How robust is the association between youth unemployment and later mental health? An analysis of longitudinal data from English schoolchildren

BACKGROUND: Several studies show that youth unemployment is associated with worse mental health later in life. However, existing studies report results for only one model, or a few models, and use regression adjustment to support causal claims. We use two novel methods to address these gaps in the literature. METHODS: We use data from Next Steps, a cohort study of English schoolchildren who entered the labour market in the aftermath of the 2008–2009 global financial crisis, and measure mental health using the 12-item General Health Questionnaire (GHQ-12) at age 25. We use specification curve analysis and a negative control outcome design (a form of placebo test) to test whether associations between youth unemployment and later GHQ-12 scores are sensitive to model specification or are likely to be confounded by unobserved factors. RESULTS: We find that the association between unemployment and later GHQ-12 is qualitatively similar across 99.96% of the 120 000 models we run. Statistically significant associations with two placebo outcomes, height and patience, are not present when regression adjustments are made. CONCLUSIONS: There is clear evidence that youth unemployment is related to later mental health, and some evidence that this cannot be easily explained by unobserved confounding.