Recent workforce surveys reveal alarmingly high levels of work-related stress worldwide. Consequently, a significant share of workers experience stress-related occupational illnesses – they burn out. This column discusses recent evidence on the risk factors of burnout, its economic implications, and the optimal design of prevention programmes, with a particular focus on the role of firms and the heterogeneous risks and effects by gender.
The labour market has undergone a structural change in recent decades. First, jobs have changed (Autor et al. 2024). While average working hours have declined, more and more jobs require workers to be ‘always on’. Additionally, while jobs have become less physically straining, they have become more mentally demanding (e.g. Autor et al. 2003) and less meaningful (Kaplan and Schulhofer-Wohl, 2018). Second, more and more households consist of dual-earners, implying a career–family trade-off for both (Goldin 2014, 2020).
Perhaps as a consequence of these factors, stress among workers appears to be on the rise. For instance, in a recent Gallup survey, 53% of American workers and 44% globally report being stressed for much of the previous day. A salient aspect of this is that a sizeable share of the population – for example, about 1% of Swedish workers per year – are severely affected by chronic work-related stress and go into burnout. In response to growing concerns about high stress levels, the World Health Organization designated burnout as a health condition caused by chronic workplace stress in 2019 (WHO 2019).
Despite this increased awareness, little is known about the interaction between burnout and the labour market. Prior research has documented the adverse long-run effects of mental illness, focusing on mental health in general (Bartel and Taubman 1998) or specific diagnoses such as bipolar disorder (Biasi et al. 2021). Another strand of the literature has looked at workplace origins of mental health. Yamamoto (2016) documents a relationship between long hours and poor mental health. Rees et al. (2024) find that becoming a physician substantially increases use of antidepressants.
In a recent paper (Nekoei et al. 2024), we study the risk factors and estimate the economic costs of burnout. Our analysis is based on the universe of sick leave records and associated medical diagnoses from 2006 to 2020 in Sweden, which we combine with various administrative registers. Sweden’s early adoption of specific criteria for diagnosing burnout in 2006 presents a unique opportunity to analyse long-run labour market outcomes for burned-out workers.
Stressful jobs and risky firms
Workplace stress is, by definition, the primary cause of burnout. We demonstrate this by studying how firms and occupations influence burnout. Figure 1 presents a binned scatter plot that compares occupation-level burnout rates with the stress tolerance requirements for occupations as delineated by O*NET data. The figure documents a robust correlation between the level of work-related stress and the incidence of burnout. To evaluate the effect of firms on burnout, disentangling the impact of worker sorting, we study workers who switch between firms with different burnout risks as measured by the burnout rate of other workers. Strikingly, moving from the least risky quartile of firms to a firm in the most risky quartile quadruples the likelihood of burnout. We detect similar effects for switches to more stressful occupations, both when switches are associated with firm moves and within firms, holding constant the firm environment.
Figure 1 Stress tolerance requirement and burnout rates
![Figure 1 Stress tolerance requirement and burnout rates](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig1.png?itok=IoA7xh3h)
![Figure 1 Stress tolerance requirement and burnout rates](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig1.png?itok=IoA7xh3h)
![Figure 1 Stress tolerance requirement and burnout rates](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig1.png?itok=ZVDye36i)
![Figure 1 Stress tolerance requirement and burnout rates](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig1.png?itok=ZVDye36i)
Note: The figure plots in solid blue dots the binned scatter plot of occupation-level burnout rate against stress-tolerance requirement according to O˚NET. In orange diamonds, the figure plots a selected subset of occupations.
Burnout risk is shaped by the interaction of workplace characteristics and workers’ personality traits. To document this, we use information on stress tolerance requirements of occupations from O*NET and individuals’ stress tolerance assessed at age 18 by psychologists during the Swedish military draft. We uncover that in high-stress occupations, workers with low stress tolerance exhibit a significantly higher likelihood of burnout compared to those with higher stress tolerance (Figure 2).
Figure 2 Mismatch: Stressful jobs and stress-tolerant workers
![Figure 2 Mismatch: Stressful jobs and stress-tolerant workers](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig2.png?itok=-QstTxvO)
![Figure 2 Mismatch: Stressful jobs and stress-tolerant workers](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig2.png?itok=-QstTxvO)
![Figure 2 Mismatch: Stressful jobs and stress-tolerant workers](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig2.png?itok=UboVT1mm)
![Figure 2 Mismatch: Stressful jobs and stress-tolerant workers](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig2.png?itok=UboVT1mm)
Note: The figure displays the interaction of occupational stress requirements and workers’ stress tolerance in the risk of burnout. Stress tolerance is assessed by psychologists around age 18 before military service. The occupational stress tolerance requirement is sourced from O*NET. We group occupations into quintiles based on this stress tolerance requirement. This figure is based entirely on male workers due to the availability of the stress tolerance measure.
Two factors may carry a counterweight in evaluating the implications of stressful jobs and firms with high burnout risk. First, workers may receive compensation for stress and the risk of burnout through higher wages. Recent research using experiments (Nagler et al. 2024) and structural models (Jolivet and Postel-Vinay 2024) has documented that workers value low-stress jobs greatly. Second, stress might be associated with higher productivity. We do not find support for either. Instead, workers in firms with higher burnout risk receive lower wages, and those jobs are less productive as measured by the value added per worker.
