Under the Hood – NGDP Sub model forecasts

Our NGDP forecast, more precisely, our estimate of the markets NGDP forecast, is not the result of a single model. As is more and more common, we keep a ‘stable’ of top models, and average the forecasts from them. The idea is that, once we narrow the basket down to only models that performed well in the past, it becomes hard to say which model is best, so we consider them all.

Our current forecast of about 4% nominal growth over the next year results from averaging forecasts across five models. The basic picture—ignoring the details—is that Model 3 and Model 4 are saying nearly the same thing (about 4%) Model 1 and Model 5 are saying “more like 3.5%”, while Model 2 is off in the clouds. Three votes against two votes, we are saying the forecast is about 4%.

Some we may reasonably ask: “why can’t you just pick one model? Which one is right?” the answer is that these models are all contenders, the rank order of which is ‘best’ changes over time. The structure of each model is also quite distinct, meaning that the models capture and emphasize patterns in the data differently. Averaging over all of them is like asking a wise counsel for a consensus.


Leave a Reply