Notes on Williams, Dziurawiec, & Heritage (2017): ERI, Burnout, and Withdrawal Intentions in University Students

Notes on the study of effort–reward imbalance, burnout, and withdrawal intentions among university students

  • Study focus and gap

    • Addresses the atheoretical nature of much student-stress research by applying Siegrist’s (1996) effort–reward imbalance (ERI) model to a university context.

    • Research questions: Do discrepancies between efforts expended and rewards obtained relate to student distress outcomes (e.g., withdrawal intentions)? Is burnout a mediator between ERI and withdrawal intentions? Do resilience and year of study moderate these relationships?

    • Educational impact emphasized: a sizeable portion of students experience detrimental ERIs, which are linked to burnout and withdrawal intentions; implications for policy and university practices.

  • Key concepts and theory

    • ERI theory (Siegrist, 1996)

    • Core idea: self-perceived discrepancies between efforts and rewards lead to distress.

    • Mechanisms: low-status control, job insecurity, limited input in the workplace (parallels to university), and acceptance of ERIs as normal/strategic for future gains.

    • Overcommitment: an intrinsic factor that can amplify ERI effects.

    • In educational contexts, ERI may manifest as perceived under-recognition, unfair marking, or limited input/choice in the learning environment.

    • Burnout (Schaufeli et al., 2002; Maslach & Jackson, 1981)

    • Components: emotional exhaustion and cynicism; linked to distress and withdrawal behaviors.

    • In this study, burnout is treated as a potential mediator between ERI and withdrawal intentions.

    • Personal resources: resilience (Smith et al., 2008)

    • Concept: the ability to bounce back from stress; hypothesized to buffer ERI effects.

    • Prior work shows resilience relates to better coping and lower burnout, but its moderating role for ERI in university samples was untested prior to this study.

  • Hypotheses (summarized)

    • ERI will be positively associated with withdrawal intentions among university students.

    • Burnout will partially mediate the ERI–withdrawal intentions link (ERI → Burnout → Withdrawal).

    • The ERI–withdrawal path and the ERI–burnout path will be strongest among first-year students.

    • Resilience will buffer (moderate) the effects of ERI on burnout and withdrawal intentions.

    • Covariates (age, gender, SES, average grade) will influence ERI, burnout, and withdrawal intentions; relatively short-term life stressors will be controlled.

  • Method: design and sample

    • Design: quantitative, cross-sectional survey with a theoretical model; acknowledges limits on causal inference.

    • Recruitment and participants

    • Sample: N = 2,451 Australian university students after data cleaning (initial consenters N = 3,805; completed data N = 2,468; final sample N = 2,451).

    • Timeframe: recruitment over 16 weeks of a typical 17-week semester, spanning two semesters.

    • Inclusion: students 17+ years old, studying in Australia; nonprobability sampling via posters, in-class briefings, and Facebook recruitment across 40 universities and multiple disciplines.

    • Incentives: relaxation CD; research-participation credit or $50 gift card raffle entry.

    • Demographics and sample characteristics

    • Age: M = 23.65, SD = 7.54; range 17–67.

    • Gender: 79.2% female; 86.9% full-time; 87.8% internal; 92.6% domestic; 89.4% undergraduate.

    • Representation across fields and states; comparable to national student profiles.

    • Ethics: approved by Murdoch University (2015/026).

  • Measures (key scales and scoring)

    • Perceived stress: four-item version of the Perceived Stress Scale (Cohen & Williamson, 1988); α = .80; range 4–20.

    • ERI (School Version; Li, Shang, Wang, & Siegrist, 2010; minor wording adjustments for university): 16 items total

    • Effort: originally 5 items, plus 2 added items (e.g., “There has been too much competition amongst classmates,” “I find that I have too much to learn everyday”). After item deletion, Effort α = .71.

    • Reward: 10 items (6 reverse-coded); Reward α = .75.

    • Scoring: 7–35 for Effort; 10–50 for Reward; higher scores indicate higher levels of the construct.

    • ERI score calculation: ERI=racEffortRewardimes1.43ERI = rac{Effort}{Reward} imes 1.43 to adjust for different item counts (Siegrist, 1996; Siegrist et al., 2004).

