Pension Effects on Land Transfer and Intra-Household Labor Allocation of Farmer Households: Evidence from China

Abstract

This article uses two waves of panel data from China Land Economic Survey (CLES) in Jiangsu Province and employs a fixed-effects two-stage least squares (FE-2SLS) approach to identify pension effects on farmers’ labor allocation and land transfer decisions. In the FE-2SLS models, pension is instrumented by the average pension of other households in the same village. The results show that pension promotes land transfer-out, reduces household farm labor input, and increases household off-farm labor input. We further identify intra-household heterogeneity behind the pension effects. Specifically, pensioners in a household tend to leave farming activities without transitioning to off-farm employment, while non-pensioners shift the labor from farm to off-farm employment. We also examine heterogeneity by household budget pressure using two grouping strategies based on shortage experience and a composite budget-constraint indicator. The results show that the pension effects are more clearly observed among households without budget shortage. The estimates for households with budget shortage are less precise. These findings suggest that pension effects are complex in driving farmers’ resource allocation in their households. However, Jiangsu Province provides a substantial number of off-farm employment opportunities and features a well-developed land transfer market. The estimated pension effect in this area may not be applicable to less developed regions.

1. Introduction

The efficient allocation of land resources is crucial for food security in many developing countries [1]. The increase in population with limited land resources has become a critical challenge for many developing countries [2]. Nevertheless, land transfer has been recognized as a potential solution to alleviate the disequilibrium between labor and land resources [3,4]. It facilitates the reallocation of farmland from less to more efficient farmers and ultimately promotes the transition from traditional smallholder farming to mechanized agricultural production [5,6]. Therefore, many measures have been implemented by the Chinese government since 1984 to develop the rural land transfer market. The 2023 Annual Statistical Report on Rural Operation and Management in China revealed that the land transfer rate reached 36.5% in 2022. However, it increased by only 0.2%, reflecting the slowest growth in decades. This is attributed to the fact that the majority of farmland in rural China is still operated by smallholders. The underdeveloped rural social security system makes land a crucial resource for rural farmers [7]. Land provides households with limited but stable income, which makes households reluctant to transfer out their land.

In the absence of pension support, household land and labor allocation decisions are closely tied to basic consumption requirements. Households must rely on current income to meet present expenditure and prepare for future needs. Although off-farm employment may yield higher returns, its income is often uncertain. In contrast, agricultural income is lower but more directly controlled through households’ own land and labor input. Under this condition, elderly individuals in many developing countries often continue farming until physical decline, and households tend to retain land in self-cultivation and maintain farm labor. In 2009, the Chinese government introduced the New Rural Pension Scheme (NRPS) to transform the rural pension insurance system. Data from China’s Ministry of Human Resources and Social Security (MOHRSS) showed that the pension insurance coverage rate in China reached 95% by the end of 2024. As a predictable source of non-labor income, pension may alleviate households’ dependence on farm income [8,9,10]. When part of basic consumption can be supported by pension, households no longer need to rely on self-cultivation to the same degree as before. This may enable elderly individuals to reduce farm work, make households more willing to shift labor between farm and off-farm activities, and enhance the likelihood of land transfer-out [11,12,13,14].

Existing studies have examined the effects of pension on either land or labor outcomes, but usually not within a unified household decision framework [9,15]. Some studies find that pension weakens the security role of land and promotes land transfer-out [8,10], while others examine their effects on farm work, retirement, or off-farm employment [15,16]. Even when pension effects on land transfer and labor allocation are discussed together, labor allocation is usually measured only at the level of aggregate household labor input [16]. Existing evidence remains limited about how land transfer and labor allocation are jointly adjusted within the same household. Pension is received by specific members, whereas land and labor decisions are adjusted at the household level [14]. Without distinguishing pensioners from non-pensioners, existing studies cannot clearly identify whether pension mainly leads to labor withdrawal, intra-household labor reallocation, or both. Moreover, heterogeneity effects of pension are often analyzed by age, gender, education, or income [10,15,16]. Such groupings do not reflect how households use pension for different purposes, and how this use shapes labor allocation and land transfer behavior [14]. The use of pension may be shaped by household budget constraints [17]. Households facing the budget shortage are likely to prioritize essential expenses and debt. Pension covers debt payments and main expenses rather than increasing disposable income. This means that the role of pension for these farm households may be limited [11,12,18].

Therefore, this study has three main objectives: (i) to develop a farmer household-level model which can combine pension, labor-input choices, and land transfer; (ii) to identify the effects of pension on farmers’ land transfer decisions and labor allocation at both the household level and the intra-household level; and (iii) to examine whether the pension effects differ across households facing different budget constraints.

This study contributes to the existing literature in the following ways. First, we bring land transfer and labor allocation into the same empirical framework. Furthermore, we classify labor input by pension eligibility and by activity type: pensioners versus non-pensioners, and farm versus off-farm. This design allows us to identify how pension reshapes labor allocation within the household. It enables us to distinguish between labor withdrawal and reallocation, which household-level aggregates cannot capture. Second, we estimate the heterogeneous effects of pension among farmers from the perspective of budget constraints. This approach shifts the focus from static demographic categories to behavioral mechanisms, by emphasizing heterogeneity in pension use at the household level.

The remainder of this article is organized as follows: Section 2 outlines the institutional background, while Section 3 develops the theoretical framework. Section 4 and Section 5 describe the data and empirical strategy. The empirical results and discussion are detailed in Section 6 whilst the last section includes the conclusion and policy implications.

2. Institutional Background

2.1. Pension System in China

In 2009, the Chinese government implemented the NRPS to restructure the rural social security framework. The NRPS employs diversified funding mechanisms comprising individual contributions, collective subsidies, and government support. The residents who participate in the NRPS make contributions to their individual pension accounts. However, rural residents aged over 60 years old who have not participated in the basic pension insurance program for urban workers are eligible to receive monthly pension without making any contributions. The local governments provide pension subsidies based on their own fiscal capacity. By the end of 2012, the NRPS had achieved nationwide coverage across all county-level administrative regions in China. In 2014, the State Council of China merged the NRPS and the Urban Residents’ Pension Scheme into a unified system known as the Basic Pension Insurance for Urban and Rural Residents (URRBPI). The reform unified the main policy framework, administration, and information systems across urban and rural areas. The merged system largely retained the institutional design of the NRPS. This reform substantially improved social security for rural residents. Compared with other sources of non-labor income including land renting income and gifts from relatives, pension is a government-administered entitlement that provides stable and predictable payments for eligible individuals. The pension system makes pension a reliable component of rural household income [19]. Therefore, pension is more likely to influence farmers’ willingness to allocate resources.

2.2. Land Transfer Market in China

For a long time, farmers had been unable to alter the size of their cultivated land. Farm production is the primary source of household income [20]. As a result, land served not only as means of production for rural households, but also as an important factor for providing social security and mitigating risks [21]. With the advancement of land reforms in agricultural production in rural China, the land transfer market has gradually emerged and enables the transfer of land operational rights among farmers. Land transfer can break the rigid “labor–land” allocation pattern in traditional rural areas, thereby facilitating the optimal allocation of land resources through market, improving production efficiency, and promoting moderate scaled agricultural operations [6].

