What is Incremental validity?
Also called Incremental prediction, Added validity
Incremental validity is the additional predictive accuracy an assessment contributes beyond the procedures already in use. It is the only validity figure that answers a purchasing question: a test that predicts performance well but measures the same thing as the interview you already run adds nothing to the decision, however strong its own coefficient.
The table the market still quotes
Schmidt and Hunter tabulated it in 1998, anchored on general mental ability at a validity of 0.51. Adding an integrity test takes the combination to 0.65 — an increment of 0.14, which they describe as a 27% increase. Adding a structured interview takes it to 0.63, an increment of 0.12 and a 24% increase. Adding a work sample also takes it to 0.63, the same increment and the same percentage.
That last coincidence is worth noticing, because it is usually omitted. In the original table a structured interview and a work sample contribute identically on top of an ability measure. A vendor citing one of those two figures and not the other is selecting from a table its own source prints side by side.
The anchor moved in 2022, and almost nobody has updated the increments
Every number above is computed against ability at 0.51. Sackett, Zhang, Berry and Lievens re-analysed the evidence in 2022 and put cognitive ability at 0.31 and structured interviews at 0.42, after correcting a systematic overcorrection for range restriction. Their 2023 follow-up revises ability down again to 0.23, on the basis of a meta-analysis of 113 twenty-first-century studies.
The 2023 paper does not merely change the numbers; it rejects the shape of the question. It observes that the earlier work tabled each predictor's incremental validity over cognitive ability, a framing that presumes ability is the anchor everything else must add to. On the revised estimates it is not: cognitive ability falls to fifth in the ranking, and to twelfth once the newest figures are used, with structured interviews at the top.
The single most quotable line in that paper is an increment run backwards. Giving cognitive ability a weight of zero in their illustrative composite reduces validity by 0.05, from 0.61 to 0.56. That is the incremental validity of cognitive ability over everything else — and it is small.
What 0.05 of validity is actually worth
Convert it. At a base rate of 50% and one hire in ten, a composite validity of 0.61 produces a success rate of 90.2% among those hired. At 0.56 it produces 87.4%. Removing cognitive ability from the composite entirely costs under three good hires per hundred.
That conversion is the whole argument of the 2023 paper in one line. If the validity cost of dropping the predictor with the largest subgroup differences is under three points, the trade-off between accuracy and adverse impact is far cheaper than the 1998 framework implied — and it is a trade-off that can now be priced rather than asserted.
Why increments are always smaller than the standalone figures suggest
Two predictors that both correlate with performance usually correlate with each other. The increment depends on that intercorrelation, not on the second instrument's own coefficient. Two cognitive tests add almost nothing to one another. A cognitive test and an integrity measure add a great deal, because they are close to uncorrelated — which is precisely why integrity produced the largest increment in the 1998 table despite a modest standalone validity.
The buying rule that follows is unintuitive and reliable: buy the predictor least like what you already have, not the one with the highest number on its own data sheet.
Two questions that settle a vendor conversation
The first is “over what?” A validity coefficient quoted with no stated comparison is a standalone figure, and standalone is not the situation any employer is in. You already interview. You already read CVs. The relevant number is what the test adds to that, and it is always smaller.
The second is the intercorrelation between the new instrument and the method you currently use. If the vendor does not have it, the increment is not zero — but it is unknown, and an unknown increment should be priced as a risk rather than presented as a benefit.
The 1998 increments, converted into hires — base rate 50%, one hire in ten
- Ability alone at 0.51 → 84.4% of those hired succeed
- Ability + integrity test at 0.65 → 92.3% (a gain of 7.9 points)
- Ability + structured interview at 0.63 → 91.3% (a gain of 6.9 points)
- Sackett et al.'s composite at 0.61 → 90.2%
- The same composite with ability weighted zero, 0.56 → 87.4%
Validities from Schmidt & Hunter (1998) Table 1 and Sackett et al. (2023); the conversion to success rates is Taylor–Russell. The 1998 increments are computed against an ability anchor of 0.51 that the 2022 and 2023 papers revise downwards, so read the first three lines as the market's inherited arithmetic rather than the current estimate.
Not the same as construct validity
Construct validity asks whether a test measures the attribute it claims to measure. Incremental validity asks whether that measurement is new information given what you already collect. An instrument can have excellent construct validity and no incremental validity at the same time — it measures the right thing, and you were already measuring it. For a buyer the second question decides the purchase.
Construct validityHow AssessAll handles it
AssessAll's free selection utility calculator takes the validity of the procedure already in place as a second input, reports the increment beside the headline success rate, and uses the difference between the two coefficients rather than the new one when it computes money (app/guides/tools/selection-utility-calculator/SelectionUtilityCalculator.tsx). A gain already being captured by an existing interview is not a gain a purchase creates, and the tool is built so the two cannot be confused.
Sources
- Schmidt, F. L. & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology. Psychological Bulletin, 124(2), 262–274 — the incremental-validity table
- Sackett, P. R., Zhang, C., Berry, C. M. & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology, 107(11), 2040–2068
- Sackett, P. R., Zhang, C., Berry, C. M. & Lievens, F. (2023). Revisiting the design of selection systems in light of new findings regarding the validity of widely used predictors. Industrial and Organizational Psychology, 16(3), 283–300 — the source of the zero-weight result (open access)
Read next
Related terms
More terms beginning with I
Check a selection process against the four-fifths rule
Free, no signup, computed in your browser — with the remedy, not just the verdict.
Last reviewed