What about the worst predictors of job performance? Studies show that measures which are irrelevant to a job are worst at predicting performance. These measures range from candidate interests, with a predictive validity of 0.10, to candidate graphology (handwriting) with a predictive validity of only 0.02. The nature of predictive validity in psychometric tests is such that predictive power increases exponentially with correlation. This means that tests able to achieve correlations of 0.5 or greater, such as a general mental ability test, are highly predictive. A predictor with a correlation of 0.5 can explain 25% of the variance between candidates. Tests with correlations as low as 0.10, on the other hand, explain less than 2% of the variance between candidates, indicating that they are very poor predictors of performance.
Want to learn more about candidate assessments? Contact Ken Kanara
, our Managing Director, for a free Strategic Needs Assessment.