Gc. Thornton et al., Higher cost, lower validity and higher utility: Comparing the utilities oftwo tests that differ in validity, costs and selectivity, INT J SEL A, 8(2), 2000, pp. 61-75
Traditional approaches to comparing the utility of two tests have not syste
matically considered the effects of different levels of selectivity that ar
e feasible and appropriate in various selection situations. For example, em
ployers who hope to avoid adverse impact often find they can be more select
ive with some tests than with others. We conducted two studies to compare t
he utilities of two tests that differ in costs, validity, and feasible leve
ls of selectivity which can be employed. First, an analytical solution was
conducted starting with a standard formula for utility. This analysis showe
d that for both fixed and variable hiring costs, a higher-cost, lower-valid
ity procedure can have higher utility than a lower-cost, higher-validity pr
ocedure when the selection ratios permissible using the two procedures are
sufficiently (yet realistically) different. Second, using a computer simula
tion method, several combinations of the critical variables were varied sys
tematically to detect the limits of this effect in a finite set of specific
selection situations. The results showed that the existence of more severe
levels of adverse impact greatly reduced the utility of a written test wit
h relatively high validity and low cost in comparison with an assessment ce
nter with lower validity and higher cost. Both studies showed that the cons
ideration of selectivity can yield surprising conclusions about the compara
tive utility of two tests. Even if one test has lower validity and higher c
ost than a second test, the first may yield higher utility if it allows the
organization to exercise stricter levels of selectivity.