Describe what a situation with equal validity between groups but unequal prediction would look like. What type of selection would this result, and what should the organization do about it?

Question

What will be an ideal response?

Answer 1

It is also possible to demonstrate equal validity in the two groups combined with unequal predictor means or criterion means and the presence or absence of adverse impact. In Figure 8.5, members of the minority group would not be as likely to be selected, even though the probability of success on the job for the two groups is essentially equal. Under these conditions, we can think of an alternative strategy involving the use of separate cut scores in each group based on predictor performance, while the expectancy of job performance success remains equal. Thus, a Hispanic candidate with a score of 65 on an interview may have a 75% chance of success on the job. A White candidate with a score of 75 might have the same 75% probability of success on the job. Although this situation might appear disturbing initially, remember that the predictor (e.g., a selection interview) is being used simply as a vehicle to forecast the likelihood of successful job performance. The primary focus is on job performance rather than on predictor performance. Even though interview scores may mean different things for different groups, as long as the expectancy of success on the job is equal for the two (or more) groups, the use of separate cut scores is justified. Indeed, the reporting of an expectancy score for each candidate is one recommendation made by a National Academy of Sciences panel with respect to the interpretation of scores on the General Aptitude Test Battery (Hartigan & Wigdor, 1989). A legal caveat exists, however. In the United States, it is illegal to use different selection rules for identifiable groups in some contexts (Sackett & Wilk, 1994). Figure 8.6 depicts a situation where, although there is no noticeable difference in predictor scores, nonminority group members tend to perform better on the job than minority group members (or vice versa). If predictions were based on the combined sample, the result would be a systematic underprediction for nonminorities and a systematic overprediction for minorities, although there is no adverse impact. Thus, in this situation, the failure to use different selection rules (which would yield more accurate prediction for both groups) may put minority group members in jobs where their probability of success is low and where their resulting performance only provides additional evidence that may help maintain prejudice (Bartlett & O’Leary, 1969). The nonminority individuals also suffer. If a test is used as a placement device, for example, since nonminority performance is systematically underpredicted, these individuals may well be placed in jobs that do not make the fullest use of their talents.