Sears Case Data Illustrations
(Oct. 10, 2012; rev. Dec. 12, 2013)
My first recognition of the pattern whereby the rarer an outcome the greater tends to be the relative difference in experiencing it and the smaller tends to be the relative difference in avoiding arose it as a result an appreciation of the way that a group that on average has weaker qualifications with respect to some outcome is increasingly represented among each segment of the population with increasingly weaker qualifications during the prosecution of EEOC v. Sears, Roebuck and Co. See the Sears Case page on this site regarding a number of matters about the case.
In the case, in which the EEOC alleged that Sears discriminatorily excluded women from commission sales position, there was considerable evidence that female sales applicants had less sales experience and fewer other characteristics related to commission sales than male sales applicants, something that was partly due to the fact that a substantially higher proportion of the male sales applicants than female sales applicants applied specifically for commission sales positions.[i] The evidence that female sales applicants were on average less qualified than male sale applicants was important to the positions of both sides. It was important to the defense to support its contentions that the EEOC’s efforts to adjust for differences in characteristics of male and female sales applicants (in accord with the discussion on the Underadjustment Issues sub-page of the Lending Disparities page of this site). It was important to the EEOC because it refuted the contention of Sears that the weaker qualifications of female commission sales hires compared with male commission sales hires indicated that, rather than discriminating against women, Sears was favoring women, That is, such data showed that because women had weaker qualifications among applicants they would tend to have weaker qualifications among hires even if they were discriminated against (as I illustrate with data based on the Sears case in Table IV of Illusions of Job Segregation (Public Interest, Fall 1988) and was illustrated in the case with a variety of exhibits including the EEOC’s Wise Exhibit 1[ii]).[iii]
Table 1 below presents the part of that exhibit showing how women comprised increasingly smaller proportions of the applicants at increasingly higher experience level.
Table 1. Female Proportion of Applicants by Level of Experience from Plaintiff’s Exhibit Wise 1
|
| Experience |
Fem%App |
| 0 |
0.685 |
| 1 |
0.587 |
| 2 |
0.416 |
| 3 or more |
0.224 |
It was shortly after the trial that I noticed an article in the Spring 1984 issue of Dissent titled “The Feminization of Poverty” by Barbara Ehrenreich and Frances Fox Piven. The article first brought to my attention the so-called “feminization of poverty,” the pattern whereby, during a period of substantial decreases in poverty, the proportion female-headed families comprised of the poor increasing substantially. Because of the understanding of patterns like that in EEOC’s Wise Exhibit 1, it was evident to me that, since women would comprise a larger proportion of the population substantially below the poverty line than they comprised of the proportion below the poverty line, decreases in poverty naturally would cause female-headed families to comprise a larger proportion of the poor than they previously did.
Review of the actual data on proportions demographic groups comprise of persons falling below various ratios of the poverty (such as that underlying the table and figures in “Can We Actually Measure Health Disparities?” (Chance, Spring 2006)) confirmed what I had assumed to be the case. But such review also caused me to recognize that reducing poverty, while increasing the relative difference in poverty rates (between female-headed families and either (a) the remainder of the population or (b) some more advantage segment of the population like married-couple families), would decrease the relative difference in rates of avoiding poverty. I then first described that pattern in “’Feminization of Poverty’ is Misunderstood” (Plain Dealer, Nov 11, 1987, reprinted in Current, May 1988, and Annual Editions: Social Problems 1988/89, 1988).
That pattern cannot be divined from the information in Table 1 itself. Because the table does not show the proportion of total applicants in each category, one cannot derive the proportion of female applicants falling within each category. Such data were available in the trial, however. If the data are among materials from the record in my possession, I may expand this item and the make this illustration similar to that in the NHANES Illustrations, Life Tables Illustrations, Income Illustrations, and Credit Score Illustrations subpages.
[i] A difficulty in the analysis of the patterns in the case involved that fact that the application contained a box for indicating interest in a sales job without distinguishing commission sales from noncommission sales, though it did allow a person to write in the type of sales desired. Only a small proportion of applicants indicated a preference but male sales applicants comprised a higher proportion of those who indicated a preference for commission sales than of sales applicants who failed to indicate an interest in commission sales. Applicant who indicated such an interest had a much higher chance of being hired for commission sales than applicants who did not (though higher selection rates associated with indication of an interest in commission sales may have been partly a function of its correlation with characteristics related to selection for commission sales rather an a function of the indication of interest itself).
[ii] The link includes Plaintiff’s Exhibits Wise 1 to 3. The third exhibit is based on the assumption that the larger relative difference in selection rates among less qualified persons reflected greater discrimination among the less qualified applicants. I believe that the data were presented like this for purposes of verisimilitude in the illustration of the way the weaker qualifications of hires from the less qualified group did not indicate that the group was not discriminated against, rather than to make a point about the implications of a greater seeming disparity among persons with weaker qualifications. But for reasons discussed in note 30 and at the end of Section E.2.b. of the Harvard University Measurement Letter, and in Section B of the University of Kansas School of Law faculty Workshop paper “The Mismeasure of Discrimination,” data like that in Plaintiff’s Wise Exhibit 3, while showing larger relative gender differences in selection among less qualified applicants than among more qualified applicants, would not support the claim that there was stronger evidence of bias among the less qualified applicants than among the more qualified applicants. That is, one commonly observes smaller relative differences in selection, but larger relative differences in rejection, among more qualified applications than less qualified applicants simply because selection rates are generally higher among more qualified than less qualified applicants. As discussed in Section C of the Kansas paper, one would need to know the actual selection rates even to effectively determine whether disparities were larger among less qualified than more qualified applicants. That would still leave open the question of whether a larger disparity among less qualified applicants is probative of anything absent reason to believe that differences in qualifications are the same among less qualified applicants as among more qualified applicants. See the Disparities – High Income subpage of the Lending Disparities page.