People quasi-randomly assigned to farm rice are more collectivistic than people assigned to farm wheat

The study tests the causal claim of the rice theory of culture: that paddy rice farming, with its intensive labor demands and interdependent irrigation systems, fosters more collectivistic tendencies than wheat farming. Prior cross-regional evidence shows correlations between rice cultivation and collectivism, but causality is hard to establish due to third-variable confounds and the impracticality of random assignment to long-term farming styles. The authors leverage a historical context in China where the government assigned people to state farms to cultivate either rice or wheat. Two such farms in Ningxia Province are geographically proximate and environmentally similar, creating conditions akin to a natural experiment. The research asks whether people assigned to rice farming exhibit more collectivistic psychological traits than those assigned to wheat farming, mirroring broader north–south differences in China.

Background work argues that paddy rice’s cooperative irrigation and higher labor requirements cultivate interdependence and collectivism, with examples across regions from Japan to Sierra Leone. Prior studies have reported rice–wheat differences in China on social norms, cognition, and behavior, as well as East–West cultural contrasts on related measures. Earlier work by the authors and others showed tighter social norms and greater holistic thought in historically rice-farming areas, and observational studies have documented rice–wheat cultural differences even among urban non-farmers (e.g., Starbucks seating behavior). A prior visit to the same Ningxia farms found more social comparison and lower happiness among rice-farm residents, paralleling broader regional patterns. These literatures motivate testing whether rice cultivation itself, rather than geography or history, causes collectivism-related psychological differences.

Design and setting: The study exploits quasi-random assignment of Chinese citizens during the mid-20th century to two state farms in Ningxia Province: Lianhu (rice) and Qukou (wheat), approximately 50–56 km apart. The farms share similar temperature, precipitation, and general topography, but the wheat farm’s higher elevation relative to the Yellow River, sandier soils, and irrigation impracticality prevented large-scale paddy rice, enabling a natural experiment. Historical records indicate the areas were largely unfarmed prior to state investment, reducing the chance that pre-existing local farming cultures drive results. Both farms were under the same land reclamation bureaucracy, reducing policy rollout differences.

Participants and recruitment: Target N was 200 (about 100 per farm), with data collection in September–October 2017 via division leaders who informed farmers, who then scheduled sessions. Total sample was 234 farmers; most (73.5%) grew up on the farms (descendants of originally assigned residents), and 96.2% were farming in the study year. Tests were administered individually on paper with a research assistant available; completion took 30–40 minutes. Compensation was a 30 Yuan phone recharge card or a family-size jug of dishwashing detergent.

Measures:

• Implicit individualism (self-inflation) via sociogram task: participants draw circles representing themselves and family (and separately friends). Larger self relative to others indicates greater implicit individualism/self-inflation.
• Loyalty/nepotism: scenario-based task comparing judgments and sanctions/rewards for a friend versus a stranger who behaved identically in a business context; preferential treatment of the friend indexes loyalty/nepotism.
• Holistic vs analytic thought: triad categorization and related perceptual tasks; selecting relational pairings (e.g., rabbit–cat) indicates holistic thought, whereas category-based pairings (e.g., rabbit–carrot) indicate analytic thought.

Controls and analyses: Demographics included age, gender identity, income, religion (Hui indicator), and maternal educational attainment. Analyses used SPSS for descriptives and reliability, and R for regressions and propensity score matching (MatchIt) to balance samples on age, gender, income, and maternal education. Assumptions (normality, homoscedasticity) were checked; p-values are two-tailed. The study was not pre-registered. Additional analyses exploited a rice-farm rotation policy starting in 2000 (about one-third of rice-farm participants not farming rice in the study year) to test whether current-year individual rice experience predicts cultural measures, versus community-level farming context.

