What is the Predictive Validity?

Predictive validity is how well a selection method's scores forecast later job performance, usually measured as the correlation between the two. For interviews, research puts structured interviews among the strongest predictors and unstructured ones well behind, so consistent questions and scoring beat instinct.

By Lee Flanagan

20th Nov. 2026  |  Last Updated: 20th Nov. 2026

Extended definition

Predictive validity answers a simple question: do the people who score well in your hiring process go on to do the job well? Researchers check it by comparing selection scores with a later measure of performance, often supervisor ratings. The result is a validity coefficient, a correlation where zero means the method tells you nothing about future performance and values closer to one mean stronger prediction.

No hiring method predicts perfectly, because performance also depends on training, management, the team and plenty of luck. So the useful comparison is which methods predict better than others, and by how much.

Predictive validity is different from face validity, which is whether a method looks relevant to candidates and interviewers, and from reliability, which is whether it gives consistent results. A method can feel convincing and still predict poorly, and an unreliable method can’t predict well.

How interview methods compare on predictive validity

The best-known summary is Frank Schmidt and John Hunter’s 1998 review of 85 years of selection research. A major update by Paul Sackett and colleagues, published in 2022, found that earlier work had overcorrected for range restriction (the fact that only hired people have performance data) and produced lower estimates for most methods. Their figures for common methods, earlier estimate first:

  • Structured interviews: .51, now .42, the highest mean validity in the update.
  • Job knowledge tests: .48, now .40.
  • Work sample tests: .54, now .33.
  • General mental ability tests: .51, now .31.
  • Unstructured interviews: .38, now .19.
  • Years of job experience: .18, now .07.

Two points stand out. Structured interviews came out on top, and unstructured interviews fell a long way, so the update puts structured interviews at more than twice the validity of unstructured ones. And these are averages: the researchers found considerable variability across settings for every method, so your own process can do better or worse.

Why predictive validity matters

It tells you which parts of your process earn their time. An interview that doesn’t predict performance is a cost with no return, and hiring managers can’t judge their own interviews’ validity from experience, because they never see how the people they rejected would have performed.

The research points to a practical answer. The methods at the top of the updated list are specific to individual jobs, such as structured interviews, job knowledge tests and work samples. For interviews, that means:

  • Base questions on what the job requires, not on favorite questions.
  • Ask every candidate the same core questions.
  • Score each answer against an interview scorecard with anchored examples.
  • Have interviewers rate independently before they discuss.
  • Compare interview scores with performance after hire, so you can check your own validity over time.

Common mistakes and misconceptions about predictive validity

  • Quoting the 1998 numbers as current — The update lowered most estimates, by around .20 for cognitive ability and .19 for unstructured interviews.
  • Assuming cognitive ability tests are the best predictor — That was the headline of the earlier work. In the update, cognitive ability is no longer the stand-out predictor.
  • Treating years of experience as a strong signal — It sat near the bottom in both reviews.
  • Confusing confidence with validity — Interviewers often feel sure after an unstructured conversation. Feeling sure isn’t evidence that the method predicts.
  • Assuming a method’s average applies to you — A structured interview only predicts well if the questions match the job and the scoring is consistent.

Train this skill: Browse the course library →

Frequently asked questions

What is the predictive validity of structured interviews?

In Schmidt and Hunter's widely cited 1998 review, structured interviews had a validity of .51. A 2022 update by Sackett and colleagues, which corrected for overcorrection in the earlier work, put the figure at .42, the highest mean validity of any method in their comparison. Unstructured interviews fell to .19 in the same update, less than half the structured figure.

What is the most valid selection method?

In the latest large review, structured interviews had the highest mean validity. The other methods near the top, such as job knowledge tests, work sample tests and empirically keyed biodata, have one thing in common: they're built for a specific job. General cognitive ability tests, once seen as the single best predictor, came out lower than earlier research suggested. Averages hide wide variation, so design quality matters too.

Why are unstructured interviews less valid?

Different candidates get different questions, interviewers rate on different things, and impressions from small talk and rapport can outweigh evidence. All of that adds noise, and noise lowers validity. The Sackett update estimated unstructured interviews at .19, compared with .42 for structured ones. Consistent questions and anchored scoring are the most direct way to close that gap.

How is predictive validity measured?

Researchers collect selection scores, such as interview ratings, for people who were hired, then later collect a measure of how well they performed, often supervisor ratings. The correlation between the two is the validity coefficient. Because only hired candidates have performance data, researchers apply statistical corrections, and how large those corrections should be is a key reason the estimates changed in the recent update.

How can you improve the predictive validity of your interviews?

Base questions on what the job actually requires, ask every candidate the same core questions, and score each answer against anchored criteria written in advance. Have interviewers score independently before the debrief. Add a work sample or job knowledge test where it fits the role. Then check your results by comparing interview scores with performance after hire, so you know whether your process predicts anything.