equation, the multiple R increases to .70. This value is similar to the final phi
coefficient of .66 found when using the multiple hurdle selection model.
The correlation coefficients for the two selection models were compared to
illustrate that the benefits associated with multiple screening procedures generalize to compensatory selection models. However, it should be clear that the
type of selection model used does not depend on these statistics. Compensatory
selection models assume that low scores on one predictor can be offset by high
scores on another. If this assumption is untenable, other selection models, such
as a multiple hurdle approach, must be used.
Bayes's theorem was used as an alternative approach to estimating the false-
positive rate for tests of employee theft and counterproductivity. Examples were
provided to illustrate how follow-up screening procedures can reduce either false-
positive or false-negative decision errors, as well as the total number of incorrect
decisions. A review of the relevant literature appears to indicate that continued
research in the area of integrity testing can provide organizations with alternative
methods of identifying dishonest applicants.
The author would like to thank Bruce Fisher, Jack Jones, Maureen McConnell, and Bob Vance for their helpful comments during the preparation of this manuscript.
In this context, positive or negative refers to the outcome of the test rather than to
an accept-reject selection decision. A positive outcome indicates that a test has classified
an applicant as dishonest (i.e., at risk for dishonest behaviors at work given the need and
opportunity). Honest applicants who are incorrectly identified as dishonest constitute false-
positive decision errors.
In practice, it is important to make a distinction between the false-positive rate and
labeling applicants as dishonest. Test administrators are trained to understand that all
psychological tests produce errors in prediction (i.e., false positives and false negatives)
and that no applicant should be labeled as dishonest.
In this context, an independent screening procedure refers to an alternative measure
that is equally effective in identifying honest and dishonest applicants. This does not
imply that the results produced by the two screenings are independent or uncorrelated.
Many honest and dishonest applicants would be correctly classified by both measures.
For this example, the two screening procedures are correlated .33.
Although this validity coefficient may appear relatively high, it was computed using
an estimate of the base rate obtained directly from base rate studies, which is expected
to be higher and more accurate than would be realized in actual criterion-related validity
studies. In addition, the meta-analysis conducted by McDaniel and
Jones ( 1988) revealed
a mean validity coefficient of .50.