How Fed policy ran amok by focusing on ‘the wrong stuff’
(AKA our ever-changing job market)
The Fed has been making policy with a strong reference to the job market. While we have heard all sorts of criticism about the flaws in GDP, some have acted as though job market data are equivalent to the Holy Grail. Several Fed officials recently have even said quite explicitly that jobs data trump GDP. In what way is that?
All data are subject to revisions of various sorts. And jobs data can be revised and can surprise as they did last month. Apart from that, job data can look friendly and familiar and yet be distant and misleading. With the labor force participation rate so low and the not-in-the-labor-force cohort so large plus myriad demographic changes who can say that we know what these current labor force metrics mean? Can you? The Fed though it could. But it couldn’t.
I am quite serious. Consider the table below:
This table presents labor market data for the labor force and unemployment rate for those age 25 and up. It provides four exhaustive categories for labor: (1) Less than a high school diploma, (2) high school diploma no college (3) college but less than a BA (4) BA and higher degrees. Roughly we have unskilled, low-skilled, semi-skilled and highly-skilled categories.
The table shows how the composition of the labor market by those four categories has shifted. When data on these metrics began in 1992, 13% of the labor force was in the lowest skill group. In May of 2016 that proportion is down to 7.6%. The top-skill, BA-plus category used to comprise 26% of the labor force, now it comprises 38.9% of it. The proportion that is low-skilled is now slightly lower and the semi-skilled group is proportionately higher.
Despite all the job openings that cannot be filled in the JOLTS report, the labor market is much more educated or skilled than it used to be. Still that is not working –apparently there is a skills mismatch. It’s like having enough silverware to set the dinner table for eight but with too many spoons and not enough knives. It doesn’t work. Even if they are the highest quality silver, a spoon is not a knife.
But, what is still working is that higher education levels are correlated with lower rates of unemployment. If our labor market were of the composition it had been in 1992 the current unemployment rate for all workers age 25 and over would be 4.4% instead of the current 3.9% - one half of one percentage point higher. It appears that our labor market gauge is no longer so easy to read. But it’s good news. It’s like having a gas gauge that is shifted so that the tank is not nearly as empty as the gauge is near ‘E’. You are not as at risk as you think you are to having to walk for more gas...
The education composition of the labor force has lowered the unemployment rate by one half of one percentage point. And this is only one demographic feature I have isolated. There are other features that are correlated with the rate of unemployment that can be analyzed in this way, among them age, sex, race and more. As we get compositional shifts in these areas either the structural unemployment rates among these various attributes will shift or the impact on the overall unemployment rate will shift or both will shift.
This is why the Fed has seemed so out of step. It has focused on and has been fearful about the rate of unemployment. Just because the rate sort of seems like the same thing it has always been does not mean that its attributes have not changed. The evidence for education is that the unemployment rate is going to look a lot lower just because of who is in the labor force now compared to who is no longer in it.
As arguments go this one sees the labor force as tilted to more skill and more education. But what it also suggests is that the less skilled and less educated have been squeezed out. We do not have the same data on the attributes of those who are not-in-the-labor–force for example so we can’t really tell.
And, as the minimum wage rises, it is likely that firms will find ways to hire more skilled workers at higher wages possibly using technology solutions to eliminate jobs and become more streamlined and that could result in an even higher skilled labor force.
Try to understand the labor force development as a dynamic process, where supply and demand combine to determine who is in it and who is out. Always bear in mind that compositional characteristics of those NOT in the labor force may be quite different from those who are in the labor force and that might keep those who are on the sidelines on the sidelines.
The table also gives us some notion of how much slack we might have left in this labor market.
For example since January 1992 the overall unemployment rate has been lower than its May 2016 level 23% of the time and it has fallen an additional 19% from its current level (minimum rate =3.8%) to its low.