    • Cut-offs: continuous ERI used in analyses; dichotomous cut-off for “detrimental ERI” defined as ERI > 1.00 (Siegrist et al., 2004). An alternative ROC-based cut-off of 0.720.72 was also reported.

    • Burnout: Maslach Burnout Inventory – Student Survey (MBI-SS; Schaufeli et al., 2002)

    • Subscales: Exhaustion (5 items) and Cynicism (4 items). α = .92.

    • Scale: 7-point Likert (1 = never to 7 = always); higher scores indicate greater burnout. A single burnout score is formed by summing items (range 9–63) because the subscales loaded onto a single factor (61.11% variance explained).

    • Withdrawal intentions: novel 9–10 item scale (monitored via PCA) developed for this study

    • Composition: 10 items (after item deletion due to poor item-total correlation, final 9 items used for analysis; α = .82).

    • Scale: 7-point Likert (1 = strongly disagree to 7 = strongly agree); higher scores indicate stronger withdrawal intentions; contains five reverse-scored items.

    • Psychometric support: Kaiser–Meyer–Olkin = .85; Bartlett’s test significant; one-factor solution explaining 43.47% of variance; loadings > .40, with several items > .60.

    • Resilience: Brief Resilience Scale (Smith et al., 2008)

    • 6 items; α = .89.

    • Five-point Likert (1–5); higher scores indicate greater resilience; reverse-worded items included.

    • Covariates and additional controls

    • Age, gender (0 = Male, 1 = Female), SES (maternal education; coded 1–6), year of study (treated as a moderator, dummy-coded 0 = first year, 1 = subsequent years), average grade, and perceived stress included as covariates in analyses.

  • Analytic approach

    • Primary analysis: moderated mediation using PROCESS (Hayes, 2013) with bootstrapping (10,000 resamples) to test indirect effects and their confidence intervals.

    • Model specification: ERI (X) predicting Withdrawal Intentions (Y) via Burnout (M), with Year of Study (V) and Resilience (W) as moderators; covariates included as described.

    • Data handling: listwise deletion for missing scales; Winsorization of univariate outliers; HC3 standard errors to address heteroscedasticity concerns.

    • Subgroup analyses: comparisons across disciplines and institutions; multilevel analyses deemed unnecessary due to very small intraclass correlations; gender-specific analyses conducted and showed consistent patterns.

    • Additional considerations: cross-sectional design acknowledged as limiting causal claims; alternative model testing included a reverse mediation check and moderation analyses using an interaction term ERI × Year (and ERI × Resilience) as robustness checks.

  • Descriptive statistics and correlations (key numbers)

    • Descriptives (after deletion/winsorization)

    • ERI: M = 0.94, SD = 0.30; range 0.20–5; αs reported above.

    • Withdrawal intentions: M = 24.95, SD = 9.49; range 9–61; α = .82.

    • Burnout: M = 31.35, SD = 11.40; range 9–63; α = .92.

    • Resilience: M = 3.06, SD = 0.84; range 1–5; α = .89.

    • Perceived stress: M = 11.83, SD = 3.10; range 4–20; α = .80.

    • Group-level prevalence

    • 37.5% of the sample exceeded the original ERI cut-off of 1.00, indicating detrimental ERIs.

    • Correlations (sample-wide; all p < .001 unless noted)

    • ERI with withdrawal intentions: r = .37 (moderate positive) [95% CI around r = [.33, .41]]

    • ERI with burnout: r = .59 (strong positive) [95% CI around r = [.56, .62]]

    • Burnout with withdrawal intentions: r = .59 (strong positive) [95% CI around r = [.57, .62]]

    • Resilience with burnout: r = −.38 (moderate negative) [95% CI around r = [−.42, −.34]]

    • ERI with resilience: r = −.33 (moderate negative) [95% CI around r = [−.37, −.29]]

    • Covariate correlations (briefly): age, perceived stress, and average grade significantly correlated with ERI, burnout, and withdrawal intentions (p < .001); SES not significantly correlated with burnout, ERI, or withdrawal intentions and therefore not included in main analyses; gender correlated with withdrawal intentions.

