Exercise confidence fuels involvement — and involvement shapes how fit students feel: evidence from a three-wave study of Chinese undergraduates

Table of Contents

1. Key Highlights
2. Introduction
3. How the study measured confidence, involvement and perceived fitness
4. Why separating within-person dynamics from between-person differences matters
5. What the data showed — main longitudinal findings and effect sizes
6. Why extracurricular exercise involvement matters (beyond attendance)
7. Practical implications for campus programs and health promotion
8. Measurement and methodological lessons for researchers
9. Limitations and directions for future research
10. A practical checklist for campus practitioners
11. Looking ahead: integrating evidence into campus policy
12. FAQ

Key Highlights

• Within individuals, higher-than-usual exercise self-efficacy predicted greater subsequent extracurricular exercise involvement; greater involvement then predicted higher perceived physical fitness.
• A small, tentative three-wave indirect effect linked initial self-efficacy to later perceived fitness through intervening involvement; perceived fitness did not reliably predict later confidence or involvement.

Introduction

University life often brings new freedoms, denser workloads, and disrupted routines. For many students this means exercise drops off or becomes irregular. Understanding not only whether students exercise but how psychological factors shape their engagement and self-evaluations matters for designing interventions that sustain activity and improve health.

A recent three-wave longitudinal study of 1,673 undergraduates across eight Sichuan universities used a random-intercept cross-lagged panel model (RI-CLPM) to separate stable individual differences from within-person fluctuations in three constructs: exercise self-efficacy (ESE), extracurricular exercise involvement (EEI), and perceived physical fitness (PPF). The analysis tracked students over roughly four months, with assessments spaced about eight weeks apart, and addressed whether temporary increases in a student’s confidence lead to higher-quality exercise involvement and whether that involvement then shifts how fit the student perceives themselves to be.

The results sharpen understanding of the dynamic links among confidence, the experiential quality of participation, and self-evaluated fitness. They offer concrete implications for campus programs and for researchers seeking to evaluate causal pathways in exercise psychology.

How the study measured confidence, involvement and perceived fitness

The investigation combined established self-report measures with a longitudinal panel design tailored to capture within-person change.

• Sample and timing: 1,673 full-time undergraduates (aged ≥18) from eight Sichuan universities completed paper surveys at three waves: early March, late April/early May, and late June 2025. Each survey asked about experiences during the previous four weeks; waves were 56 days apart. Retention was high: 92.2% at T2 and 87.0% at T3.
• Exercise self-efficacy (ESE): Measured with a 10-item Physical Activity Self-Efficacy Scale for College Students (Jiang et al., 2018). Items assess confidence in maintaining exercise when facing barriers (e.g., bad weather, fatigue) on a 1–5 Likert scale. Cronbach’s alpha was ≈0.88.
• Extracurricular exercise involvement (EEI): Measured with Dong’s 20-item Physical Exercise Involvement Scale, which indexes vigor/persistence, absorption/satisfaction, value cognition, and participation autonomy. The scale captures motivational and experiential quality rather than objective frequency or duration. Reliability was excellent (α ≈0.92).
• Perceived physical fitness (PPF): Measured with the five-domain International Fitness Scale (IFIS): general fitness, cardiorespiratory fitness, muscular strength, speed/agility, flexibility. The domains were averaged when at least four were present. Reliability was acceptable (α ≈0.78).
• Analysis approach: The RI-CLPM decomposed each construct into a stable between-person component (random intercept) and time-specific within-person deviations, allowing cross-lagged paths to reflect whether a student’s upswing or downturn relative to their own norm predicted later ups or downs in another construct. Models were estimated with robust maximum likelihood and used full-information maximum likelihood to handle missing data.

The study blended careful measurement with a modeling strategy designed to align statistical inference with developmental questions about intraindividual change.

Why separating within-person dynamics from between-person differences matters

People differ stably in confidence, motivation, and perceived fitness. Cross-sectional or conventional cross-lagged analyses confound these stable differences with short-term fluctuations, making it unclear whether an observed association reflects: (a) people who are generally more confident also being more active, or (b) a temporary increase in confidence leading to a later increase in involvement within the same person.

