Separating Accuracy from Forecast Certainty: A Modified Miscalibration Measure
13 Pages Posted: 19 May 2017 Last revised: 20 Jul 2018
Date Written: May 19, 2017
Abstract
The standard interval forecasting task is modified, asking subjects to provide point predictions for future returns and assess the likelihood of fixed length intervals around their point estimates. The difference between the subjective likelihood estimates and the realized hit rate is advanced as an improved miscalibration measure. Class and take-home studies reveal that 140 of 169 finance students and experts overestimate their hit rates, confirming the overprecision hypothesis. A comparative study additionally shows that the revised task weakly decreases the miscalibration rates from 50% to 38% (p=0.16).
Keywords: Miscalibration, overprecision, interval forecasting, accuracy
JEL Classification: C53, D84, C90
Suggested Citation: Suggested Citation