The land transfer process requires two conditions—a farmer’s willingness to transfer out the land and a counterparty ready to accept it. However, these conditions are not always fulfilled. In many rural areas, finding a suitable counterparty is difficult due to the low farm profitability, aging population, and insecure land property rights [21,22]. Land transfer is exacerbated by the lack of large and capable agricultural enterprises and professional large-scale households [6]. Consequently, many farmers continue to exploit their lands or leave them fallow. The local governments may promote or inhibit land transfer. Compared with many less-developed inland areas, Jiangsu Province has more abundant off-farm employment opportunities and a more developed land transfer market. In addition, village collectives and related organizations in Jiangsu often play an active role in land coordination, service provision, and the organization of larger-scale operations. These conditions help reduce transaction frictions in land transfer, expand the pool of potential counterparties, and make it easier for households to turn transfer willingness into actual transfer behavior [23]. They also mean that changes in household security brought by pension are more likely to be reflected in observable adjustments in land transfer and labor allocation, rather than being blocked by weak markets or limited local coordination. Therefore, the Jiangsu setting provides a suitable context for identifying the effect of pension on land transfer behavior, although the regional specificity of the sample should be kept in mind when interpreting the findings.

3. Theoretical Analysis

We construct a theoretical model to analyze the impact of pension on farmers’ labor allocation and land transfer decisions. Traditional rural economies lack stable non-labor income such as pension. Households meet both current consumption needs and essential future expenditures through their own labor and fixed land endowment [24,25]. Farmers’ decisions regarding allocation of labor input, leisure consumption, and land use exhibit a highly conservative equilibrium [12]. Household consumption largely depends on farm output. With rising population and fixed lands, the marginal utility of consumption remains high [26]. In other words, each additional unit of income plays a critical role in fulfilling basic subsistence needs and managing unexpected shocks. Consequently, farmers tend to retain and exploit lands since they expect that cultivating their lands can yield any positive net returns. In addition, they are more incentivized to extend their working hours since they expect that the additional unit of labor time may generate a positive marginal return.

3.1. Model Specification

For analytical convenience, we adopt the following assumptions: (1) The household lives two periods. In the first period, the household participates in agricultural production and off-farm labor to generate income ${\mathrm{y}}_1$. This income is allocated to first-period consumption ${\mathrm{c}}_1$, with any remaining amount $\mathrm{S}$ saved for consumption ${\mathrm{c}}_2$ in the second period. In the second period, basic expenditures are entirely financed by savings $\mathrm{S}$ and pension $\mathrm{P}$. (2) The household faces a binding minimum consumption constraint ${\mathrm{c}}_{\mathrm{t}}\ge \overline{\mathrm{c}}$ for $\mathrm{t}\in \{1,2\}$. Pension is insufficient to meet this constraint alone, thereby necessitating positive precautionary savings ($\mathrm{S}$ > 0). (3) The household is endowed with farmland of size ${\mathrm{L}}_{\mathrm{a}\mathrm{l}\mathrm{l}}$, of which $\mathrm{L}$ is cultivated land and $\mathrm{X}$ is leased out at a rental price $\mathrm{r}$. The total time endowment available to the household is $\mathrm{T}$, which is allocated to agricultural labor ${\mathrm{l}}_{\mathrm{a}}$, off-farm labor ${\mathrm{l}}_{\mathrm{n}}$, and leisure $\mathrm{h}$. (4) Household income comes from farm income $\mathrm{p}{\mathrm{y}}_{\mathrm{a}}$, off-farm income ${\tilde{\mathrm{w}}\mathrm{l}}_{\mathrm{n}}$, land rental income $\mathrm{r}\mathrm{X}$, and pension P. Farm output is governed by a Cobb–Douglas production function, ${\mathrm{y}}_{\mathrm{a}}=\mathrm{A}{{\mathrm{l}}_{\mathrm{a}}}^{\mathsf{\alpha }}{\mathrm{L}}^{1-\mathsf{\alpha }}$ with $0<\mathsf{\alpha }<1$. The off-farm wage $\tilde{\mathrm{w}}$ is random, with expectation $\mathrm{E}\left[\tilde{\mathrm{w}}\right].$ This reflects the instability of off-farm employment. Farm income, land rental income, and pension are deterministic. (5) The household maximizes expected utility over two periods, which is derived from both consumption and leisure. Expected utility is $\mathrm{E}\left[\mathrm{u}\left({\mathrm{c}}_1-\overline{\mathrm{c}},\mathrm{h}\right)+\mathsf{\beta }\mathrm{u}\left({\mathrm{c}}_2-\overline{\mathrm{c}},\mathrm{h}\right)\right].$ The period utility $\mathrm{u}$ is continuous, twice differentiable, strictly increasing (${\mathrm{u}}_{\mathrm{c}}>0,{\mathrm{u}}_{\mathrm{c}\mathrm{c}}>0)$, and strictly concave (${\mathrm{u}}_{\mathrm{h}}{<0,\mathrm{u}}_{\mathrm{h}\mathrm{h}}<0)$. The problem can be specified as follows:

$$\begin{array}{c}\mathrm{M}\mathrm{a}\mathrm{x} \mathrm{E}[\mathrm{u}\left({\mathrm{c}}_1 - \overline{\mathrm{c}},\mathrm{h}\right) + \mathsf{\beta }\mathrm{u}\left({\mathrm{c}}_2 - \overline{\mathrm{c}},\mathrm{h}\right)]\\ \mathrm{S}\mathrm{u}\mathrm{b}\mathrm{j}\mathrm{e}\mathrm{c}\mathrm{t} \mathrm{t}\mathrm{o} {\mathrm{c}}_1+\mathrm{s} = \mathrm{p}\mathrm{A}{{\mathrm{l}}_{\mathrm{a}}}^{\mathsf{\alpha }}{\mathrm{L}}^{1-\mathsf{\alpha }} + {\mathrm{l}}_{\mathrm{n}}\tilde{\mathrm{w}} + \mathrm{r}({\mathrm{L}}_{\mathrm{a}\mathrm{l}\mathrm{l}} - \mathrm{L}) + \mathrm{P}\\ {\mathrm{c}}_2 \le \left(1 + {\mathrm{r}}_{\mathrm{f}}\right)\mathrm{S} + \mathrm{P}\\ \mathrm{T} = {\mathrm{l}}_{\mathrm{a}} + {\mathrm{l}}_{\mathrm{n}} + \mathrm{h}\end{array}$$

(1)

