January 03, 2004

The Median Duration of Unemployment

Half of all the unemployed have been unemployed for more than X weeks:

The number of unemployed at any point in time is the product of (a) how many people become unemployed in a typical week times (b) how long they stay unemployed. This statistic is a measure of how lousy the labor market feels to those of us who have lost their jobs (or who are just entering the labor force and starting to search).

Posted by DeLong at January 3, 2004 09:48 AM | TrackBack


You can have ten percent unemployment in a million ways inclduing the xtremes (a) every body unemployed ten percent of the time (like extended vacation for all) and (b) ten percent of folks unemployed all the time -- that's bad, that's long term unemployment.

Bill Clinton, if I sensed it right, given a certain level of unemployment, tried conciously to have labor market work more on the (a) rather than (b)extreme.

Bush administration does not seem to have any notion of that kind of thing.

Posted by: Bulent Sayin on January 3, 2004 10:26 AM


How would this be different if all the people that lose their benefits, and are therefore no longer "unemployed", were still counted?

Posted by: slcINny on January 3, 2004 12:38 PM


Why would those dips happen so regularly around the turn of a decade?

Posted by: Patrick Taylor on January 3, 2004 03:58 PM


"How would this be different if all the people that lose their benefits, and are therefore no longer "unemployed", were still counted?"

Are they not? The unemployment number is based on a survey where people have to label themselves as employed, unemployed, seeking work, etc. According to the BLS, it has nothing to do with whether they receive benefits.

"Why would those dips happen so regularly around the turn of a decade?"

Without offering much of a real explanation, because the political/economic system has feedback loops, and in a "cybernetic" sense resembles a control loop, which swings around some (not precisely defined) goal function. Economic performance, political and social environment, and people's activities are not independent from each other, and have causal effects with delays on each other.

One parameter is that a presidential term is 4 years, but that seems at odds with the 10-year cycle.

Posted by: cm on January 3, 2004 04:59 PM


The early and mid nineties look sort of strange in this graph as well. What would happen if Congress didn't keep voting to extend unemployment benefits?

Posted by: Jason on January 3, 2004 06:28 PM


Where can the figures you state in yr equation actually be found? With so many numbers being cooked by so many forces, eg. can we rely on the Bureau of Labor Statistics in an Administration whose primary product seems to be spin?

Posted by: Allen Thomas on January 5, 2004 07:42 AM


The beginning of the decade is when population controls are changed in the household survey due to new information from the census.

Posted by: direlink on January 5, 2004 10:48 AM


Population controls in the household survey are updated soon after the census data is available every decade giving the unbenchmarked data in the survey some sort of anchoring which soon losses accuracy. (As this is done only once and then is updated by estimates in-between, everyone should be familiar with hideous census data estimates)

Posted by: direlink on January 5, 2004 10:58 AM


Post a comment