Which hiring / selection methods actually predict job performance?

This is a rigorous SCIENCE-vein note, and it is frequently reported wrong online. The current best evidence is the peer-reviewed reanalysis by Sackett, Zhang, Berry & Lievens (2022) in the Journal of Applied Psychology, which corrected a systematic statistical error — overcorrection for range restriction — that had inflated validity estimates in earlier meta-analyses. After correction, the revised operational validity estimates re-rank the methods. The headline ranked figures from the paper’s revised table (Table 3) are: structured interviews .42; job-knowledge tests .40; empirically-keyed biodata .38; work-sample tests .33; cognitive-ability (GMA) tests .31; integrity tests .31; assessment centres .29; situational-judgment tests .26; contextualized conscientiousness .25; interests .24; unstructured interviews .19; overall conscientiousness .19; years of job experience .07. Sackett et al. explicitly note that this “suggests a reframing: Although Schmidt and Hunter (1998) positioned cognitive ability as the focal predictor, with others evaluated in terms of their incremental validity over cognitive ability, one might propose structured interviews as the focal predictor against which others are evaluated.”

The crucial correction to flag: content that parrots the old 1998 “GMA-on-top” ranking is outdated. Schmidt & Hunter (1998) reported cognitive ability at .51; Sackett et al. (2022) put it at .31 — in their words, “this is our estimate of cognitive ability test validity, in contrast with the .51 value presented by Schmidt and Hunter (1998).” Cognitive ability is no longer the single best standalone predictor and now ranks behind several job-specific methods. A separate 21st-century reanalysis (Griebe et al., 2022) suggests GMA’s validity may be lower still (~.23) in modern, service- and team-oriented jobs — a more recent and separate estimate, not the 2022 paper’s headline figure. The structured vs. unstructured interview gap is long-established: McDaniel et al. (1994) estimated structured interviews at roughly .31 vs. .23 for unstructured (overall uncorrected mean ~.26), and structured interviews consistently outperform unstructured ones across a dozen meta-analyses.

Two rigor caveats Newman should preserve. First, validity has meaningful spread: structured interviews’ .42 carries an 80% credibility interval of roughly .18 to .66 — better read as “about .42, plus or minus .24” — because a structured interview can be designed well or badly. Second, Sackett et al. (2023) caution the revised numbers are “not a mandate for preferring one predictor over another regardless of circumstances”; cost, time, subgroup/adverse-impact differences, and applicant reactions all factor in, and predictors are normally combined (composites raise validity above any single method). The practical, evidence-based takeaway for executive search and recruitment: anchor selection on a structured, job-analysis-based interview, add job-knowledge or work-sample assessment, and treat cognitive testing as one input among several rather than the deciding factor. Labelled verified, with the 2022 correction as the current state of the science.

Pillar anchor: Executive search + recruitment (selection science). Links down to the NUMBERS vein for cost-of-a-bad-hire.