On Thursday 1 March, Fabian Krüger, University of Heidelberg, will give a talk in the Social, Economic, and Decision Psychology Colloquium.

Forecast evaluation: the role(s) of the scoring function

Forecasts specify the value of a functional (such as the mean or a quantile), conditional on a given information set. For example, a theory may predict the probability of a person preferring option A over option B in a lab experiment. Alternatively, economists aim to predict the mean rate of inflation as a function of relevant background variables. When evaluating such forecasts, the choice of loss function (or scoring function) is crucial. From an ex ante perspective, it sets the incentive to provide honest and accurate forecasts. From an ex post perspective, it allows to compare and rank alternative forecasters or statistical models. This talk presents an overview of the recent statistical literature on the subject, discuss some recent results under model misspecification, and points to applications across the disciplines.