  • Results: moderated mediation and key findings

    • Overall model test

    • In the full moderated-mediation model, the direct effect of ERI on withdrawal intentions was not significant after controlling for burnout: p = .741.

    • There was a significant interaction of ERI × Year of Study on the direct path to withdrawal (c5 = −4.45, p = .010), indicating year of study moderated the direct ERI→withdrawal path. This prompted further probing of direct effects by year group.

    • Indirect effect (through burnout) of ERI on withdrawal intentions was significant: ab = 7.40; 95% CI [6.54, 8.37]. This indicates burnout fully mediates the ERI–withdrawal link in the presence of year and resilience as moderators, with resilience not significantly altering the indirect path (a3 = −0.41, p = .525; b2 = 0.08, p = .106).

    • Model statistics: R² for withdrawal intentions ≈ 0.385 (38.5% of variance explained); R² for burnout ≈ 0.445 (44.5% of burnout variance explained) for the full model. This demonstrates substantial explanatory power.

    • Parsimonious mediation model (no direct effects)

    • Given the non-significant direct effects by year and the non-significant moderation of the indirect path, a parsimonious model was reported where burnout remains a robust mediator between ERI and withdrawal intentions across years.

    • In this reduced model: ERI is positively related to burnout (a = 16.65); burnout is positively related to withdrawal intentions (b = 0.44).

    • Indirect effect remains large and significant: ab = 7.40; 95% CI [6.54, 8.37].

    • The completely standardized indirect effect accounted for 13.71% of the variance in withdrawal intentions and 91.67% of the total effect (CI for standardized effects provided in the paper).

    • Year-of-study moderation details

    • The moderation effect was significant on the direct ERI→withdrawal path (c5 = −4.45, p = .010) but did not alter the indirect effect via burnout.

    • Conditional direct effects (Table 5):

      • First-year students: Direct effect c’ = 2.15, SE = 1.22, p = .079; 95% CI = [−0.24, 4.54].

      • Subsequent-year students: Direct effect c’ = −0.46, SE = 0.70, p = .517; 95% CI = [−1.84, 0.92].

    • These results show no statistically significant direct effect of ERI on withdrawal intentions for either year, once burnout is accounted for.

    • Moderation by resilience

    • Resilience did not significantly buffer (moderate) either the direct or indirect effects in the full model (a3 and b2 were not statistically significant).

    • Alternative models and sensitivity checks

    • An ERI formulation using an interaction term (Effort × Reward) yielded non-significant moderation by rewards on the ERI–burnout and ERI–withdrawal paths (p = .068 and p = .487, respectively).

    • A reverse mediation test (burnout as an outcome, ERI as the mediator) yielded ab = .02 with 95% CI [−.01, .04], not significant, supporting the directional ERI → Burnout → Withdrawal interpretation over the reverse path.

    • Subgroup analyses

    • Across 13 of 14 disciplinary groups and 14 of 16 institutions, the pattern of a significant indirect effect via burnout with no direct ERI effect persisted; no major deviations observed.

    • Analyses controlled for age, gender, average grade, year of study, resilience, and perceived stress and yielded consistent results.

  • How to interpret the core findings

    • Direct ERI–withdrawal link not significant when burnout is included as mediator; ERI appears to influence withdrawal intentions primarily through burnout.

    • Burnout plays a central mediating role in translating ERI experiences into withdrawal considerations for both first-year and later-year students.

    • The lack of a significant moderation by resilience suggests that short-term coping resources may not be sufficient to buffer chronic, education-specific ERIs; longer-term resources or different protective factors may be needed.

    • Year of study moderates the direct ERI–withdrawal relationship (a small effect) but does not alter the core indirect pathway through burnout.

    • Investment theory (Rusbult) helps explain why first-year students may still show a direct effect pattern: they may already be invested in their studies, which reduces the likelihood that ERI alone drives withdrawal decisions in the direct pathway, though burnout remains a meaningful mediator.

  • Theoretical and practical implications

    • The ERI model generalizes to tertiary education settings and to academic outcomes like withdrawal intentions, broadening the applicability of work-stress theory beyond occupational contexts.

    • Burnout as a mediator suggests that interventions targeting burnout could mitigate ERI-related withdrawal risks even when ERI itself is difficult to modify directly.

