Stigma, Discrimination, and Employment Outcomes among Persons with Mental Health Disabilities

Abstract

In major cultures around the world mental disorders are common and associated with substantial levels of disability. According to projections of the World Health Organization, depression-related disorders will be the single leading cause of global disease burden by the year 2020 (Murray and Lopez 1996 ). The 12-month prevalence of serious mental disorders in the U.S. is 6.5%, with higher rates among women than men (Kessler et al. 2008 ). Persons with mental disorders comprise the largest single diagnostic category of persons receiving SSDI or SSI (McAlpine and Warner 2002 ), and tend to be among the most severely disabled recipients (Estroff et al. 1997 ).

Appendix

Binary Model

Define the binary variableM _i such that:

$$ {M}_{i}\text{ = 0 if the }i\text{th individual is in the comparison group}\text(C)$$

$$ {M}_{i}=1\text{ if the }i\text{th individual is a person with a mental disorder}\left(\text{MD}\right)$$

Also define a vector of variables,Z _i , that includes controls for human capital, job characteristics, demographics, and functional limitations. Parameters of the employment function for the pooled sample can be estimated using a logistic regression model (where the dependent variable equals one if employed, zero otherwise). Predicted probabilities of employment,$\stackrel{⌢}{P}$, for each group are:

$$ {\hat{P}_{\text{MD}}} = \left[ {\frac{1}{{{n_{\text{MD}}}}}\sum\limits_{i = 1}^{{n_{\text{MD}}}} {F({\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{\beta }}{Z_i}} + {\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{\alpha }}{M_i})} \right]\quad \left( {\text{group with mental disorders}} \right) $$

(3.1)

The expression$\stackrel{⌢}{b}$represents estimated coefficients of the control variables in the employment function, that is, determinants of the relationships between an additional year of education, experience, etc., and the predicted employment rate. The expressions$F(\stackrel{⌢}{b}{Z}_{i})$and$F(\stackrel{⌢}{b}\text{\hspace{0.05em}}{Z}_{i}+\stackrel{}{\stackrel{⌢}{\alpha }{M}_{i}})$are predicted probabilities of employment for theith individual in the comparison and disadvantaged groups, respectively. So the predicted probabilities for each group$({\stackrel{⌢}{P}}_{C},{\stackrel{⌢}{P}}_{MD})$, are simply the averages across the (n _C ,n _MD) numbers of individual probabilities in the group.

If discrimination reduces the probability of employment for persons with mental disorders beyond what would be expected given their human capital, job characteristics, and functional limitations, then$\widehat{\alpha}<0$. In this case, the estimated effect of discrimination is the difference${\tilde{P}}_{\text{MD}}-{\widehat{P}}_{\text{MD}}$, where

$$ {\tilde{P}}_{\text{MD}}=\left[\frac{1}{{n}_{\text{MD}}}{\displaystyle \sum _{i=1}^{{n}_{\text{MD}}}F}(\stackrel{⌢}{b}\text{}\text{\hspace{0.05em}}{Z}_{i})\right],$$

the probability of employment for the disadvantaged group without the stigmatizing penalty associated with a mental disorder.

Decomposition Model

Employment functions are estimated separately for persons with and without mental disorders with the same control variables as described above, except that the binary variable identifying the disadvantaged group is omitted. The predicted employment rate for each group$({\stackrel{⌢}{P}}_{j})$can be expressed as:

$$ {\widehat{P}}_{j}=\left[\frac{1}{{n}_{j}}{\displaystyle \sum _{i=1}^{{n}_{j}}F\text(\stackrel{⌢}{b}\text{}\text{\hspace{0.05em}}{Z}_{i})}\right],$$

(3.2)

wheren _j is the number of persons in the group (j = Cor MD). After some simple algebraic manipulation, the difference in employment rates can be decomposed into an “explained” part attributed to differences in human capital, job characteristics, or functional limitations; and an “unexplained” part attributed, in some part, to discrimination against persons with mental disorders:

$$ \begin{array}{l}{\widehat{P}}_{C}-{\widehat{P}}_{\text{MD}}=[\frac{1}{{n}_{C}}\sum _{i=1}^{{n}_{c}}F({\stackrel{?}{b}}_{C}\text{}\text{\hspace{0.05em}}{Z}_{i}{}_{C})-\frac{1}{{n}_{\text{MD}}}\sum _{i=1}^{{n}_{MD}}F({\stackrel{?}{b}}_{C}\text{\hspace{0.05em}}{Z}_{i}{}_{M})]\\ \text+\text[\frac{1}{{n}_{\text{MD}}}\sum _{i=1}^{{n}_{MD}}F({\stackrel{?}{b}}_{C}\text{\hspace{0.05em}}{Z}_{i}{}_{MD})\text-\frac{1}{{n}_{\text{MD}}}\sum _{i=1}^{{n}_{MD}}F({\stackrel{?}{b}}_{MD}\text{\hspace{0.05em}}{Z}_{i}{}_{MD})]\end{array} $$

(3.3)

To see this, note that the first term on the right hand side of (3.3) reflects differences in employment rates attributed to differences in productivity-related characteristics (differences in the vectorsZ _i ) between the comparison group and persons with mental disorders (the “explained” part of the difference in employment rates). The second term reflects differences in employment rates attributed to differences in the structure of employment functions (the coefficients$\stackrel{⌢}{b}$) for the two groups (the “unexplained” part).