Scarring effects and spillovers of burnout
Workers who burn out suffer severe earnings losses. Compared to a control group of similar workers that either never burn out or burn out later, burning out has a negative, substantial, and persistent effect on labour income, similar to the well-documented displacement effect of layoff (Davis and von Wachter 2011). As depicted in Figure 3, annual income drops by nearly 15% at impact, primarily attributed to sick leave days. Seven years later, the income drop stabilises at 12%, with half resulting from labour market exit and the other half from transitions into part-time employment. While social insurance reduces immediate earnings losses through sick pay and other transfers, disposable income falls by more than 6% in the long run.
Figure 3 Scarring effect on labour income
![Figure 3 Scarring effect on labour income](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig3.png?itok=sF2vHfZC)
![Figure 3 Scarring effect on labour income](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig3.png?itok=sF2vHfZC)
![Figure 3 Scarring effect on labour income](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig3.png?itok=EdF8dmmc)
![Figure 3 Scarring effect on labour income](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig3.png?itok=EdF8dmmc)
Note: The figure shows the proportional effect of burnout on labor income estimated from a dynamic matched difference-in-difference model. Pre-treatment average incomes in 10k SEK and proportional effects are reported in the upper right corner.
We further document how the repercussions of burnout transcend the individual worker and extend to their families, affecting their spouse’s career and their children’s human capital. Female spouses experience an immediate and persistent drop in labour income of 4.4%. The corresponding drop for male spouses is much smaller (1.1%). Parental burnout during children’s schooling years also reduces their children’s enrolment by 2.5 percentage points, or 8%, compared to a control group whose parents burn out when children have passed college-decision age. Similarly, parental burnout negatively affects performance on high-stakes exams.
Aggregating across channels – the direct effect of sick leave and the career consequences for those affected and their families – we estimate a 2.3% loss of aggregate labour income due to burnout in 2019. This figure likely underestimates the long-term loss, given that our diagnosed burnout data only extend back to 2006. We project that this loss will escalate to 3.5% in a steady state if conditions remain unchanged as of 2019.
Women at risk
Women are disproportionately hurt by work-related stress, in line with a general gender discrepancy in mental health (Boneva et al. 2024). They are three times more likely to burn out than men (1.85% versus 0.54%), implying that by the age of 40, one in every seven women has experienced burnout. Among women, single mothers, women who earn more than their husbands, and women with rising careers are at most risk of burnout.
While the severity of income loss is, however, unrelated to gender, women are disproportionally affected in other dimensions. For example, we estimate a permanent reduction in fertility among women who burn out, while for men, the reduction is only transitory.
Prediction and prevention
Preventing burnout is more effective than addressing it post-occurrence, akin to other mental health issues (Tetrick and Winslow 2015, Aust et al. 2023). While the development of preventive programmes remains a dynamic area of research (Bouskill et al. 2022), practical implementation necessitates identifying individuals most at risk to ensure effective targeting. Additionally, understanding the costs associated with burnout is crucial for optimising the scope of these programmes (Demerouti et al. 2021).
Using machine learning, we use the ‘kitchen sink’ of administrative data to accurately predict higher burnout risk in 81% of comparisons between pairs of individuals, one experiencing burnout and the other not. However, we find that self-reported stress from a nationally representative work environment survey contains additional information beyond what is captured by the kitchen sink of administrative data, suggesting the existence of private information (Hendren 2017). Figure 4 visualises the strong correlation between self-reported subjective work stress and burnout in the subsequent year.
Figure 4 Self-reported stress and burnout rate in the subsequent year
![Figure 4 Self-reported stress and burnout rate in the subsequent year](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig4.png?itok=9k5Zc9YV)
![Figure 4 Self-reported stress and burnout rate in the subsequent year](https://cepr.org/sites/default/files/styles/flexible_wysiwyg/public/2024-06/nekoei22junefig4.png?itok=9k5Zc9YV)
![Figure 4 Self-reported stress and burnout rate in the subsequent year](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig4.png?itok=QGau0a6X)
![Figure 4 Self-reported stress and burnout rate in the subsequent year](https://cepr.org/sites/default/files/styles/popup_small/public/2024-06/nekoei22junefig4.png?itok=QGau0a6X)
Note: The figure documents the relationship between the self-reported work condition in the “Work Environment Survey” (AMU) and the burnout rate of the same worker in the year after the survey. The survey asks workers to judge how mentally stressful they find their job. The distribution of responses is reported on the second (right) y-axis. The regression controls for gender – education – age – family type fixed effects and year fixed effects.
We use our cost estimates and predictions to evaluate the optimal scope and targeting of preventive programmes. In particular, a cost-benefit analysis shows that the optimal size of the programme is reduced by half, and its net gain is 2.5 times larger, once we use the survey, in addition to demographic information, to target those at risk.
Conclusion
The current evidence highlights the significant economic impact of burnout, emphasising the urgent need for targeted interventions, especially for at-risk groups like women. We invite fellow economists to explore this critical issue further, contributing to a deeper understanding and the development of effective prevention strategies.
Source : VOXeu