Pierre K. Alexandre1, Michael T. French. 1. University of Miami (D93), Highland Professional Bldg., 1801 NW 9th Avenue, Third Floor, Miami, Florida 33136, USA - Tel.:+1-305-243-3482 - Fax:+1-305-243-2149, p.alexandre@miami.edu
Abstract
BACKGROUND: Depression represents one of the most common behavioral health problems among the workforce in the United States, with about 1 in every 20 employees experiencing this condition. A recent study estimated that in 1990 the economic costs of depressive disorders in the American workplace amounted to as much as $43 billion, with absenteeism alone accounting for $12 billion. Recently, economists have been focusing attention on the relationship between mental health and labor supply, but a lack of quality data sets containing detailed information on mental health and labor market variables represents a significant barrier to rigorous research. AIMS OF THE STUDY: The primary aims of the present study were to (i) examine the relationship between depression and employment, (ii) conditional on being employed, estimate the effect of depression on annual weeks worked, and (iii) examine the stability of the model estimates to the co-morbid effects of substance use (illicit drugs and alcohol), which has been consistently found to be a correlate of depression. DATA: The study used a unique set of survey data collected between 1996 and 1997 in crime-ridden and low-income neighborhoods of Miami-Dade County, Florida. A targeted sampling strategy was used to recruit chronic drug users (including injection drug users) and non-drug users to examine local health care delivery system characteristics in relation to the population of substance users. The final analysis sample for the present study included 1,274 adults, aged 18 to 65. Depression status was measured from the 20-item Zung Self-Rating Depression Scale (SDS) that classified 384 individuals as depressed and 890 as non-depressed. According to the definition developed by the U.S. Office of National Drug Control Policy for chronic drug use (CDU), about 46 percent of the depressed individuals were found to be CDUs compared to 30 percent of the non-depressed sample. The survey instrument collected information on alcohol use and problem drinking as defined by the 10-item Michigan Alcoholism Screening Test (MAST-10). Based on criteria defined in the MAST-10, 26 percent of the depressed individuals were problematic alcohol users (PAUs) compared to about 16 percent of the non-depressed sample. METHODS: The labor supply measures included employment in the past 30 days and number of weeks worked in the past 12 months. The analysis estimated a univariate probit model of employment as well as a bivariate probit model of depression and employment, which accounted for the possible correlation between the unobserved determinants of depression and employment. The annual weeks worked specification was estimated by a standard Tobit model as well as an instrumental variable (IV) Tobit model, which, in addition to the censoring of the observations, accounted for the possible endogeneity of depression. The stability of the estimated effects of depression to comorbid illicit drug and alcohol use was assessed, by controlling for CDU and PAU in these models. RESULTS: Results from both the univariate probit and the bivariate probit models indicate that depression significantly decreased the probability of being employed. Specifically, depression reduced the probability of employment by an average of 19 percentage points in both models, from a sample average of 43 percent for the non- depressed to 24 percent for the depressed. Estimates from the Tobit models revealed that depression also significantly reduced the number of weeks worked. Conditional on being employed, depressed individuals worked an average of 7 fewer annual weeks than the non-depressed sample in the univariate Tobit model and 8 fewer weeks in the IV Tobit. The findings also showed that the effects of depression on employment and annual weeks worked may be over-estimated if the analysis does not account for the comorbid influence of substance use. IMPLICATIONS FOR HEALTH CARE PROVISION AND USE: The results suggest that prevention and/or treatment of mental health problems such as depression may yield economic benefits by promoting employment and enhancing labor supply. While expansion of public mental health services may not lead to overall increases in employment, it may be justified on social grounds given the high unemployment rate in low-income and crime-ridden neighborhoods. Further insights can be gained by estimating these models with national and international data if one applies appropriate econometric tools to account for complex sample designs.
BACKGROUND:Depression represents one of the most common behavioral health problems among the workforce in the United States, with about 1 in every 20 employees experiencing this condition. A recent study estimated that in 1990 the economic costs of depressive disorders in the American workplace amounted to as much as $43 billion, with absenteeism alone accounting for $12 billion. Recently, economists have been focusing attention on the relationship between mental health and labor supply, but a lack of quality data sets containing detailed information on mental health and labor market variables represents a significant barrier to rigorous research. AIMS OF THE STUDY: The primary aims of the present study were to (i) examine the relationship between depression and employment, (ii) conditional on being employed, estimate the effect of depression on annual weeks worked, and (iii) examine the stability of the model estimates to the co-morbid effects of substance use (illicit drugs and alcohol), which has been consistently found to be a correlate of depression. DATA: The study used a unique set of survey data collected between 1996 and 1997 in crime-ridden and low-income neighborhoods of Miami-Dade County, Florida. A targeted sampling strategy was used to recruit chronic drug users (including injection drug users) and non-drug users to examine local health care delivery system characteristics in relation to the population of substance users. The final analysis sample for the present study included 1,274 adults, aged 18 to 65. Depression status was measured from the 20-item Zung Self-Rating Depression Scale (SDS) that classified 384 individuals as depressed and 890 as non-depressed. According to the definition developed by the U.S. Office of National Drug Control Policy for chronic drug use (CDU), about 46 percent of the depressed individuals were found to be CDUs compared to 30 percent of the non-depressed sample. The survey instrument collected information on alcohol use and problem drinking as defined by the 10-item Michigan Alcoholism Screening Test (MAST-10). Based on criteria defined in the MAST-10, 26 percent of the depressed individuals were problematic alcohol users (PAUs) compared to about 16 percent of the non-depressed sample. METHODS: The labor supply measures included employment in the past 30 days and number of weeks worked in the past 12 months. The analysis estimated a univariate probit model of employment as well as a bivariate probit model of depression and employment, which accounted for the possible correlation between the unobserved determinants of depression and employment. The annual weeks worked specification was estimated by a standard Tobit model as well as an instrumental variable (IV) Tobit model, which, in addition to the censoring of the observations, accounted for the possible endogeneity of depression. The stability of the estimated effects of depression to comorbid illicit drug and alcohol use was assessed, by controlling for CDU and PAU in these models. RESULTS: Results from both the univariate probit and the bivariate probit models indicate that depression significantly decreased the probability of being employed. Specifically, depression reduced the probability of employment by an average of 19 percentage points in both models, from a sample average of 43 percent for the non- depressed to 24 percent for the depressed. Estimates from the Tobit models revealed that depression also significantly reduced the number of weeks worked. Conditional on being employed, depressed individuals worked an average of 7 fewer annual weeks than the non-depressed sample in the univariate Tobit model and 8 fewer weeks in the IV Tobit. The findings also showed that the effects of depression on employment and annual weeks worked may be over-estimated if the analysis does not account for the comorbid influence of substance use. IMPLICATIONS FOR HEALTH CARE PROVISION AND USE: The results suggest that prevention and/or treatment of mental health problems such as depression may yield economic benefits by promoting employment and enhancing labor supply. While expansion of public mental health services may not lead to overall increases in employment, it may be justified on social grounds given the high unemployment rate in low-income and crime-ridden neighborhoods. Further insights can be gained by estimating these models with national and international data if one applies appropriate econometric tools to account for complex sample designs.