In 2014, the American Psychological Association published research from five renowned academicians who extracted 147,328 scientific results in the form of effect sizes (correlation coefficients) from the Journal of Applied Psychology and Personnel Psychology published during a 30-year period beginning in 1980 and ending in 2010.[1]
One of the purposes of the study was to provide the scientific community and the world with a more accurate understanding of what small, medium, and large effect sizes (correlations symbolized by the letter “r”) are in applied Psychology.
They found that J. Cohen’s expectations (referred to here as benchmarks) published in 1988 regarding the typical magnitude of validity results[2] are substantially inflated. This means that referring to Cohen’s correlational benchmarks, many normal study results were judged incorrectly too small to be appreciated by the scientific community.
Frank A. Bosco, Herman Aguinis, Kulraj Singh, James G. Field, and Charles A. Pierce determined that the median uncorrected correlational effect size of all the 1,660 studies explained between two (2) and three (3) percent of variability in their dependent variable. A typical 50th percentile effect size for an assessment that measures knowledge, skills, and/or abilities (KSAs) explains 6.5% variability in performance. An assessment measuring psychological characteristic (such as personality traits) explains 4.1% of variability in performance.
4.1% and 6.5% might not seem like noteworthy magnitudes, but if the assessment correlates with general performance in a mission-critical sales or service role, the utility can equate to tens of thousands to millions of dollars of additional revenue thanks to increased quality of hire.
Another finding of this research summarizing 30 years of published science is that KSAs are generally a better predictor of performance than psychological characteristics. Ree and Earles wrote in 1992[3] that Intelligence is the best predictor of job performance.
Finally, success in explaining variance in performance is roughly double that of predicting employee movement (for example, turnover). Predicting turnover requires larger sample sizes. Thirty years of research confirms the value of psychometric assessments to predict performance, although typical results are modest.
[1] Download the paper here: http://www.hermanaguinis.com/JAP2015.pdf
[2] Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, Jn: Erlbaum.
[3] Ree, M.J. & Earles, J.A. (1992). Intelligence is the best predictor of job performance. Current Directions in Psychological Science, 1, 86-89. Doi:10.1111/1467-8721.ep10768746
