Arnold Kling  

Why No Mini-Recessions?

PRINT
Proving You're Qualified Rothbard on Stigler and Friedm...

Scott Sumner writes,


Now here's one of the most striking facts about US business cycles. When the unemployment rate does rise by more than 0.6%, it keeps going up and up and up. With the exception of the 1959 steel strike, there are no mini-recessions in the US. The smallest recession occurred in 1980, when the unemployment rate rose 2.2% above the Carter expansion lows. That's a huge gap, almost nothing between 0.6% and 2.2%.

It is possible that the cause of this is the way that recessions are defined. A loose definition is two consecutive quarters of negative GDP growth. (The actual definition is that "The NBER will know it when it sees it," but the loose definition works well enough.)

The point here is that a recession is defined in terms of a shortfall of growth from zero. In an alternate universe, you could define a recession in terms of a shortfall from trend. Last time I looked at this, which was more than two decades ago, in the alternate universe recessions are no longer all short and steep and recoveries are no longer all long and steady. Instead, relative to trend, the behavior of GDP is fairly symmetric. Periods of above-trend growth are about as long as periods of below-trend growth.

In the alternate universe, you would define a recession as two quarters of below-trend growth. Some of these alternate-universe recessions probably would have increases in unemployment of small magnitudes. In some sense, the real-universe definition of a recession tilts the statistical playing field to make it more likely that one will observe only "big" (in terms of unemployment) recessions.

In short, I suspect that the absence of mini-recessions is not informative about our macroeconomic theories. It is informative about the way in which our use of zero rather than trend growth as a baseline affects the statistical properties of recessions.

UPDATE: See Dave Schuler's graph. Looking for years when GDP growth was below the trend line but we were not in an official recession, I see mini-recessions in 1968, 1994, and 1996. If you think we should see more mini-recessions, then remember that these data are smoothed by looking at annual growth rates. Quarterly data might show more mini-recessions, and if you had reliable monthly data they might show even more. I will grant that we don't see the mini-recession phenomenon in either payroll employment or household unemployment rates. To me, that says that employment relationships are sticky--there is a fixed cost to breaking them and a fixed cost to creating them.


Comments and Sharing





COMMENTS (10 to date)
Martin writes:

"In an alternate universe, you could define a recession in terms of a shortfall from trend."

I believe this is being done by Prescott and others?

"Instead, relative to trend, the behavior of GDP is fairly symmetric. Periods of above-trend growth are about as long as periods of below-trend growth."

This stuck me as odd. Won't periods of above trend growth have to roughly equal periods of below tread growth, almost by necessity?

If one was out of whack with the other, that would just define a new trend line.

Chris Koresko writes:

Steve Fritzginer: Won't periods of above trend growth have to roughly equal periods of below tread growth, almost by necessity?

Not necessarily. I think you're making an implicit assumption that the growth rate is distributed symmetrically about the mean. But it doesn't have to be. Example: Imagine that recessions are six months long, occur every 4th year, and have zero growth, while during the non-recession times the economy grows at a 3% annual rate. Then the GDP growth trend defined by the average growth rate will be 2.6%.

PrometheeFeu writes:

@Steve Fritzginer:

Not necessarily. Let's say growth is in % 1, 1, 1, 1, 101. Trend line is 21. But the below-trend periods are much closer to trend than the above-trend period. In other words, the distribution has a high skew.

Scott Sumner writes:

That doesn't explain it. Those periods that might be called recessions, but are not, fail to see "medium-sized" rises in unemployment. It's all or nothing. There are no mini-recessions, except the 1959 steel strike.

All you need to do is look at a graph of the unemployment rate. It's way too "smooth" looking, compared to things like graphs of stock prices, where big changes are less frequent then medium changes, which are less frequent that small changes.

There really isn't a single post-war case where it's debatable if a recession occurred, which is shocking, as a priori you'd expect lots of disputes. That's because unemployment barely budges, or it rises a whole lot.

Fralupo writes:

Isn't the "above-trend/below-trend" calculation you're proposing something relatively easy for an economics professor to do? Why do you need to guess?

Chris Koresko writes:

This kind of observation is quite interesting.

If someone will point me to the appropriate input data, I'd be pleased to take a crack at them.

PrometheeFeu writes:

@Scott Sumner:

Could this be explained by the economy being made up of lots of asymmetrical chaotic systems? Here goes the theory: At any one point in time, there are many weaknesses in parts of the economy that are currently being propped up by overall good economic performance. As soon as something breaks though, the performance is no longer enough to hide all the other flaws which also break. So you can't have a moderate recession. If such a thing happened, it would immediately trigger all the other lurking moderate recessions which add up to a deep recession. In other words, the current down-turn is not because the housing market tanked. It's because failure in the housing market caused the failure of many other sectors that were themselves only looking healthy thanks to the housing market bubble.

Costard writes:

Sumner's right. Unemployment has not been a random walk... it waxes between periods of stability and periods of rapid change. How you measure GDP has nothing to do with this.

Tom Cullis writes:

Scott-

A rise in UE is a person with a job losing it and not being able to find another for a period of time. A mini-recession by definition is going to affect the marginally attached workers who don't show up in the Ue rate. The UE rate is in fact calculated to look "smooth" as it knocks out the most sensitive portions of the population. If you wanted to look for mini-recessions you would look at LFPR or hours worked or some combination.

Comments for this entry have been closed
Return to top