And yet, there’s also no question that nothing like such spirals are showing themselves either in the actual data, as shown below, or in forward-looking expectations.
As noted, economists associate this concept of “full employment” with a job market that’s tight enough to generate wage and price pressures. But the concept also includes a strong bargaining power component. When there’s too much slack in the job market, as has been the case for most of the past 30 years, workers lack the clout they need to claim a greater share of the growth they’re helping to produce.
The problem is, economists are unable to accurately pin down the unemployment rate commensurate with full employment (let’s call it the FEUR). The Federal Reserve guesses it’s between 5.2 and 5.5 percent, meaning we’re already butting up against its upper bound.
That observation is awfully hard to square with the first figure below. It plots the unemployment rate, along with the yearly growth of the Fed’s favorite inflation gauge and the average wage. I’ve also drawn a line for the Fed’s FEUR: 5.35 percent being the mid-point of the range. As is quite clear, although the unemployment rate has fallen to a level just above the Fed’s FEUR, we see no pressure on wage or price growth.
So what’s going on here? The next figure provides what I’d argue is an important answer to that question. The blue line is just nominal wage growth, same as in figure 1. It has been stuck at about 2 percent since 2010. The red line comes from a simple model that predicts wage growth based on unemployment alone (see data note below).
I ran the model through 2012 and then predicted wage growth going forward based on the actual decline of unemployment since then. As you see, as unemployment gets close to the Fed’s full employment rate, the model predicts wage acceleration.
But that prediction is wrong. Following work by economists Adam Posen and David Blanchflower, I added another variable to the model: the labor force participation rate. The idea for its inclusion is simple: The job market is less tight than you’d think from just looking at the unemployment rate, because a lot of people gave up the search when job creation was negative or weak.
As the pace of job growth has picked up, the labor force has stabilized, and that’s a good sign. But the participation rate is still low and some economists, including me, believe there are significant numbers of people who are not officially in the labor force — they’re neither working nor looking for work — but would come back if decent jobs were available.
If that’s true, then two other things should be true: 1) Part of the decline in the labor force is just slack that’s not measured by the unemployment rate, and 2) putting the labor force rate in my model should improve its out-of-sample prediction of the wage data.
And that’s just what you see in the green line in the figure. Adding the labor force participation rate to the same model, stopping the estimation in 2012 and predicting wage growth going forward, this version shows no acceleration and cuts right through the actual wage trend.
As Fed Chairman Janet Yellen has observed, “Some ‘retirements’ are not voluntary, and some of these workers may rejoin the labor force in a stronger economy . . . a significant amount of the decline in participation during the recovery is due to slack.”
So, you want to know why wages aren’t rising yet? It’s because there’s still too much slack in the job market. Yes, there are more jobs, but there are still either too many workers chasing them or waiting in the wings to do so that employers don’t have to bid wages up to get the workforce they need.
What to do about that will be the subject of my next post later this week. For now, I’ll say this: The Fed needs to either update its FEUR to match reality or clarify why these variables are not behaving the way they should be if their FEUR is correct.
Data note: The model regresses yearly percent changes of nominal average hourly wage growth on unemployment with one lag and four lags of the dependent variable. The second model adds the labor force participation rate, also with one monthly lag. The estimation goes through Dec. 2012 and then forecasts using actual values for unemployment and lfpr.