When more is not better: understanding the potential nonlinear relationship between intelligence and rating accuracy

Schade, Marizanne

When more is not better: understanding the potential nonlinear relationship between intelligence and rating accuracy

Master Thesis

2022

Abstract

Employers rely on judges or raters to accurately rate the potential or performance of candidates through interviews or assessment centre evaluations. As the judgment process places heavy demands on information processing, cognitive ability (of raters) is important to detect and interpret behavioural cues presented by those being rated. A consistent empirical finding is that intelligence is the strongest predictor of rating accuracy, but prior research has largely been based on linear models. However, researchers have yet to investigate whether these variables could be nonlinearly related. By studying nonlinear models in judgment and accuracy, we can not only deepen our understanding of the ‘good judge' in HRM, but we may further enhance methods to select and train raters in applied practice. This secondary research study re-analysed data from a prior published study to evaluate the relationship between rater intelligence and accuracy of interview ratings provided by 146 South African managers. The predictiveness of an ordinary least squares (OLS) linear regression model was compared to two nonlinear models (quadratic and cubic) to determine which statistical approach explained the most variance in rating accuracy scores. Findings provided further support of a linear relationship between intelligence and rating accuracy suggesting no quadratic or cubic interactions. Judges, therefore, produced more accurate ratings at higher levels of intelligence. Possible explanations of the findings include the sample size and task complexity. Study limitations and recommendations for future research are discussed in detail