It’s long been understood by anyone trying to assess the labor market that the unemployment rate is, by itself, not up to the task. Most importantly, it leaves out those not looking for work, but it’s also not adjusted for demographic change, nor does it factor in those who are working fewer hours than they’d like. It combines racial groups with persistently different levels of unemployment. At times like now, these shortcomings can lead this premiere indicator to underestimate the extent of slack in the job market.
This WSJ article from yesterday–“For decoding labor market, unemployment rate may not do the job”–is but the latest salvo in this healthy discussion about the need for a dashboard, not a single dial.
And yet, most of us, when trying to provide a quick overview of economic conditions, still cite the top-line rate, a practice I’d like to defend here, with at least moderate conviction, based on the correlation matrix below. The data run from 1994 through now, and the variables are the u-6 underemployment rate, the prime-age employment rate, the Richmond Fed’s non-employment index, both with and w/out those involuntary part-timers, and the black unemployment rate.
As you see, if we’re comparing levels, the unemployment rate correlates highly–close to unity in most cases–with the other variables in the table. Even the non-employment index, “an alternative to the standard unemployment rate that includes all non-employed individuals and accounts for persistent differences in their labor market attachment,” correlates with unemployment at 0.99.
Of course, we don’t just look at levels. We also pay a lot of attention to changes in these variables, and, as is always the case, change correlations are a lot lower than level correlations. We also see some interesting variation. When it comes to both prime-age epops and the black unemployment rate, changes carry different information relative to the topline jobless rate than do the levels. For African-Americans, this is due to their “high-Beta” relationship with the overall rate: a one-point change in unemployment correlates with a 1.5 change in the black rate. That’s a great elasticity to tap in high-pressure labor markets, and a hugely negative one in recessions.
Feel free to add other variables to the comparison, of course, as this is surely an incomplete list. And let me be unequivocal about the need for a dashboard of indicators, including very importantly, “price” variables like wage and price growth (see the subhead of the WSJ piece: “wage growth has been muted and inflation weak, leading economists to come up with new measures of joblessness”). The fact of moderate-at-best wage growth in recent months is one of the main reasons I suspect there’s still some slack in the U.S. job market, even at 3.7 percent unemployment.
But we shouldn’t be ashamed to cite the unemployment rate as a key indicator of labor market health. There’s no single, perfectly informative measure, but the good news is that all the imperfect ones are pretty highly correlated.