In their book, Understanding Economic Forecasts, David Hendry and Neil Ericsson acknowledge the weaknesses of the econometric models in use. The problem, they say, is that they tend to be based on two key assumptions: that the model is a good representation of the economy and that the structure of the economy will remain relatively unchanged.
In reality, the models are mis- specified and the economy is subject to unexpected shifts. “Thus, the failure to make accurate predictions is relatively common.”
Problems of mis-specification, including the use of wrong variables and/or mathematical misrepresentations, are relatively easy to fix and do not necessarily produce bad forecasts. What really wrecks a forecast is a structural break, some underlying parameter or event that has changed in a way that wasn’t foreseen.
An interesting observation: You only have to remember the famous lament by Goldman Sachs’ chief financial officer when the credit crisis broke in 2008, that “we were seeing things that were 25- standard deviation moves, several days in a row”. When, as Hendry points out, one day should have been enough to recognize that the world had changed.