The relationship between stress and academic burnout in college students: evidence from longitudinal data on indirect effects

The study investigates how stress contributes to academic burnout among college students and explores indirect mechanisms involving social support and self-esteem. Academic burnout is a heterogeneous negative emotional state characterized by lack of enthusiasm, negative attitudes, and disengagement, linked to poorer academic performance and mental health problems. Prior research shows a positive correlation between academic stress and burnout, with the COVID-19 pandemic exacerbating stressors (school closures, online learning, economic hardship, social isolation). Prolonged burnout can have psychological, physiological, and social consequences. Stress affects cognition, emotion, and behavior, potentially initiating a cycle that increases burnout. Understanding indirect mechanisms can inform targeted interventions. The authors propose four hypotheses: (H1) Stress positively predicts academic burnout. (H2) Social support indirectly links stress to academic burnout (buffering theory). (H3) Self-esteem mediates the stress–burnout relationship (self-evaluation and social-cognitive perspectives). (H4) Social support and self-esteem form a chain of indirect effects between stress and academic burnout.

The literature highlights that social support—emotional, tangible, and informational—buffers the negative effects of stress (buffering theory; social support theory), and lower social support is associated with higher academic burnout. Stress can reduce engagement with others, diminishing perceived support. Self-esteem, reflecting evaluations of self-worth, relates positively to mental health and adaptation; stress and negative social comparisons can lower self-esteem. Lower self-esteem is linked to greater burnout, while higher self-esteem relates to better engagement and performance. Studies suggest social support can enhance self-esteem over time, implying a sequential pathway where stress undermines support and self-esteem, increasing burnout. The authors integrate these perspectives into a model proposing independent and chained mediations via social support and self-esteem.

Design and participants: A longitudinal design with three waves spaced 3 months apart. Cluster sampling from 30 institutions in Wuhu and Bozhou (Anhui Province) and Hangzhou (Zhejiang Province) yielded three participating universities: Anhui Normal University, Tourism College of Zhejiang, and Bozhou University. First- to third-year students were included; fourth-years were excluded due to internships. Wave 1 (May 16, 2023): stress measured; 437 collected. Wave 2 (Aug 19, 2023): social support and self-esteem measured; 442 collected. Wave 3 (Nov 21, 2023): academic burnout measured; 433 collected. Inclusion required completion of all waves; after excluding invalid responses, N=428 (male 25.5%, female 74.5%; rural 81.1%, urban 18.9%). Procedures followed the Declaration of Helsinki; ethics approval from the Tourism College of Zhejiang (ZT83682900); informed consent obtained. Measures: (1) College Student Stress Scale (Li & Mei, 2002): 30 items, 3 factors (academic distress, personal distress, negative life events), 4-point Likert (0–3); higher scores indicate greater stress; α=0.96. (2) Academic Burnout Scale (Wu et al., 2010): 16 items, but this study used two dimensions—emotional exhaustion and academic alienation—due to heterogeneity in low achievement; 5-point Likert (1–5); higher scores indicate greater burnout; α=0.89 (two-dimension version). (3) Social Support Scale (Xiao): 10 items across objective, subjective, and utilization dimensions; mixed scoring with some counted sources; α=0.97. (4) Rosenberg Self-Esteem Scale (10 items; 4-point Likert; total 10–40; higher scores=greater self-esteem); α=0.78. Analytic strategy: SPSS 25.0 for descriptive statistics, t-tests, ANOVAs, and Pearson correlations. Common method bias assessed with Harman’s single-factor test: 11 factors with eigenvalues>1 explained 67.13% variance; first factor 32.08% (<40%), indicating limited bias. Confirmatory factor analyses and structural equation modeling conducted in Mplus 7.0. Given many items, item packaging was used. Control variables included gender, age, and household registration. Model fit criteria: RMSEA<0.10, SRMR<0.10, TLI>0.90, CFI>0.90. SEM proceeded in two steps: (1) direct effect of stress on burnout; (2) mediation model with social support and self-esteem (independent and chain mediations). Indirect effects tested via bootstrap (1,000 resamples; 95% CIs).

