The Strength of Weak Enforcement: The Impact of Discrimination Charges, Legal Environments, and Organizational Conditions on Workplace Segregation

s for each year from 1990 to 2002. To measure attention to race discrimination, I include a similar count of the number of articles in the ABI Inform database that include “race or racial” and “discrimination, bias, or harassment” in their titles or abstracts for each year.10 These are annual measures that gauge attention to sex and race discrimination in the business literature. I restrict media coverage to industry and trade publications—as opposed to relying on popular national media outlets—to 254—–AMERICAN SOCIOLOGICAL REVIEW 8 An additional measure of the extent of direct EEO enforcement would be a binary variable indicating whether EEOC investigators verified complainants’ charges. However, because the vast majority of charges verified by investigators (i.e., deemed to have “reasonable cause”) result in monetary settlements or mandated policy changes, the inclusion of this measure introduces multicollinearity among the charge outcome measures. I thus assess the impact of charges by measuring the monetary and policy scope of their resolutions, rather than the EEOC’s investigative decision. In additional analyses (available on request), I found a binary measure of verified claims to produce results similar to those of the monetary settlement and policy change measures. 9 These industry-level enforcement measures are based on my sample data; thus, there is sampling error associated with their measurement. Obtaining population data on charges at the industry level would entail manually matching all discrimination charges filed with the EEOC from 1990 to 2002 to workplaces that are charged, and then aggregating to the industry level. 10 I scanned all articles to ensure they dealt with sex and race discrimination in the context of employment. Delivered by Ingenta to : Cornell University Library Tue, 15 Sep 2009 21:42:14 isolate the political importance of sex and race discrimination to the business community. ORGANIZATIONAL FACTORS. Given the theoretical importance of organizational characteristics for predicting workplace change and potentially mediating the effect of EEO enforcement efforts, I include five measures of organizational conditions thought to affect occupational segregation. First, I measure establishment size as the total number of employees (fulland part-time). Because I expect establishment size to affect occupational segregation at a decelerating rate, I employ the log transformation. Second, to gauge the influence of female and minority leadership, I include binary variables to identify establishments with considerable representation of women and minorities among managers. Specifically, establishments in the top 20 percent of the sample with respect to the proportion of managers who are women or nonwhite are coded 1; all other are coded 0.11 Third, to account for the relative location of sex and race groups in the occupational hierarchy and how this might facilitate or constrain sex and race desegregation, I include measures of women’s and racial minority groups’ occupational distance from white males. To compute these measures, I f irst identif ied the core (modal) occupation of white males, women, blacks, Hispanics, and Asians in each establishment. I then assigned rankings to each core occupation based on the EEO-1 hierarchical occupational schema: officials and managers = 1, professionals = 2, technicians = 3, sales workers = 4, office and clerical workers = 5, craft workers = 6; operatives = 7, laborers = 8, and service workers = 9.12 Finally, I computed the vertical occupational distance between the core occupational level, k, of white males and each focal group, j, in each establishment, according to the following formula: –1 * (kwm – kj) where kwm equals the core occupational level of white males and kj equals the core occupational level of focal group j, where j = {women, Blacks, Hispanics, Asians}. The measure ranges from 8 to –8. Positive values indicate that the core occupation of white males is of higher ranking than that of the focal group; negative values indicate that the core occupation of white males is lower than that of the focal group. For instance, in a given establishment, if the core occupational group of white males is officials and managers, and the core occupational group of women is office and clerical workers, the occupational distance between females and white males is 4 levels: [–1 * (1 – 5)] = 4. For models predicting sex segregation, I include the measure of females’ occupational distance from white males; for models predicting race segregation, I include measures of blacks’, Hispanics’, and Asians’ occupational distance from white males. The inclusion of these measures allows me to account for the fact that employers can more easily integrate groups if they are connected by shorter job ladders. Next, because the degree of occupational differentiation within an establishment can affect establishments’ capacity to alter divisions of labor, I include a control for establishments’ occupational heterogeneity. EEO-1 reports ask employers to categorize workers into nine specific occupational categories, but establishments’employment across these categories can vary and thus impact segregation processes. For instance, an establishment with substantial employment across the nine occupational categories is likely to have a more segregated workforce than would an establishment that employs workers in only four categories, simply because the employer with nine categories can distribute workers across a larger number of occupations. Following Tomaskovic-Devey and DISCRIMINATION CHARGES AND WORKPLACE SEGREGATION—–255 11 Because levels of occupational segregation and the presence of female and minority managers are definitionally dependent (i.e., more female or minority managers will reduce occupational sex and race segregation), I use a binary variable to identify establishments in the top 20 percent in terms of female/minority management, rather than a simple measure of the proportion of women or racial minorities among managers. Using this binary variable allows me to better isolate the effect of female/minority leadership on occupational segregation. 12 It is important to note that these rankings capture the ease with which employers might move workers across occupations, rather than the social status of occupations. For instance, the ranking of office and clerical workers reflects the idea that it is easier to move an office worker than a craft worker into a sales position. This is not to suggest that office workers are of higher social status than craft workers. Delivered by Ingenta to : Cornell University Library Tue, 15 Sep 2009 21:42:14 colleagues (2006:575–76), I control for such variation in occupational differentiation within establishments using the Gibbs-Martin index of heterogeneity (Gibbs and Martin 1962): 1 – ( (P0e1–92))/ (T2) 100 where ( (P0e1–92)) is the sum of the total employment in each occupational category squared and T2 is the total establishment employment squared. To control for internal and external labor supply, I include measures of the proportion of female and nonwhite employees in an establishment and an industry, for sex and race models respectively, as well as a measure of the log of the total employment in an industry. To test for nonlinear effects of workplace composition, I also include the proportion of female and nonwhite employees in the establishment squared. Finally, I include a control for time to estimate establishments’ general trend toward desegregation. The measure is coded 1 through 13, representing the observation years 1990 to 2002. The inclusion of this measure controls for general trends in segregation and thus allows me to isolate the effects of EEO enforcement, legal environments, and organizational conditions on segregation, net of the general time trend. Table 1 presents the means, standard deviations, and data sources of all variables (for a correlation matrix of the independent variables, see Table S2 in the Online Supplement on the ASR Web site: http://www2.asanet.org/journals/asr/ 2009/toc068.html).13