Current Unemployment, Unemployment History, and Mental Health: A Fixed-Effects Model Approach

Abstract

Poor mental health among the unemployed-the long-term unemployed in particular-is established, but these associations may be driven by confounding from unobserved, time-invariant characteristics such as past experiences and personality. Using longitudinal register data on 2,720,431 residents aged 30-60 years, we assessed how current unemployment and unemployment history predict visits to specialized care due to psychiatric conditions and self-harm in Finland in 2008-2018. We used linear ordinary-least-squares and fixed-effects models. Prior to adjusting for time-invariant characteristics, current unemployment was associated with poor mental health, and the risk increased with longer unemployment histories. Accounting for all time-invariant characteristics with the fixed-effects models, these associations attenuated by approximately 70%, yet current unemployment was still associated with a 0.51 (95% confidence interval: 0.48, 0.53) percentage-point increase in the probability of poor mental health among men and women. Longer unemployment histories increased the probability among men in their 30s but not among older men or among women. The results indicate that selection by stable characteristics may explain a major part of the worse mental health among the unemployed and especially the long-term unemployed. However, even when controlling for this selection, current unemployment remains associated with mental health.

Keywords: hospitalization; longitudinal study; mental health; unemployment.

Figure 1

Predicted probability of visits to specialized care due to psychiatric conditions and self-harm among men and women aged 30–60 years, Finland, 2008–2018. Results are derived from an ordinary-least-squares (OLS) model adjusted for year, quarter, and birth year and a fixed-effects (FE) model additionally adjusted for year, quarter, partnership status, and coresident children: A) men, OLS model; B) men, FE model; C) women, OLS model; D) women, FE model.

Figure 2

Predicted probability of visits to specialized care due to psychiatric conditions and self-harm among men according to 10-year age group, Finland, 2008–2018. Results are derived from an ordinary-least-squares (OLS) model adjusted for year, quarter, and birth year, and a fixed-effects (FE) model additionally adjusted for year, quarter, partnership status, and coresident children: A) OLS model, ages 30–39 years; B) FE model, ages 30–39 years; C) OLS model, ages 40–49 years; D) FE model, ages 40–49 years; E) OLS model, ages 50–60 years; F) FE model, ages 50–60 years.

Figure 3

Predicted probability of visits to specialized care due to psychiatric conditions and self-harm among women according to 10-year age group, Finland, 2008–2018. Results are derived from an ordinary-least-squares (OLS) model adjusted for year, quarter, and birth year, and a fixed-effects (FE) model additionally adjusted for year, quarter, partnership status, and coresident children: A) OLS model, ages 30–39 years; B) FE model, ages 30–39 years; C) OLS model, ages 40–49 years; D) FE model, ages 40–49 years; E) OLS model, ages 50–60 years; F) FE model, ages 50–60 years.