Anticipating Friday's Q3 Advance Estimate And the Long Road That Follows
October 23, 2012
by Doug Short
Friday we get the Advance Estimate for Q3 GDP. But, seriously, how useful is the Advance Estimate? We will get a Second Estimate a month later and a Third Estimate a month after that. And then next July, the Q2 GDP release will be accompanied by annual revisions for all the quarters of the previous few years. And periodically the annual revisions go back several years. The July 2011 update modified the data back to Q2 2003. And the July 2009 modifications involved a major overhaul of GDP back to its 1929 origin.
Before tackling the question of usefulness of the Advance Estimate, let's review analysts' expectations. Yesterday I posted a commentary that looked the Wall Street Journal's survey of economists on their forecasts for Q3 GDP. The mean (average) of the responses was 1.7 percent, which was also the median (middle) of the 48 survey respondents. Briefing.com's consensus is 1.9 percent, and Forexpros' consensus is 1.8 percent. So I'll take the 1.8 percent median and mean of the three to illustrate the how this estimate stacks up with GDP history since the turn of the century. For a complete snapshot of quarterly GDP, see my more recent update of the quarterly series here. Note: The 10-year moving average is included to highlight the longer-term historical trend.
About Those GDP Revisions
With all the drama of anticipation surrounding Friday's GDP release and our consensus estimates in mind, let's now take a moment to consider how extensive the future GDP revision process can be. To illustrate my point, I've picked two quarters (admittedly dramatic ones) during the last recession, Q3 and Q4 of 2008. This was the period of the scariest days of the Financial Crisis, including the collapse of Lehman Brothers in September of that year.
In the chart above, Q3 2008 went from -0.3 in the Advance Estimate to -0.5 in the Second Estimate, was unchanged in the Third Estimate and then slashed to -2.7 in that 2009 benchmark overhaul of GDP. The July 2010 annual revision made another dramatic cut, this time to -4.0. Then the 2011 annual revisions adjusted Q3 2008 upward to -3.7.
The Q4 2008 revisions were even more dramatic. For this quarter, not surprisingly, the downward revisions came faster. The Advance Estimate of -3.8 was slashed to -6.2 in the Second Estimate, -6.3 in the Third Estimate. But then the 2009 benchmark revision resulted in an upward revision to 5.4. But then a year later the BEA used the 2010 annual revision as an opportunity to slash Q4 to -6.8. Game over? Not quite. The 2011 annual revision further cut Q4 GDP to a gruesome -8.9. That's the second lowest quarterly GDP in US history, the lowest being -10.4 in Q1 1958, toward the end of the brief but savage 8-month recession in 1957-1958.
After our close look at the revisions for two quarters, I've posted below a chart of five consecutive quarters, including the two just discussed, beginning with Q4 2007, the quarter when the Great Recession started (December 2007).
Other than pointing out one curious anomaly, I'll let the chart above speak for itself. The anomaly? Note Q2 2008, the green line in the chart above. We were in months 5-7 of the Great Recession during that quarter, although, of course, the NBER didn't make its official recession determination until November 28th of that year. So, at five months into the recession and just three months before Lehman's collapse, we have a positive GDP: 1.9 in the Advance Estimate, adjusted to an amazing 3.3 in the Second Estimate and then trimmed to 2.8 in the Third Estimate. The 2009 benchmark revisions nearly halved the Q2 2008 Third Estimate to 1.5, and the 2010 annual revisions cut it further to 0.6. But then 2011 annual revisions lifted it to 1.3. Crapshoot, anyone?
The table below is my best effort to show the dynamics of the BEA's GDP revisions since 2005. It shows all individual Advance, Second and Third Estimates. The next-to-last column shows the current official GDP for the timeframe (at least until next year's annual revision). And the last column shows the change from the Third Estimate created by the annual revisions.
Unless we dig into the historical data, most people have no idea how extensively the BEA changes previously reported GDP – not only the revisions between the Advance, Second and Third estimates, but the sometimes astonishing annual revisions. And the periodic benchmark revisions (as in 2009) can tweak GDP all the way back to 1929.
In the larger historical context, Friday's GDP release is important. It can have an immediate impact on business and investor sentiment going into the 2012 holiday season. But we must also bear in mind that Q3 GDP will be subject to revisions for years to come. It is, ultimately, a "rear-view mirror" look at the previous quarter. At best it is a very long lagging indicator of where we were a few months ago.
Note from dshort : The raw data on which I based this commentary is available here in the Federal Reserve's ALFRED data repository.