December 3, 2008

NBER's Anomalous Recession Calls

By Randall Hoven

The National Bureau of Economic Research, the official caller of recessions, recently said we are now in a recession that started one year ago, in December 2007.

"The committee determined that a peak in economic activity occurred in the U.S. economy in December 2007. The peak marks the end of the expansion that began in November 2001 and the beginning of a recession. The expansion lasted 73 months; the previous expansion of the 1990s lasted 120 months."

This struck me as odd.  Not that I don't believe we are in a recession now, but the starting date of December 2007 just seemed too early to me.  To see if this made sense, I looked at some underlying economic data and the NBER's explanation, such as it is.  Why this might be important, I'll explain later.

The rule of thumb for defining a recession is two consecutive quarters of negative real growth in GDP.  This is now the second recession called by the NBER in the two terms of President George W. Bush, yet in neither case were there two such consecutive quarters.  In fact, at no time in Bush's Presidency were there two such quarters.

Of all 11 NBER-called recessions since 1947, only one other involved no two consecutive quarters of negative real growth.  That was the recession of April 1960 to February 1961.  However, that recession involved one quarter with significant negative growth, -5.1% annualized, and a cumulative -1.0% growth for a whole year

Compare that to Bush's two "recessions."  In 2001

  • No two consecutive quarters of negative growth.
  • Worst single quarter: -1.4% annualized.
  • Year-to-year: +0.2% (positive real growth, 4Q2000 to 4Q2001).

In 2007

  • No two consecutive quarters of negative growth.
  • Last four quarters: -0.2%, +0.9%, +2.8%, -0.5%.
  • Year-to-year: +0.7% (positive real growth, 3Q2007 to 3Q2008).

In all the other nine recessions since 1947, the NBER-called recession involved at least one quarter of year-over-year negative real growth.

President Bush deserves some sort of prize for getting two recessions assigned to him, yet never presiding over either (1) two consecutive quarters of negative real growth, or (2) year-over-year negative real growth.  I think that's a first.  It certainly is in the last 60 years.

But the NBER does not use the "rule of thumb."  Here is how the NBER explains its method.

"The committee's procedure for identifying turning points differs from the two-quarter rule in a number of ways. First, we do not identify economic activity solely with real GDP, but use a range of indicators. Second, we place considerable emphasis on monthly indicators in arriving at a monthly chronology. Third, we consider the depth of the decline in economic activity. Recall that our definition includes the phrase, "a significant decline in activity."  Fourth, in examining the behavior of domestic production, we consider not only the conventional product-side GDP estimates, but also the conceptually equivalent income-side GDI estimates.  The differences between these two sets of estimates were particularly evident in 2007 and 2008."

Get it?  I don't.  I mean I sort of understand it, but I could never duplicate the NBER's results with that explanation.  No one could.  It lacks transparency.  If the NBER explains its method elsewhere, I could not find it and no such link was provided in its FAQ on the matter.

For example, in the 2001 "recession", why does the NBER say it started in March, under Bush, and not in 2000, under Clinton?  The first quarter of negative real growth was actually the third quarter of 2000, under President Clinton.  It showed -0.5% growth contraction, annualized.  But the NBER said no recession.  When it again showed -0.5% growth six months later, under Bush, the NBER said recession.

In 2007, the final revision of the estimate of fourth quarter growth was slightly negative: -0.2%.  The NBER now says that was a recession.  One quarter of -0.5% under Clinton, not a recession.  One quarter of -0.2% under Bush, a recession.

Maybe unemployment was more of a factor in the NBER's analysis.

When the NBER said the recession of 2001 started, the unemployment rate was 4.3%.  That's pretty low.  In fact, the unemployment rate was 4.3% or higher in every single month of President Clinton's first term, and every single month of his second term until March of 1999.  No recession that whole time.

What about in December 2007, the beginning of our current "recession"?  The unemployment rate was 5.0%.  Then it dropped below that for the next two months and still stood at 5.0% in April of 2008.  Again, the unemployment rate was at or above 5.0% in every single month of Clinton's recession-free first term.  It did not go below 5.0% until May of 1997.

Well, neither real GDP nor the unemployment rate quite explains the NBER's method.  The NBER said it looks at the "income side."  So let's try Disposable Personal Income in real dollars.