To simplify the analysis, we assume that the household’s second period consumption is bounded at $\overline{\mathrm{c}}$. The second period utility, $\mathsf{\beta }\mathrm{u}\left({\mathrm{c}}_2-\overline{\mathrm{c}},\mathrm{h}\right)$, becomes constant and can thus be excluded from the optimization problem. This simplification does not completely remove the intertemporal linkage, because the household must still choose first-period saving S to satisfy second-period consumption $\overline{\mathrm{c}}$. Under this simplified setting, the household’s decision problem reduces to select the optimal levels of agricultural labor ${\mathrm{l}}_{\mathrm{a}}$, non-agricultural labor ${\mathrm{l}}_{\mathrm{n}}$, and land input L in period 1 to maximize expected utility. A stable pension income reduces the amount of saving that the household must set aside to secure future consumption, thereby freeing up more current-period resources. These resources, in turn, affect the household’s first-period labor supply and land input decisions. Hence, the model captures how pension expectations influence current factor allocation decisions. The household’s problem can be stated as follows:

$$\mathrm{M}\mathrm{a}\mathrm{x} \mathrm{E}\left[\mathrm{u}\left(\mathrm{p}\mathrm{A}{{\mathrm{l}}_{\mathrm{a}}}^{\mathsf{\alpha }}{\mathrm{L}}^{1-\mathsf{\alpha }}+{\mathrm{l}}_{\mathrm{n}}\tilde{\mathrm{w}}+\mathrm{r}({\mathrm{L}}_{\mathrm{a}\mathrm{l}\mathrm{l}}-\mathrm{L})+\mathrm{P}-{\mathrm{S}}^{*}-\overline{\mathrm{c}},\mathrm{T}-{\mathrm{l}}_{\mathrm{a}}-{\mathrm{l}}_{\mathrm{n}}\right)\right]$$

(2)

where ${\mathrm{S}}^{*}={\displaystyle \frac{\overline{\mathrm{c}}-\mathrm{P}}{1+{\mathrm{r}}_{\mathrm{f}}}}>0$ represents the mandatory savings required to meet the future consumption constraints.

Solving the above maximization problem leads to first-order conditions that characterize the household’s optimal decision-making, that is:

$$\mathrm{p}\mathrm{A}\left(1-\mathsf{\alpha }\right){{\mathrm{l}}_{\mathrm{a}}}^{\mathsf{\alpha }}{\mathrm{L}}^{-\mathsf{\alpha }}=\mathrm{r}$$

(3)

$$\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\cdot \left(\mathrm{p}\mathrm{A}\mathsf{\alpha }{{\mathrm{l}}_{\mathrm{a}}}^{\mathsf{\alpha }-1}{\mathrm{L}}^{1-\mathsf{\alpha }}\right)\right]=\mathrm{E}\left[{\mathrm{u}}_{\mathrm{h}}\left(·\right)\right]$$

(4)

$$\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\cdot \tilde{\mathrm{w}}\right]=\mathrm{E}\left[{\mathrm{u}}_{\mathrm{h}}\left(·\right)\right]$$

(5)

Equation (3) implies that the land allocation decision within household is independent of income uncertainty. Households allocate land to the point where its marginal value product equals the exogenous rental rate. This yields the optimal land demand function:

$${\mathrm{L}}^{*}=\mathsf{\kappa }{\mathrm{l}}_{\mathrm{a}}$$

(6)

where $\mathsf{\kappa }={\left({\displaystyle \frac{\mathrm{p}\mathrm{A}\left(1-\mathsf{\alpha }\right)}{\mathrm{r}}}\right)}^{{\displaystyle \frac{1}{\mathsf{\alpha }}}}>0$ is a constant.

Substituting (6) into (4), we obtain:

$$\mathrm{E}\left[{\mathrm{u}}_{\mathrm{h}}\left(·\right)\right]=\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\mathrm{p}\mathrm{A}\mathsf{\alpha }{\mathsf{\kappa }}^{1-\mathsf{\alpha }}\right]$$

(7)

Combining (5) and (7), we obtain:

$$\begin{array}{cc}\hfill \mathrm{p}\mathrm{A}\mathsf{\alpha }{\mathsf{\kappa }}^{1-\mathsf{\alpha }}\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\right]& =\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\cdot \tilde{\mathrm{w}}\right]\hfill \\ & =\mathrm{C}\mathrm{o}\mathrm{v}({\mathrm{u}}_{\mathrm{c}}\left(·\right),\tilde{\mathrm{w}})+\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\right]\cdot \mathrm{E}\left[\tilde{\mathrm{w}}\right]\hfill \end{array}$$

(8)

Subsequently, Equation (8) can be further rearranged as follows:

$$\mathrm{E}\left[\tilde{\mathrm{w}}\right]+{\displaystyle \frac{\mathrm{C}\mathrm{o}\mathrm{v}\left({\mathrm{u}}_{\mathrm{c}}\left(·\right),\check{\mathrm{w}}\right)}{\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\right]}}=\mathrm{p}\mathrm{A}\mathsf{\alpha }{\mathsf{\kappa }}^{1-\mathsf{\alpha }}$$

(9)

Equation (9) is the key condition governing optimal labor allocation.

The left side represents the risk-adjusted expected return to non-farm labor. It comprises the expected wage $\mathrm{E}\left[\tilde{\mathrm{w}}\right]$ plus a risk premium term, ${\displaystyle \frac{\mathrm{C}\mathrm{o}\mathrm{v}({\mathrm{u}}_{\mathrm{c}}\left(·\right),\tilde{\mathrm{w}})}{\mathrm{E}\left[{\mathrm{u}}_{\mathrm{c}}\left(·\right)\right]}}$. Since ${\mathrm{u}}_{\mathrm{c}}\left(·\right)$ is negatively correlated with $\tilde{\mathrm{w}}$ as the marginal utility of consumption tends to be lower during periods of high income. This covariance is negative, resulting in a negative risk premium. This negative adjustment reflects a “penalty” on risky off-farm income, which arises due to the agent’s risk aversion. The right side of the equation represents the risk-free marginal return to agricultural labor—the value of additional income from one more unit of farm work.

3.2. Comparative Static Analysis of Pension Effect

Based on the optimal conditions derived in Section 3.1, the effect of pension can be obtained by comparative statics with respect to P. Since the minimum-consumption constraint in period 2 is binding, mandatory savings are pinned down by that constraint, thus ${\mathrm{S}}^{*}={\displaystyle \frac{\overline{\mathrm{c}}-\mathrm{P}}{1+{\mathrm{r}}_{\mathrm{f}}}}$. A higher pension, therefore, reduces the amount of savings that need to be set aside for period 2 and increases current consumption ${\mathrm{c}}_1$. Given that the period utility function is strictly concave, the marginal utility of current consumption decreases as current consumption increases, which implies

$${\displaystyle \frac{\partial {{\mathrm{u}}_{{\mathrm{c}}_1}}^{*}}{\partial \mathrm{P}}}<0$$

(10)

From Equation (7), we observe that ${\mathrm{u}}_{\mathrm{h}}\left(·\right)$ moves with ${\mathrm{u}}_{\mathrm{c}}\left(·\right)$. Since the utility function is strictly increasing and strictly concave in leisure, a lower marginal utility of leisure implies a higher optimal level of leisure. The total time is fixed, so expanding leisure time necessarily reduces total labor supply, and thus

$${\displaystyle \frac{\partial ({{\mathrm{l}}_{\mathrm{a}}}^{*}+{{\mathrm{l}}_{\mathrm{n}}}^{*})}{\partial \mathrm{P}}}<0$$

(11)

This income effect weakens the incentive to supply labor in both farm and off-farm works. When hired labor or mechanization are expensive substitutes for own labor, reducing own-labor input makes land transfer optimal.

