Good, I have enough. Here is the article:
Studies You Should Know Personality Test Validity: Does the Stakes Level Matter?
The Question
A persistent and practically important debate in personnel selection concerns faking: when job applicants complete personality assessments knowing the results will influence hiring decisions, do they distort their responses in ways that reduce the validity of the test? Some researchers argue that faking inflates scores on desirable traits but leaves validity largely intact because faking is systematic and consistent. Others argue it introduces noise that genuinely attenuates the relationship between personality scores and job performance. Loy and colleagues (2025) conducted a meta-analysis to settle the question empirically by directly comparing personality test validity in low-stakes and high-stakes employment settings.
The Study
The meta-analysis compared personality test validity across two types of contexts: low-stakes settings, such as employee assessments where results carry no immediate selection consequences, and high-stakes settings, such as applicant testing where scores directly influence hiring decisions (Loy et al., 2025). This distinction is the key methodological lever: if faking does not meaningfully affect validity, the two settings should produce comparable validity coefficients. If it does, validity should be systematically lower in high-stakes settings.
What They Found
The results consistently supported the faking-matters position. Personality test validity was higher in low-stakes settings than in high-stakes settings across both unmatched and matched sample comparisons. In unmatched studies, validity in low-stakes settings was r = .17 compared with r = .13 in high-stakes settings (Loy et al., 2025). The matched sample analyses, which control for other differences between studies by comparing the same personality measures across both contexts, reinforced this pattern.
What This Means
The difference between .17 and .13 may appear modest in absolute terms, but its implications are not. The validity coefficients used to justify personality testing in personnel selection are predominantly drawn from research conducted in low-stakes settings, using incumbent samples or research participants who have no incentive to present themselves strategically. When those coefficients are applied to justify using personality tests in actual hiring, the transfer assumes the validity will hold. This meta-analysis provides systematic evidence that it does not hold fully (Loy et al., 2025).
This finding connects directly to the conscientiousness validity review by Watrin and colleagues (2023) covered earlier in this series, which found that only around 12 percent of conscientiousness validity research used real applicant samples in predictive designs. The present study provides the complementary finding: when you do compare low-stakes and high-stakes settings directly, validity is measurably lower in the conditions that actually matter for hiring.
The Broader Implication
The debate about faking has sometimes been framed as whether individual fakers can be detected or corrected for. This study shifts the frame usefully: even without identifying individual fakers, the aggregate effect of high-stakes conditions on validity is real and detectable at the meta-analytic level. The field cannot assume that validity established in low-stakes research transfers intact to high-stakes application (Loy et al., 2025).
For organisations using personality tests as part of selection, this is a practical concern. The validity they are implicitly relying on, drawn from the published literature, is likely somewhat higher than what their own high-stakes assessments would produce. How much higher depends on the specific measure, context, and stakes involved, but the direction of the effect is consistent and clear.
Reference
Loy, R. W., Christiansen, N. D., Tett, R. P., Klein, K., & Toich, M. (2025). Personality test validity differs between low-stakes and high-stakes employment settings. International Journal of Selection and Assessment, 33, e70018. https://doi.org/10.1111/ijsa.70018
