Combining qualitative forecasts using logit

This paper introduces a computationally–convenient means of combining qualitative forecasts, through use of logit regression applied to training set data, applicable in dichotomous, polychotomous and ordered polychotomous contexts. It can be employed in the cases of combining probability forecasts, combining qualitative forecasts which have no associated probability forecasts, and combining both of these types of forecasts, a case for which no combining method currently exists. This methodology offers insights into the suitability of equal-weight averaging of probability forecasts, yields an existing method as a special case, and facilitates associated hypothesis testing.