In three of the last four months of 2000, all under President Clinton, real DPI declined.  NBER said no recession.  In the next three months, or the first three months of 2001 and mostly under President Bush, real DPI increased in each month.  NBER said recession.  Decline in three of four months, no recession.  Increase in three of three months, recession.

Just for fun, I looked at quarterly GDP numbers since 1947 and all 11 NBER recessions called since that year.  Here are a couple of interesting observations.

(1) Every time there was at least one quarter of year-to-year negative real GDP growth, there was a recession associated with it.  There were no recessions without such negative year-to-year growth, with two exceptions.

(2) With simple rules based on real GDP numbers alone, a recession as well as its beginning and ending quarters could be called.  Every recession called by these rules was also called by the NBER.  Every recession called by the NBER was also called by these rules, with two exceptions.  For every recession these rules called that matched the NBER recessions (9 of the 11), the starting and ending quarters matched within one quarter, at worst.  What were these simple rules?

  • A recession starts in a given quarter when that quarter-to-next-quarter's growth is negative and the total growth over the two quarters combined is also negative.  (A somewhat weaker version of the "two quarters" rule.)
  • A recession ends as many quarters after that beginning quarter as there remains cumulative negative growth.

That is, without trying really hard, using real GDP data easily available from the St. Louis Fed only, and programming simple rules in a spreadsheet, I was able to match 9 of the 11 NBER-called recessions, with no false alarms and with, at most, one quarter mis-match in timing.  The only two exceptions in any of this?  The two recessions under George W. Bush.

My simple rules said no recession in either case (yet called all other recessions, with no false alarms).

The year-to-year negative growth rule said no recession in either case (yet called all other recessions, with no false alarms).

The "two quarters" in a row rule said no recession in either case (yet called all but one other recession, with no false alarms).

(It's still possible, by any of these rules, that we are in a recession now, but one that started in the third quarter of 2008, meaning July at the earliest.  But for any of these rules to kick into effect, we need the fourth quarter's data.)

I'm sure the NBER has good reasons for calling and timing the two Bush recessions. But even it would have to admit that those two recessions are anomalous -- oddballs among the 11 recessions in the last 60 years.

Why would this matter?  Why would it matter whether the US recession started in December of 2007 or July of 2008, for example?  After all, President Bush presided in either case.

Here is why.  Europe is now in recession, and it started in the second quarter measured the old-fashioned way: two consecutive quarters of negative real growth in 2008.  The US did not have its first quarter of negative growth until the third quarter of 2008.  The question is whether the recession spread from Europe to the US or vice versa.

If the US recession started in 2007, as the NBER states, Europe caught our cold.  If we go by the simple rule of two negative quarters in a row, we are catching Europe's cold.  It's all about who gets the blame, at least plausibly so.

In 2001, President Bush got the blame instead of President Clinton, per the NBER.

In both cases, blame Bush.  In this second case, blame Bush not only for a US recession, but a global recession.  If Iceland is bankrupt, it must be Bush's fault.

Yet in both cases, we don't know exactly how the NBER did it.

Here is what the NBER says about itself:

Committee members are: Robert Hall, Stanford University (chair); Martin Feldstein, Harvard University and NBER President Emeritus; Jeffrey Frankel, Harvard University; Robert Gordon, Northwestern University; James Poterba, MIT and NBER President; David Romer, University of California, Berkeley; and Victor Zarnowitz, the Conference Board. Christina Romer of the University of California, Berkeley, resigned from the committee on November 25, 2008, and did not participate in its deliberations of November 28.

Christina Romer (formerly on the committee) was just designated as the chair of Barack Obama's economic advisors.  She is married to David Romer (on the committee).

Here's how the NBER might help: tell us exactly the formula for calling and timing a recession and give us the input data so that we can reproduce its results.  If it can't, or won't, it should not be considered the "official" caller of recessions in my opinion.

In my opinion, there should be both transparency and clear objectivity in calling and timing recessions.  The method should be repeatable and based on publicly available data.  It should be more than simply the considered, consensus opinion of a panel of seven experts.  Otherwise we invite public doubt -- public doubt in the area of cause and effect of economic downturns.  This is important stuff -- or should be, in a democracy.

Data sources:


Randall Hoven can be contacted at randall.hoven@gmail.com or  via his web site, kulak.worldbreak.com.
If you experience technical problems, please write to helpdesk@americanthinker.com