Database Reference
In-Depth Information
(Balsa et al., 2005; Mcguire et al., 2008) or whether “expert” patients obtain a more
favorable treatment (Grytten et al., 2011). Another strand of statistical discrimina-
tion models studies how negative rational stereotypes of employers differentiates
firms' hirings and wages, and workers' investments, e.g., in education. (Lang &
Lehmann, 2011) call this class as
rational stereotyping
models. Finally, we refer to
the survey (Fang & Moro, 2010) for a theoretical discussion of models of statistical
discrimination and affirmative actions.
6.4
(Quasi-)Experimental Perspective
A recurring problem in discrimination analysis is the collection of controlled data,
as opposed to observational data, for which the results of analytical and statistical
techniques can be interpreted without any concern for external or confounding fac-
tors. This has been tackled through quasi-experimental and experimental methods,
that we review in the next two subsections.
6.4.1
Auditing
Auditing, also known as
field experiments
, follows a quasi-experimental approach
to investigate for the presence of discrimination by controlling the factors that may
influence decision outcomes. The basic idea consists of using pairs of
testers
(also
called
auditors
), who have been matched to be similar on all characteristics that
may influence the outcome except race, gender, or other grounds of possible dis-
crimination. The tester pairs are then sent into one or more situations in which
discrimination is suspected, e.g., to rent an apartment or to apply for a job, and
the decision outcome is recorded. The difference in the outcomes among the paired
groups provides then a measure of discrimination. A summary of recent audit stud-
ies in employment discrimination is due to (Pager, 2007). (Riach & Rich, 2002)
review and compare the statistical significance of field experiments on racial, sex,
and disability discrimination in employment, and on discrimination on housing sale
and rental. Criticism of the conclusions drawn from audit methods is discussed in
(Heckman & Siegelman, 1993) and (Heckman, 1998), while (Riach & Rich, 2004)
comment on ethical implications of such methods. (Quillian, 2006) discusses how
the measurement of discrimination through audit methods should incorporate recent
advancement in psychological theories of prejudice.
We categorize three different approaches in detecting discrimination by auditing.
Situation testing
occurs when the testers come in contact with the decision maker.
This is the case, for instance, of job interviews involving human testers, who are se-
lected and trained in advance to act similar each other (Bendick et al., 2010; Moreno
et al., 2004; Pager & Quillian, 2005; Pager et al., 2009; Turner & Ross, 2005; Turner
et al., 2002). A strong point in favor of situation testing is that testers can record the
cause of discrimination, such as prejudice or stereotypes, hence allowing for a causal
analysis of the discrimination cases. A limitation of situation testing is that the phase
of data collection is expensive. In addition, situation testing cannot be applied at all
Search WWH ::
Custom Search