Max Sawicky follows up our Econoblog sessions with a challenging post that puts some real meat on the bones of his contention that labor force participation rates are abnormally low. Max was wise enough to issue a geek alert at the outset of his post, so I will do the same: Proceed only if you are really interested in some inside baseball.
Here is the essence of Max's demonstration, courtesy of the Economic Policy Institute's Jared Bernstein:
What’s needed is a believable counterfactual against which to compare actual trends in LFPRs. We devise this by modeling LFPRs for two groups, men and women, age 25-54.
The model is constructed using STAMP, the structural time series software which employs methods associated with Andrew Harvey and others (one advantage in this context is that STAMP has the capacity to control for the cyclical component of a time-series).
This is exactly the kind of analysis that opens the door to some real progress on the question, and I for one am grateful to Max and Jared for the contribution. Grateful, of course, does not quite equal convinced.
There is a deep problem associated with using this sort of trend/cycle breakdown to measure "slack" or "participation gaps." Think of labor force participation over the course of a year, and suppose that for some demographic group the participation rate is rising over the period of time. Despite a positive trend, you will find that the participation rate will still fluctuate over the course of the year. Over the seasons, for example, as summer vacations and winter holidays kick in. Or every weekend. We would never think of calling these sorts of fluctuations "gaps." But if this is so for a weekly or seasonal frequency, why not over the span of a business cycle?
To put it another way, efficient changes in labor force participation can have both temporary and permanent components. Temporary does not equal "bad" or "perverse" or "suboptimal" or "inefficient." Yet calculating gaps as deviations from a trend treats them just that way. (Those of you familiar with this line of thought will note, of course, that I am just reiterating the real business cycle logic I invoked in the Econoblog exchange.)
But even excepting the legitimacy of identifying the cyclical part of a time-series representation with slack, there is the inevitable ambiguity in exactly how that representation ought to be constructed. I'm not sure exactly how STAMP works, but let me try another cycle/trend decomposition, based on a statistical tool known as the "Hodrick-Prescott filter". Here is a graph of the actual time-series of the total labor-force participation rate (men and women, all ages) along with the secular trend estimated by the Hodrick-Prescott filter:
The analog to Max's measure of slack in this picture would be the difference between the trend line and the actual series. You will probably have to enlarge the picture (by clicking on it) to see it, but this particular time-series calculation suggests that the current overall labor force participation rate is actually somewhat above its estimated trend. You will get the same story if you do the experiment with male or female participation rates separately. (I'll show you if you force me.)
To be fair, Max and Jared do the more interesting experiment of estimating recent labor-market slack on the basis of a trend forecast, but as they note a simple extrapolation of trend won't do:
One important point in this discussion is the importance of separating secular from cyclical movements in LFPRs. This third figure, for prime-age women, shows what happens if you simply extrapolate off of the secular, positive trend in women’s LFPRs. In this case, the counterfactual LFPRs are too high and the gap estimate is too large. But once we control for this possibility, we still end up with a cyclical gap.
I'm not sure how they control for that possibility -- I would be interested in the details -- but that does bring me to my final point. I don't believe that this question can be settled by the application of purely statistical models. What we need is a theoretical framework sophisticated enough to allow serious quantitative analysis on the question of which labor market outcomes are associated with "normal" fluctuations in economic opportunities, and which ones can be attributed to policy failures of some sort.
We have a way to go on that agenda. I am a macroeconomist by trade, and operate in the realm of the usual macro models that are, in the end, not much different than they type you would find in a typical intermediate-level macro-theory textbook. (Like this one, for example.) The fundamental shortcoming of these models is that they treat the labor market as if it is, more or less, a spot market. They make no allowance for the long-run nature of employment relationships, the zero/one nature of labor force participation, and the like. Furthermore, these are models in which the key variable is hours -- there is no distinction to be made between changes in hours that result from the number of hours an average employee works and the number of people working (the latter being with what labor force participation and unemployment rates are solely concerned).
All of this may be OK if what we are focused on is forecasting GDP growth or inflation. It is obviously a disaster if we are trying to forecast unemployment or some measure of slack that depends on labor force participation rates. To put it another way, I think macroeconomists know even less about labor-market slack than we do about most of the other things that we nonetheless pontificate about at great length. And that is saying something.
UPDATE: pgl weighs in, and looks to tepid compensation growth as evidence that the labor market is weak. Although I maintain that "weak" does not necessarily imply "slack" (for the reasons identified above and in the Econoblog discussion), I do agree that the slower real wage expansion (inclusive of benefits) makes for a more convincing case than low participation rates.
UPDATE II: Tim Duy tries to disentangle things by looking at Oregon participation rates, at Economist's View.