Employer Stigma as a Mediator Between Past and Future Hiring Behavior

Participants

Research participants for this study consisted of a representative sample of employers recruited from Chicago, Beijing, and Hong Kong (Rao, Angell, Lam, & Corrigan, 2009). Employers were randomly selected from each city from comprehensive lists of small businesses. The criteria for small businesses were that the businesses consist of less than 3 to 50 employees, and they should not have a designated human resources division. These lists of small employers were obtained from Dun and Bradstreet for Chicago, the Labor Department of Hong Kong, and the Industry and Commerce Bureau in Beijing. In total, 879 participants were recruited for the overall study (293 from Chicago, 302 from Beijing, and 284 from Hong Kong) to examine the effects of stigma associated with behaviorally driven health conditions on employment-related discrimination. These three cities were chosen due to the fact that they belong to countries boasting the largest economies in the world (the United States and China). In addition, the potential for investigating cultural differences across these three cities in terms of employment-related stigma and discrimination was a factor in this recruitment decision. The results of such cross-cultural investigations have been reported in previous studies (Corrigan, Tsang, Shi, Lam, & Larson, 2010; Rao, Horton, Tsang, Shi, & Corrigan, 2010; Tsang et al., 2007). Findings from Corrigan et al. (2010) suggest that employers from the United States endorsed onset and offset attributions related to behaviorally driven health conditions less so than employers from China, and recognized more benefits of hiring people with various health conditions. Rao et al. (2010) found that employers in Chicago, Beijing, and Hong Kong all endorsed concerns with regard to hiring people with HIV; however, cross-cultural differences were noted regarding the nature of these concerns (for example social contagion was of particular concern to employers in Beijing and Hong Kong compared with Chicago). Tsang et al. (2007) specifically examined employer attitudes toward individuals with serious psychiatric disabilities cross-culturally, finding many similarities between employers from Chicago, Beijing, and Hong Kong in terms of the types of concerns endorsed, and also finding that Chinese employers were significantly more likely to report concerns that people with psychiatric disabilities would exhibit a weaker work ethic and less loyalty to the company. As these data have been explored cross-culturally in these earlier studies, cross-cultural differences were not the focus of the analyses described here.

Within the context of the survey, employers were randomly assigned to one of five health conditions (described in greater detail in later discussion of the measures employed in this study). For the purpose of this study, we were only interested in stigma and discrimination related to potential employees with psychiatric disabilities and therefore selected only participants assigned to this condition (N = 184). Participants were approximately evenly split among the three cities (Chicago = 35.1%, Hong Kong = 32.2%, and Beijing = 32.8%). There were slightly more women (53.4%) than men (46.6%). Within the U.S. sample, participants were predominantly Caucasian (29.5%) and Asian (29.5%), with 24.6% being Black or African American and 16.4% Hispanic/Latino. All employers within the sample from Beijing and Hong Kong identified as Chinese. The mean age of the sample included in the analyses was 47 years (SD = 12.78) and 55.6% of the sample reported having at least a bachelor’s degree.

Procedure

This research was reviewed and approved by the Institutional Review Boards (IRBs) of the University of Chicago, Northwestern University, Hong Kong Polytechnic University, the Chinese Academy of Science, and Beijing Normal University. The entire study was conducted face-to-face with employers, with trained research personnel administering all survey questions. Informed consent was obtained from all participants. Participants were compensated for completing the study. Following participation, all participants were debriefed to the study.

Measures

All measures were developed based on qualitative interviews with employers from small firms with 3 to 100 employees and without a human resource department. All interviewees were owners of their firms or personally charged with making hiring decisions. An expert panel of researchers from Beijing, Chicago, and Hong Kong (N = 11) with expertise in rehabilitation psychology (important for decisions about health conditions) and industrial/organizational psychology (for work-related decisions) informed the development of the interview guide. A penultimate draft of the interview guide was written in English and revised into Cantonese and Mandarin through translation, back-translation, and reconciliation of items by the expert panel. Transcripts were then coded using a grounded theory approach (Miles & Huberman, 1994) and from these data a quantitative measure was developed. A more complete discussion of the qualitative findings can be found in Corrigan et al. (2008). All measures and vignettes were presented to participants in the language appropriate to them (English, Cantonese, and Mandarin).

Participants were first presented with a vignette in which they imagined they had posted an advertisement in the newspaper for an entry-level position with their business. They were then asked to rank a group of 10 fictitious job applicants with behaviorally driven health conditions on their willingness to offer the applicants a job interview. Behaviorally driven health conditions are defined here as those conditions for which onset or offset of the condition are viewed as being under the control of the individual with the condition. The behaviorally driven health conditions included in the survey were alcohol dependence, drug dependence, bone cancer, psychiatric disability, and HIV/AIDS. For a discussion regarding why these particular conditions were chosen to represent those that are behaviorally driven, see Corrigan et al. (2010). As the focus of the analyses described here is on individuals with psychiatric disabilities, it is important to point out that previous work has demonstrated public attributions about the stability and controllability of psychiatric disabilities and how such attributions contribute to stigmatizing attitudes (Corrigan, River, Lundin, & Uphoff Wasowski, 2000; Weiner, Perry, & Magnusson, 1988). This work has demonstrated that individuals with psychiatric disabilities are often viewed as being responsible for their disability (related to the controllability factor) and incapable of recovery (the stability factor). These stigmatizing attributions regarding individuals with psychiatric disabilities potentially contribute to discrimination in important life domains, such as employment.