    • Specific subfacets of effort/reward may yield different moderation effects; the study’s ratio approach was robust but may overlook facet-level dynamics (e.g., feedback quality, marking transparency, esteem, future prospects).

    • Practical implications include improving marking practices, feedback transparency, and wellbeing programs to bolster coping resources, particularly for first-year students.

    • Policy implications span three levels (combining individual, university, and government action):

    • Individual: skills for coping with ERIs and cognitive strategies.

    • University: reduce ambiguous feedback, improve fairness in marking, and increase recognition from staff; explore learning communities to reduce competitive pressures.

    • Government: consider ERI as a component of university rating systems and potentially offset tuition costs to reduce the effort burden.

  • Strengths and limitations

    • Strengths

    • Large, diverse sample across 40 universities, multiple disciplines, and two semesters; enhanced generalizability within the Australian context.

    • Use of a theoretically anchored ERI model in education, with a validated school-based ERI measure adapted for university settings.

    • Robust statistical approach: bootstrapping, HC3 standard errors, and parallel testing of direct/indirect effects.

    • Limitations

    • Cross-sectional design limits causal claims; longitudinal studies needed to confirm directionality and to link ERI and burnout to actual attrition.

    • Convenience sampling and Facebook recruitment may introduce self-selection or nonresponse biases, though the sample appears broadly representative.

    • ERI was analyzed using a ratio form (and supplementary moderation analyses with a multiplicative form were reported); facet-level analyses of specific subcomponents could uncover more nuanced effects.

  • Summary conclusions

    • More than one-third of university students experience detrimental ERIs; these imbalances relate to higher burnout and withdrawal intentions.

    • Burnout fully mediates the ERI–withdrawal relationship across first-year and subsequent-year students; resilience did not buffer these effects in this study.

    • The findings support applying work-stress theories to higher education and highlight burnout as a key leverage point for interventions aimed at reducing student distress and attrition.

  • Key numerical references for quick review (all values in extLaTeXext{LaTeX})

    • Sample and data

    • N = 2,451 (final sample)

    • Age: M = 23.65, SD = 7.54

    • Completion rate: 64.86% of consenting participants provided complete data

    • ERI and cutoffs

    • ERI mean: ERIˉ=0.94,<br>extSD=0.30\bar{ERI} = 0.94,<br>ext{SD} = 0.30

    • Proportion with detrimental ERI: 37.5%37.5\%

    • ERI ratio calculation: ERI=racEffortRewardimes1.43ERI = rac{Effort}{Reward} imes 1.43

    • Cut-offs: ERI > 1.00 (standard); alternative: ERI > 0.72 (ROC-based)

    • Reliability

    • Effort: 0˘3b1Effort=0.71\u03b1_{Effort} = 0.71

    • Reward:  = 0.75

    • Burnout:  = 0.92

    • Withdrawal:  = 0.82

    • Resilience:  = 0.89

    • Perceived stress:  = 0.80

    • Correlations (selected)

    • ERI with Withdrawal: r=0.37r = 0.37

    • ERI with Burnout: r=0.59r = 0.59

    • Burnout with Withdrawal: r=0.59r = 0.59

    • Resilience with Burnout: r=−0.38r = -0.38

    • ERI with Resilience: r=−0.33r = -0.33

    • Mediation results (full model)

    • Indirect effect: ab=7.40,<br>95ext−CI=[6.54,8.37]ab = 7.40,<br>95 ext{-CI}=[6.54, 8.37]

    • Direct ERI → Withdrawal (controlling for Burnout): not significant,
      p=ext0.741p= ext{0.741}

    • Moderation: Year of Study on direct path: c5=−4.45,<br>p=.010c5 = -4.45,<br>p = .010

    • Parsimonious model results

    • Burnout predicting Withdrawal: b=0.44b = 0.44

    • ERI predicting Burnout: a=16.65a = 16.65

    • Indirect effect: ab=7.40,<br>95ext−CI=[6.54,8.37]ab = 7.40,<br>95 ext{-CI}=[6.54, 8.37]

    • Variance explained

    • Full moderated-mediation: R2<em>withdrawal ≈ 0.385R^2<em>{withdrawal} \,\approx \,0.385; R2</em>burnout ≈ 0.445R^2</em>{burnout} \,\approx \,0.445

    • Completely standardized indirect effect:
      extindirecteffectstd=0.34 extSDchangeinwithdrawalper1SDERIext{indirect effect}_{std} = 0.34 \, ext{SD change in withdrawal per 1 SD ERI}

    • The paper notes: a 0.29 change in ERI (one SD) corresponds to a 0.34 SD change in withdrawal intentions.

