Estimating Earnings Losses Due to Mental Illness: a Quantile Regression Appoach

In this paper, we examine the effects of mental illness on earnings by recognizing that effects may vary across the distribution of earnings. Using data from the National Comorbidity Survey, we employ a quantile regression estimator to identify the effects at key points in the conditional earnings distribution. We find that earnings effects vary importantly across the distribution. While average effects are often not large, mental illness more commonly imposes earnings losses at the lower tail of the conditional earnings distribution, especially for women. Consequently, mental illness can have larger negative impacts on economic outcomes than previously estimated, even if those effects are not uniform. INTRODUCTION Tens of millions of American workers suffer from mental illness every year. During the past decade, we have come to better understand the effects of mental illness on the economic lives of the afflicted. In general, mental illness has relatively large employment effects. However, the extent to which mental illness has negative effects on earnings has been found to be less uniform. There has been a substantial amount of research published in the past few decades estimating the earnings effects of mental illness. Much of that research, especially the most recent, has devoted significant attention to developing instrumental variables (IV) estimators to control for unobserved heterogeneity between workers who suffer from mental illness and workers who do not. Still, much remains to be understood about the effects of mental illness on workers’ earnings. Not only may workers afflicted with mental illness differ from their healthy peers in ways that are hard to measure, but once afflicted it is likely that a separate non-random process plays a role in determining who remains employed or how substantially illness impedes work. Several factors shape the extent to which illness impairs workers’ abilities to maintain employment or work effectively. First, and most importantly, there is substantial variation in access to treatment. During the past three decades, there have been remarkable advances in treatment. So, disparities in access can result in important differences in the consequences of illness. Second, employment contracts vary in the extent to which mental illness might be accommodated in the workplace. Salaried workers and those with generous leave policies may 1 Estimates of the 12-month prevalence of mental disorders in the United States (excluding alcohol/substance abuse or dependence) are about 22 to 30 per 100 persons in the adult population (see Regier et al. (1993) for estimates based on the Epidemiological Catchment Area Study and Kessler et al. (1994) for estimates from the National Comorbidity Survey). 2 be more likely to maintain employment and earnings even if afflicted with an episode of illness. Those paid hourly rates or with little leave may not fare as well. Access to health care and the nature of the working environment play important roles in determining the economic consequences of mental illness. in access In considering the earnings effects of mental illness, it is important to recognize that there is a substantial amount of variation to health care and sick leave and other employment flexibilities across the earnings distribution. As a result, focusing on average earnings losses may provide insufficient information on the impact of mental illness in the labor market. Rather, this may mean that the extent to which a worker’s ability to work, and how much his/her earnings from such work are impeded depend upon his/her position in the earnings distribution. In this paper, we reexamine the effects of mental illness on earnings. We consider whether the traditional focus on mean effects provides too limited a set of information about the consequences of mental illness on earnings. We contend that such effects may vary across the earnings distribution, and that focusing on mean effects may mask important earnings losses associated with mental illness. We employ a quantile regression approach to estimate the effects of various mental illnesses at key points in the earnings distribution (conditional on the values of the independent variables in the analysis). We find that earnings effects vary substantially across the conditional distribution. In general, we find negative earnings effects to be larger at the bottom of the conditional distribution. In only one case do we find an illness to have negative effects across the conditional distribution. Below, we briefly review what is known about the labor market effects of psychiatric 3 disorders. We then turn to the general estimation problems confronting researchers in this area, and to our estimation model. Finally, we present our results and discuss their