Today the BEA released its Third Estimate for Second Quarter GDP. The third (final) estimate bumped up the prior estimate from 1.1% to 1.4%.
The BEA changed the name from “final estimate” to “third estimate” because GDP is subject to revisions years or even decades later.
I am curious how this would impact the Atlanta Fed GDPNow forecast and the FRBNY Nowcast estimate both due out tomorrow. I have a detailed answer from Pat Higgins at the Atlanta Fed, creator of GDPNow.
Hi Mish,
I can give you a more detailed answer tomorrow after the GDPNow update is released. We don’t do an update today because the monthly real PCE data consistent with the revised Q2 real PCE growth rate are not released until tomorrow.
Except for consumer spending, each GDPNow subcomponent forecast is a weighted average of two forecasts. One forecast from a quarterly BVAR [Bayesian Vector AutoRegression] model that uses the GDP subcomponents and one forecast that uses the monthly source data with bridge equations.
The forecasts from the quarterly BVAR model can change when the quarterly subcomponents are revised. Subcomponents that put a nontrivial weight on the quarterly BVAR forecasts, like intellectual property products investment, can change somewhat after a GDP revision. Subcomponents that do not, like residential investment, will probably not be impacted much by the revision.
The forecast of Q3 real consumer spending can change because of the revisions to monthly PCE generally released the business day after the GDP report. And the forecast for the change in inventory investment can change as well. GDPNow forecasts the revision to inventory investment for the previous quarter (Q2 currently) up until the third GDP report released today. Going forward, GDPNow will use the revised Q2 inventory investment number released today when forecasting the change in Q3 inventory investment because that Q2 number will not be revised again until next year.
Today the “Advance Economic Indicators” report was released.
Tomorrow’s GDPNow update will also take that data into account. That could make it a tricky to determine how much, if any, revision to the GDPNow forecast is due to the GDP revision, the monthly PCE revisions and data for August, and the advance economic indicators release.
Best regards,
Pat
Thanks Pat!
I learn a lot from these discussions, and what I typically learn is that things are not as straight forward as it seems.
I have a great deal of respect for research analysts like Pat Higgins and corresponding contacts at the New York Fed.
I also have a lot of respect for the BLS economists that I have talked with. The BLS does not even have access to the data they need to do their job because of privacy restrictions. There are seemingly easy solutions (encrypted names and social security numbers for example), but nothing is as easy as it seems in large bureaucracies.
Regardless, it’s important to distinguish researchers from those who make policy decisions based on that research.
Mike “Mish” Shedlock
Mish, I hope someday you can find out how much of the official GDP number is government spending. As you said, I’m sure it is quite complicated, like all government bureaucracy stuff.
I wonder what intermediate and long term broadening tops in the Dow 30 will have on GDP.
Who makes these rules that restrict their data access? That might be the real question here.
“Bayesian Vector AutoRegression” Google Translate sucks.
It is the goal of the ESTABLISHMENT for us to focus on the GDP numbers while they have little to do with the health of the economy.
I want your take on two questions related to GDP.
(1) The GDP is a summary of a (unknown to me) number of measurements, each with its own error of measurement. What is the overall summation error statistic (standard deviation) that applies to the summation? It seems to me likely to be larger than the actual quarterly GDP changes currently being reported.
(2) Since our government lies to us daily on a wide variety of issues, why does anyone take their economic statistics as factually based?