Abstract

The Group Calibration Index (GCI) provides a means of assessing the quality of forecasters’ predictions in situations that lack external feedback or outcome data. The GCI replaces the missing outcome data with aggregated ratings of a well-defined reference group. A simulation study and two experiments show how the GCI classifies forecaster performance and distinguishes between forecasters with restricted information and those with complete information. The results also show that under certain circumstances, where members of the reference group have high-quality information, the new GCI will outperform expert classification that is based on traditional calibration indices.