At the same time, the predictable stream of pension hedges against off-farm wage risk and reduces income volatility. ${\mathrm{u}}_{\mathrm{c}}\left(·\right)$ becomes less sensitive to shocks in $\tilde{\mathrm{w}}$. $\mathrm{C}\mathrm{o}\mathrm{v}\left({\mathrm{u}}_{\mathrm{c}}\left(·\right),\check{\mathrm{w}}\right)$ becomes less negative. A smaller covariance term lowers the risk premium on off-farm work, thereby raising its certainty-equivalent wage. Off-farm work becomes relatively more attractive.

Jointly, the income effect reduces total labor supply, while the risk effect shifts labor toward off-farm work, that is

$${\displaystyle \frac{\partial {{\mathrm{l}}_{\mathrm{a}}}^{*}}{\partial \mathrm{P}}}<0$$

(12)

The decline in farm labor further changes land allocation through the optimal land demand function in Equation (6). The optimal operated land ${\mathrm{L}}^{*}$ is increasing in farm labor input. Land transfer-out is equal to total land endowment minus operated land. Therefore,

$${\displaystyle \frac{\partial \left(\overline{\mathrm{A}}-{\mathrm{L}}^{*}\right)}{\partial \mathrm{P}}}=-{\displaystyle \frac{\partial {\mathrm{L}}^{*}}{\partial {\mathrm{l}}_{\mathrm{a}}^{*}}}\cdot {\displaystyle \frac{\partial {\mathrm{l}}_{\mathrm{a}}^{*}}{\partial \mathrm{P}}}>0$$

(13)

This means that a higher pension reduces farm labor, lowers the demand for self-cultivated land, and increases land transfer-out. By contrast, the net impact on off-farm work depends on the relative strength of these two forces.

3.3. Intra-Household Heterogeneity

Although the model is not formally built at the household member level, the household-level mechanism can still be extended descriptively by distinguishing between pensioners and non-pensioners. The comparative statics above describe how pension changes farm labor, off-farm labor, and land transfer at the household level. In practice, however, these adjustments must be carried out by specific household members. Pensioners and non-pensioners may respond differently because they face different labor-market opportunities, labor capacities, and labor-leisure trade-offs.

Pensioners receive pension income directly. As the income security of pensioners improves, the marginal utility of consumption declines. The value of leisure increases. Compared with non-pensioners, they often face weaker off-farm labor-market conditions. They also face stronger age-related barriers, and their expected returns are lower. As a result, they may be more likely to reduce farm labor without a corresponding increase in off-farm employment [27,28]. By contrast, non-pensioners do not receive pension directly. However, they are usually more able to access off-farm work opportunities [29,30,31]. Therefore, when pension reduces the household’s dependence on self-cultivation, non-pensioners may be more likely to shift labor from farming to off-farm employment rather than to leisure.

3.4. Heterogeneous Effects of Pension Under Budget Shortage

The utilization of pension among rural households varies according to their budget constraints, leading to distinct labor allocation and land transfer decisions. When households face higher budget constraints, their cash flow is insufficient to cover debts and essential expenditures. Any new income is primarily allocated to debt payments and essential expenses rather than increasing discretionary spending [32]. Since consumption level ${\mathrm{c}}_1$ is kept below the high threshold $\overline{\mathrm{c}}$, the marginal utility of consumption ${\mathrm{u}}_{\mathrm{c}}\left(·\right)$ stays very high. This means that although pension increases household wealth, it has little effect on reducing marginal utility. Subsequently, households cannot afford leisure and do not reduce labor supply because every unit of income is essential. In addition, they are often reluctant to transfer out their contracted land since the agricultural activities continue to provide a steady cash flow. Therefore, for the households with higher budget constraints, the income effect of pension primarily lies in alleviating financial pressure, rather than substituting the agricultural labor. On the contrary, households with lower budget constraints can cover their main expenditures through deposits and stable income [33]. Pension serves as an additional source of income that can be used to enhance the quality of life and increases leisure consumption. When weighing the marginal benefits of labor against those of leisure, such households are more likely to perceive the marginal benefit of continuing to farm as lower than the marginal benefit of relinquishing land in exchange for greater leisure and comfort [12,34]. Therefore, they are more inclined to sub-lease large areas of contracted land to major growers or agricultural enterprises.

4. Data and Variables

4.1. Data Source

The data used in this article are from the China Land Economic Survey (CLES), conducted in 2020 and 2021. The baseline survey in 2020 employed a probability-proportional-to-size sampling method and randomly selected 2600 households across 52 villages in 26 counties of Jiangsu Province. The 2021 wave followed up with these households, with some attrition and additional samples. After dropping observations with missing or incorrect data, we retained households interviewed in the two years, resulting in a balanced panel of 1468 households.

4.2. Variables and Measurements

The key independent variable refers to pension. Since the survey does not allow us to verify individual pension receipts, this variable represents the total pension received by each household per year. It may mechanically covary with household demographic composition, especially the number of older members. To reduce this concern, we control for household member structure in the regressions. We define pension as a continuous variable, measuring annual pension in units of ten thousand yuan. The main dependent variable is land transfer-out, measured as a total area of land transferred out within a household. For labor input variables, household members reported the days that they spent on farm and off-farm activities in each year. We aggregate these individual reports to the household level, categorized by activity type (farm and off-farm) and by pension eligibility (age ≥ 60 and age < 60). Following pension eligibility, we classify household members aged ≥60 as pensioners and those aged <60 as non-pensioners. Given the high pension coverage during the sample period, this eligibility-based grouping provides an approximation to actual pension receipt at the member level, although some measurement error may still remain. In total, we obtain six labor variables: farm work (household), off-farm work (household), farm work (pensioner), off-farm work (pensioner), farm work (non-pensioner), and off-farm work (non-pensioner). We select control variables that capture household, land-related and village characteristics. For the categorical variable budget shortage, farmers were asked in the 2020 panel survey whether they faced insufficient funds for agricultural production or daily consumption. We also construct a simple budget constraint indicator using shortage amount, household income, household expenditure, and year-end savings. Based on the concept that household financial conditions mirror both current budget constraints and liquidity buffers, and that multidimensional financial information can be integrated into a single indicator [35,36], the budget–constraint indicator is defined as:

$${BC}_i={\displaystyle \frac{1}{2}}\left[{\displaystyle \frac{{SA}_i}{{Exp}_i}}+ {\displaystyle \frac{\max \left({Exp}_i-{Inc}_i\right)}{{Exp}_i}}\right]-{\displaystyle \frac{{Sav}_i}{{Exp}_i}}$$

(14)

where ${SA}_i$ is the shortage amount of household $i$, ${Exp}_i$ is total household expenditure, ${Inc}_i$ is household income, and ${Sav}_i$ is year-end household savings. The first term captures the realized shortage intensity. The second term reflects the expenditure-income gap in relation to expenditure, and the third term measures the household’s liquidity buffer.

