If you just extrapolate the dynamics of past economic expansions, you’d say that a recession within the next few years is quite possible but by no means certain. The question is how much weight you want to attach to some of the other factors.
Barry Ritholtz of the Big Picture had a comment last week that caught my eye:
A possible recession in the 2006-07 time frame is hardly a stretch. Consider the past few contractions, and then fill in the blank: 1990, 1994, 2000, ______.
Barry clarified that although 1994 was not a recession, he included it because it was a bad year for the stock market.
The idea that recessions occur at regular intervals has long intrigued economists. I described a couple of ways one might look for evidence of periodicity in a paper that just came out in the Federal Reserve Bank of St. Louis Review. One approach uses spectral analysis, which basically tries to decompose the value of a variable like the unemployment rate at date t into a series of sine and cosine functions that are undulating up and down at different frequencies. If there was a specific periodicity to the business cycle– for example, if recessions usually arrive approximately every 5 years– then we’d expect to see strong weights in that decomposition attached to cycles that had a 5-year periodicity. In fact, when you estimate the spectrum for a series like the U.S. unemployment rate, you don’t find any extra weight on what would logically correspond to a business cycle periodicity.
|Start of |
|End of |
|Length of |
|Length of |
|November 1948(IV)||October 1949(IV)||11||37|
|July 1953(II)||May 1954(II)||10||45|
|August 1957(III)||April 1958(II)||8||39|
|April 1960(II)||February 1961(I)||10||24|
|December 1969(IV)||November 1970(IV)||11||106|
|November 1973(IV)||March 1975(I)||16||36|
|January 1980(I)||July 1980(III)||6||58|
|July 1981(III)||November 1982(IV)||16||12|
|July 1990(III)||March 1991(I)||8||92|
|March 2001(I)||November 2001(IV)||8||120|
Another thing you can do is try to estimate an equation summarizing the dynamics of a variable like the unemployment rate, and then examine that equation to see whether it implies any periodic or cyclic behavior in response to a shock. Again, when you estimate such a relation, you don’t find those cyclical dynamics.
The table at the right summarizes the dates at which U.S. recessions began and ended according to the National Bureau of Economic Research, along with the duration in months of the contraction and preceding expansion. One can perhaps see why the statistical methods described above fail to find much evidence of periodicity. U.S. recessions have sometimes come as close together as one year or as far apart as ten years.
Another question that has been investigated in academic studies is the following. Suppose that the economy has already been in an expansion (or a contraction) for n quarters. What is the probability of remaining in expansion (or contraction) in quarter n+1? If that probability turns out to depend on n, economists refer to that as “duration dependence.”
One of the first papers to study duration dependence was by Federal Reserve researcher Dan Sichel, published in the Review of Economics and Statistics in 1991. Sichel found evidence of duration dependence in the contractions since World War II– the longer a contraction has been continuing, the more likely we are to see the economy recover from it next quarter. This might be interpreted as evidence that the economy has natural mechanisms to get itself out of recession, or that Federal Reserve responses to recessions tend to be effective. On the other hand, Sichel found no evidence of duration dependence for expansions– it seems that no matter how long an expansion has been proceeding, that doesn’t change the probability of seeing a downturn soon.
Subsequent analyses with more data by
Chang-Jin Kim and Charles Nelson of the University of Washington in the Review of Economics and Statistics in 1998 and Thomas Zuehlke in the
Journal of Business and Economic Statistics in 2003 found some indication of duration dependence in expansions, though any such effect is very subtle if the economy is less than 5 years into an expansion.
To get one gauge on Ritholtz’s prediction of a recession sometime in 2006-2007, suppose you just picked a quarter at random between 1947:II and 2002:IV, and made a prediction that some time between the next 3 to 11 quarters the U.S. would experience a recession. If you made such a prediction out of the blue, historically you would have been right 48% of the time, simply by virtue of the fact that recessions occur fairly frequently and two years is a significant chunk of time.
At this point the U.S. has been in an expansion for 14 consecutive quarters. There are 70 quarters over the last half century for which the economy had been in an expansion for at least that long. If you again picked one of these completely at random, and predicted that the U.S. would be in a recession some time between the next 3 to 11 quarters, you would have been right 54% of the time, a manifestation of the very slight duration dependence mentioned above.
Of course, Ritholtz bases his prediction of a recession in 2006-2007 on more than just the regularity of business downturns. One factor is the ever-rising price of oil, which has often been a precursor of economic recessions. A second is the narrowing spread between the yields on short- and long-term securities, another common warning that a recession may be coming, as both Barry and I have emphasized. And a third is the unstable situation in the real estate market, particularly in conjunction with statements like the one made yesterday by Michael Moskow, president of the Federal Reserve Bank of Chicago:
If we do not remove that accommodation, or raise rates, then you risk significantly higher inflation in the economy.
Berkeley researchers Christina and David Romer came up with a particular set of historical episodes in which the U.S. Federal Reserve intentionally adopted a contractionary monetary policy. A large number of academic studies have used these Romer and Romer dates because they seem to have strong predictive power for whether the U.S. experienced a recession. What is sometimes forgotten is that the way that an episode came to be included in the Romer and Romer dates was when Fed officials made exactly the kinds of statements like the one quoted from Moskow above.
Now, one can always argue that this time things are different. For example, the oil shocks may be different this time because they have been caused by demand growth rather than supply reductions, so they will not be as economically harmful. Or the flattening yield curve may have a different connotation this time because it is the result of foreign purchases of long-term U.S. securities. And the Fed contraction may be different this time because– somebody help me out with that one, I’m stumped for the moment.
Maybe Barry’s prediction of a recession in 2006 or 2007 isn’t so reckless.