  • Figures referenced (described verbally)

    • Figure 1: ERI model for educational contexts (adapted from Li et al., Siegrist, 2010/2014).

    • Figure 2: Hypothesized model with ERI (X) predicting Withdrawal Intentions (Y) via Burnout (Mi), moderated by Year of Study (V) and Resilience (W).

    • Figure 3: Full moderated-mediation diagram showing paths from ERI to Burnout to Withdrawal Intentions, with moderation by Year and Resilience and covariates.

    • Figure 4: Parsimonious model illustrating the indirect effect (ER I → Burnout → Withdrawal) with year covariation.

  • Final takeaway for exam-style understanding

    • ERI can be applied to university contexts; in this study, ERI relates to withdrawal intentions primarily through burnout rather than directly.

    • Burnout serves as a robust mediator between ERI and withdrawal intentions across first-year and later-year students; resilience did not buffer this pathway in this cross-sectional analysis.

    • Year of study moderates the direct ERI–withdrawal path but does not change the main indirect effect via burnout, underscoring the importance of addressing burnout in interventions aimed at reducing student attrition.

Notes on the study of effort–reward imbalance, burnout, and withdrawal intentions among university students

  • Study focus and gap

    • Addresses the atheoretical nature of much student-stress research by applying Siegrist’s (1996) effort–reward imbalance (ERI) model to a university context, recognizing that students face similar stressors to employees.

    • Research questions:

    • Do discrepancies between efforts expended and rewards obtained relate to student distress outcomes (e.g., withdrawal intentions)?

    • Is burnout a mediator between ERI and withdrawal intentions?

    • Do resilience and year of study moderate these relationships, providing insights into protective factors and distinct student experiences?

    • Educational impact emphasized: a sizeable portion of students experience detrimental ERIs, which are linked to burnout and withdrawal intentions; implications for policy and university practices aim to mitigate these negative outcomes and support student well-being.

  • Key concepts and theory

    • ERI theory (Siegrist, 1996)

    • Core idea: self-perceived discrepancies between high efforts and low rewards lead to distress and health decline.

    • Mechanisms: originally applied to occupational settings, it highlights situations like low-status control, job insecurity, and limited input in the workplace. These mechanisms find parallels in the university setting through perceived lack of academic control, uncertainty about future career prospects, limited input in course design, and acceptance of ERIs as normal/strategic for future gains (e.g., tolerating high effort for a good degree).

    • Overcommitment: an intrinsic factor (personality trait) characterized by an inability to withdraw from work demands, which can amplify ERI effects and lead to chronic stress.

    • In educational contexts, ERI may manifest as perceived under-recognition for academic achievements, unfair marking practices, or limited input/choice in the learning environment, despite significant time and energy investment.

    • Burnout (Schaufeli et al., 2002; Maslach & Jackson, 1981)

    • Components: emotional exhaustion (feeling drained of emotional and physical resources) and cynicism (a detached, indifferent attitude towards one's studies or university experience); linked to distress, academic disengagement, and withdrawal behaviors.

    • In this study, burnout is treated as a potential mediator between ERI and withdrawal intentions, implying that ERI leads to burnout, which in turn drives withdrawal.

    • Personal resources: resilience (Smith et al., 2008)

    • Concept: the ability to bounce back or recover from stress and adversity; hypothesized to buffer (weaken) ERI effects on distress outcomes.

    • Prior work shows resilience relates to better coping strategies and lower burnout, but its moderating role for ERI in university samples was untested prior to this study, making it a novel contribution.

  • Hypotheses (summarized)

    • ERI will be positively associated with withdrawal intentions among university students, meaning higher ERI will predict stronger intentions to leave.

