Nowcasting Errors – Atlanta vs. St. Louis

Reader Steven Kopitz seems to think that the GDPNow of the Federal Reserve Bank of Atlanta is the only current forecast. There are actually quite a few of them, from GDP tracking from consulting firms (e.g. IHS Markit, formerly Macroeconomic Advisors) to Goldman Sachs, Deutsche Bank, etc. However, for illustrative purposes, comparison can be easily made here (since the data is in FRED) nowcasting forecasts immediately prior to advance release.

If we subtract the growth quarters associated with the start of the pandemic (Q1, Q2, Q3), GDPNow and St. The mean Louis errors for a total sample of 33 observations are 0.12 and -0.82 ppt, respectively (SAAR). RMSFE is 1.4 and 1.8 ppt respectively. In other words, GDPNow is slightly underestimating the forecast, while St. Index Louis exaggerates a bit. The offset adjustment means that the Q2 growth rate implied for GDPNow is -1.7 p.p. and +3.2 p.p. for St. Louis Fed.

Note that I used the final revised growth rate in my calculations, not the lead growth rate (because otherwise I would have to calculate all the lead growth rates in the ALFRED spreadsheet), so consider this a quick and dirty analysis. And there is the more fundamental question of whether pre-pandemic data is useful in this comparison of predictive ability (which is why I used data from the last three years in the previous report). mail).