Screening Discrimination in a Broader Context

• Date: 01 October 2016
• DOI: http://doi.org/10.1002/isaf.1388
• Author: James Fain
• Published date: 01 October 2016

SCREENING DISCRIMINATION IN A BROADER CONTEXT

JAMES FAIN*

Oklahoma State University, Stillwater, OK,USA

SUMMARY

I employ simulations to investigate the impact of screening discrimination, which addresses discrimination in

hiring. I extend a well-known model of screening discrimination by including minority firms, wages, worker pref-

erences and competition among the firms for workers. The Gale–Shapley algorithm is used to find a stable

matching of heterogeneous workers to heterogeneous firms. Screening discrimination gives majority applicants

an advantage in the hiring process, but this advantage is reduced by the presence of minority firms. In this broader

context, screening discrimination produces segregated firms but has little impact on median wages or employment

probabilities. Copyright © 2016 John Wiley & Sons, Ltd.

Keywords: Gale-Shapley; labour economics; screening discrimination in a broader context; screening discrimina-

tion; simulations

1. INTRODUCTION

Cornell and Welch (1996) introduced the concept of screeningdiscrimination. In the traditionof statistical

discrimination,screening discrimination offersan explanation of differential labourmarket outcomes that

is based on information asymmetries, rather than a dislike for a certain group. Screening discrimination

focuses on the hiring decision. In Cornell and Welch’smodel,firms are more able to discern quality

differences in applicants of their own race or type than they are able to discern quality differences in

applicants of different races or types. This greater ability to discern quality is modelled asfirms receiving

a greater number of signals from applicants of their same race. This gives same-race applicantsa distinct

advantage in the hiring process, as the signal-receiving differential means that a same-race worker is

almost always the highest ranked worker. The fact that a same-race applicant is almost always the lowest

ranked applicantas well is irrelevant,as the firms are assumed to hire the highestranked applicant. Cornell

and Welch demonstratethat differential hiring outcomescan persist for a long time: if interviewersare very

likely to hire people of the same race, then this result is likely to repeat itself in the next iteration.

This study has spawned a substantial literature, which I review below. However, it suffers from

several deficits. First, in Cornell and Welch’s main model there is only one firm, so there is no compe-

tition among firms for workers. The single firm simply evaluates its candidates and makes a job offer to

the best candidate, who accepts with a probability of one. The notion that firms are always able to hire

their most preferred applicant is not realistic. It is fairly common for a firm to make an offer to a certain

worker, only to be outbid by another firm. The fact that a firm’s top-ranked applicant is almost always

of the same type as the firm is less relevant when firms cannot automatically hire that applicant.

* Correspondence to: James Fain, Economics and Legal Studies in Business, Oklahoma State University, Stillwater, OK 74078,

USA. E-mail: jim.fain@okstate.edu

Copyright © 2016 John Wiley & Sons, Ltd.

INTELLIGENT SYSTEMS IN ACCOUNTING, FINANCE AND MANAGEMENT

Intell. Sys. Acc. Fin. Mgmt. 23, 276–294 (2016)

Published online 9 February 2016 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/isaf.1388

A second deficit is that the model explicitly considers only majority firms; the formal model does not

admit the possibility of minority firms that will almost certainly hire minority workers because they too

can better discern quality differences in individuals of their same race. Cornell and Welch recognize this

limitation and offer the following statement relatively late in the paper: ‘Also, we would expect to see

not only majority interviewers discriminate against minority applicants but also minority interviewers

discriminate against majority applicants.’(Cornell & Welch, 1996: 562). The lack of minority firms

in the formal model skews the analysis. In a fuller model the labour market impact of majority firms

hiring mostly majority workers would be mitigated by minority firms hiring mostly minority workers.

While there certainly would be segregation, it does not follow that there would necessarily be signifi-

cant differences in incomes and employment opportunities.

A third deficit is that Cornell and Welch assume that the only information available to firms is a

difficult-to-read signal. This assumption produces problems when that signal plays a prominent role in

the hiring process. It is arguablymore realistic to assume that every employercan read certain items, such

as years of education and years of experience,from a resumé without error. This suggests that employers

have some important information about job applicants that is straightforward and easy to process. More

subtle signals undoubtedly exist as well. While the subtle signals may be important at the margin as the

firm decides between two candidates, they are not the only information the firm has about candidates.

In this paper I extend Cornell and Welch’s analysis to a broader context. This broader context was

suggested by the authors in their fourth footnote, in which they speculate that their model could be

embedded in a full two-sided matching model. In this paper I perform this embedding, with a few

alterations to Cornell and Welch’s model. Applicants have some signals that all firms are able to

read correctly and other signals that are more difficult to read. I allow both majority and minority

firms to exist, with the assumption that both groups are more adept at reading the more subtle sig-

nals in applicants of their own type than they are at reading the more subtle signals in applicants of

the other type. Firms evaluate each applicant, and different firms may reach a different estimate of an

applicant’s productivity. Firms post wage offers and screen out workers who do not meet their min-

imum productivity standards. A two-sided matching model matches workers to firms; this process

explicitly allows for competition among firms for workers. Applicants may receive multiple job

offers, and they are assumed to accept the offer with the highest wage. Owing to the competition

for workers, firms are not automatically able to hire their highest ranked applicant. As a result,

the fact that the highest ranked applicant is likely to be of the same type as the firm is mitigated.

In this context I explore the degree to which screening discrimination affects hiring decisions. I also

consider the impact of screening discrimination on wages and employment. I find that screening

discrimination produces firm-level segregation of the workers. If majority firms are better able to screen

majority workers and minority firms are better able to screen minority workers, then one should observe

segregation by firms as the two types of firms utilize their respective advantages. I find that the degree

of segregation increases as the information advantage becomes more pronounced. I also find screening

discrimination has little to no impact on minority employment prospects or median wages. The latter

results are important because they suggest that screening discrimination alone is not able to explain

some important labour market outcomes.

The outline of the paper is as follows. Section 2 contains a review the relevant literature regarding

screening discrimination and Section 3 provides details about the relationship between signal quality

and inferred ability. In Section 4 I review the concept of stable matchings, while Section 5 discusses

screening workers when some signals are hard to read. In Section 6 I explain how I construct the

simulations. Section 7 ties my results to some of Cornell and Welch’s results, and Section 8 presents

the main simulation results.

Screening Discrimination in a Broader Context 277

Copyright © 2016 John Wiley & Sons, Ltd. Intell. Sys. Acc. Fin. Mgmt., 23, 276–294 (2016)

DOI: 10.1002/isaf