by Calculated Risk on 1/06/2008 03:07:00 PM

Professor Nouriel Roubini suggests the debate has shifted from whether there will be a recession following the housing bust, to the severity of the recession. Roubini argues the recession will be severe:

“As argued here before, at this point the debate is not about soft land or hard landing; rather it is about how hard the hard landing will be. … This author’s assessment is … of a … severe and painful recession – lasting at least four quarters...”

Others think it is still possible for the economy to avoid recession, but even then it will probably feel like one. As Goldman Sachs noted last week:

“the economy [may] skirt a technical recession, but in many respects this distinction may feel like an academic one.”

This raises the question: What is the difference between a mild and a severe recession? Looking back at the last ten recessions, perhaps we can define a severe recession as lasting a year or more, with unemployment rising above 8%, and real GDP falling 2.5% or more from peak to trough.

By that definition, the U.S. has had two severe recession in the last 60 years:

1) Nov-73 to Mar-75:

Duration: 16 months
Peak unemployment: 9%
Real GDP declined 3.1%

2) Jul-81 to Nov-82:

Duration: 16 months
Peak unemployment: 10.8%
Real GDP declined 2.9%

We could use other measures for employment, such as the change in the unemployment rate (from expansion trough to recession peak) or the year-over-year change in total employment.

This graph shows the unemployment rate and the year-over-year change in employment vs. recessions for the last 60 years.

Click on graph for larger image.

Back in the '40s and '50s, it was common for the YoY total employment to decline by significantly more than 2%. This was because of the large swings in manufacturing employment. Now a YoY decline of 2% would be severe.

Also the recession with the highest unemployment rates started from pretty high levels ('70s and early '80s). So maybe the change in unemployment, from expansion trough to contraction peak, would be a better measure to gauge the severity of a recession than the absolute unemployment rate.


The second graph shows manufacturing and construction employment as a percent of total employment. The smaller percentage of manufacturing employment - compared to the '40 and '50s - is one of the reasons the economy hasn't experienced the large swings in employment characteristic of recession in those earlier periods.

Construction employment could fall back to 4.5% of total employment, with the loss of over 1 million construction jobs, but manufacturing probably won't see sharp layoffs like earlier periods. Note that Bernard Markstein, director of forecasting at the National Association of Home Builders, recently suggested the loss of 1 million construction jobs was possible.

If we assume the loss of 1 million construction jobs, 0.5 million retail jobs, and another 0.5 million jobs elsewhere, the unemployment rate would only rise to 6.3% (probably less because the participation rate would fall). And under most definitions that probably isn't a severe recession.

Perhaps other areas of the economy will shed more jobs and Roubini will be proven correct, but my expectation right now is for a recession, but not severe (the unemployment rate will stay below 8%). I also expect that the eventual recovery will be sluggish, especially for employment. For housing related industries, the depression will continue for some time.