Unemployment

Hey! We got an e-mail:

Guys-

Can you straighten me out on unemployment statistics. I dimly remember a commentary from someone about different versions of the unemployment statistics. The commentator seemed to imply that whatever administration was in charge always picked the version of the statistics that made them look best. Am I remembering this correctly? If we use a consistent measurement for the last 20 years, how does our current unemployment trend look?

We're going to use the tag-team approach to answer this, because blogging is our life.  It's true that the Bureau of Labor Statistics (the BLS) does two separate surveys of employment/unemployment, in addition to all the private surveys that are out there.

The first measure is known as the Establishment Survey.  This is a measure of employment, not unemployment.  Every month about 150,000 businesses large and small are asked to fill out a survey about how many people they employ.  This generates headlines like "Economy Adds 150,000 jobs in June," or "Economy loses 200,000 jobs in July."  The numbers are revised twice in the months following the initial release - always on the 1st Friday of the month, 8:30 AM Eastern time.  Later on, there are long term revisions to both the surveys themselves and to the benchmarks used to calibrate them.

The second measure is call the Household Survey.  This is a survey of a certain number of households, and it is a measure of employment as a percentage of the total workforce.  From this comes the unemployment rate, which is 1 - (# employed / # in workforce).  From here come the headlines like "Unemployment Rate rises to 5.7%" and such-like.

The two surveys generally match up pretty well over the long run, but in the short term they can deviate dramatically.  Part of the reason is that the household survey can be somewhat subjective; participants are asked if they have a job, and if not, whether they are actively looking for work.  Some people don't tell the truth, and others decide to describe themselves as 'self-employed' even though they aren't doing anything to generate an income.  Others might claim to be looking for work when they really aren't.

The key difference between the two surveys lies in the concept of the total size of the workforce.  In the establishment survey, there is no such concept.  All the survey measures is the number of people employed, and breaks it down into various categories like industry or size of company.  There are also seasonal adjustments, and fairly odd treatment of people like teachers, who are not counted as employed when school is out, and the military and farm workers, who are not counted at all - technically the series is known as the 'Civilian non-Farm Payroll Report.

The household survey measure the size of the workforce in addition to the number of people employed.  If people get so discouraged that they give up looking for work, the unemployment rate can actually fall because the size of the workforce shrinks.  Similarly, when the job situation starts to pick up after a downturn, the unemployment rate can rise despite lots of people getting new jobs, because formerly discouraged workers rejoin the workforce. 

So which number do we pay attention to?  The press likes the household survey, because it produces a number that people think they understand - the unemployment rate.  The government will certainly attempt to spin the report toward whichever version gives the answer they like the most.  Wall Street wavers back and forth - for most of my career, the establishment survey was favored, but more recently people have become suspicious of it because of its large seasonal adjustments and massive revisions and re-revisions, and the household survey has come into vogue.  The answer really is that no one number can sum up the employment situation or the overall economy as a whole.  Far more important than any one number is the direction the series is going - is it steadily improving, steadily getting worse, staying about the same? Also important is the aggregation of more data than just unemployment.  Are workers getting paid more or less?  Are they working longer hours or shorter?  Is industry adding jobs, or is it just the government hiring more people?  And on and on.

So that's the basic idea... stay tuned for more on the ins and outs of unemployment stats.  And if you're really a geek, you can get the raw data at www.bls.gov/ces.