GDP Puzzles Even the Best Economists

Two days ago, the Federal Reserve Bank of San Francisco issued an Economic Letter entitled "The Puzzle of Weak First-Quarter GDP Growth." In the letter, economists from the bank claimed that first quarter gross domestic product was substantially higher than reported by the U.S. Department of Commerce's Bureau of Economic Analysis in the Advance Estimate provided on April 29.

The BEA announced that the annualized rate of GDP growth expansion was 0.2%, which is almost exactly what the Federal Reserve Bank of Atlanta forecasted with the banks GDPNow model.

The BEA's advance estimate and the GDPNow's accurate forecast of GDP were substantially lower than the consensus expectation as announced by multiple other private sector and academic economists which were closer to a rate of about 2%. The discrepancy between the two groups is fantastically large and did not reflect well upon the economists who wrongly estimated GDP.

Ever since, analysts have been trying to figure out why the private sector estimates were wrong. FRBSF economists claim that they were not wrong and that it was the BEA and the economists at the FRBATLANTA that were in fact incorrect, maintaining that GDP growth was closer to the private sector estimates. 

The justification provided by the FRBSF economists for this was that the average Q1 GDP growth rates of the past 16 years has been lower than that of the 1980's and 90's and that after applying a "residual seasonality" adjustment to account for this, the Q1 GDP growth rate for this year moved to a positive 1.8%. 

There is nothing wrong with the concept of residual seasonality adjustment, and it's a regular part of all economic analysis. However, it should be applied only when there is a legitimate expectation that a new seasonality issue has arisen that may reasonably be expected to repeat on a regular basis moving forward.

That is not the case here.

The period in question, between the years 2000 and 2015, include the two largest capital asset market shocks since 1929, the NASD bubble collapse of 2000 and the subprime/housing, and Lehman crises culminating in 2008.

In both instances, the economic responses to these shocks could been seen most in the years in which they occurred and the following year.

Due to space constraints here, I will analyze what the average quarterly growth rates are for the 16 year period minus the years 2000, 2001, 2008, and 2009, using the most recent estimates from each year and the advance estimate for Q1 2015, which is the only one available now.

This will provide an average for 12 quarters of Q1 data and 11 quarters for Q2, Q3, and Q4.

The average annualized rate of growth for the 12 observations of Q1 data is 1.74%. For the 11 quarterly observations for Q2, Q3, and Q4, it's 2.75%, 3.14%, and 2.58% respectively.   

This results in a quarterly GDP pattern that much more closely aligns with the traditional pattern of increased production during the spring and summer months in the U.S.

More importantly, by removing the financial crisis years, the annualized GDP growth rate as averaged across the 12 first quarter periods considered here almost doubles that of the aggregated average rate of growth achieved in those first quarters. Simultaneously, there is little impact on the average growth rates of Q2, Q3, and Q4 across the 11 quarters measured.

The most logical implication of this is that the traditional and known seasonal decline in production that occurs during the winter months was magnified in the crisis years because of the financial market disruption that preceded the winter months.

Applying a residual seasonality adjustment for distortions caused by financial market crises that impact the real economy requires an institutional belief and expectation that financial crises of the magnitude experienced in 2000, 2001, 2008, and 2009 are now cyclical events that may be anticipated to occur every eight years on average and will always have the greatest impact on economic activity during the first quarter of the year following the crisis peak.

I do not believe this to be a proper use or function of residual seasonality.

The more important issue for investors to be mindful of and that the letter by the FRBSF makes clear is that not only can data be manipulated to produce a desired and predetermined outcome, but it also may be retroactively altered to meet a more desirable one even after an outcome has been observed.