Page 346 - Discrimination at Work The Psychological and Organizational Bases

313
13. HRM PRACTICES
multiple-choice tests—instead of the construct. The mechanism of these
effects is via processes of test perception, namely differences in test anxiety
and stereotype threat, perceived fairness, perceived low face validity or
job relatedness, and lack of motivation (Chan & Schmitt, 1997; Chan et al.,
1997; Hough et al., 2001; McKay & Doverspike, 2001; Ryan, 2001; Steele,
1997). To reduce adverse impact, alternative test formats have been used,
including performance tests (e.g., Chan & Schmitt, 1997), assessment cen­
ters (e.g., Goldstein et al., 1998), video-based tests with an oral presentation
of test items only (e.g., Chan & Schmitt, 1997; Weekley & Jones, 1999), oral
presentation of both items and responses (e.g., Rand, 1987; Schmitt & Mills,
2001), and constructed response tests (e.g., Arthur et al., 2002). In addition
to engendering more favorable attitudes and perceptions, the use of these
nonmultiple-choice test formats also minimizes the nonjob-related reading
demands (Arthur et al., 2002). At the present time, alternative formats do
appear to be a practical method of reducing adverse impact. However, em­
pirical research delineating the psychological mechanisms responsible for
these reductions is needed. Furthermore, in many real-world situations,
alternative formats may not be practical.
In spite of its potential advantages, a major limitation of the method-
change approach has been the confounding of method (techniques or pro­
cedures) and content (constructs, Arthur et al., 2002; Schmitt et al., 1996).
For instance, if the levels of adverse impact for performance tests and
paper-and-pencil multiple-choice tests are to be compared, then to obtain
interpretable results, both test formats should be measuring the same con­
struct or content. Another limitation with the preceding body of research
is that there is a glaring absence of criterion-related validity data. Thus, it
is still unclear whether the reductions in subgroup differences associated
with alternative methods are associated with lower or higher criterion-
related validity (cf. Edwards & Arthur, 2004).
A variation of the method-change approach is to use design techniques
to minimize the inherent subjectivity and potential for bias and discrim­
inatory behaviors associated with some test methods such as interviews.
Specifically, in the case of interviews, which are one of the most widely
used selection tools (Dipboye, 1997), they could be so highly structured as
to completely eliminate any discretionary input on the part of the inter­
viewer and consequently also eliminate the potential for subjective bias on
the part of the interviewer or rater. This level of structure is what Huffcutt
and Arthur (1994) designate as Level 4 structure in which all candidates
are asked the exact same questions in the exact same order with no choice
or follow-up, and each individual response to each question is evaluated
according to preestablished answers. At this level, the interview function­
ally operates like an orally administered and scored objective test. This