Monday, August 13, 2012

Stop Using the Unemployment Rate

The standard unemployment rate is a deeply flawed measurement of macroeconomic conditions. It is so flawed that a decline in the unemployment rate as we measure it is not necessarily good news, nor is an increase in the unemployment rate surely bad news. Movements in the unemployment rate are -- or perhaps have become -- ambiguous signals of economic improvement or worsening.

What's wrong with the unemployment rate? The problem comes down to definitions. We know that the unemployment rate is the fraction of unemployed workers in the labor force. To be counted as "unemployed" by the surveyors of the Bureau of Labor Statistics, you must "have actively looked for work in the past 4 weeks," and to be in the labor force, you must either match that definition of unemployment or have a job.

Normally these definitions work well. A stay-at-home mother or father who is not looking for work, for a simple instance, should not be counted as "unemployed" in a way which is relevant for public policy. The Bureau of Labor Statistics' definitions of unemployment and labor force participation, however, become deeply problematic when economic conditions are so poor that they mischaracterize unemployed workers as having left the labor force, or people who should have joined the labor force as non-participants.

This is breaking the unemployment rate today. As I've pointed out in prior posts on this blog, the labor force stopped growing in 2008. That is largely the result of unemployed workers giving up and dropping out of the labor force, as well as would-be entrants to the labor force not doing so. Only a small fraction of this can be explained by changes in population or workforce, such as the retirement of the Baby Boom generation. These individuals who are not being counted as members of the labor force are, in every sense which is meaningful for public policy, unemployed. They number roughly 5.6 million as of July 2012.

This is why the numbers are so misleading, and why I think we cannot keep talking about falling unemployment as good news. We're saying, in effect, this: Good news, America! Another million workers gave up hope this morning! Don't you feel better about the labor market now!

Relatedly, there has been an incipient return of labor force growth in the last few months. I think that is among the major causes of the recent increases in the unemployment rate to 8.3 percent. Yes, hiring has slowed down, but labor force growth is a much larger phenomenon -- we're talking about a change in the tens of thousands in hiring against a change in the hundreds of thousands, if not millions, in the labor force. So is it a bad thing, or a good thing, that the unemployment rate is rising?

The math may be instructive. As of July 2012, the BLS counted 12.8 million Americans as unemployed out of a labor force of 155 million. Divide 12.8 million by 155 million: voilĂ , 8.3 percent unemployment.

When we add 5.6 million to the standard counts of the unemployed and the labor force, we get 18.6 million unemployed out of a labor force of 160.6 million. Repeat the division, and we get 11.5 percent unemployment.

What if we assume that half of the 5.6 million have dropped out of the labor force for demographic, not economic, reasons? (That is, in my opinion, that is the upper bound on reasonable estimates.) 5.6 million divided by two is 2.8 million; adding that to our numbers, we get 15.6 million unemployed out of a labor force of 157.8 million. That is 9.9 percent unemployment.

Let's look at this on a graph. Because, you know, everything is better when it comes with a graph.Two things should be obvious from this graph:

(1) Using a more accurate measure of the labor force, we see that unemployment has not improved at all since the end of the recession. This assumes a constant rate of labor force participation for demographic reasons.

(2) Even with the most aggressive demographic adjustment, unemployment remains far higher, and has improved far slower, than the headline statistics say it has.

[Sidenote: I will make the labor force adjusted unemployment rate data available on the "Data" sidebar tonight via FRED direct download.]

A labor force adjusted unemployment rate bears some similarities to the BLS' broader measures of unemployment -- the so-called "U4," "U5," and "U6" -- which include discouraged workers, the broader set of marginally attached workers, and the former plus workers who are part-time and for economic reasons cannot find full-time employment. (Less common but preferable is the employment-population ratio.) To be counted as "discouraged," workers have to tell the BLS that they have not looked for work within the last 4 weeks, but have in the past 12 months, for one of the following reasons:
They believe no job is available to them in their line of work or area.
They had previously been unable to find work.
They lack the necessary schooling, training, skills, or experience.
Employers think they are too young or too old, or they face some other type of discrimination.
To be counted as "marginally attached," workers must "indicate that they currently want a job, have looked for work in the last 12 months (or since they last worked if they worked within the last 12 months), and are available for work," according to the BLS.

One of the problems with these broader measures is that they aren't directly comparable to the unemployment rate, whereas mine is. I also worry that the nature of these survey questions have the same problem as those which create tighter definitions of unemployment and labor force participation. People who by rights should be in the labor force in working will be excluded if they say they do not want a job, if they have not looked for work in a year, or if they could not be immediately available for work within a week.

People who follow the economy and especially those who set the tone of conversation -- economists, reporters, news pundits, bloggers, etc. -- need to understand how seriously the unemployment rate is a misleading economic indicator. The problem is exacerbated by the extent to which this indicator is built into our political and economic dialogues. We need to stop using it or start adjusting for labor force dropouts and non-entrants.

8 comments:

  1. I looked into a related issue a couple of years ago: http://contraryneal.blogspot.com/2010/09/measures-of-unemployment-closely-track.html

    Perhaps this has disintegrated since I checked, but all of the Us are very tightly correlated to each other. This is a point in favor of using your measurement, since one does not get any new information by using the broader U4, U5, and U6 in place of U3.

    Perhaps I will go back and update that post with pictures. How do you embed those neat graphs in your posts?

    ReplyDelete
  2. Solid analysis, Evan.

    I tend to use the employment/population ratio.

    ReplyDelete
  3. I rechecked. The major point is that U1-U6 are very tightly correlated, so your LFA unemployment rate includes information not readily available from any of the BLS rates, simply by virtue of not being tightly correlated to U3.

    There are also some interesting minor points. For instance, U1 and U2 are almost exactly the same now, after almost two decades of proportionality.

    ReplyDelete
    Replies
    1. Right. This is why I think labor force adjustment is crucial, besides the fact that economists have developed an intuition about U3 -- not U4 and U5 -- as to what constitutes full employment or improvement...all of which will miss all of this labor force behavior.

      Delete
  4. Good post. I agree, which is why I (like Brad DeLong, with whom I am not personally acquainted) prefer to use the employment-population ratio. At least then you aren't excluding people who may participate in the labor force during better times.

    ReplyDelete
    Replies
    1. The only problems with employment-population are: (1) long-term change in workforce participation, like women in the 1970s, (2) demographic change to the age distribution. We can fix (2) by using a prime-age employment-population ratio.

      Delete
  5. I would like to see you responding directly to Karl Smith defense of the U3:

    http://modeledbehavior.com/2012/05/08/in-defense-of-u3/

    ReplyDelete
  6. Unemployment all around the globe should really be eradicated. It is good to have a safety net like income insurance but being able to cut the root cause of the problem is even better.

    ReplyDelete