The impact of recent mental health changes on employment: new evidence from longitudinal data

ABSTRACT

This study uses longitudinal data and four different measures of mental health to tease out the impact of psychiatric disorder onsets and recoveries on employment outcomes. Results suggest that developing a mental health problem leads to a significant increase in the probability of transitioning to non-employment, while a recovery increases the probability of return to work among the not employed with a mental health problem. No consistent effect was found on hours worked and earnings. Research and policy attention is needed with respect to early interventions such as job retention programmes to help workers with mental health problems remain employed as well as interventions that may lead to recovery and return to work. More research is needed especially with data and models that can differentiate between the effects of mental health onsets and recoveries on employment exit and return to work transitions.

KEYWORDS:

• Employment
• labour market outcomes
• mental illness
• psychiatric disorders
• substance abuse

JEL CLASSIFICATION:

• I10
• J20

Acknowledgement

The authors are grateful for expert research assistance by Ginette Carvalho, Yunfei Song and Dr Brandon Vick, and for comments from seminar participants at Fordham University, the City University of New York Graduate Center Microeconomics seminar, the National Institute of Mental Health Conference in Mental Health Economics and the 2013 Journées d’Economie Publique Louis-André Gérard-Varet.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Other IVs used in the literature include religiosity/spirituality (Alexandre and French 2001; Chatterji et al. 2007), social support (Hamilton, Merrigan, and Dufresne 1997; Ojeda et al. 2010), parents’ mental health and substance abuse behaviour (Chatterji, Alegria, and Takeuchi 2011; Ettner, Frank, and Kessler 1997; Renna 2008), exposure to daylight (Tefft 2012) and the local availability of health or mental health service providers (Alexandre and French 2001)

² The HCC identifies three mental health disorders (major depression, dysthymia, general anxiety disorder) that can be considered short-term mental health conditions with questions that query about the last 12 months. It also identifies three mental health disorders (panic disorder, psychosis and mania) that can be considered as long-term measures of MH and therefore should be present at both waves. For instance, in the panic screener, the first question is as follows: ‘In your entire lifetime, have you ever had an attack when all of a sudden you felt frightened, anxious or very uneasy?’ In HCC, we find some time-inconsistent variation in the prevalence of these long-term mental health disorders, e.g. reporting one in the first wave, but not in the second. This variation may result from measurement error, say, due to lapsed memory or confusion as to the meaning of ‘lifetime’ versus ‘in last 12 months’ questions. We therefore do not study the impact of the onset of panic disorder, psychosis and mania.

³ The 10 screening questions for having an alcohol problem were as follows: (1) how often drank alcohol past 12 months; (2) number of alcohol drinks per day; (3) how often had more than six drinks at once; (4) how often unable to stop drinking; (5) how often failed to do thing due to drinking; (6) how often need a drink in morning; (7) how often felt guilt/remorse after drinking; (8) how often unable to remember after drinking; (9) had anybody been injured due to your drinking; and (10) is somebody concerned about your drinking.

⁴ The screening questions for having a drug problem asked if in the past 12 months the person used sedatives, tranquilizers, amphetamines, analgesics, inhalants, marijuana/hashish, cocaine/crack/free base, LSD/hallucinogen or heroin, any other drug and needed larger amount. It also asks if the person had emotional or psychological problems from using drugs and the number of days in the past 30 days that the person used barbiturates, tranquilizers, amphetamines, analgesics, inhalants, marijuana, cocaine, LSD/hallucinogen or heroin.

⁵ Given that we have the dates of the interviews for both waves, we replaced the difference in age between the two waves with the number of months elapsed between the two waves for each individual. It varies from a minimum of 21.3 months to a maximum of 45.2 months.

⁶ When first-stage F-statistics are small, two-stage least-square estimates and confidence intervals are unreliable (Staiger and Stock 1997).

⁷ Descriptive statistics are not given for males and females separately. In brief, women are more likely to have depression/dysthymia/anxiety than men (15% versus 9%), mostly due to their greater prevalence of major depression disorder. As for employment outcomes, a smaller proportion of women were employed (74% versus 88%) with lower median work hours per week (40 versus 45) and median earnings ($23,000 versus $35,000) compared to men among the employed. Characteristics were similar for women and men.

⁸ Chronic health conditions for which information is available are as follows: asthma, diabetes, hypertension, arthritis, physical disability, trouble breathing, cancer, neurological disorders, stroke, angina, back problem, ulcer, liver, migraine, bladder, gynaecological disorders and chronic pain.

⁹ The remaining 10% are from a supplemental national sample.

¹⁰ There are two possible concerns. First, observable factors that result in attrition may be correlated with the error term in the specification of interest. Second, unobservable variables in the attrition equation may be correlated with the unobservable variable in the empirical specification of interest (Fitzgerald, Gottschalk, and Moffitt 1998).

Additional information

Funding

The authors gratefully acknowledge funding from the National Institute of Mental Health [MH078188].