How Outsourced Farming Services Are Quietly Reshaping Who Farms China’s Wheat Land

Across the wheat belts of Shandong, Hebei, Henan, and Shaanxi, a quiet organizational revolution is underway. Instead of buying tractors, sprayers, and harvesters, a growing share of smallholder households are hiring specialized providers to handle land preparation, mechanical sowing, aerial crop protection, harvesting, and technical guidance. New research published in Frontiers in Sustainable Food Systems suggests that this outsourcing of fieldwork is not merely a convenience: it is statistically linked to how much land wheat-growing households operate, how much they rent in, and how much they keep rather than rent out. The study, based on 1,178 valid household questionnaires collected across four major wheat-producing provinces, offers one of the more granular portraits to date of how service markets and land rental markets move together in China’s countryside.

The research team, led by Yihang Jiang of Qingdao University and Yue Sun of Ningbo University of Finance and Economics, set out to answer a deceptively simple question: when households purchase agricultural production services, do their farmland holdings change? The question matters because China’s agricultural transformation is constrained by distinctive institutions. Household contract farming remains the basic organizational form, operated plots are small and dispersed, and farmland rental markets are shaped by transaction relationships, incomplete contracts, and information frictions. Simply renting more land raises search and bargaining costs and does not automatically provide the machinery, technology, or managerial capacity that a larger operation demands. Expanding farmland area and improving production capacity, the authors emphasize, are not the same process.

Agricultural production services offer an alternative organizational arrangement. By outsourcing labor-intensive and equipment-specific stages of production, households can access machinery and skills without making one-time investments in highly specific equipment. Service providers pool demand across multiple farm households, spreading equipment and technology costs over a larger operating area. Farm households, in turn, substitute variable expenditure for fixed investment, lowering the capital barrier to scaling up. The authors call this service-based scaling: an increase or maintenance of operational capacity through purchased production-stage services rather than through household ownership of machinery. It is narrower than general outsourcing, distinct from mechanization ownership because the capital asset remains with the provider, and complementary to land-based scaling because it relaxes nonland constraints while land rental reallocates the land input itself.

The empirical strategy is notable for its caution. The team measured service use in two ways: by expenditure on wheat production services, which averaged 2,345.78 yuan per household, and by the number of production stages served, which averaged 3.48 and is less mechanically related to cultivated area. They also deployed two-part models to distinguish the decision to participate in land rental from the area rented conditional on participation, a distinction that matters because rented-in and rented-out areas contain many zeros. As a complementary check, they drew on an independent three-wave panel from China’s national rural fixed observation points covering 2019 to 2021. Households in the cross-sectional sample operated an average of 13.20 mu of farmland, about 0.88 hectares, rented in an average of 6.13 mu, and rented out only 0.87 mu.

The headline results are strikingly consistent. In the baseline ordinary least squares specifications, service expenditure was positively associated with operated farmland area and rented-in area, and negatively associated with rented-out area. The coefficients were 0.707 for operated area, 1.438 for rented-in area, and minus 0.531 for rented-out area, all statistically significant at the one percent level. Supplementary instrumental-variable estimates, using the leave-one-out mean of service use among other sampled households in the same village, produced similar signs and slightly larger magnitudes. The first-stage F statistic of 449.62 supported instrument relevance, though the authors are careful to note that instrument relevance alone cannot establish the exclusion restriction, since village-level service supply may also directly affect local land markets.

Robustness checks reinforced the core pattern while revealing its limits. When service expenditure was replaced by the number of service stages, the positive associations with operated area and rented-in area persisted, but the negative rented-out association became statistically insignificant. Two-part models added a further nuance: service breadth was positively associated with the probability of renting in land but not with the conditional rented-in area among participants, and it was not significantly associated with the probability of renting out land, while the conditional rented-out area was smaller among participants. In other words, the extensive and intensive margins of land rental need not move together. The independent panel validation, meanwhile, found that wheat service expenditure per mu was positively associated with both operated area and wheat-growing area within households over time, with coefficients of roughly 0.218 and 0.407 respectively.

Why would purchasing services be linked to land reallocation? The authors propose two channels, both tested only in correlational form. The first is labor substitution. Wheat production is highly seasonal, and rural labor out-migration and population aging intensify constraints during critical farming periods. Machinery operations and aerial crop-protection services substitute capital and specialized skills for household labor, reducing labor input per mu and allowing household workers to reallocate between agricultural and nonagricultural employment. The data are consistent with this channel: service expenditure was negatively associated with labor input per mu, which in turn was negatively associated with operated area. The second channel operates through expected returns. Services may improve households’ assessments of farming returns through more timely operations, more standardized technologies, and more stable risk management. Here the evidence is weaker: service use was positively associated with expected returns, and expected returns were positively associated with operated area, but including expected returns barely moved the service coefficient.

Perhaps the most policy-relevant findings concern heterogeneity. Formal interaction tests showed that the service associations differed significantly by initial farm size. For households operating less than 13 mu, the threshold used in the analysis, services were linked to a positive renting-in response and a significant reduction in renting out, suggesting an operation-stabilizing margin. For households operating at least 13 mu, the rented-in association was much stronger, pointing to an expansion margin, while the rented-out association was insignificant. Age produced a different split: younger households, with heads under 56, showed a stronger rented-in association, whereas older households showed a comparatively stronger reduction in renting out, consistent with greater reliance on services to substitute for household labor. The same service, in short, may generate an expansion effect for some households and an operation-stabilizing effect for others.

The authors are explicit about the boundaries of their claims. The main survey is cross-sectional and purposively limited to active wheat growers in four provinces; households that had completely rented out their land, exited wheat production, or withdrawn from farming were not represented. The mechanism estimates cannot satisfy the temporal-ordering conditions required for causal mediation, and the panel validation cannot remove time-varying confounding. The study also does not measure carbon emissions, pollution reduction, or digital-technology intensity, so no claims are made about environmental performance or digital rural transformation. The findings, the authors stress, should not be generalized beyond comparable wheat-producing settings.

Even with those caveats, the implications are considerable. For comparable wheat-producing areas, the results support cautious policy steps: improving smallholders’ access to multi-stage production services through transparent pricing, quality standards, and contract enforcement; coordinating service provision with local land-market institutions, since services and rental decisions appear to be complementary margins of farm adjustment; and linking service-operation records with panel measures of inputs, yields, income, and environmental outcomes before drawing conclusions about digital transformation or carbon mitigation. As China’s wheat growers age and migrate, the study suggests that who farms the land may be determined less by who owns the machinery than by who can hire it, and that service markets and land rental markets are best understood, and best governed, as two sides of the same farm-scale equation.

Subject of Research: Associations between agricultural production service use and farmland reallocation among wheat-growing households in China

Article Title: Agricultural production services and farmland reallocation among wheat-growing households in China

Article References: Jiang, Y., & Sun, Y. (2026). Agricultural production services and farmland reallocation among wheat-growing households in China. Frontiers in Sustainable Food Systems, 10, Article 1968687. https://doi.org/10.3389/fsufs.2026.1968687

Image Credits: AI Generated

DOI: 10.3389/fsufs.2026.1968687

Keywords: agricultural production services, farmland reallocation, land rental, wheat production, smallholders, China, farm scale, labor substitution, mechanization, rural panel data, Agricultural, production