We use the Banco de la República expectations survey of external economic analysts to study whether fixed-event individual forecasts of inflation and exchange rate are updated efficiently when new information becomes available. To this end, we test for weak-form and strong-form efficiency. The novel aspects of this paper are that we relax the individual homogeneity assumption, and consider a forecasters’ information set that contains a large number of empirically relevant variables. We address model selection using two of the most popular methods available in the penalised regression literature, and a new form of multiple testing. Our results show that more than half of the analysts’ revisions are independent of one another (weakly efficient). Also, conditional on passing weak efficiency, we find that analysts use past values of inflation and exchange rate changes to revise their forecasts and a broader array of variables during periods of market instability.