The specific definitions and measurements of variables employed in the analysis are detailed in

Table 1. Variable definitions and descriptive statistics.

Variables	Definition	Mean	SD
Dependent variables
Transfer-out land	Area of land transferred out by farmers (mu)	2.344	3.193
Farm work (household)	Days spent by all household members in agricultural production per year (days)	135.346	191.678
Off-farm work (household)	Days spent by all household members in off-farm work per year (days)	433.686	364.097
Farm work (pensioner)	Days spent by pension-eligible members (≥60) in farm work production per year (days)	68.381	129.239
Off-farm work (pensioner)	Days spent by pension-eligible members (≥60) in off-farm work per year (days)	46.758	115.351
Farm work (non-pensioner)	Days spent by ineligible members (<60) in the farm work production per year (days)	66.965	147.738
Off-farm work (non-pensioner)	Days spent by ineligible members (<60) in the off-farm work per year (days)	386.284	348.686
Independent variables
Pension	The annual pension received by the household (10,000 CNY)	0.854	1.610
Control variables
Male share	Share of male household members (%)	0.510	0.163
Age	Average age of all household members (year)	50.159	12.887
Education level	Average years of schooling of household members (years)	7.470	2.768
Health status	Average self-reported health status of the households (1 = Lost labor ability; 2 = Poor; 3 = Fair; 4 = Good; 5 = Excellent)	4.188	0.829
Number of pensioners	Number of household members aged 60 and above	1.343	0.874
Number of non-pensioners	Number of household members aged below 60	2.763	1.897
Farmland	Area of contracted farmland under cultivation (mu)	6.357	21.177
Irrigation	Irrigable area within contracted farmland (mu)	0.900	0.300
Land fertility	Fertility of contracted farmland (1 = Poor; 2 = Fair; 3 = Good)	2.476	0.613
Production assets	Original value of household agricultural production assets (10,000 CNY)	1.110	5.694
Village off-farm business presence	Whether the village had rental, transfer, or shareholding activities involving rural construction land for production and business purposes over the past year	0.136	0.342
Instrumental variable
Average villagers’ pension	Annual pension of other households within the same village (10,000 CNY)	0.838	0.691
Classified variable
Budget shortage	Household experienced a shortage of money for agricultural production or daily consumption in the past year (1 = Yes; 0 = No)	0.186	0.389
Budget constraint index	Composite indicator of household budget constraint constructed from shortage amount, household income, household expenditure, and year-end savings	1.500	0.500
Observations		2936

Note: (1 mu = 1/15 ha); CNY is Chinese currency (1 USD = 7.12 CNY in 2024).

.

4.3. Descriptive Statistics

Table 1 presents the definitions and summary statistics of the variables used in the econometric analysis. Specifically, the average land transfer-out per household is estimated at 2.344 mu. The average farm and off-farm labor for all household members are 135.346 and 433.686 days per year, respectively, while they are 68.381 and 46.758 days per year for pensioners. Comparatively, the average farm and off-farm labor times are 66.965 and 386.284 days per year for non-pensioners. The mean pension of respondents was 0.854 (10,000 CNY), indicating that pensions account for only a small share of household income and thus provide limited security. Finally, the sample households that have experienced a budget shortage for agricultural production or daily consumption are approximately 20%, suggesting that a considerable proportion of households face budget shortage, which in turn, restrict their resource allocation decisions.

5. Empirical Models

To identify the impact of pension on farmers’ labor allocation and land transfer behavior, we employ a fixed-effects two-stage least squares method with instrumental variable estimation (FE-2SLS).

A central challenge in our identification strategy lies in mitigating potential bias due to endogeneity. Although household fixed effects can absorb time-invariant unobserved heterogeneity, the estimated relationship between pension, land transfer, and labor allocation may still be biased by time-varying omitted factors and household-level self-selection. Omitted variable bias arises when unobservable or difficult-to-measure variables that vary over time are excluded from the analysis. These variables influence households’ land transfer decisions and are consequently captured in the error term. Examples include time-varying village-level and household-level characteristics. In addition, household pension may partly reflect differences in contribution choices and other household characteristics. As a result, the fixed effects model may not fully identify the relationship of these variables.

To address these potential endogeneity issues, we employ the FE-2SLS estimation. We instrument pension with the average pension of other households in the same village. The underlying idea is that households in the same village face a common local pension environment, so pension received by other households is informative about a household’s own pension. In the empirical specification, the instrument is used together with household fixed effects, year fixed effects, and observed time-varying household controls.

The first stage of our two-stage IV model is expressed as follows:

$${\mathrm{P}}_{\mathrm{i},\mathrm{y}}={\mathsf{\kappa }}_{\mathrm{o}}+{\mathsf{\kappa }}_1{\mathrm{Z}}_{\mathrm{i},\mathrm{y}}+{{\mathsf{\kappa }}_2\mathrm{X}}_{\mathrm{i},\mathrm{y}}+{\mathsf{\alpha }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{y}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{y}}$$

(15)

where P_i,y is the pension of the household i in year y = 2020 and 2021; Z_i,y denotes the instrument for the pension. X_i,y is a vector of demographic and socio-economic time-varying controls that are likely to influence land transfer and labor allocation behaviors. The model incorporates household-specific fixed effects, denoted as ${\mathsf{\alpha }}_{\mathrm{i}}$, to account for unobserved and time-invariant heterogeneity among farmers. Such heterogeneity may arise from differences in expertise, access to information, risk preferences, and household characteristics that tend to remain stable. The model incorporates time dummy variables ${\mathsf{\gamma }}_{\mathrm{y}}$ to control for temporal variations or common annual shocks. Error terms ${\mathsf{\epsilon }}_{\mathrm{i},\mathrm{y}}$ are clustered at the household level.

The second stage is expressed as follows:

$${\mathrm{Y}}_{\mathrm{i},\mathrm{y}}={\mathsf{\beta }}_{\mathrm{o}}+{\mathsf{\beta }}_1{\overline{P}}_{i,y}+{{\mathsf{\beta }}_2\mathrm{X}}_{\mathrm{i},\mathrm{y}}+{\mathsf{\alpha }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{y}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{y}}$$

(16)

where Y_i,y denotes the area of land transferred out and labor input of household i in year y (2020 or 2021); $\overline{P_{i,y}}$ is the fitted value of household pension income obtained from the first stage; X_i,y is a vector of demographic and socio-economic time-varying controls that are likely to influence land transfer and labor allocation behaviors, and ${\mathsf{\alpha }}_{\mathrm{i}}$, ${\mathsf{\gamma }}_{\mathrm{y}}$, ${\mathsf{\epsilon }}_{\mathrm{i},\mathrm{y}}$ are household fixed effects, time dummies and error terms, respectively.