    • Burnout will partially mediate the ERI–withdrawal intentions link (ERI "," Burnout "," Withdrawal), indicating that ERI's effect on withdrawal is partly explained by its impact on burnout.

    • The ERI–withdrawal path and the ERI–burnout path will be strongest among first-year students, reflecting the unique transitional stress and lack of established coping mechanisms in the initial stages of university.

    • Resilience will buffer (moderate) the effects of ERI on burnout and withdrawal intentions, suggesting high resilience will mitigate the negative impact of ERI.

    • Covariates (age, gender, SES, average grade) will influence ERI, burnout, and withdrawal intentions; relatively short-term life stressors will be controlled statistically to isolate the study's primary relationships.

  • Method: design and sample

    • Design: quantitative, cross-sectional survey with a theoretical model; acknowledges limits on causal inference due to simultaneous data collection, though the theoretical model proposes temporal order.

    • Recruitment and participants

    • Sample: N = 2,451 Australian university students after comprehensive data cleaning (initial consenters N = 3,805; completed data N = 2,468; final sample N = 2,451). This robust sample size allows for advanced statistical analyses and good generalizability.

    • Timeframe: recruitment over 16 weeks of a typical 17-week semester, spanning two semesters to capture a diverse student experience and minimize temporal biases.

    • Inclusion: students 17+ years old, studying in Australia; nonprobability sampling via posters, in-class briefings, and Facebook recruitment across 40 universities and multiple disciplines to maximize reach and representativeness.

    • Incentives: relaxation CD; research-participation credit (for psychology students) or $50 gift card raffle entry (for other students) to encourage participation.

    • Demographics and sample characteristics

    • Age: M = 23.65, SD = 7.54; range 17–67, indicating a wide age distribution including mature-age students.

    • Gender: 79.2% female; 86.9% full-time; 87.8% internal (on-campus); 92.6% domestic; 89.4% undergraduate. The high proportion of female students is typical for psychology-related studies and reflects broader trends in Australian higher education.

    • Representation across fields and states; comparable to national student profiles for demographics (e.g., age, gender), enhancing external validity.

    • Ethics: approved by Murdoch University (2015/026), ensuring adherence to ethical research guidelines.

  • Measures (key scales and scoring)

    • Perceived stress: four-item version of the Perceived Stress Scale (Cohen & Williamson, 1988); α = .80, indicating good internal consistency; range 4–20.

    • ERI (School Version; Li, Shang, Wang, & Siegrist, 2010; minor wording adjustments for university): 16 items total.

    • Effort: originally 5 items, plus 2 added items (e.g., “There has been too much competition amongst classmates,” “I find that I have too much to learn everyday”). After item deletion to improve reliability, Effort α = .71, acceptable for research purposes.

    • Reward: 10 items (6 reverse-coded to prevent response bias); Reward α = .75, demonstrating good internal consistency.

    • Scoring: 7–35 for Effort; 10–50 for Reward; higher scores indicate higher levels of the construct.

    • ERI score calculation: ERI=EffortReward×1.43ERI =\frac{Effort}{Reward} \times 1.43 to adjust for different item counts between the Effort and Reward subscales, as recommended by Siegrist (1996; Siegrist et al., 2004) to ensure comparable weighting.

    • Cut-offs: continuous ERI used in analyses; dichotomous cut-off for “detrimental ERI” defined as ERI > 1.00 (Siegrist et al., 2004), meaning experienced effort is greater than perceived reward. An alternative ROC-based cut-off of 0.720.72 was also reported, providing a more sensitive threshold for identifying students at risk.

    • Burnout: Maslach Burnout Inventory – Student Survey (MBI-SS; Schaufeli et al., 2002)

    • Subscales: Exhaustion (5 items) and Cynicism (4 items); α = .92, indicating excellent internal consistency for the overall burnout construct.

    • Scale: 7-point Likert (1 = never to 7 = always); higher scores indicate greater burnout. A single burnout score is formed by summing items (range 9–63) because the subscales loaded onto a single factor (61.11% variance explained), justifying its use as a unidimensional construct in this sample.

