Article by Fiona Steele and Robert French (LEMMA 3, University of Bristol), and Mel Bartley (ESRC International Centre for Lifecourse Studies, UCL). This article appears in the Spring 2013 issue of MethodsNews newsletter (opens a .pdf file).
Podcast with Fiona Steele: 'Relationship between employment transitions and mental health among British men'
Â
Previous research suggests that there is a strong association between unemployment and mental health, but can we infer a causal effect of unemployment on subsequent ill health? To what extent can the relationship be explained by selection of less healthy men into unemployment or by unmeasured characteristics that affect both health and the risk of unemployment?
There is substantial interest in understanding the association between labour force participation and mental health. Longitudinal data from panel studies allow detailed examination of how changes in employment status relate to changes in health over time, but selection bias remains a serious threat to causal inference. Selection into unemployment may be conceptualised as occurring by two distinct mechanisms: direct and indirect selection. Direct selection asserts that poor health leads to difficulty in securing and maintaining employment, which leads to an over-representation among the unemployed of less psychologically healthy individuals. Indirect selection, on the other hand, occurs when there are unmeasured individual characteristics that influence both health and the risk of unemployment. The standard approach taken to allow for direct selection is to adjust for health measured before the experience of unemployment. Adjustment for indirect selection is usually achieved by including indicators of family background, but such information is limited in panel studies which began in adulthood.
We have developed a framework for studying the relationship between men's employment transitions and their mental health that allows us to test for both direct and indirect selection effects. We used annual data from the British Household Panel Survey for 1991-2009. Mental health was measured using the General Health Questionnaire (GHQ) and employment status was classified as employed, economically inactive or unemployed. Employment transitions were identified by comparing an individual's status at two consecutive interviews. We estimate two models simultaneously: a model for the effect of annual change in employment status on GHQ at the start of the following year, and another for the effect of GHQ on subsequent employment change. The two models are linked to allow for unmeasured individual factors that affect both employment and GHQ. We also consider simpler models to explore the impact of successive adjustments for selection bias on estimates of the impact of employment change on GHQ.
We find that moving out of unemployment into employment is strongly associated with an improvement in mental health, while becoming unemployed is detrimental for mental health. Although these effects are slightly inflated if indirect selection is ignored, they remain substantial following adjustment. For example, job loss is associated with half a standard deviation change in the GHQ score. There is a much stronger impact of allowing for indirect selection on estimates of the effect of moving between employment and economic inactivity, which is expected due to the high proportion of inactive men who are long-term sick or disabled. There is little evidence of direct selection by poorer pre-existing mental health; GHQ has a weak effect on any type of employment transition.
Future work should seek to understand the pathways through which the experience of unemployment influences mental health. For example, is most of the effect due to financial hardship following job loss? It is also important to consider differential effects of employment change. Previous research has found that men with higher status jobs suffer a worse reaction to unemployment than those in less prestigious occupations. The effect of employment change may also differ for men and women.
Â
Forthcoming: Adjusting for selection bias in longitudinal analyses of the relationship between employment transitions and mental health using simultaneous equations modelling, Epidemiology (in press).