In this post, I am going to present some measures of slack that are based on Blanchflower and Levin’s “employment gap” analysis, referenced in my last post. The main point I raise here is that there appears now to be a bit more slack than was evident before the CBO took up its estimate of the “potential” labor force participation rate in January. Still, we are probably within about a year of achieving full employment, at least as measured by this particular form of analysis.
The CBO’s decision to revise up its estimate of the potential labor force is a bit ironic. B/L’s labor market analysis is largely driven by those estimates and had given a correctly-dovish signal for monetary policy throughout the entire recovery from the Great Recession. Where others had heralded full employment and the need for the Fed to get going, they had emphasized remaining slack.
The irony is that as soon as their own metric got close to signaling full employment, the CBO changed its estimates and thereby “created” some renewed slack. One wonders, then, if these guys are ever going to be able to get hawkish. I am not sure about Levin, but following Blanchflower on Twitter I can tell that he at least is showing no signs of being eager! 😉
Before turning to the data, I want to make two quick points of clarification. First, this is an approximation of their approach, as I skip over one aspect that requires some data to which I don’t have easy access and which is very probably not a big deal, particularly now.
Second, this data analysis cannot itself answer the question of whether the effort is itself even useful. Does the NAIRU exist? Should we even care about slack? For the so-called heterodox doves, this point has become more pressing over the past few years, because the labor market looks less “loose” than it did a couple years ago, even with the CBO revisions I highlight here. For the heterodox doves, the slack measures are no longer harmless because by some accounts they now present a case for tightening — or will soon. *
I am a wet on the raging debate about slack, and will leave that to a follow-up post. Here I just want to get myself updated on what we are even arguing about. Before getting into whether it is relevant, I would like to see what it is. That might not be best practice, but I do this as a hobby. And today, I feel like crunching numbers.
Ok, so B/L claim that the employment gap results from the sum of three factors:
- The extent to which the unemployment rate is above estimates of the natural rate, as estimated by CBO.
- The extent to which the labor force participation rate is cyclically depressed or below its “potential” rate, as estimated by CBO. In the charts U-BL, is the unemployment rate that B/L would calculate with the participation rate at its potential and employment where it actually is in the current period. Logically U-BL it is comparable to NAIRU, which allows us to create the first measure of slack, Gap: 1.
- The extent to which the share of total employment working part-time for economic reasons (PTER) is above the median that held over 1994-2007, which B/L claims was a period of normalcy. Adding this effect to Gap: 1 allows us to calculate Gap: 2.
According to B/L the first two factors are the subject of greatest controversy while the third is most straightforward. Still, it is on the third that I most need to explain my own approach.
They show part-time for economic reasons as a ratio to total employment, rather than potential labor force. And they convert this into an unemployment equivalence by looking at the ratio of hours worked among involuntary part-timers to hours worked among full-timers. That requires data I do not have easy access to, so I just use a constant of 1/3, based on the current-period data mentioned in the B/L paper of 2015. This will be close enough, particularly given that PTER has fallen steeply in recent years and is not very far above assumed “normal” now.
Ok, so here is a chart summarizing that analysis based on employment data to January and the CBO’s estimates of the potential labor force participation rate (time series) as of August, i.e. before CBO revised up its (time series) estimate of the potential labor force.
The top row of the chart shows gap analysis without considering PTER. And the lower row of the chart shows PTER as well as the summary measure of total gap including it. As you can see, on this perspective, the labor market looked – as I have been calling it – “indistinguishable” from being at full employment. I was not following B/L specifically, but it fit.
Now let’s consider how the data look incorporating CBO’s revision of its estimate of the potential labor force participation rate. The estimated potential participation rate for 2017 has risen from 62.7% to 63.2%, a gain of 0.8%. On B/L’s approach, this had made the labor market look a bit less “indistinguishable” from full employment. ** Note too that the share of employment taken by involuntary part-timers has collapsed almost to “normal” as defined by B/L. As a result, any slippage associated with my relation of that to effective unemployment equivalence is just not important, at least for the current period.
Presuming the Fed takes this sort of analysis seriously, which I do, the following question arises. Roughly how long might it take for the B/L estimated “employment gap” to close, assuming the CBO does not change its own inputs any further. I figure it is reasonable to project about a year, with employment slowing just slightly from its recent pace, as indicated in the simulation pictured immediately below. With nonfarm jobs running at about 185k a month for about a year, the unemployment rate dips marginally below the estimated NAIRU by next January and the PTER ratio falls almost back to “normal.” The two tiny gaps there are offsetting and Gap:2 goes exactly to zero. I picked the jobs gain figure to achieve that result.
I guess we can now resume arguing about whether this sort of analysis is even relevant. I will have some of my own – wet – thoughts on that later.
* Incidentally, in the second half of the Blanchflower and Levin paper on which this chart presentation is based, the authors argue that their slack metric is quite relevant for policy and that as of 2015 it was giving a quite dovish signal. Things have moved on since then, although slightly less than we might have assumed before the CBO revisions. Not to say that CBO has a monopoly on truth, but just to recognize that CBO estimates are direct inputs into the B/L analysis.
** Compared to how far we have come from the lows of crisis, these revisions are trivial. And in certain contexts, e.g. the question of the wisdom of fiscal stimulus here, that would be the main point to make. But for those who buy the logic of slack and its link to monetary policy and want to argue about whether we are yet quite there, the revisions may be somewhat relevant.