    • Withdrawal intentions: novel 9–10 item scale (monitored via PCA) developed specifically for this study due to the lack of existing suitable measures for student withdrawal in this context.

    • Composition: 10 items initially, after item deletion due to poor item-total correlation, final 9 items used for analysis; α = .82, showing good internal consistency.

    • Scale: 7-point Likert (1 = strongly disagree to 7 = strongly agree); higher scores indicate stronger withdrawal intentions; contains five reverse-scored items to minimize acquiescence bias.

    • Psychometric support: Kaiser–Meyer–Olkin = .85 (meritorious); Bartlett’s test significant (p < .001); one-factor solution explaining 43.47% of variance (acceptable for a novel scale); loadings > .40, with several items > .60, confirming its unidimensionality and good item performance.

    • Resilience: Brief Resilience Scale (Smith et al., 2008)

    • 6 items; α = .89, demonstrating very good internal consistency.

    • Five-point Likert (1–5); higher scores indicate greater resilience; reverse-worded items included to ensure careful reading and reduce response set bias.

    • Covariates and additional controls

    • Age, gender (0 = Male, 1 = Female), SES (maternal education; coded 1–6 reflecting increasing education levels), year of study (treated as a moderator, dummy-coded 0 = first year, 1 = subsequent years to compare initial and later student experiences), average grade, and perceived stress included as covariates in analyses to statistically control for their potential influence on the main relationships.

  • Analytic approach

    • Primary analysis: moderated mediation using PROCESS (Hayes, 2013) with bootstrapping (10,000 resamples) to test indirect effects and their confidence intervals, which is robust for non-normal data and complex models.

    • Model specification: ERI (X) predicting Withdrawal Intentions (Y) via Burnout (M), with Year of Study (V) and Resilience (W) as moderators, allowing for nuanced understanding of how these factors influence the ERI-burnout-withdrawal pathway; covariates included as described to enhance model precision.

    • Data handling: listwise deletion for missing scales (if a participant missed responses to an entire scale); Winsorization of univariate outliers (replacing extreme values with less extreme ones) to reduce their undue influence on statistical estimates; HC3 standard errors to address heteroscedasticity concerns (unequal variance of residuals) and provide more accurate inference.

    • Subgroup analyses: comparisons across disciplines and institutions to check for consistency of findings; multilevel analyses deemed unnecessary due to very small intraclass correlations (indicating minimal variance explained by grouping variables like university or discipline); gender-specific analyses conducted and showed consistent patterns, suggesting the model is generalizable across genders.

    • Additional considerations: cross-sectional design acknowledged as limiting causal claims (cannot definitively establish cause-and-effect); alternative model testing included a reverse mediation check (e.g., Withdrawal " Burnout " ERI) and moderation analyses using an interaction term ERI "," Year (and ERI "," Resilience) as robustness checks to ensure the primary model's fit and validity.

  • Descriptive statistics and correlations (key numbers)

    • Descriptives (after deletion/winsorization)

    • ERI: M = 0.94, SD = 0.30; range 0.20–5; αs reported above.

    • Withdrawal intentions: M = 24.95, SD = 9.49; range 9–61; α = .82.

    • Burnout: M = 31.35, SD = 11.40; range 9–63; α = .92.

    • Resilience: M = 3.06, SD = 0.84; range 1–5; α = .89.

    • Perceived stress: M = 11.83, SD = 3.10; range 4–20; α = .80.

    • Group-level prevalence

    • 37.5% of the sample exceeded the original ERI cut-off of 1.00, indicating a significant proportion of students experienced detrimental effort-reward imbalances, highlighting the practical relevance of the study.

    • Correlations (sample-wide; all p < 0.001)

    • ERI was positively correlated with burnout (r = .55), withdrawal intentions (r = .49), and perceived stress (r = .51).

    • Burnout was positively correlated with withdrawal intentions (r = .65), and perceived stress (r = .68).

    • Withdrawal intentions were positively correlated with perceived stress (r = .59).

    • Resilience was negatively correlated with ERI (r = -.31), burnout (r = -.47), withdrawal intentions (r = -.41), and perceived stress (r = -.52), indicating its protective role. Gender was weakly but significantly correlated with some variables.