Multidimensional opportunity constraints were associated with a clear population-level gradient in subsequent weekly exercise participation, though the null fixed-effects result indicates this gradient reflects between-person inequality rather than short-term within-person causal association.
Key Findings
Results
Weekly exercise participation rates showed a clear gradient across constraint groups, ranging from 43.94% in the low-constraint group to 19.24% in the high-constraint group.
Subsequent weekly exercise participation was 43.94% in the low-constraint group
Participation was 27.65% in the moderate-constraint group
Participation was 19.24% in the high-constraint group
The analytic sample included 32,306 person-period observations from 19,956 respondents
Data came from CFPS waves 2018, 2020, and 2022, with two lagged person-period intervals: 2018-2020 and 2020-2022
Results
Moderate and high opportunity constraints were each independently associated with lower odds of subsequent weekly exercise participation compared to low constraints.
Moderate constraints: OR = 0.555, 95% CI: 0.524–0.588, p < 0.001
High constraints: OR = 0.369, 95% CI: 0.340–0.401, p < 0.001
Models were adjusted for prior exercise participation, age, sex, marital status, period fixed effects, and province fixed effects
Standard errors were clustered at the individual level
Results
The continuous opportunity-constraint score showed an inverse association with subsequent weekly exercise participation in adjusted lagged logistic models.
OR = 0.764, 95% CI: 0.750–0.779, p < 0.001 per unit increase in the continuous score
The score ranged from 0 to 9 based on nine binary indicators
Indicators spanned four domains: resource conditions, time conditions, health feasibility, and information access
Results
Resource, time, and information-access constraints were each inversely associated with subsequent exercise participation, but health-feasibility constraints were not independently associated after mutual adjustment.
Health-feasibility constraints showed no independent association with the outcome after adjusting for the other constraint dimensions
Resource, time, and information-access constraint dimensions each retained inverse associations in the mutually adjusted models
The exposure was constructed from nine binary indicators across the four domains
Results
An individual fixed-effects logit sensitivity analysis found no association between the continuous constraint score and subsequent exercise participation, indicating the population-level gradient reflects between-person inequality rather than within-person causal effects.
Fixed-effects logit OR = 0.996, 95% CI: 0.927–1.070, p = 0.908
The authors state: 'The null fixed-effects result indicates that this gradient should be interpreted as a between-person inequality rather than evidence of a short-term within-person causal association'
This result contrasts with the strong associations observed in the lagged population-level models
Methods
The study constructed a standardized opportunity-constraint score aggregating nine binary indicators across four domains to capture cumulative conditions enabling or restricting exercise.
The four domains were: resource conditions, time conditions, health feasibility, and information access
The score ranged from 0 to 9, with higher scores indicating greater constraints
Previous literature modeled education, residence, work, health, and internet access as separate covariates rather than cumulative conditions
Individual-level longitudinal data from CFPS 2018, 2020, and 2022 were used
What This Means
This research suggests that people who face multiple overlapping barriers to exercise — such as limited financial resources, lack of free time, poor health, and limited internet access — are much less likely to exercise regularly than those with fewer barriers. Using data from nearly 20,000 Chinese adults tracked over four years, the researchers found that those with the most constraints had roughly half the exercise participation rate (19%) compared to those with the fewest constraints (44%). These associations held up even after accounting for people's prior exercise habits and other personal characteristics.
However, the study also found an important nuance: when the researchers used a statistical method that only looks at changes within the same person over time (individual fixed-effects analysis), the association between constraints and exercise disappeared. This means the gradient in exercise participation is better explained by long-standing differences between people — such as those who have always faced more barriers versus those who have always had more opportunities — rather than by short-term changes in a person's circumstances causing them to exercise more or less.
This research suggests that exercise inequality in China is largely a structural, between-person phenomenon rooted in cumulative disadvantage across multiple life domains, rather than something that shifts quickly as individual circumstances change. Addressing low physical activity rates may therefore require systemic approaches that reduce multiple overlapping barriers simultaneously, rather than targeting any single factor in isolation.
Check Your Own Numbers
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Zhang X, Li W. (2026). Opportunity constraints and social inequalities in exercise participation: a lagged analysis using CFPS 2018-2022.. Frontiers in public health. https://doi.org/10.3389/fpubh.2026.1889919