Exercise Was Linked to Higher Subjective Wellbeing Across 30,247 Chinese Adults

Adults in China who exercised at least several times a month reported higher subjective wellbeing than those who rarely moved, across five waves of a national survey. The size of that subjective wellbeing advantage shifted with age, differed between men and women, and narrowed across birth cohorts.

Subjective wellbeing and leisure-time exercise were analyzed together in 30,247 Chinese adults in a study published on November 2, 2026 in Frontiers in Public Health.

The analysis pooled repeated cross-sectional data from five waves of the China General Social Survey, conducted in 2012, 2015, 2018, 2021 and 2023. Participants ranged from 18 to 97 years of age. The work was carried out by Bingyang Zhai of the College of Physical Education at Kunsan National University in Gunsan, Republic of Korea, and Mingyuan Li of the Department of Military and Physical Education at the Jiangxi Institute of Fashion Technology in Nanchang, China.

Exercise was captured with one survey item asking about the frequency of leisure-time physical exercise during the previous year. Respondents exercising several times a month or more were coded as exercisers. Those reporting several times a year or less, including never, were coded as non-exercisers. About 59.8 percent of the pooled sample fell into the exerciser group.

Subjective wellbeing rested on a single question: overall, how happy do you feel about your life. Answers ran on a five-point scale from very unhappy to very happy. In the first adjusted model, exercisers scored 0.1810 points higher on that scale, with a standard error of 0.0104. The authors describe the finding plainly: exercisers consistently exhibited higher predicted subjective wellbeing than non-exercisers across the age range.

What the Survey Actually Asked

Both of the study’s central variables came from one question each. That matters for how far the result travels. The exercise item recorded frequency only, not duration, intensity or type, so a weekly badminton game and a daily training session can land in the same category. Any subjective wellbeing difference the model finds is therefore attached to how often people moved, nothing finer.

The happiness item is similarly compact. A five-point scale anchored on a single word captures a general mood judgment rather than a profile of life satisfaction, purpose or psychological distress. Readers should treat subjective wellbeing here as self-reported happiness in one number, and the authors list both measurement choices among the study’s limits.

Why the Subjective Wellbeing Gap Shrank in Middle Age

The age pattern is the most distinctive part of the analysis. Rather than a flat benefit, the model found a curve. The interaction between exercise and age was negative at minus 0.00300, while the interaction with age squared was positive at 0.000075. Both reached statistical significance below the 0.001 level.

In plain terms, the subjective wellbeing advantage attached to exercise was pronounced in young adulthood, attenuated through middle adulthood, and strengthened again in later life. The authors do not name a turning-point age, and no single age should be read into the shape.

Middle age is also the stretch of life where work hours, caregiving and income pressures weigh most heavily on reported happiness. A narrower subjective wellbeing gap there does not mean exercise stopped mattering; it may mean other forces crowded into the same single-item measure.

Men Showed a Larger Association Than Women

Gender changed the picture as well. With men coded as one, the exercise-by-gender interaction came to 0.0420 with a standard error of 0.0155, significant below the 0.01 level. At the mean-centered reference age, the subjective wellbeing advantage was significantly stronger among men than among women.

A three-way term combining exercise, gender and age added 0.00130 with a standard error of 0.00040, meaning the male and female subjective wellbeing patterns also diverged as people aged. The paper does not test why, and the authors say the proposed mechanisms behind the heterogeneity were not directly examined.

The Cohort Pattern Points to a Moving Baseline

Because the survey repeats, the authors could separate birth cohorts from survey years. Cohort-level variation in the exercise slope was statistically significant, with a random-slope variance of 0.000147 in one model. Variation across survey periods was not significant, at 0.000036.

The subjective wellbeing gap between exercisers and non-exercisers was largest for the earliest cohort and narrowed across later ones. It reached a comparatively low level among those born during 1990 to 1999, then recovered modestly among those born in 2000 or later. The authors caution that this trajectory should be read as a descriptive pattern rather than a formally confirmed nonlinear relationship.

Where This Subjective Wellbeing Signal Sits Among Other Evidence

Trials that put people on an exercise program and track mood for weeks generally report improvements, but they recruit volunteers and run on small samples. Surveys like this one trade that control for scale, and the trade shows: 30,247 respondents give a stable national picture while saying nothing about what a program would do.

The two approaches answer different questions. A trial asks whether assigning exercise changes mood. This analysis asks whether people who already exercise report different subjective wellbeing, and finds that they do, by about 0.18 points. Neither result substitutes for the other, and the subjective wellbeing literature advances when both are read side by side.

Two Reported Test Statistics Do Not Match

One inconsistency deserves flagging. The results section reports the joint test of the gender terms twice, as a chi-squared value of 6.217 with a p-value of 0.045, and also as 10.611 with a p-value below 0.01. The accompanying table lists 10.611.

Both versions point the same direction, so the gender conclusion does not hinge on which is correct. Still, readers comparing this paper with later work should note the discrepancy rather than quote a single figure with confidence.

How Much a Repeated Cross-Section Can Carry

Each survey wave interviewed a different sample of people. That design describes a population at five points in time; it never follows the same person twice. The authors write that it does not capture within-person changes over time and therefore cannot establish causal relationships, and they add that the observed association may also be bidirectional.

Bidirectional matters here. People who feel better may find it easier to exercise, and people who exercise may feel better. A single snapshot cannot separate those two stories, and the authors say plainly that these observational findings do not establish that exercise causes higher happiness.

What Regular Exercisers Can Take From the Subjective Wellbeing Findings

A coefficient of 0.18 on a five-point scale is a population-level signal, not a personal forecast. It describes an average difference between two large groups, and it says nothing about how any individual will feel after a given workout. Treated that way, the study still carries a few usable points.

• Frequency was the only exercise variable measured, so the result supports a habit of moving several times a month or more rather than any particular program.
• The activities behind the data are ordinary leisure pursuits, not structured training, which is part of why the subjective wellbeing signal appears at all.
• The later-life widening is consistent with other work on activity and healthy aging, but this analysis cannot confirm a benefit for any age group on its own.
• Anyone exercising specifically to address low mood should treat that as a conversation for a qualified clinician, not a conclusion from survey data.

Readers following this evidence base may also want our coverage of a separate national dataset in how exercise participation tracked relative health in a Chinese survey, which uses a comparable self-report structure. A much larger sleep analysis appears in our report on regular exercise and restful sleep across 702,000 adults.

For the short-term mood question that survey designs cannot reach, see our write-up of ten minutes of slow running and its effect on mood and executive function. And for the later-life end of the curve, our summary of leisure-time exercise and dementia risk covers what longer-horizon reviews have found.