In addition to the five fictitious candidates with behaviorally driven health conditions, five additional fictitious candidates were included who also have commonly stigmatized identities (i.e., homosexual identity, criminal background, learning disability, stroke survivor, and “life problems”). All of the applicants were described as having had a period of unemployment in their personal history during the past 3 months due to their condition (e.g., “During the past 3 months, I was in a psychiatric hospital because I was receiving treatment for psychosis.”). The word psychosis was used to represent psychiatric disability throughout the survey as the employer focus groups revealed that employers felt this would be the best way to communicate serious psychiatric disability. No additional information was provided about any of the applicants during this part of the survey. The vignettes were intentionally kept short so that the employers’ responses would reflect their fundamental reactions to the labeled condition; the participants were expected to reveal their stereotypes more when given only brief vignettes. Employers sorted the 10 applicants based on the order in which they would invite them for a face-to-face interview with 1 being most preferred. Employers were also informed that applicants ranked 5 through 10 would not be offered an interview.

Testing Mediated Effects

We compared our hypothesized, partially mediated model with a fully mediated model. Structural equation modeling (SEM) using maximum likelihood estimation in the AMOS 20 program was used to examine these alternative models. Five indices were used to assess goodness of fit of the models. These included chi-square (χ²), comparative fit index (CFI), non-normed fit index/Tucker–Lewis index (NNFI/TLI), root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). Of these indices, only chi-square provides evidence of statistical significance. Unlike most tests of statistical significance associated with chi-square, however, support for a proposed model is demonstrated with a nonsignificant value. Specifically, a nonsignificant value (p > .05) for a particular model indicates that the observed and reproduced variance–covariance matrices are not significantly different from one another. The remaining indices are interpreted relative to common “rules of thumb” (Kline, 2011). For the present study, these rules of thumb were CFI and NNFI/TLI (.95 or greater), RMSEA (.05 or less), and SRMR (.08 or less).

Because the maximum likelihood procedures used to test the hypothesized model assumes normality, we examined the multivariate normality of the observed variables. Specifically, we focused on values of multivariate kurtosis as multivariate kurtotic data are seen to be particularly problematic for SEM (Byrne, 2010; Raykov & Marcoulides, 2010). West, Finch, and Curran (1995) consider rescaled β² values equal to or greater than 7 to indicate early departure from normality. Using this value as a guide, our data did not violate the assumption of multivariate normality.

Model testing

We predicted that the effect of previous hiring behavior related to individuals with serious psychiatric disabilities on future hiring of individuals with serious psychiatric disabilities would be mediated by employer stigma toward individuals with serious psychiatric disabilities. The structural model used to test this hypothesis (see Figure 1) resulted in a significant chi-square; however, the corresponding fit indices showed reasonable fit to the data, χ²(43) = 83.90, p < .001, CFI = .95, NNFI/TLI = .94, SRMR = .04, RMSEA = .07. The standardized path coefficients for Model A can be found in Figure 1. The direct effects for this model indicate that previous employer hiring behavior related to individuals with serious psychiatric disabilities significantly predicted employer stigma (β = −.28, p < .001), which in turn predicted future employer hiring behavior of individuals with serious psychiatric disabilities (β = −.23, p < .001). These results indicate that for employers with more previous experience in hiring, accommodating, promoting, and interviewing people with known psychiatric disabilities, stigma levels are lower, and for those employers with lower levels of stigma, likelihood of offering an interview to an individual known to have a serious psychiatric disability is greater. It is important to note that the parameter representing the direct effect of previous hiring behavior on future hiring behavior was not significant.

OPEN IN VIEWER

Figure 1.

Partially mediated model of the relationship between previous and future hiring.

Because the direct path between previous hiring behavior and future hiring behavior was nonsignificant, this justified constraining this path to zero in the estimation of our alternative model, the fully mediated model (see Figure 2). The chi-square difference test was then used to compare these two nested models to determine which model best fit our data. The fully mediated model resulted in a nonsignificant chi-square; moreover, the corresponding fit indices indicated reasonable fit to the data, χ²(44) = 85.82, p < .001, CFI = .95, NNFI/TLI = .94, SRMR = .04, and RMSEA = .074. The standardized path coefficients for Model B can be found in Figure 2. A nonsignificant chi-square difference, Δχ²(1) = 1.92, p = .20, indicated no differences between these two models. On the basis of the principle of parsimony, the fully mediated model (Model B) was selected as the best model. The coefficients associated with this model indicate that previous employer hiring behavior related to individuals with serious psychiatric disabilities significantly predicted employer stigma (β = −.28, p < .001), which in turn predicted future employer hiring behavior of individuals with serious psychiatric disabilities (β = −.26, p < .001). The final, best fitting model is therefore a fully mediated model where employer stigma explains the relationship between employers’ previous hiring behavior with regard to individuals with serious psychiatric disabilities, and employers’ future likelihood of offering a person with a serious psychiatric disability an interview (future hiring behavior). The indirect effect of previous hiring behavior on future hiring behavior as estimated by this model is .07. The Sobel test (Sobel, 1986) was used to test whether the indirect effect of the independent variable in this model (previous hiring behavior) on the dependent variable (future hiring behavior) was significantly different from zero. The resulting Sobel test value of 2.54 (p < .05) indicated that this was the case.

OPEN IN VIEWER

Figure 2.

A fully mediated model of the relationship between previous and future hiring.