Robust Forecasting

We use a decision-theoretic framework to study the problem of forecasting discrete outcomes when the forecaster is unable to discriminate among a set of plausible forecast distributions because of partial identi?cation or concerns about model misspeci?cation or structural breaks. We derive “robust” forecasts which minimize maximum risk or regret over the set of forecast distributions. We show that for a large class of models including semiparametric panel data models for dynamic discrete choice, the robust forecasts depend in a natural way on a small number of convex optimization problems which can be simpli?ed using duality methods. Finally, we derive “e?cient robust” forecasts to deal with the problem of ?rst having to estimate the set of forecast distributions and develop a suitable asymptotic e?ciency theory.