demand and supply of agricultural productive services at different aging stages.

(3)

Promote differentiated development paths for agricultural productive services based on regional heterogeneity in factor substitution capacity.

For the northeastern region, the core task is “entity reconstruction”. Focusing on the human capital reconstruction of agricultural business entities, policies such as talent attraction subsidies and entrepreneurship support should be used to inject high-quality human capital into large-scale land operations, breaking the path dependence of traditional extensive scale management.

For the eastern region, characterized by abundant capital and technology endowments, efforts should be made to upgrade agricultural productive services from single-task outsourcing to an integrated full-chain service system. Policy should focus on cultivating high-value-added service formats such as smart agricultural services and brand agricultural operation services, unleashing the upgrading and enabling effects of capital and technology factors on agricultural productive services.

For the central and western regions, where the agricultural production base is relatively weak and factor endowments are relatively insufficient, the layout and improvement in basic agricultural socialized service networks should be strengthened. Combined with regional topographical characteristics and the development needs of specialty agriculture, small-scale, specialized agricultural productive service formats with strong adaptability should be cultivated, consolidating the fundamental foundation for the development of agricultural productive services.

(4)

Build a dual support system of “technology adaptation for the elderly” and “service adaptation for the elderly”.

On the technology supply side, research institutions and agricultural enterprises should be guided to conduct targeted R&D, focusing on the actual production needs of aging business entities to develop aging-adapted agricultural machinery and technologies that are easy to operate, cost-effective, and low-maintenance, promoting the adaptive integration of modern agricultural technology with the production capacity of aging farmers.

On the technology extension side, the grassroots agricultural technology extension system should be optimized. The traditional “lecture-style” one-way training model should be abandoned, and the rich agricultural production experience of aging farmers should be organically integrated with modern productive service technologies, achieving experience compensation for human capital decline, and transforming the real challenge of population aging into a development opportunity that integrates agricultural production experience with modern science and technology.

6. Discussion

Different from existing studies on the relationship between rural labor aging and agricultural productive services, this paper constructs an analytical framework that incorporates a nonlinear threshold and a behavioral transmission mechanism, and empirically tests it using provincial panel data. Previous studies have mostly listed the two opposing effects in parallel or focused on only one of them, failing to explain how these two effects dynamically evolve as the degree of aging deepens. By employing a panel threshold regression model, this paper identifies for the first time a critical value of 8.17%, dividing the impact of aging on service development into a “dividend period” and a “challenge period”. This reveals that the relative strength of the two effects reverses as aging deepens. This finding not only reconciles the seemingly contradictory conclusions of “promotion” and “inhibition” in the existing literature, but also provides a clear stage-specific target for policy making.

Furthermore, unlike the common view that cultivated land scale management inevitably promotes agricultural modernization, our mediation analysis shows that the increase in “cultivated land area per laborer” induced by rural labor aging actually inhibits the development of agricultural productive services. This seemingly counter-intuitive conclusion precisely reveals the existence of the “inefficient scale management” trap: when land concentration is not accompanied by synergistic upgrades in human capital, management capacity, and technological innovation, scale expansion only solidifies extensive farming practices and weakens the positive effects of services. However, because our measure captures scale rather than efficiency directly, this interpretation remains inferential and should be validated in future research using more direct measures of management quality. Our heterogeneity analysis further shows that this inhibitory effect is particularly pronounced in northeastern China, while it is insignificant in the eastern region where capital and technology are abundant. This indirectly confirms that “factor substitution capacity” is a key moderating variable determining whether scale management is efficient.

Nevertheless, this study has several limitations. First, this study relies on provincial-level panel data, which, while appropriate for capturing macro-level structural relationships and inter-provincial policy variations, inevitably masks heterogeneity at the prefecture or farm household level within provinces. For example, within a single province such as Henan or Sichuan, there may be substantial intra-provincial variation: aging rates tend to be higher in labor-exporting mountainous counties than in plains counties with more diversified economies; the adoption of productive services may be concentrated in major grain-producing counties with better infrastructure, while peripheral counties lag far behind. Such intra-provincial disparities are averaged out in provincial-level measures, potentially attenuating the estimated relationships and limiting the precision of our recommendations for local policy design. Moreover, provincial-level aggregation cannot capture the behavioral heterogeneity among individual farm households—e.g., older farmers with different land endowments, family structures, or off-farm income sources may respond to aging pressures in opposite ways. This limitation implies that our threshold estimate (8.17%) should be interpreted as a macro-level reference point rather than a universally applicable micro-level behavioral threshold, and future research using county-level or farm-level microdata is needed to validate our findings.

Second, this paper measures cultivated land scale management using a single indicator—cultivated land area per laborer—which captures the scale dimension of land management but does not directly measure its efficiency or quality (e.g., whether modern management practices are adopted, whether green technologies are applied, whether operators have received professional training). Consequently, our empirical test of the ‘inefficient scale management’ mechanism is indirect: we infer inefficiency from the negative mediating effect of scale expansion under aging conditions, rather than observing inefficiency directly. While this interpretation is theoretically grounded—as argued in Section 2.2—we acknowledge that a more direct measure of management inefficiency (e.g., combining scale data with data on technical efficiency, adoption of best practices, or operator education levels) would provide stronger evidence. This remains an important direction for future research.

Future research can be extended in the following directions. First, using micro-level data at the county or farm household level to test the robustness of the threshold effect identified in this paper and to explore more detailed heterogeneity patterns. A finer geographical scale would allow for a more precise identification of the threshold and better capture the intra-provincial variations that provincial-level data inevitably obscure. Second, constructing a multi-dimensional evaluation system for “inefficient scale management” that incorporates technical efficiency, management capacity, and service matching degree into the measurement. Third, introducing spatial econometric models to examine the spatial spillover effects of rural labor aging and agricultural productive services, and whether inefficient scale management exhibits diffusion or contagion across regions. Fourth, conducting cross-country comparative studies to test whether the theoretical framework of this paper is applicable to other developing countries facing similar aging challenges. Given that many developing economies—including India, Vietnam, Thailand, and Brazil—are experiencing rapid rural aging alongside ongoing agricultural transformation, the analytical framework proposed in this paper could be extended to these contexts to examine whether the threshold effects and mediating mechanisms we identify are generalizable beyond China. Such cross-country comparisons would not only test the external validity of our findings but also provide valuable insights for global agricultural policy design in the face of population aging. Through these extensions, a more comprehensive understanding of the complex mechanisms of agricultural servitization transformation in the context of population aging can be achieved, providing theoretical support for more precise agricultural policies.