- Group differences: By gender, stress differed significantly (t = −2.13, p < 0.05), but academic burnout, social support, and self-esteem did not (all p > 0.05). By household registration (urban vs. rural), no significant differences across variables (all p > 0.05). By grade, stress differed (F = 3.99, p < 0.05), but burnout, social support, and self-esteem did not (all p > 0.05). - Correlations (N=428): Stress correlated positively with academic burnout (r = 0.55, p < 0.01) and negatively with social support (r = −0.26, p < 0.01) and self-esteem (r = −0.45, p < 0.01). Social support correlated negatively with burnout (r = −0.30, p < 0.01) and positively with self-esteem (r = 0.40, p < 0.01). Self-esteem correlated negatively with burnout (r = −0.49, p < 0.01). - SEM direct-effect model: Stress positively predicted academic burnout (β = 0.50, p < 0.001); excellent fit: RMSEA = 0.00, SRMR = 0.00, TLI = 1.00, CFI = 1.00. - SEM mediation model fit: RMSEA = 0.06, SRMR = 0.04, TLI = 0.94, CFI = 0.95. Paths: Stress → Academic Burnout (β = 0.43, p < 0.001); Stress → Social Support (β = −0.35, p < 0.001); Stress → Self-Esteem (β = −0.31, p < 0.001); Social Support → Self-Esteem (β = 0.42, p < 0.001); Social Support → Academic Burnout (β = −0.16, p < 0.05); Self-Esteem → Academic Burnout (β = 0.28, p < 0.001). - Indirect effects (bootstrap, 1,000 resamples; 95% CI): Total indirect effect = 0.06. Specific paths: Stress → Social Support → Academic Burnout = 0.02 (CI: 0.00–0.05); Stress → Self-Esteem → Academic Burnout = 0.03 (CI: 0.01–0.05); Stress → Social Support → Self-Esteem → Academic Burnout = 0.01 (CI: 0.00–0.03). All CIs excluded zero, indicating significance. Findings support H1–H4.

Findings confirm that higher stress predicts greater academic burnout, aligning with stress-adaptation models where stress disrupts cognitive, emotional, and behavioral functioning, thereby undermining learning and engagement. Consistent with buffering and social support theories, social support both reduces burnout directly and enhances self-esteem, which in turn relates to burnout. The chain mediation indicates that stress undermines perceived social support, which lowers self-esteem, ultimately increasing burnout—a Dual-Buffering pathway involving external (support) and internal (self-esteem) resources. An unexpected SEM result showed self-esteem positively predicting burnout, which the authors suggest may be due to unstable self-esteem, potentially impairing adaptability under stress and contributing to burnout. Contextualized within Chinese educational pressures, the results underscore the need to reduce academic stress, increase supportive environments (family, peers, teachers), limit behaviors that erode social connection (e.g., excessive gaming), and foster stable self-esteem to mitigate burnout.

Stress directly and indirectly predicts academic burnout among college students. Indirect pathways operate via external psychological buffering resources (social support) and internal resources (self-esteem), forming a Dual-Buffering Pathway Model of Academic Burnout. Practical implications in the Chinese context include reducing excessive academic pressure, broadening evaluation criteria beyond test scores, enhancing social support networks, and cultivating healthy, stable self-esteem through affirmation and positive feedback. Future research should expand samples across regions and institutions, assess longer observation windows or different intervals to capture psychological change, and examine stability versus level of self-esteem as potential moderators.

- Sample size (N=428) from three universities may limit generalizability. - The 3-month interval between waves, chosen based on precedent, may not optimally capture changes in psychological states. - Use of self-report measures, despite checks for common method bias, may introduce shared method variance.