Noah Smith On DSGE Models

Over at Noahpinion is a very good overview of the shortcomings of DSGE models, which are basically the workhorses of policymakers and central bankers.

He also links to an interesting criticism of these models by a private firm that does macro forecasting.

Here’s my take: DSGE models are one of the most conceited forms of intellectual over-extension.

The R2 correlations for the models Noah highlights are very low – as in less than 0.1 in all circumstances for all forecasting attempts.

This means that they are worthless in terms of foreacasting.

But they are exactly what Treasury and the Reserve Bank use to perform forecasting.

Which throws into question, why do we forecast again?

Oh, that’s to help policymakers deliver better policy.

Which they have not been able to do, which isn’t surprising when their mathematical representations of the world do not even represent the world unlike simpler micro models which are closer to how individuals and firms actually behave.

It would be more intellectually honest for people attempting to forecast the future to simply continue the last observation with some trend factor. In an Excel spreadsheet. With no fake mathematics attached.

Economics is not physics. Models are important – but it is hilarious how econometricians have failed to apply the basic 100-level concept of diminishing returns to their DSGE models.

Read more:
6 Charts You Need To See From The Latest Reserve Bank Bulletin

The Reserve Bank Bulletin is always an interesting read. The June issue delivers with "The last financial cycle and the...

Zero Marginal Benefit Of Surveillance (Does MB Really Equal Zero?)

If privacy is the opportunity cost of surveillance, however the government wants to spin it, is there an efficient level...

Close