Exercise Linked to Higher Fertility Intentions in Chinese Women, Survey Analysis Finds

Women of childbearing age who exercise more often report wanting more children, according to a new analysis of nationally representative Chinese survey data published in Frontiers in Public Health. The study, based on the 2021 Chinese General Social Survey, found a statistically significant positive association between how frequently women engaged in physical exercise and the number of children they said they would ideally like to have. The researchers, led by Mengmeng Man and Junhua Zhou of Ludong University, argue that their findings point to physical activity as a potentially low-cost, non-economic factor that could complement existing fertility support policies — while carefully stopping short of claiming that exercise causes higher fertility.

The demographic backdrop motivating the work is stark. Data from the Organization for Economic Co-operation and Development show that the total fertility rate across its 38 member countries fell from 3.3 in 1960 to 1.5 in 2022, and projections cited in the paper suggest that by 2100, 183 countries will have fertility rates below replacement level. China’s Seventh National Census recorded a total fertility rate of 1.3, and the National Bureau of Statistics reported only 9.02 million births in 2023, with a natural population growth rate of −1.48 — the second consecutive year of negative population growth. Despite the introduction of the three-child policy and accompanying support measures in 2021, fertility has continued to decline, prompting researchers to look beyond economic incentives toward individual-level, non-economic influences on the desire to have children.

To investigate one such influence, the team drew on the 2021 wave of the Chinese General Social Survey, a stratified four-stage probability sample covering 21 provinces, autonomous regions and municipalities. Starting from 8,148 respondents, the researchers sequentially excluded men, individuals with divorced or non-standard marital status, and people outside the 20-to-50-year age range, then dropped observations with ambiguous answers on fertility intention or exercise frequency. The final analytical sample comprised 1,230 women. The authors report that the sample’s age composition (mean age 36.1 versus 36.0 in the full survey) and urban–rural ratio (33.2 percent urban versus 33.0 percent) closely mirror the national distribution, supporting the generalizability of the descriptive findings.

Fertility intention was measured with a single questionnaire item asking respondents how many children they would like to have in the absence of policy restrictions, with desired sons and daughters summed into a score from 0 to 10 that was recoded to a maximum of 6. Physical exercise was captured by a separate item asking how often respondents had exercised in their spare time over the past year, coded from 0 (never) to 4 (daily). Health status and subjective well-being, the two candidate mediating variables, were each measured on five-point self-report scales. Controls included age, household registration, education, marital status, social frequency, social interaction, social class and socioeconomic status. The authors acknowledge that the fertility measure captures ideal rather than planned childbearing, a distinction that matters because intentions do not always translate into behavior.

The benchmark analysis used ordinary least squares regression with heteroscedasticity-robust standard errors. Without controls, each one-step increase on the exercise frequency scale was associated with a 0.046 increase in ideal fertility (p < 0.01). With the full set of controls, the coefficient rose to 0.087 (p < 0.01), nearly double the unadjusted estimate. The authors attribute this amplification not to multicollinearity — variance inflation factors ranged only from 1.16 to 1.90, well below conventional thresholds — but to the correction of omitted-variable bias. Education level and urban household registration, which correlated positively with both exercise and fertility intention in simple bivariate comparisons, turned significantly negative once age and marital status were held constant, suggesting they had acted as masking factors in the unadjusted model.

Because the data are cross-sectional and observational, the researchers invested heavily in robustness checks. They redefined exercise as a binary treatment variable and applied propensity score matching, using 1:1 nearest-neighbor matching with a caliper of 0.05. Matching successfully eliminated significant covariate imbalances, reducing standardized mean differences to below 10 percent, and regression on the matched sample of 429 women yielded a coefficient of 0.069 (p < 0.01). Kernel matching (ATT = 0.142) and radius matching with calipers of 0.01 and 0.05 (ATT = 0.155 and 0.141) produced moderately larger and statistically significant estimates. An Ordered Probit model, which treats the 0-to-6 fertility intention score as an ordered category rather than a continuous variable, confirmed the pattern with coefficients of 0.063 and 0.139 in the parsimonious and full specifications respectively.

The team also quantified how sensitive the matching results might be to unmeasured confounding using Rosenbaum bounds. The critical gamma values ranged from 1.3 to 1.4, meaning an unobserved variable would need to raise the odds of exercising by roughly 30 to 40 percent to erase the significant association. The authors flag this as moderate sensitivity: a plausible hidden confounder such as general health consciousness, which could plausibly drive both regular exercise and proactive fertility preparation, might operate at or near this magnitude. They therefore interpret the matching results as suggestive evidence of association rather than definitive causal identification, and note that reverse causation remains a highly persuasive alternative explanation — women who already want children may simply invest more in preconception health behaviors, including exercise.

To probe mechanisms, the researchers fitted a parallel mediation model that simultaneously estimated indirect pathways through health status and subjective well-being. Exercise was positively associated with self-rated health (coefficient 0.068, p < 0.001) and with subjective well-being (coefficient 0.045, p < 0.01), and both mediators independently predicted fertility intention when entered together. The total indirect effect was 0.0085 (95 percent CI 0.0035 to 0.0135), with the health pathway contributing 0.0045 and the well-being pathway 0.0040 — together accounting for just 9.7 percent of the total association, while the direct effect accounted for the remaining 90.3 percent. The significant residual covariance between the two mediators (0.114, p < 0.001) suggested they are correlated but not redundant, capturing complementary physiological and psychological dimensions. The authors stress that the small share of the association explained by these pathways implies other unmeasured mechanisms — partner dynamics, social support, broader lifestyle factors — likely dominate.

Heterogeneity analysis produced a more cautious picture. Descriptively, the exercise coefficient was larger among women aged 30 or younger (0.106) than among older women (0.084), and larger among urban women (0.114) than rural women (0.077). But formal interaction tests found no statistically significant differences between these subgroups: the age interaction term yielded p = 0.465 and the residence interaction p = 0.260. The authors explicitly caution that any interpretation of these descriptive patterns — such as younger women being more physiologically responsive to exercise or urban women benefiting from group-based sports and denser social networks — is exploratory and awaits validation through longitudinal or experimental designs with larger samples.

The study carries clear limitations. Exercise was measured by a single frequency item without distinguishing intensity, duration or type, even though prior evidence suggests dose and intensity matter: a meta-analysis cited in the paper found no significant association between moderate-intensity exercise and fertility among European subgroups, while high-intensity exercise was negatively correlated. The mediators were global five-point self-reports that cannot validate specific physiological or neuroendocrine processes, and the cross-sectional design cannot establish the chronological order of the variables. With these caveats in place, the authors conclude that physical exercise merits further investigation as a potential adjunct to fertility-friendly policy — perhaps integrated into comprehensive preconception care rather than promoted as a standalone intervention — but that any population-level impact would likely be small, and policy implementation should await evidence from interventional or quasi-experimental studies.

Subject of Research: The association between physical exercise and fertility intention among Chinese women of childbearing age

Article Title: The positive correlation between physical exercise and the fertility intention of women of childbearing age: an empirical analysis based on the Chinese General Social Survey

Article References: Man, M., Zhang, K., Zhao, C., Chu, Y., & Zhou, J. (2026). The positive correlation between physical exercise and the fertility intention of women of childbearing age: an empirical analysis based on the Chinese General Social Survey. Frontiers in Public Health, 14, 1927726. https://doi.org/10.3389/fpubh.2026.1927726

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1927726

Keywords: fertility intention, physical exercise, China, Chinese General Social Survey, health status, subjective well-being, mediation analysis, propensity score matching, low fertility, women of childbearing age, public health, demography