Beware of Standard Prediction Intervals for Causal Models

We're all well aware that point forecasts are subject to a degree of error, and so we frequently report the forecast with a margin for error around it; that is, we present a prediction interval (PI). Much has been written about our prediction intervals often being too narrow to reflect the confidence we have in the forecast- for several reasons-and this is especially so when we forecast from regression and other causal models. For these models-those that forecast future values of a dependent variable based on assumed drivers of that variable (the explanatory variables)-the prediction intervals are calculated under the assumption that future values of the drivers are known or can be controlled. When this assumption is unjustified, these prediction intervals will be erroneously narrow. Here I explain why, and show from a case study just how serious the problem can be. Copyright International Institute of Forecasters, 2018