Personnel selection is the process of choosing individuals for employment based on their predicted job performance. It involves identifying relevant predictors (tests, interviews, work samples), measuring candidates on those predictors, and making hiring decisions using some combination rule. Effective selection requires that predictors are both valid (actually related to job performance) and legally defensible (based on job-relevant criteria established through job analysis). The core challenge is prediction under uncertainty: using limited information gathered before hiring to forecast performance that will unfold over months and years.
Personnel selection sits at the intersection of measurement science and practical decision-making. The foundational idea is simple: organizations want to hire people who will perform well, and they use various assessment tools to predict who those people will be. But the execution is anything but simple, because prediction of human behavior is inherently uncertain and the consequences of poor prediction — bad hires, legal liability, wasted training resources — are substantial.
The predictor-criterion framework organizes all of selection psychology. Predictors are the assessments administered to candidates: cognitive ability tests, personality inventories, structured interviews, work samples, assessment centers, biographical data, and more. Criteria are the outcomes the organization cares about: job performance ratings, sales figures, tenure, counterproductive behavior, and other indicators of success. The validation question is whether predictor scores actually predict criterion scores — and how well.
Different predictors have different validity profiles. Meta-analyses by Schmidt and Hunter (1998) and subsequent updates have established a rough hierarchy: general cognitive ability (GCA) is the single strongest predictor of job performance across nearly all jobs, with validity coefficients around .50-.65 for complex jobs. Structured interviews, work sample tests, and integrity tests also show strong validity. Unstructured interviews, years of experience, and graphology perform much worse. But validity is only one consideration — organizations also care about applicant reactions, adverse impact (differential selection rates across demographic groups), cost, and legal defensibility.
A critical concept is incremental validity: the extent to which a new predictor improves prediction beyond what existing predictors already provide. Because many predictors are correlated with each other (cognitive ability correlates with work sample performance, for instance), adding a second predictor does not simply add its validity to the first. The gain depends on how much unique variance it captures. This is why optimal selection batteries typically combine predictors that tap different constructs — for example, cognitive ability plus conscientiousness plus a structured interview — rather than stacking similar measures.
Selection decisions also require a combination rule for integrating information across multiple predictors. Compensatory models (like multiple regression) allow high scores on one predictor to compensate for low scores on another. Non-compensatory models (like multiple hurdles) set minimum cutoffs on each predictor — failing any one is disqualifying regardless of other scores. The choice between these approaches depends on the job: a firefighter must meet minimum physical requirements regardless of cognitive ability, suggesting a multiple-hurdle approach, while many office jobs allow compensation across dimensions.