6. Results and Discussion

6.1. Impacts of Pension on Farmer’s Land Transfer and Labor Allocation Behavior on Household Level

This study employs fixed-effects two-stage least squares (FE-2SLS) approach to identify pension effects on farmers’ labor allocation and land transfer decisions. In the FE-2SLS models, pension is instrumented by the average pension of other households in the same village. Column 1 of

Table 2. Effects of pension on household land transfer-out and labor allocation.

	(1) FE-2SLS Model (1st Stage)	(2) FE-2SLS Model (2nd Stage)	(3) FE-2SLS Model (2nd Stage)	(4) FE-2SLS Model (2nd Stage)
	Pension	Transfer-Out Land	Farm Work (Household)	Off-Farm Work (Household)
Pension	—	0.885 * (0.468)	−110.472 *** (33.858)	95.273 ** (43.583)
Average villagers’ pension	0.545 *** (0.105)	—	—	—
Male share	0.134 (0.239)	0.157 (0.587)	1.462 (42.461)	0.139 (54.658)
Age	0.019 *** (0.006)	−0.009 (0.017)	2.229 * (1.254)	0.910 (1.614)
Education level	0.069 *** (0.016)	−0.051 (0.051)	11.749 *** (3.721)	15.992 *** (4.790)
Health status	−0.020 (0.046)	−0.052 (0.112)	23.597 *** (8.094)	18.608 * (10.419)
Number of pensioners	0.454 *** (0.061)	−0.158 (0.272)	76.405 *** (19.655)	31.389 (25.301)
Number of non-pensioners	0.078 ** (0.037)	0.027 (0.096)	8.603 (6.916)	147.290 *** (8.903)
Farmland	−0.000 (0.001)	0.006 ** (0.003)	0.117 (0.222)	−0.630 ** (0.286)
Irrigation	−0.026 (0.104)	1.006 *** (0.253)	−34.854 * (18.284)	10.723 (23.536)
Land fertility	−0.009 (0.051)	0.161 (0.125)	−2.435 (9.083)	8.661 (11.692)
Production assets	0.004 (0.007)	−0.005 (0.018)	2.035 (1.326)	0.082 (1.707)
Village off-farm environment	0.024 (0.084)	0.238 (0.204)	5.092 (14.785)	−10.663 (19.032)
Year FE	−0.028 (0.039)	0.404 *** (0.096)	−12.192 * (6.963)	6.680 (8.963)
Observations	2936	2936	2936	2936
F-test	27.21	—	—	—
Endogeneity test	—	3.487	16.595	6.875
p-value	—	0.062	0.000	0.009

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Standard errors are in the parentheses. The first-stage F-statistic is 27.210. The Anderson canonical correlation LM statistic is 26.946 (p = 0.0000). The Cragg–Donald Wald F statistic is 27.206, which is above the Stock-Yogo 10% critical value of 16.38.

reports the first-stage results of the FE-2SLS model. The estimates show that the average villagers’ pension as instrument strongly predicts household pension. Moreover, the F-test is greater than 10 and the Cragg–Donald Wald F statistics are 27.545, suggesting the rejection of the hypothesis about the weakness of the instruments. The underidentification test rejects the null hypothesis of underidentification, indicating that the model is properly identified. The endogeneity test further indicates that pension should be regarded as an endogenous variable. Therefore, the following analysis mainly relies on the FE-2SLS estimates.

The results in Table 2 show that pension has significant effects on household land transfer and labor allocation. Pension is positively associated with land transfer-out and household off-farm labor, and negatively associated with household farm labor. The estimates indicate that pension promotes land transfer-out, reduces household agricultural labor input, and increases household off-farm labor input, with statistical significance across these outcomes. Pension does not lead households to withdraw from the labor market as a whole. Instead, it shifts household resource allocation away from farming and toward off-farm activities, while also leading to a partial withdrawal from land operation through land transfer-out. This finding is consistent with previous studies of Ji et al. [16] and Shu [37]. Pension, as a stable and predictable source of non-labor income, changes households’ original income constraints and production arrangements. It reduces households’ dependence on agricultural income and self-cultivated land, thereby weakening the need to maintain the original scale of agricultural operation and labor input. On this basis, households reduce farm labor input, reallocate part of their labor to off-farm activities, and transfer out part of their land. These changes reflect a joint adjustment in land operation and labor allocation after pension enters the household.

Regarding control variables, larger contracted farmland is positively linked to land transfer-out behavior and negatively linked to household off-farm labor, indicating that households with more land resources are more likely to participate in land transfer while still needing to retain some labor for agricultural production. The number of non-pensioners significantly increases household off-farm labor days, indicating that non-pensioners remain the main contributors to off-farm labor input.

6.2. The Effects of Pension on Labor Allocation Among Different Household Members

While the previous section focuses on the overall household labor allocation, this section examines how labor adjustments differ between pensioners and non-pensioners. The identification tests are consistent with those in the previous section, suggesting that the instrument remains relevant and that weak-instrument concerns are limited.

The results in

Table 3. Second-stage FE-2SLS estimates of pension effects on intra-household labor allocation.

	(1)	(2)	(3)	(4)
	Farm Work (Pensioner)	Off-Farm Work (Pensioner)	Farm Work (Non-Pensioner)	Off-Farm Work (Non-Pensioner)
Pension	−35.665 * (20.129)	24.791 (17.090)	−74.807 *** (22.801)	67.701 * (37.182)
Male share	−11.746 (25.244)	17.343 (21.432)	13.208 (28.595)	−17.758 (46.629)
Age	−1.396 * (0.745)	−1.687 *** (0.633)	3.625 *** (0.844)	2.763 *** (1.377)
Education level	6.727 *** (2.212)	−1.928 (1.878)	5.022 ** (2.506)	18.220 *** (4.086)
Health status	15.283 *** (4.812)	0.089 (4.086)	8.314 (5.451)	19.039 ** (8.889)
Number of pensioners	85.950 *** (12.208)	29.979 *** (9.921)	−8.782 (13.236)	2.512 (21.585)
Number of non-pensioners	−13.330 *** (4.112)	−10.713 *** (3.491)	21.933 *** (4.658)	158.099 *** (7.595)
Farmland	−0.017 (0.132)	−0.209 * (0.112)	0.134 (0.150)	−0.420 * (0.244)
Irrigation	−17.474 (10.870)	13.599 (9.229)	−17.380 (12.313)	0.828 (20.079)
Land fertility	−4.160 (5.400)	6.089 (4.585)	1.725 (6.117)	2.273 (9.975)
Production assets	−0.713 (0.788)	1.183 * (0.669)	2.749 *** (0.893)	−1.075 (1.456)
Village off-farm environment	4.083 (8.790)	−13.787 (7.463)	1.009 (9.957)	3.542 (16.236)
Year FE	−7.544 (4.140)	2.659 (3.515)	−4.648 (4.689)	2.737 (7.646)
Observations	2936	2936	2936	2936

Note: This table reports only the second-stage FE-2SLS estimates. First-stage and identification statistics are reported in Table 2. ***, ** and * test statistical significance at 1%, 5% and 10%; Standard errors in parentheses.

