As OTEers know, I track various measures of wage and compensation growth. Five—count ‘em!—show up in the wage mashup, recently updated.
One increasingly popular series I’ve not included is the Atlanta’s Fed Wage Tracker (FWT). It’s an interesting and useful contribution, so let me make a few points about this series: what it shows, why it doesn’t belong in the mashup, and importantly, given that it’s accelerating faster than other series, what it says about wage-push inflation.
The figure below shows the most recent version of the FWT. While my mashup was last seen growing at around 2.3%, the FWT is popping along at 3.4%. Whussup widdat?
In fact, it’s a very different beast from all the other series, with two major differences (I first saw this approach in a Jesse Rothstein paper from a few years back). First, while all the other series take snapshots of wages in period 1 and compare them with snapshots from period 2, the FWT measures the growth of the same workers over 12 months. If Jane earned $10 in April of 2015 and $10.50 in April of 2016, her wage growth is recorded as 5%. In the “snapshot” series that we’re more used to, Jane’s wage growth is not tracked. We’re just comparing averages or medians of a sample of all wage earners (or blue-collar workers, or whatever) in two different periods.
By looking at continuously employed workers (actually, workers employed in both time periods), the FWT will typically show more growth than other series, since it’s including an “experience premium,” the bump to wages some workers enjoy as they age (or, similarly, as they add another year of tenure at their job).
The FWT is also, and this is an important attribute, less susceptible to compositional or demographic changes. If an improving job market pulls in a bunch of low-wage workers, the snapshots will reflect slower wage growth than the FWT, which excludes new entrants. This recent analysis by economists at the SF Fed shows this cyclical dynamic, along with a secular one of aging boomers leaving the job market (which also tends to push down wage growth since these leavers have higher wages), to be in play in the current wage-growth story.
The second unusual aspect of this wage series is that the median plotted above is not the median wage. It’s the median growth rate. They calculate the wage growth of people like Jane who were working in the survey over the course of a year, and construct a distribution of growth rates from which they plot (a 3-mo. moving average of) the median. That means the median growth rate in any given month could be for low-, high-, or mid-wage workers (that’s why I don’t think it belongs in the mashup).
So, besides the findings from the SF Fed re the impact of secular and cyclical trends holding back the broad measures of wage growth with which we’re more familiar, is there anything else to learn from this series? We know that continuously employed workers, especially full-timers, are more likely to experience wage growth than others, so the fact that this series grows faster than the mashup isn’t surprising.
But is columnist Robert Samuelson correct to conclude that these findings should exert “pressure on the Fed to raise interest rates,” as they reveal a tighter job market goosing wages more than we thought?
The SF Fed authors suggest maybe not: “As long as employers can keep their wage bills low by replacing or expanding staff with lower-paid workers, labor cost pressures for higher price inflation could remain muted for some time.” They do, however, note that if low-wage entrants are “less productive,” that could create inflationary pressures.
I examined such linkages in a recent post wherein I argued that any pass-through from wage growth to price growth was awfully hard to tease out of the data, even controlling for slower productivity growth. Using a simple method I describe in the post, “I simulate an increase in the wage variable and estimate the impact on price growth.”
Let’s apply that same test to the FWT.
The first figure shows that a one standard deviation spike in the FWT wage growth series has no impact on core PCE inflation over the course of the next 12 months. The second figure shows the same result for full-time workers.
This shouldn’t surprise you. A look at the first figure above shows that even FWT wage growth remains below where it was in the full employment 90s and even the less full 2000’s. If anything, this measure shows we’re still pretty far from full employment. Moreover, there’s increasing evidence that wage-push inflation isn’t so much of a thing as it used to be.
So, happy to add the FWT to the mix, but a) it’s not signaling inflationary pressures, and b) once you extract its built-in premiums, I don’t think it’s telling a very different story than the mashup series. The job market’s tightening, and that’s giving workers a bit more bargaining clout to push for wage gains. To which I say: it’s about time!
UPDATE: See Dean Baker’s smart take on this series (the one used in the SF Fed piece). He stresses a composition effect embedded in that series, stemming from an increasing selection bias.
Data note: The statistical analysis uses monthly, year-over-year percent changes in the core PCE deflator and the FWT. These are run through a VAR (vector autoregression) with 6 lags of the price and wage variables, and one control variable: the annual change in the broad trade-weighted index of the dollar. The figures show impulse-response functions of a one standard deviation shock in the wage variables on the price index (PCEC), with confidence bands of two standard errors.