Page 344 - Discrimination at Work The Psychological and Organizational Bases

13. HRM PRACTICES
311
"internal item bias analysis methods." However, because of the tendency
of the public to associate bias with discrimination, such techniques are
now often referred to as methods for estimating measurement equivalence
or "differential item functioning (DIF)." Common methods (Crocker &
Algina, 1986; Raju, Laffitte, & Byrne, 2002) for investigating DIF include
(a) differences in means after controlling for the total score (the ANOVA
technique), (b) differences in item-total correlations, (c) differences in item
characteristic curves as identified through item response theory, (d) differ­
ences in the proportion of correct responses after controlling for the total
score (chi-square methods including the Mantel-Haenszel technique), and
(e) differences in factor structures as identified through exploratory or con­
firmatory factor analysis. To date, there have been a number of problems
with the application of internal bias analysis techniques. First, the use of
different methods may not always lead to the identification of the same
items as biased. Second, it is often difficult to determine why a particu­
lar item is identified as biased. Third, given the limited number of items
identified as biased on current tests, the removal of biased items has only
a small effect, if any, on subgroup differences.
Test Taking Motivation The focus on test taking motivation is based on
the premise that minorities' less favorable test perceptions such as per­
ceived unfairness, perceived low face validity, perceived low job related­
ness, test anxiety, and stereotype threat (Arvey, Strickland, Drauden, &
Martin, 1990; Chan & Schmitt, 1997; Chan, Schmitt, DeShon, Clause, &
Delbridge, 1997; Edwards & Arthur, 2004; Hough et al, 2001; McKay
& Doverspike, 2001; Ryan, 2001; Steele, 1997) translates into lower moti­
vation, which subsequently partially accounts for the observed subgroup
differences. Research such as that cited above tends to support the view
that differences in test-taking motivation can lead to test score differences.
Although the amount of variance explained is typically small, these dif­
ferences can translate into important consequences, especially when the
selection ratios are very low.
The possibility that differences in test-taking motivation explains
minority-majority test score differences leads to the practical possibility
that changes in test-taking motivation could reduce adverse impact. At­
tempts to address these perceptual and motivational issues have been made
through the use of various types of training programs including test-taking
skills training, which should (a) encourage an overall positive attitude to­
ward the testing process; (b) include some type of training on handling
test anxiety (Goldstein, Yusko, Braverman, Smith, & Chung, 1998); and (c)
attempt to increase participant motivation.
Altering the Selection Criteria Altering the selection criteria in an attempt
to reduce subgroup differences and adverse impact has had several foci