The RI-CLPM isolates the latter. It answers questions such as: if a given student is more confident than usual at this occasion, are they more likely to report greater involvement at the next occasion? That intraindividual focus better maps onto intervention logic: programs aim to shift students’ states or processes, not just sort students by baseline traits.

Methodologically, the RI-CLPM accomplishes this by:

• Modeling a random intercept for each construct (the person’s mean across waves).
• Estimating autoregressive and cross-lagged paths among the within-person residuals (deviations from personal means).
• Allowing between-person correlations among the random intercepts to capture stable associations.

This distinction changed the substantive picture: sizable between-person correlations existed (students who, on average, had higher ESE also tended to have higher EEI and PPF), but the RI-CLPM revealed that within-person increases in confidence predicted later within-person increases in involvement—an effect that cross-sectional data alone could not confirm.

What the data showed — main longitudinal findings and effect sizes

The free-lag RI-CLPM (allowed lag effects to vary between intervals) provided excellent model fit and became the primary basis for interpretation. Key within-person results:

• ESE → EEI (adjacent intervals): Higher-than-usual ESE at one wave predicted higher-than-usual EEI at the next wave at both intervals:
  • T1→T2: standardized β = 0.100 (p = 0.023)
  • T2→T3: standardized β = 0.179 (p < 0.001) These are modest effects, but consistent across occasions.
• EEI → PPF (adjacent intervals): Higher-than-usual EEI predicted higher-than-usual PPF at the following wave:
  • T1→T2: standardized β = 0.101 (p = 0.029)
  • T2→T3: standardized β = 0.205 (p < 0.001)
• EEI → ESE (feedback): A significant reverse effect appeared in the second interval:
  • T2→T3: standardized β = 0.129 (p = 0.004) The T1→T2 EEI→ESE path was not significant, and between-interval differences were not significant after multiplicity adjustment; nonetheless, the second-interval feedback is theoretically meaningful.
• PPF → ESE / EEI: Perceived fitness did not significantly predict later ESE or EEI at either interval; coefficients were small and statistically compatible with near-zero effects.
• Autoregressive within-person stability: ESE, EEI, and PPF each showed modest within-person carryover across adjacent waves (βs ranged roughly 0.15–0.28), consistent with person-centered modeling where trait variance is separated out.
• Three-wave indirect effect (ESE_T1 → EEI_T2 → PPF_T3): The unstandardized product b = 0.02250 (standardized = 0.0204). The Wald test was marginal (p ≈ 0.055), but a bias-corrected bootstrap (2,000 draws) yielded a 95% CI that excluded zero ([0.004, 0.051]). The effect is small and inferentially sensitive, offering tentative support for the hypothesized mediation where within-person gains in self-efficacy lead to greater involvement that then yields higher perceived fitness.
• Between-person associations: Random-intercept correlations were substantial (ESE–EEI r = 0.600; ESE–PPF r = 0.618; EEI–PPF r = 0.558), indicating strong, stable individual differences alongside the within-person dynamics.

Sensitivity analyses adjusting for baseline demographic covariates, attrition patterns, and complete-case models preserved the direction and significance patterns of the main within-person effects, with only small changes in magnitude.

Why extracurricular exercise involvement matters (beyond attendance)

EEI was conceptualized and measured as the motivational and experiential quality of voluntary exercise—persistence, absorption, perceived value, and autonomy—rather than frequency or minutes exercised. That distinction is central to the study’s main message.

How EEI functions:

• EEI translates confidence into sustained action. Higher ESE makes students more willing to begin or persist in voluntary exercise; when participation feels absorbing and self-endorsed, students are more likely to sustain behavior in ways that produce meaningful bodily feedback.
• EEI shapes self-evaluative change. When participation is persistent and satisfying, students notice functional gains (endurance, strength, flexibility) and update their perceived fitness. The EEI→PPF paths were consistently positive across waves and larger in the later interval, suggesting accumulation matters.