show clear differences in the effects of pension on labor allocation across household members. Pension significantly reduces pensioners’ farm labor (Column 1) but has no significant effect on pensioners’ off-farm labor (Column 2). The evidence indicates that, after pension income enters the household, the labor adjustment of pensioners mainly takes the form of withdrawing from farm work rather than significantly moving into off-farm employment. This finding is in line with the finding of Ning et al. [38], who find that the NRPS decreases pensioners’ participation in farm activities and encourages non-market activities, with little increase in off-farm work. For non-pensioners, pension significantly increases off-farm labor (Column 4), with a negative and significant effect on farm labor (Column 3), suggesting that non-pensioners reallocate labor away from farm activities and increase off-farm participation. In other words, the household responds by shifting work from pensioners to non-pensioners, with the latter reallocating time toward off-farm opportunities. Pensioners exhibit lower labor productivity due to the decline of both physical capacity and professional skills. A stable source of pension further reduces the marginal utility of labor, thereby increasing their preference for leisure. Compared with non-pensioners, pensioners face greater obstacles in accessing non-agricultural employment, including skill deficiencies, age-related hiring biases, and health constraints. Consequently, they are more inclined to transition from agricultural work to leisure, rather than from off-farm employment. Non-pensioners are generally in the phase of wealth accumulation and career development within their life cycle, during which they tend to prioritize absolute income growth over income stability.

Regarding control variables, education has significantly positive effects on both agricultural and off-farm labor among non-pensioners, suggesting that a higher education level strengthens their labor allocation capacity, especially their ability to move into off-farm work. Health status has a significantly positive effect on pensioners’ agricultural labor and non-pensioners’ off-farm labor, indicating that labor capacity remains an important factor shaping labor input.

Taken together, the evidence indicates a common mechanism of intra-household labor reallocation. Both pensioners and non-pensioners reduce their engagement in farm labor, but the increase in off-farm labor is concentrated among non-pensioners. This pattern aligns with Huang and Zhang [39]. These findings support the intra-household reallocation interpretation.

6.3. Heterogeneous Effects of Pension on Farmer’s Land Transfer and Labor Allocation Behaviors

Higher pension allows farmers to reduce farm labor, increases off-farm labor, and ultimately promotes land transfers. These changes reflect how farmers re-optimize resource such as land and labor allocation. However, all farmers are not able to make such adjustments under the budget shortage. To examine this issue,

Table 4. Heterogeneity by budget constraints.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	LTO	FW (HH)	OFW (HH)	FW (P)	OFW (P)	FW (NP)	OFW (NP)
Panel A. Without budget shortage
Pension	1.166 *** (0.444)	−92.655 *** (39.407)	88.537 ** (39.407)	−34.505 ** (18.678)	19.244 (15.472)	−58.149 *** (18.749)	66.325 ** (33.572)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2388	2388	2388	2388	2388	2388	2388
Panel B. Budget shortage
Pension	22.093 (59.466)	1224.467 (3325.812)	−1143.52 (3186.437)	360.3494 (1071.713)	−520.792 (1431.929)	864.118 (2329.972)	−622.728 (1890.419)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	546	546	546	546	546	546	546
Panel C. Low Budget Constraint
Pension	1.069 * (0.551)	−101.347 ** (37.523)	70.655 (48.495)	−42.809 ** (22.480)	21.480 (18.305)	−58.538 ** (24.418)	50.555 (42.453)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1468	1468	1468	1468	1468	1468	1468
Panel D. High Budget Constraint
Pension	0.779 (0.727)	−122.838 (57.622)	112.312 (72.525)	−29.947 (29.214)	27.734 (41.164)	−92.891 (410.812)	77.202 (59.822)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1468	1468	1468	1468	1468	1468	1468

Note: ***, ** and * test statistical significance at 1%, 5% and 10%; Standard errors in parentheses. LTO = land transfer-out; FW = farm work; OFW = off-farm work; HH = Household; P = pensioner; NP = non-pensioner.

reports heterogeneity results based on two alternative grouping strategies. The first strategy uses households’ self-reported shortage experience and separates the sample into households with and without budget shortage. The second strategy uses the composite budget-constraint indicator constructed from shortage amount, household income, and household expenditure, and then divides the sample into low- and high-budget-constraint groups by the median.

As shown in Panel A, for the households without budget shortage, the coefficient of pension is significantly and positively associated with land transfer-out and off-farm labor among household and non-pensioners, and negatively associated with farm labor at both. The impact of pension on land transfer and labor allocation remains consistent across the full sample and the group of households without budget shortages. A similar pattern is also observed in Panel C, where the lower budget constraints group shows the positive effects of pension on land transfer-out and the negative effects of pension on farm labor, although the estimated effect on off-farm labor is less stable. This indicates that pension provides stable disposable income for less constrained households, reducing the need for savings and reliance on land as retirement security. As a result, farmers are more inclined to reduce low-return farm work, leading to greater land transfer.

By contrast, as shown in Panels B and D of Table 4, in the shortage experience group and the high budget constraint group, most estimated coefficients are not statistically significant, and the first-stage results show that the explanatory power of the instrument for household pension becomes notably weaker. Under the grouping based on shortage experience, households with reported shortage account for only about 20% of the sample, and the uneven sample distribution substantially reduces estimation precision. Under the grouping based on the composite budget-constraint indicator, the subgroup sizes are more balanced and there is no obvious sample imbalance. However, the estimates for households with higher budget constraints remain broadly insignificant and the first-stage identification is still weak. This suggests that the instability of the results among households with higher budget constraints is not simply driven by uneven sample distribution.

Across the two grouping strategies, the effects of pension on households’ land and labor allocation are more clearly observed among households with lower budget constraints. In these groups, pension is more often associated with greater land transfer-out and lower farm labor input, and in some specifications also with higher off-farm labor. By contrast, for households with higher budget constraints, the estimates are generally imprecise and remain inconclusive under the current sample and identification conditions. This pattern is broadly consistent with Cheng et al. [40], who report that pension effects are more evident among relatively advantaged groups.

6.4. Robustness Check

To further assess the robustness of the results, we conduct additional checks from the perspectives of identification validity and alternative variable measurement. As shown in

Table A1. Robustness check: controlling for village-specific trends.

	(1)	(2)	(3)	(4)
	FE-2SLS Model (1st Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)
	Pension	Transfer-Out Land	Farm Work (Household)	Off-Farm Work (Household)
Average villagers’ pension	0.513 *** (0.108)	—	—	—
Pension	—	0.851 * (0.506)	−145.001 *** (42.017)	98.850 ** (48.550)
Year FE	0.085 (0.241)	−0.083 (0.576)	125.090 (47.831)	145.463 (55.267)
Control variables	Yes	Yes	Yes	Yes
Baseline village characteristics × Year FE	Yes	Yes	Yes	Yes
Observations	2936	2936	2936	2936
F-test	22.00	—	—	—

Note: ***, ** and * test statistical significance at 1%, 5% and 10%; Standard errors in parentheses. Control variables are the same with the baseline regression. Baseline village characteristics interacted with the year dummy are included in all specifications to absorb differential village-specific trends.

