Just how much slack remains in the labor market?
It’s a key question for the Fed, of course, which is explicitly gauging many indicators—not just unemployment—for signs of inflationary pressures. Thus far, such pressures have been noticeably absent from their favored measure: the core PCE, which if anything, has decelerated of late. Chair Janet Yellen has been particularly explicit on these points, stressing these dynamics as her rationale for continuing to up-weight the full employment side of their mandate.
Which indicators do Yellen&Co. watch? Ylan Mui goes through some of that here, citing some nice research by Cornerstone Macro. But their work just made me want to get out the big pasta pot and boil up a bunch of the labor market statistics into a principal components analysis (here’s a pretty simple explanation, but it’s really just a way a reducing a large number of variable down to a few factors that summarize much of their variance).
I used unemployment and employment rates, labor force participation, long-term unemployment (as share of labor force), initial claims, payroll and wage growth (yr/yr; production, non-supervisory wage), and involuntary part-timers (as share of employment, adjustment for big measurement change in 1994). I left out some of the JOLTs measures—vacancy and quit rates—because even though the Fed watches them, they don’t go back very far and I wanted to look at this mix over a few cycles (I start in 1979).
The first figure below just plots the summary slack measure (the first principal component, which explains 60% of the total variance of the eight variables) against the actual unemployment rate. They’re very similar, of course, but not the same in some interesting ways (don’t worry about the scale of the summary measure). For example, even though unemployment rate looks almost as low at the end of the 2000s cycle as the 1990s one, such that a bunch of folks think of them as similar episodes of full employment, the summary measure shows that there was considerably less slack in the late 1990s (compared to the unemployment series, the ‘pc1’ series late 90s trough is lower than its late 2000 trough).
Source: BLS, my analysis
But it’s the next figure that bears on the current dynamics. Here, I’ve indexed both series to 0 at the peak of the last cycle, Dec07. Both unemployment and the 8-variable-summary-measure have fallen about the same amount, but the summary measure went up further and is thus still more highly elevated.
Source: BLS, my analysis
Anyway, you don’t have to look much further than the low and decelerating core PCE price index to understand where Chair Yellen and others are coming from on this, but my little data compression exercise shows that if the eye is trained solely on the unemployment rate, there’s more slack than meets the eye.