Concrete campus examples that illustrate EEI’s role:

• A recreational running group that emphasizes graded goals, enjoyment, and peer support tends to produce higher EEI than a compulsory lap program that measures only attendance. Members who feel absorbed and find personal value in the runs are more likely to keep going and to report improvements in perceived stamina.
• Small-group strength-training workshops that let beginners choose manageable loads, track visible progress, and celebrate milestones produce mastery feedback that both raises involvement quality and enhances perceived strength.
• Low-barrier activities (walking clubs, dance classes, non-competitive sports) that allow choice and autonomy appeal to students who might avoid gym culture; these formats can increase EEI by aligning activity with personal identity and values.

EEI captures the “why and how” of exercise. Programs that focus solely on getting students through the door risk modest behavioral change unless the experiential quality supports persistence and absorption.

Practical implications for campus programs and health promotion

The study identifies two promising intervention targets: exercise self-efficacy and the quality of extracurricular exercise involvement. Practical program elements follow directly.

1. Strengthen exercise self-efficacy with structured, evidence-based tactics
   • Scaffolded mastery experiences: design progressive skill-building modules (e.g., four-week incremental strength or endurance programs) where early successes are likely and clearly attributable to students’ efforts.
   • Barrier-focused planning: teach simple coping plans for common obstacles—time constraints during exams, poor weather, minor fatigue—so students practice overcoming those barriers in advance.
   • Peer modeling and mentoring: use near-peer demonstrators or student ambassadors who share realistic starting points and show attainable progress.
   • Self-monitoring with guided interpretation: provide tracking tools (logs, apps, wearables) paired with prompts that interpret small gains as meaningful.
2. Design activities to maximize EEI (quality of involvement)
   • Emphasize autonomy: offer activity choice rather than a one-size-fits-all roster. Allow students to select times, formats, and intensity levels.
   • Build value and personal relevance: connect activities to student goals—stress relief, social connection, improved stamina for active commuting—rather than only health metrics.
   • Promote absorption and satisfaction: structure sessions to be engaging (varied drills, music, goal-oriented tasks) and to include short reflective moments for noticing progress.
   • Foster persistence and accountability: use small teams, buddy systems, or low-stakes attendance commitments to support routine without coercion.
3. Pair fitness feedback with actionable guidance
   • Combine subjective and objective feedback: if offering fitness testing (e.g., timed runs, strength tests), follow with short plans that translate results into next steps matched to the student’s current level and preferences.
   • Avoid raw comparisons: frame results around individual improvement trajectories and effort-based attributions to avoid discouraging social comparisons.
4. Evaluate interventions experimentally
   • Randomized designs should test whether programs that intentionally boost ESE and EEI produce both sustained exercise behavior and improvements in perceived and objective fitness.
   • Use repeated measures and within-person analytic strategies (RI-CLPM or latent change models) to test whether interventions shift intraindividual dynamics in the predicted direction.

Real-world pilot model: a twelve-week “Student Fit Lab” program could randomly assign consenting students to (a) mastery-supported group (goal-setting, graded tasks, mentor feedback), (b) standard exercise referral (attendance-based classes), or (c) wait-list. Pre/post and monthly within-person measures of ESE, EEI, perceived fitness, and objective markers (e.g., 3-minute step test, grip strength) would test whether quality-focused programming outperforms attendance-focused delivery.

Measurement and methodological lessons for researchers

This study offers methodological takeaways relevant to exercise psychology and behavioral research more broadly.