, after adding interactions between baseline village characteristics and the year dummy to absorb differential trends associated with initial village conditions, the estimated effects of pension on land transfer-out, household farm work, and household off-farm work remain stable in both sign and significance. This suggests that the baseline results are unlikely to be mainly driven by village-level common trends. We also use a fixed-effects model in the subsample of households without pensioners to examine whether the average village pension affects land transfer and labor allocation. The results are reported in

Table A2. Placebo test: households without pensioners.

	(1) Transfer-Out Land	(2) Farm Work (Household)	(3) Off-Farm Work (Household)
Average villagers’ pension	0.328 (0.597)	−41.174 (47.260)	7.679 (67.790)
Year FE	0.297 * (0.153)	−3.146 (12.216)	−1.609 (17.523)
Control variables	Yes	Yes	Yes
Observations	692	692	692

Note: * test statistical significance at 10%; Standard errors in parentheses. Control variables are the same with the baseline regression.

. In this sample, the average villagers’ pension does not significantly affect land transfer-out, household farm work, or household off-farm work. This suggests that the instrument does not predict land or labor allocation in the absence of the household pension channel, which provides additional support for the identifying assumption.

We also examine whether the main findings depend on the measurement of key variables. As shown in

Table A3. Robustness check: alternative measure of pension.

	(1)	(2)	(3)	(4)
	FE-2SLS Model (1st Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)
	Pension	Transfer-Out Land	Farm Work (Household)	Off-Farm Work (Household)
Average villagers’ pension	0.271 *** (0.059)	—	—	—
Pension	—	1.778 * (0.957)	−221.812 *** (71.453)	191.294 ** (89.545)
Year FE	−0.014 (0.022)	0.405 *** (0.098)	−12.231 (7.320)	6.714 (9.173)
Control variables	Yes	Yes	Yes	Yes
Observations	2936	2936	2936	2936
F-test	21.04	—	—	—

Note: ***, ** and * test statistical significance at 1%, 5% and 10%; Standard errors in parentheses. Control variables are the same with the baseline regression.

, when household pension is replaced by pension measured on a per-pensioners basis, the core results remain unchanged. Pension still promotes land transfer-out, reduces household farm work, and increases household off-farm work. This suggests that the baseline findings do not rely on a specific definition of pension.

Table A4. Robustness check: Winsorized pension.

	(1)	(2)	(3)	(4)
	FE-2SLS Model (1st Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)
	Pension	Transfer-Out Land	Farm Work (Household)	Off-Farm Work (Household)
Average villagers’ pension	0.479 *** (0.099)	—	—	—
Pension	—	1.008 * (0.536)	−125.811 *** (39.753)	108.502 ** (50.413)
Year FE	−0.029 (0.038)	0.408 (0.097)	−12.690 (7.194)	7.110 (9.124)
Control variables	Yes	Yes	Yes	Yes
Observations	2936	2936	2936	2936
F-test	23.52	—	—	—

Note: ***, ** and * test statistical significance at 1%, 5% and 10%; Standard errors in parentheses. Control variables are the same with the baseline regression.

reports the results after Winsorizing household pension at the 99th percentile. The coefficients on land transfer-out, household farm work, and household off-farm work remain similar in sign, magnitude, and significance to the baseline estimates, suggesting that the main findings are not driven by a small number of extreme observations.

Table A5. Robustness check: alternative measures of land transfer-out.

	(1)	(2)	(3)	(4)
	FE-2SLS Model (1st Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)	FE-2SLS Model (2nd Stage)
	Pension	Transfer-Out Land	Percentage of Land Transferred Out	Area of Land Transferred
Average villagers’ pension	0.545 *** (0.105)	—	—	—
Pension	—	0.322 *** (0.084)	0.207 *** (0.065)	0.446 *** (0.136)
Year FE	−0.028 (0.039)	0.054 (0.017)	0.057 (0.013)	0.114 (0.028)
Control variables	Yes	Yes	Yes	Yes
Observations	2936	2936	2936	2936
F-test	27.21	—	—	—

Note: *** test statistical significance at 1%; Standard errors in parentheses. Control variables are the same with the baseline regression.

further replaces the land transfer-out variable with alternative measures, including a dummy for whether land is transferred out, the share of transferred-out land in total contracted land, and the logarithm of transferred-out land area. The estimated effects remain positive and statistically significant across specifications, indicating that the conclusion on land adjustment is robust to alternative measures of land transfer-out. These robustness checks show that the conclusion that pension promotes land transfer-out, reduces household farm labor input, and facilitates household off-farm labor allocation is robust to alternative specifications and variable definitions.

7. Conclusions and Policy Implications

Efficient allocation of land and labor resources is key to mitigating land scarcity and improve agricultural productivity under large rural populations in many developing countries. Pension is a candidate policy instrument in this scenario. In this article, we employ the FE-2SLS approach to identify pension effects on farmers’ land transfer and labor allocation decisions. We also estimate the heterogeneous effects of pension among farmers from the perspective of budget constraints. Three main findings can be drawn from the present study. First, we find that pension has a positive impact on farmers’ land transfer decision and promotes a reallocation of labor away from farming and toward off-farm activities. This suggests that pension changes households’ overall production and labor allocation decisions. Second, we identified an intra-household heterogeneity behind the pension effects. Pensioners primarily reduce their participation in farm work without reallocating to off-farm work, while non-pensioners reallocate from farming to off-farm work. Third, across the two grouping strategies, the effects of pension are more clearly observed among households facing lower budget constraints. In these groups, pension is associated with greater land transfer-out and lower farm labor input, and in some specifications also with higher off-farm labor. By contrast, the estimated effects are generally imprecise among households with higher budget constraints. These results suggest that the evidence supports the effects of pension on land transfer and labor allocation among households with lower budget constraints, while further examination is still needed for households with higher budget constraints.

The findings of this study provide important policy implications. First, policymakers can pay more attention to the role of pension in supporting land transfer, but this effect should not be viewed as uniform across all households, since the results show that pension effects are more evident among households without serious budget shortage. Improving the stability and predictability of pension benefits is more likely to support land transfer where households already face relatively limited budget constraints. Second, policymakers can support better-functioning land transfer and labor markets to ensure efficient allocation of land and labor resources. For example, the government can establish a standardized, low-cost land transfer platform to facilitate trade with each other. Finally, special measures are needed for households with higher budget constraints. The result indicates that solely enhancing pension income might not be sufficient to generate reallocation responses among this group. In such settings, pension policy may need to be complemented by measures that reduce immediate financial pressure and improve access to more stable income opportunities.

This study has the following limitations. First, our two-period panel dataset limits our ability to capture the long-term dynamic effects of pension on land transfer and labor reallocation behaviors. Our findings should be interpreted as short- to medium-term average effects. Future research could extend the tracking period to better understand impacts in the long run. Second, we measure budget shortage with a binary indicator and a composite budget-constraint indicator. Although these two measures help capture household budget constraints from different angles, they still cannot fully reflect the intensity and dynamic changes in budget constraints. Future research could use richer balance-sheet information and longer panel data to improve the measurement of household financial constraints and the identification of heterogeneity.