Direct Answer
A predictor-criterion relationship is the statistical link between a selection tool — a test, an interview, a reference check — and a measure of job performance. The strength of this relationship, expressed as a validity coefficient, tells you how well the tool actually predicts how someone will perform on the job. It is the single most important piece of evidence for deciding whether a hiring tool is worth using.
Why It Matters
Every hiring decision is a prediction. When a company uses a cognitive ability test, a structured interview, or a reference check, it is betting that scores on that tool will correspond to future performance. The predictor-criterion relationship is the evidence that determines whether that bet is justified.
Without this evidence, a hiring tool is just a guess dressed up as a process. With it, you can compare different tools, combine them intelligently, and build a selection system that meaningfully improves the quality of your hires.
The Science Behind It
The concept is grounded in what researchers call criterion-related validity — one of the primary forms of evidence for the predictive inference that is central to personnel selection (Van Iddekinge & Ployhart, 2008). As Van Iddekinge, Lievens, and Sackett (2023) explained, a “sign” strategy relies on statistical evidence that individuals with higher predictor scores subsequently perform better on a criterion of interest. The empirical relationship itself is the basis for supporting the predictive inference.
The strength of predictor-criterion relationships varies considerably across selection methods. Hermelin and Robertson (2001) applied standardized correction procedures across 20 meta-analytic validity coefficients for six selection methods and found that structured interviews and cognitive ability tests demonstrated the highest operational validity for predicting job performance. Cognitive ability tests show operational validities of approximately .40–.52 depending on job complexity (Roth et al., 2011), while personality measures such as Conscientiousness show uncorrected validities in the range of .15–.30 (Berry & Sackett, 2009).
Structured reference checks occupy a meaningful position in this landscape. Hedricks et al. (2013) found that structured, web-based multisource reference checks achieved criterion-related validity of r = .35 with supervisory performance ratings (p < .001, N = 223) — a level that rivals many established selection methods and exceeds unstructured references, which show validity of approximately ρ = .26 (Hunter & Hunter, 1984).
Crucially, different predictors can be combined to improve overall prediction. Because reference checks assess constructs that are partially distinct from what cognitive ability tests or personality questionnaires measure — specifically, observer-rated personality and actual workplace behavior — they can add incremental validity to a selection battery. The goal is not to find the single best predictor, but to assemble a set of complementary tools that together predict performance more accurately than any single tool alone (Van Iddekinge et al., 2023).
Common Misconceptions
A common error is assuming that a higher validity coefficient always means a better selection tool in every context. In practice, the most valid predictor may also produce the greatest adverse impact. For example, cognitive ability tests have among the highest validities but also the largest group differences (Landers et al., 2023). The predictor-criterion relationship must be evaluated alongside fairness, applicant reactions, and practical constraints — not in isolation.
How This Connects to Better Hiring
Understanding predictor-criterion relationships is what allows you to move beyond intuition toward evidence-based hiring. When you know that structured reference checks predict performance at r = .35, that observer-rated personality doubles the predictive power of self-reports, and that different tools capture different aspects of performance, you can design a hiring process grounded in evidence rather than tradition. Every other concept in this glossary — from criterion-related validity to content validity — feeds into this central idea: the strength of the link between what you measure and what you are trying to predict.