Economic forecasts are critical inputs to policymakers and private-sector agents. They are usually made with different models and techniques. At one extreme are judgmental methods that rely on the expertise of individual forecasters to fine-tune forecasts produced by a suite of models; at the other end are dynamic stochastic general equilibrium (DSGE) models that use modern economic theory to produce a forecast disciplined by economic principles.
Whether because of an increasing perception that economic problems could be controlled, or perhaps simply because there was a widespread sentiment that democratic governments should at least try to do so, by the middle of the 20th century there was a growing consensus in most countries that government had some responsibility for economic prosperity. This prompted many attempts to create precise forecasting “models” that could guide policy choices.
While it is difficult to estimate the quantitative accuracy of a given model, it is possible to evaluate different models at the same time. This is done by comparing their ability to make accurate predictions about output growth over various horizons. It is also helpful to see how they perform in real time, that is, how accurately the predictions they produce are verified by the actual outcomes observed.
In this article we compare model-based forecasts of output growth to the judgmental forecasts of Blue Chip Indicators (the Philadelphia Fed Greenbook). We find that DFMS and AR(2) models and their combination predictors are able to outperform Blue Chip forecasts at both short- and medium-run horizons.