Sunday, December 06, 2009
Setting cutoff scores on personality tests
What's the best way to set a cutoff score for a personality test, knowing that some candidates inflate their score? It all depends on your goal. Are you trying to maximize validity or minimize the impact of inflation?
According to a research study by Berry & Sackett published in the Winter '09 issue of Personnel Psychology, if your goal is to maximize validity, your best bet is to wait until applicants have taken the exam, then set your cut-score (e.g., the top two-thirds); this was particularly true when selection ratios are small (i.e., organization is very selective).
If your goal is to minimize the number of deserving applicants who are displaced by "fakers", you're better off establishing the cut point ahead of time, by using a non-applicant derived sample (e.g., job incumbents, research group). The results were generated using a Monte Carlo simulation.
Interestingly, the authors also replicated the work of other researchers who have shown that the impact of faking on the criterion-related validity of personality measures is relatively low. There are a few other very good points made in this article:
- Expert judgment methods of establishing pass points (e.g., Angoff method) may be difficult to use for personality tests since experts may find it difficult to judge individual items. Methods used to select a certain number of applicants or methods based on a criterion-related validity study (both used as variables in this study) are more appropriate for personality tests.
- There is no consensus of how prevalent faking on personality exams is; estimates range from 5-71%. It likely depends on the situation and how motivated test takers are to engage in impression management.
- Some recommend setting a very low cutoff score for personality tests, which would exclude only those likely not suitable for the position (and not faking), while others prefer a more stringent cutoff to maximize utility.
- A reasonable range of d-values for score inflation on personality inventories is .5-1.0 (used in this study).
- There exists very little research on the skewness of faking score increases. A positively-skewed distribution (meaning most people faked a small amount) was used in this study. (I would think this would also vary on the situation)
So bottom line: where--and how--you set your cutoff score on personality inventories depends on whether you want to maximize the predictive validity or minimize the number of deserving applicants that get left out of the process.
Other good reads in this issue:
- Police officer applicants reactions to promotional assessment methods
- The impact of diversity climate on retail store sales
- The construct validity of multisource performance ratings
- Labor market influences on CEO compensation