Intelligence tests do (not) belong in personnel selection


The fact that employees are the key to a company’s success is fortunately no longer a new attitude. However, the methods used to select suitable employees and their effectiveness have been the subject of heated debate for decades. The central question that many HR managers are constantly asking is this: Do we really need an intelligence test for good personnel selection? In the following article, we get to the bottom of this question on the basis of current research and many years of practical experience.

By Annika Olofsson and Marie Mylord

For many years, the discussion was influenced by the meta-analysis by Schmidt and Hunter (1998), which compiled the predictive power of various selection procedures in relation to later work or education/training performance from 85 years of research. Even then, it was assumed that intelligence as a generic aptitude characteristic was a good predictor of career success. Almost 20 years later, Schmidt (2016) himself published a revision which, despite some corrections, overwhelmingly confirmed the central result of the superiority of cognitive test procedures in predicting professional performance over all other methods. With a predictive power of r=.68, there is actually no way around their use – and in combination with integrity tests or structured interviews, variance explanations of up to 60 % can be achieved according to Schmidt (2016). However, Sackett et al. (2022) take a critical look at the results of Schmidt and Hunter (1998). In their revised version of the meta-analysis, they were able to show that the high validity of cognitive tests was due to specific conditions. Specifically, Sackett et al. (2022) criticize the (data) quality of the meta-analyses used (which often date from the middle of the last century), the selection of the samples and the corrective measures of the validity estimates. For example, the cognitive tests were carried out in the entire population, while the results of the assessment centers were only available from candidates who had already undergone pre-selection. This restriction of variance in the AC sample reduces the correlation found.


Sackett et al. (2022) attempt to do justice to the shortcomings found and consider more recent meta-analyses in their revision, exclude those with poor data quality and analyze the variance restriction of the samples more systematically, which leads to a more consistent correction of the validities. Overall, the predictive power of all methods is reduced. The reweighting of the data also revealed that cognitive ability tests lose their outstanding position, but with r=.31 they can still make a good prediction of career success. Structured job interviews increase in validity to r=.42, which corresponds to a good to very good prediction. It should be noted here that Sackett et al. (2022) do not analyze the incremental predictive values of the selection procedures beyond intelligence, as Schmidt and Hunter (1998) do. Since they only examine individual validities and it can be assumed that intelligence and structured interviews only partially overlap in their variance clarification (Kanning, 2018), it can be assumed that a combination of an intelligence test and a structured attitudinal interview still has a very high predictive accuracy.


What do the results mean for the practice of personnel selection?

As things stand today, it can be said that high-quality personnel selection procedures are very useful, as they can predict job performance well. All established aptitude diagnostic procedures are valid, but their overall predictive power appears to be somewhat lower than assumed by Schmidt and Hunter (1998). According to Sackett et al. (2022), structured, requirement-based interviews represent the new gold standard. Intelligence continues to have a high, albeit somewhat lower, predictive power than previously assumed (Richardson et al., 2015) and, as a potential factor, provides information about later career success in particular if the challenges (of the future) cannot yet be clearly defined. Fittingly, the predictive power of intelligence increases with the increasing complexity of the profession. Nevertheless, the use of intelligence tests must be examined critically in each individual case. In particular, cognitive tests should only be used with a specific job requirement, as often not all intelligence facets are relevant for a job and the acceptance of the tests among candidates increases with the perceived job requirement (Weinert, 2015). Transparent communication throughout the entire selection process and comprehensible feedback are further factors that increase the acceptance of intelligence tests and selection procedures in general (Weinert, 2015). The required level of cognitive ability must also be considered. For example, a person who is “too intelligent” could possibly lose motivation due to underchallenge, which could also have a negative impact on career success (Duckworth et al., 2011). In addition, contextualized/occupational procedures appear to be more valid overall (Sackett et al., 2022). In conclusion, this means that intelligence tests still belong in personnel selection. However, their use does not – as previously assumed – generally increase the quality of the selection process. If an organization decides to use intelligence tests, it is important to check specifically required intelligence facets and to make the test implementation and feedback fair and as transparent as possible.