Back on March 10, the New York Fed’s attempt at real-time GDP forecasting predicted that the Q1 2017 estimate would be 3.2%. That would have qualified as another decent quarter, the second out of the past three and somewhat in keeping with “reflation.” As we know today, the advance figure calculated by the Commerce Department amounted to just 0.69% growth in Q1.
The point is not to cherrypick the highest quarterly prediction and make fun of FRBNY. At the same time in mid-March the Atlanta Fed’s GDPNow competing model had already collapsed below 1%. Like the New York version, the Atlanta tracker had early on projected better than 3% growth for the quarter. It was at one time in early February just shy of 3.5%, higher than at any point for the other one.
The New York Fed’s parsimonious statement today tersely explained:
Today’s advance estimate of GDP growth for 2017:Q1 from the Commerce Department was 0.7%, substantially weaker than the latest FRBNY Staff Nowcast of 2.7%.
While that was of no value, pointing out merely the obvious, the regular weekly report gives us a clue ironically in trying to explain why we should readily trust its methods.
Extensive back-testing of the model, research, and practical experience have shown that the platform is able to approximate best practices in macroeconomic forecasts. The model produces forecasts that are as accurate as, and strongly correlated with, predictions based on best judgment.
There is a whole lot to that statement covered underneath statistical jargon and buzzwords. That they attempt to “approximate best practices” rather than the results of those practices is especially significant. The New York Fed is trying to model the economy that “should be” rather that the Atlanta Fed’s model of the economy as it actually is. Very important to the former is sentiment, one reason why monetary policy builds itself philosophically around rational expectations and the institutional expectation for successfully manipulating them.
Leave A Comment