Cognitive Skills and Confidence: Interrelations With Overestimation, Overplacement and Overprecision

Abstract

This experimental study measures the three major types of judgmental overconfidence in a withinsubjects design. Performance-based overestimation and overplacement are elicited in a Raven Progressive Matrices test for general intelligence. Calibration-based overprecision is evaluated in a forecasting by confidence intervals task. This study takes a closer look at the interrelations of these different types. Moreover, as the main focus, it considers the effect of cognitive abilities on overconfidence. These are quantified in a cognitive reflection test. I find that cognitive skills are substantially interrelated with subjects' confidence levels. Although overconfidence in absolute terms (overestimation) is not affected by cognitive abilities, the effect on overconfidence in relative terms (overplacement) is positive and significant. Overprecision, the calibration-based type of overconfidence, is found to be significantly affected by cognitive capacity as well. Interval forecasts of cognitively more able subjects were much better calibrated than those of their peers who displayed substantial overconfidence in the precision of their forecasts.

Cognitive Abilities, Analytic Cognitive Style and Overconfidence: A Commentary on Duttle (2016)

“…We agree with Duttle that tasks such as intelligence tests are a great way to capture overconfidence, as ‘ the effect of cognitive abilities to be most informative in an environment where they actually are an important factor of success ’ (Duttle , p. 53). Individuals are more likely to have a well‐calibrated self‐image in the context of such important factors, as compared to self‐image in less important tasks (e.g., those measuring general knowledge unrelated to everyday problems).…”

“…Our rethinking of Duttle's () experiment suggests that there is no substantial evidence that cognitive abilities have an impact on overconfidence. Duttle's interpretation would concur with the current psychological literature proposing that cognitive biases arise from the general architecture of the human mind rather than low cognitive ability (De Neys, ; Stanovich and West, ).…”

“…The arguments presented in Section suggest that CRT scores’ relationship with overconfidence cannot be interpreted in the manner done by Duttle (), namely that this relationship evidences the impact of cognitive abilities on overconfidence. However, Duttle provided enough data to test his hypotheses, which state that RPM scores are a valid measure of cognitive abilities and that the interval estimation task provides a valid measure of overconfidence (i.e., overplacement and overprecision).…”

“…Even Duttle himself admits that CRT is used as a measure of analytic cognitive style rather than of cognitive abilities; however, he concludes that ‘the CRT is utilized in a number of studies that investigate cognitive skills […] since despite its simplicity CRT scores quite accurately predict cognitive abilities, in particular with respect to mathematical skills. As a matter of simplicity, in the remainder of the text the CRT therefore is referred to as a measure of cognitive skills ’ (Duttle, , p. 46). Although a similar approach has been taken in some other economical works (Oechssler et al ., ; Hoppe and Kusterer, ; Brañas‐Garza et al ., ), we cannot find any justification for substituting one cognitive notion with another psychologically distinct concept (especially if these measures correlate merely at r = .4, as reported in the aforementioned work).…”

“…Thus, they are ‘mounted’ or ‘hardwired’ in the human mind (De Neys, ; Evans and Frankish, ). Contrary to this view, in his recent paper, Duttle () reported a study of cognitive abilities’ impact on overconfidence, endorsing the idea that higher cognitive abilities promote better calibration of confidence. This claim is thought provoking and theoretically intriguing.…”