Biased Forecasts and Voting: The Brexit Referendum Case

This paper explores whether professional macroeconomic forecasters manipulate their forecasts to influence voting outcomes. We model a referendum in which a voter chooses between two policies. The voter relies on a forecaster to learn about the macroeconomic consequences of the policies. The forecaster favours one of the policies and faces a choice between lying to influence the vote and maintaining a reputation for honesty. The model yields three predictions. First, the forecaster is more likely to bias the forecast associated with the policy that is (i) less likely to be selected in the referendum and (ii) associated with greater macroeconomic uncertainty. Second, as the influence of the forecast on the voter’s decision increases, so does the likelihood that the forecaster lies. Third, the forecaster sticks to her biased forecasts even after the referendum, at least for some time. We show that these predictions are empirically supported in the context of the Brexit referendum.