Recent writings underscore an important hole in economists’ knowledge base: we know neither the natural rate of unemployment nor the potential level of GDP. I mean, we’ve got estimates for days, but an honest confidence interval around them renders them useless as policy guidelines.
As Alan Blinder recently put it:
“For [the natural rate] to be useful you have to have at least a little confidence you know the number. You don’t need to know it to two decimal places, but within a reasonable range. If your range is 2.5 to 7, that doesn’t tell you anything.”
This post is about a related number that we don’t quite have right, either, one I wrote about a couple of years back: the jobs breakeven level (BL), or the monthly, net change in payroll employment that’s consistent with a stable unemployment rate. Though not as big a deal as the “natural” rate of unemployment, to which it’s closely related, the BL is an important piece of datum that provides insight into one of the most important questions in econ policy today: how much more room-to-run exists in the job market?
Back in mid-2016, when I was writing about this, economists thought the BL was roughly between 50,000-100,000 (note that this SF Fed analysis uses a natural rate of 5%, clearly too high, and one way to underestimate BLs). In fact, I wrote my piece to warn people not to be too worried if monthly payroll gains started coming in well below 200,000. Though I hedged my bet in a way I’ll get to in a moment, the dominant argument was that gains of around 100,000 were consistent with stable unemployment in the mid-4’s—the lowest rate thought to be consistent with stable inflation—along with labor force participation and working-age population growth around where they were back then.
This prediction now looks off. For the past six months, the jobless rate has held at 4.1% while payrolls are up an average of 211,000 per month, on net. Why hasn’t faster job growth led to lower unemployment, as most economists would have predicted a few years back?
The answer must be that there’s more labor market capacity than folks thought there was. Here’s one, simple way to look at it.
PR=PR/EMP * EMP/LF * LF/Pop * Pop
…where PR is payrolls, EMP is employment from the Household Survey, LF is the labor force (so EMP/LF = 1 – unemp rate; because EMP/LF + UN/LF = 1), and Pop is the working-age population (LF/Pop is thus the participation rate). Cancel everything out and you’re left with payrolls, which you can difference to get your monthly BLs.
If you think we’re pretty much at capacity, then your BL is solely a function of your expectations about population growth. The first line of the table below shows those numbers in May 2016, when I wrote my post. These generated a BL of about 100K.
|PR/EMP||EMP/LF (or 1-un rate)||LF/Pop||Pop*|
*Annualized population growth rates; the top number is what I plugged in; the bottom number is the pace of pop growth between these two dates.
But if you think the jobless rate might fall further or the labor force participation rate might rise, then you’d predict higher BLs. In one scenario we ran from the earlier post, we predicted BLs of 200,000, based on an unemployment rate of 4% and LFPR of 63.5%.
Before I pat myself on the back for that prescient forecast, recognize, that as the 2nd line in the table above reveals, while the LFPR is up, it’s still well below 63.5%. The population growth rate is a little faster, so that makes a difference, as does the first factor in the table, which is just a conversation factor to go from the HH survey numbers to the payroll ones.
But there’s another important capacity change, unforeseen by many: the climbing of the prime-age (25-54) employment rate and LFPR. Neither are back to their pre-recession peaks, but especially the prime-age employment rate is clawing its way back. In fact, prime-agers have recovered 4.4 out of 5.5 percentage points, or 80%, of their decline over the course of the recession. Prime-age men, whose employment rates have suffered a longer-term decline, have made back 76% of their loss; women have done better, clawing back 90%.
So, economists need to update their BLs to accommodate some unknown degree of labor supply that we formerly discounted.
OK, caveat-time. As noted, while the cyclical part of the prime-age guys employment rate looks better than expected, the structural decline is real (see figure below). That said, I’m increasingly off the mindset that separating structural from cyclical is yet another area where economists are fuzzy (this great Yagan paper underscored that point for me). Be careful about writing people off; the reach of really strong labor demand may pull more people in than we tend to think.
Also, while payrolls continue to chug along posting numbers that are about 2x of most economists BLs from a few years back, in percentage terms, their growth is decelerating, from around 2% back in 2015 to around 1.5% now, much as we’d expect as we close in on full employment, whatever that much-sought-after state looks like.
But the punchline remains: the fact that we’ve been adding an average of ~200K jobs a month, while unemployment sticks around 4%, along with, importantly, tame wage and price outcomes, means that we must not yet be at full employment.