• Align model to question: when interest centers on intraindividual processes (does a change in X precede a change in Y within person?), use models that separate within- and between-person variance. The RI-CLPM does this for panel data with three or more waves.
• Time lag matters: psychological states and behavioral responses operate on differing timescales. ESE may fluctuate weekly, EEI may accumulate over several weeks, and objective fitness changes may require longer. Researchers should choose lags based on theory and, where feasible, test multiple time scales (e.g., weekly diaries, intensive sampling, or longer panels).
• Measurement precision matters: the present analysis used manifest scale means. Multiple-indicator RI-CLPMs (latent variable RI-CLPM) can further guard against measurement error, but they require strong longitudinal measurement invariance and larger samples.
• Combine subjective and objective outcomes: perceived fitness is meaningful, but it is not a substitute for objective performance or physiological measures. Triangulating self-report and objective data clarifies whether perceived changes reflect real physiological adaptation or changes in interpretation and self-concept.
• Address attrition: even with high retention, selective dropout affects inference. Here, students with higher baseline ESE and females were slightly less likely to complete all waves. Attrition-informed sensitivity models help probe robustness, but experimental designs and strategies to minimize dropout are preferable.
• Report effect sizes and uncertainty: small standardized mediation estimates are common in behavioral science. Resampling methods (bootstrap) for indirect effects are advisable; interpret effect sizes in practical context rather than focusing exclusively on p-values.

Limitations and directions for future research

The study’s strengths (large multisite sample, RI-CLPM) coexist with limitations that frame next steps.

• Observational design: despite temporal ordering, the study cannot establish causality. Time-varying confounders (exam periods, injuries, access to facilities, weather) might influence within-person fluctuations. Quasi-experimental and randomized interventions are needed.
• Three waves and the chosen eight-week spacing: three assessment points limit modeling of complex nonlinear or cumulative processes. The eight-week lag suited semester-level changes but may miss faster or slower dynamics. Future work should test alternative lags (weekly diaries, monthly panels, or longer-term follow-ups).
• Reliance on self-report: all focal constructs were subjectively reported. Perceived fitness is inherently subjective, but combining it with objective fitness tests and accelerometer data on activity volume would strengthen inference.
• Measurement invariance and manifest scoring: EEI and PPF achieved full scalar invariance under approximate fit criteria; ESE required partial scalar invariance. A multiple-indicator latent RI-CLPM could parse measurement error, though it demands larger samples and robust invariance.
• Regional sample: participants were from eight universities in Sichuan Province, China. Cultural, institutional, and facility differences may limit generalizability to other regions or countries. Replication across diverse contexts is necessary.
• Small indirect effect: the three-wave mediation was small and inference depended on bootstrap versus normal-theory tests. Pre-registration, larger samples, or more intensive repeated measures can clarify replicability.

Suggested research priorities:

• RCTs that manipulate ESE and/or EEI with repeated within-person measurement to test causality.
• Intensive longitudinal studies (ecological momentary assessment or weekly diaries) to map short-term reciprocal dynamics and habit formation.
• Mixed-methods work exploring how students interpret fitness feedback and what attributions (effort vs luck vs comparison) moderate PPF→behavior links.
• Cross-cultural replication and subgroup analyses (gender, baseline fitness, major disciplines) to identify boundary conditions.

A practical checklist for campus practitioners

• Design activities that prioritize autonomy and personal relevance (offer choices, flexible schedules).
• Build progressive, mastery-based modules with clear, achievable milestones and visible markers of improvement.
• Train mentors or peer leaders to provide effort-focused feedback and model small, realistic progress.
• Pair subjective feedback (IFIS-style ratings) with objective or performance indicators and immediately translate results into short-term action plans.
• Monitor involvement quality (not just attendance) through simple EEI-style items (e.g., “I feel absorbed during my sessions,” “I stick to the plan even when busy”).
• Pilot programs with built-in repeated measurement and randomized assignment where feasible; use within-person analytic strategies to detect change processes.
• Evaluate both perceived and objective fitness outcomes and collect information on barriers (time, facilities, injury) to adapt programs responsively.

Looking ahead: integrating evidence into campus policy

Universities aiming to boost student activity should invest in interventions that do more than get bodies into gyms. Programs that build confidence, make participation meaningful, and generate repeated mastery experiences are likeliest to change how students feel about their fitness—and to sustain behavior. Pairing practical supports (flexible scheduling, low-cost access, peer support) with deliberate psychological targets (self-efficacy, autonomy, absorption) aligns program design with the intraindividual dynamics that predict durable change.

