Page 307 - Discrimination at Work The Psychological and Organizational Bases

STONE-ROMERO
274
THE CRITERION-RELATED VALIDITY
OF PERSONALITY MEASURES
There is yet another important reason to eschew the use of many types
of personality measures for selection and many other personnel decision-
making purposes. More specifically, the extant evidence (e.g., Barrick &
Mount, 1991; Tett et al., 1991) shows that they explain only trivial amounts of
variance in various criterion measures. For example, Barrick and Mount's
(1991) meta-analysis revealed that measures of emotional stability had an
average observed correlation of .05 with a number of performance mea­
sures. A subsequent meta-analysis by Tett et al. (1991) was alleged to be
superior to that of Barrick and Mount (1991) because it dealt with theory-
relevant (confirmatory) predictions between personality measures and var­
ious outcomes. It showed an average correlation of -.15 between neuroti­
cism and various criteria. Finally, a recent meta-analysis by Bobko, Roth,
and Potosky (1999) showed that measures of conscientiousness had an
average validity of only .18 with job performance.
Taken together, the just-noted meta-analytic studies paint a very dismal
picture of the value of using measures of NA or other dimensions of per­
sonality for personnel selection purposes. Thus, selection should be based
upon predictors that have lesser potential for unfair discrimination than do
measures of personality. One alternative is work samples. Research shows
that they have much higher criterion-related validity than do personality
measures (Hunter & Hunter, 1984). In addition, they often have lower lev­
els of adverse impact than do traditional preemployment tests (Schmitt&
Mills, 2001). However, it is important to note that in some circumstances,
work samples may have fairly high levels of adverse impact. For example,
this would be true in situations in which they called for applicants to ex­
hibit technical skills and knowledge that were not critical to job success at
the time of hire, because all persons hired would receive the training needed
to develop the same skills and abilities.

CONCLUSIONS CONCERNING THE USE OF PERSONALITY
PREDICTORS OF VARIOUS CRITERIA
As the foregoing clearly shows, there are serious construct validity prob­
lems with many measures of so-called personality traits (e.g., emotional
stability, NA, extra version). In addition, the use of such measures for selec­
tion purposes appears to open the door to indirect forms of unfair discrimi­
nation against both minority group members and majority group members
who have scores on personality measures that differ from organizational