• Main rice–wheat differences (propensity score–matched samples; Table 3):
  • Self-inflation (family sociogram): Rice farm B = -1.92, SE = 0.73, t = -2.62, P = 0.010, 95% CI [-3.36, -0.47] (less implicit individualism on rice farm).
  • Self-inflation (friends sociogram): Rice farm B = -1.00, SE = 0.73, t = -1.37, P = 0.173, 95% CI [-2.09, 0.79] (not significant).
  • Loyalty/nepotism: Rice farm B = 2.63, SE = 0.63, t = 4.16, P < 0.001, 95% CI [1.46, 3.80] (greater preferential treatment of friends on rice farm).
  • Holistic thought: Rice farm B = 0.28, SE = 0.13, t = 2.24, P = 0.025, 95% CI [0.04, 0.53] (more holistic cognition on rice farm).
• Descriptive figures align with these results: rice farmers showed smaller self-inflation, higher loyalty to friends versus strangers (e.g., larger monetary reward difference), and higher rates of relational (holistic) categorizations (e.g., ~90% vs ~80%). Samples were N ≈ 196 overall (N = 188 for thought task due to missing data).
• Demographic effects: Some demographics related to outcomes (e.g., maternal education associated with less loyalty/nepotism; age related to holistic thought in some models), but rice–wheat differences remained after controlling for demographics and after propensity score matching.
• Robustness to current-year experience (Table 4): Individual current-year rice farming did not predict cultural measures; farm-level rice–wheat differences remained when adding individual farming experience. This suggests community-level rice ecology, not just immediate individual experience, underlies differences.
• No evidence of self-selection or policy differences: Historical records and interviews indicate assignments were effectively random, farms were founded on largely unfarmed land, and shared administration led to parallel policy timing, reducing major confounds like temperature, latitude, prior local culture, or policy variation.

Findings show that farmers on the rice state farm exhibit more collectivistic hallmarks than those on the nearby wheat farm: stronger loyalty/nepotism toward friends, more holistic cognition, and lower implicit individualism. The quasi-random assignment across environmentally similar farms strengthens causal inference, effectively ruling out confounds such as latitude, temperature, and divergent historical events that typically complicate regional comparisons. Persistence of differences among rice-farm residents not farming rice in the study year suggests community norms and accumulated experience maintain these cultural patterns beyond immediate tasks. Results mirror broader rice–wheat contrasts within China and classic East–West differences on analogous measures, supporting the hypothesis that subsistence ecology—particularly interdependent rice farming—contributes causally to collectivism and related cognitive styles. The natural experiment context also diminishes the plausibility of genetic explanations for these farm-level differences. Together, the findings indicate that key cultural differences can emerge within a single generation of exposure to rice-farming ecology and are likely amplified over longer periods by social institutions.

This study provides rare quasi-experimental evidence that rice farming causally fosters more collectivistic tendencies than wheat farming. Leveraging two closely matched state farms with quasi-random assignment, the authors demonstrate that rice-farm residents show greater loyalty/nepotism toward close others, more holistic thought, and lower implicit individualism. Differences persist net of demographics and even among those not currently farming rice, indicating community-level cultural processes. These results substantiate the rice theory of culture and help explain broader rice–wheat and East–West differences. Future research should extend this framework to other subsistence styles (e.g., herding, fishing), investigate how social institutions (e.g., clan structures) magnify or attenuate ecological effects over generations, and test mechanisms linking interdependent production demands to specific social-cognitive adaptations.

• Sampling and demographics: Participants were not recruited via probability sampling in these remote settings, and demographic differences (e.g., age, maternal education) existed between farms. The authors mitigated this via measuring demographics, regression controls, and propensity score matching, but residual confounding is possible.
• Task suitability and reliability: Some tasks initially designed for college samples were challenging for farmers. Difficulties recalling friends on the sociogram led to a family-version adaptation. The relational mobility questionnaire showed low reliability and was discarded, limiting breadth of measures.
• Generalizability and scope: The comparison involves two specific state farms with unique historical contexts; while quasi-experimental, effects may differ in regions with entrenched social institutions (e.g., family clans) that could amplify differences over longer timescales.