The modest size of the indirect effect reported here does not undermine its practical value. Small within-person shifts, accumulated across many students and sustained over time, can yield meaningful public health benefits. The next step is routine experimentation: test which combinations of mastery tasks, autonomy-support, feedback, and social structures most effectively raise EEI and whether those increases translate into objective fitness and long-term maintenance.

FAQ

Q: What is the difference between extracurricular exercise involvement (EEI) and simply exercising more? A: EEI emphasizes quality—persistence, absorption, satisfaction, perceived value, and autonomy—rather than only frequency, duration, or intensity. Two students attending the same number of sessions can differ strongly on EEI: one may feel absorbed and autonomous, the other may attend out of obligation. The study shows EEI predicts later perceived fitness above and beyond stable individual differences.

Q: Does feeling more fit make students more likely to exercise later? A: In this study perceived fitness (PPF) did not significantly predict later exercise self-efficacy or involvement across eight-week intervals. The pattern supports an interpretation where involvement tends to shape perceived fitness—students notice bodily changes after repeated, meaningful engagement—rather than the reverse, at least on the tested timescale.

Q: Can interventions that boost self-efficacy lead to better fitness? A: The within-person pathway is plausible: temporary increases in self-efficacy predicted later increases in involvement, and involvement predicted later perceived fitness. The hypothesized mediation (self-efficacy → involvement → perceived fitness) received tentative support. Experimental trials that manipulate self-efficacy and measure downstream involvement and fitness are needed to confirm causality.

Q: How should campus programs apply these findings? A: Prioritize activities that combine confidence-building with enjoyable, autonomous experiences. Offer graded mastery tasks, options for personal choice, peer modeling, and feedback that highlights progress attributable to students’ effort. Track involvement quality as well as attendance. Pair subjective feedback with actionable guidance.

Q: Are the effects large enough to matter for policy? A: The within-person standardized effects are modest, and the indirect pathway is small. Small effects can nevertheless accumulate when programs reach many students or when changes are sustained. Policymakers should focus on interventions with demonstrated ability to increase involvement quality and on scaling approaches that maintain fidelity to EEI-enhancing principles.

Q: Would objective measures (e.g., accelerometers, fitness tests) likely show the same patterns? A: Perceived fitness captures self-evaluative changes and may respond faster or differently than objective markers. The study’s reliance on self-report limits conclusions about physiological adaptation. Combining objective fitness measures and device-based activity monitoring with subjective measures is recommended for future research to clarify whether increased EEI translates into measurable improvements in performance or health.

Q: Does the eight-week interval influence the findings? A: Yes. Time lags shape what processes become visible. Eight-week intervals aimed to capture short-term accumulation and self-evaluative change across a semester. Shorter or longer intervals might reveal different dynamics—weekly fluctuations in self-efficacy or longer-term physiological adaptation—so future work should test multiple timescales.

Q: Who benefits most from programs based on these findings? A: While the study did not test differential effects across subgroups, programs that emphasize approachable mastery, autonomy, and social support likely help students who lack confidence or who have irregular exercise habits. Future research should test whether effects differ by gender, baseline fitness, major, or socioeconomic background.

Q: What are the next steps for researchers? A: Run randomized or quasi-experimental interventions that manipulate self-efficacy and involvement components, incorporate objective fitness and activity measures, use multiple time scales (including intensive longitudinal designs), and estimate latent versions of RI-CLPM when measurement invariance permits. Pre-registration and replication will strengthen causal claims.

Q: Where can campus health teams start tomorrow? A: Implement small pilots that focus on graded mastery and choice—e.g., a six-week, coach-led, small-group initiative where students set personal goals, receive weekly brief feedback, and log short reflective entries about what felt meaningful. Measure ESE and EEI at baseline and weekly, and evaluate changes in perceived fitness and attendance trends. Use pilot findings to refine and scale programs.