The physical activity paradox in later life: does leisure-time exercise modify the association between agricultural labor and self-rated health?

Abstract Background Healthy ageing requires more than total physical activity volume; the context, voluntariness, recovery opportunity, and social meaning of activity may shape how activity relates to health in later life. Guided by the physical activity paradox, this study examined whether leisure-time exercise moderates the longitudinal association between agricultural labor and self-rated health among older Chinese adults. Methods Data were drawn from five waves of the China Family Panel Studies (2010, 2012, 2014, 2016, and 2018). The analytical sample included 18,050 person-wave observations from 9,449 adults aged 60 years and older. Self-rated health was modeled as a five-point outcome in individual fixed-effects linear models. Standard errors were clustered at the individual level. Agricultural labor, leisure-time exercise, and their interaction were the principal explanatory variables. Results In the fully adjusted interaction model, agricultural labor was associated with lower self-rated health among respondents who did not exercise (β = −0.753, 95% CI − 0.856 to − 0.651, p < 0.001). Leisure-time exercise showed a small but statistically non-significant association with self-rated health among respondents not engaged in agricultural labor (β = 0.047, 95% CI − 0.006 to 0.099, p = 0.082). The agricultural labor × leisure-time exercise interaction was positive (β = 0.898, 95% CI 0.793 to 1.002, p < 0.001), consistent with a less negative agricultural labor-SRH pattern among respondents who exercised. Age-stratified interaction estimates were numerically larger among adults aged ≥ 70 years than among those aged 60–69 years; however, formal between-group differences were not tested. Conclusions Agricultural labor and leisure-time exercise were not interchangeable activity domains in this longitudinal sample. Agricultural labor was related to lower self-rated health, the general main association of exercise was not statistically significant in the fully adjusted interaction model, and the interaction suggested that the exercise-health association was contingent on agricultural labor status. Rural healthy-ageing strategies should distinguish physically demanding work from voluntary, recovery-sensitive exercise while avoiding causal claims from observational associations.