For each of our age cohorts the unemployment rate has been lower: over 20.1% of the time for the lowest skills group, 48.5% of the time for the second lowest skill group, 37.5% of the time for the semi-skilled and 41.0% of the time for the high skill group. Interestingly despite the focus on skills shortages the highly educated group does not have a really low rate of unemployment relative to its own historic standards-especially relative to the other categories- it is in fact the least tight on this metric.
The percentages of time that each cohort’s rate has been lower also echoes the next set of statistics which concern how much farther these cohorts’ unemployment rates can fall to get back to their respective lows (since 1992). The answer is that unskilled unemployment can fall the least, by only 18.3%, and all the other categories can fall in a relatively tight range of from 37.3% to 38.5% further-all about the same order of magnitude. Oddly it is the unemployment rate among the unskilled that is the relative lowest (has the least far to fall to hit its all-time low).
This prompts us to ask why? Are these ‘unskilled’ workers truly scare? Or are they so low-paid that they more easily leave the labor force and become no longer counted as unemployed because government support programs may look more attractive than the low paid wage?
Apply 1992 weights to all the historic data and we find that the lowest age 25 and up unemployment rate allows for a 30.4% fall while the lowest 2016 weigh pattern would foresee a further drop of 32.1%. The 1992 reweighed unemployment rates are lower than their actual May 2016 level 25.2% of the time while the 2016-weighted data have been lower 26.9% of the time. Under the up-to-date weights the unemployment rate can fall by more.
Any way you slice it the new-view of the labor market does not seem quite so tight. There is still some further 30% or more that unemployment rates can fall in order to get back to previous cycle lows looking at a fixed compositional weight for the labor market.
Moreover, we can look at a decomposed Phillips curve over the last three cycles to gain some further insight on what risks we may face in terms of wage repercussions.
In the chart above we show wage changes (one year % change in Average Hourly Earnings) Vs the unemployment rate for this cycle and for the average of the last two cycles. We present these data by arraying them by lining them up starting with when the unemployment rate first reached 6% in each cycle, possibly creating some wage tension. We color code the two cycles with unemployment dark blue and wages light blue in the two previous ‘averaged’ cycles and with unemployment dark green and wage light green in the current cycle. What we see is that the unemployment rate has been falling MUCH FASTER in this cycle after it reached 6% compared to the last two cycles. But that at this lower unemployment rate the wage gains in the last cycle would have been much higher at 4% instead of around 2.5%, currently.
Of course there is still evidence here that the Phillips Curve is ‘working.’ The unemployment rate is falling and wages are rising, but the relationship is much diminished. Since the Fed was basing its policy decisions on this relationship the fact that it has been muted had caused Fed policy plans to become overly aggressive.
In short the Fed was simply too worried too soon about the level of the unemployment rate. And we have given several reasons above why this is might be so.
The macroeconomics of the situation has been just too unstable to work. That is an admission that there are too many compositional shifts in the labor force for us to treat all changes in the unemployment rate as equivalent in nature to WHAT THEY HAVE BEEN IN THE PAST.
But there are macroeconomic diagnostics to reveal this too. It’s just that the Fed has scrupulously (unscrupulously?) avoided presenting them.
The Beveridge Curve
What: The Beveridge curve shows a SHIFTED relationship between job openings and the unemployment rate.
Meaning: it now takes MORE job openings to reduce the rate of unemployment
Missed analysis: While everyone has been ‘oohing and aahing’ over the high count of job openings, economists have missed the fact that it takes more job openings to reduce the rate of unemployment than it used to (and to create a hire). In other words ‘openings’ are worth less than they used to be.
This macroeconomic diagnostic, The Beveridge Curve, has been ignored
Economics usually does not dirty its hands with such things. But because there are so many changes in the economy it is no longer wise to deal in such aggregated data. Yet that is exactly what macroeconomic models do. It is much wiser to decompose the data to see what the macro data really mean in times such as these and not to simply swallow the model forecast hook line and sinker. When we take the time to do that we find that the job market data are really not as reliable as many assumed.
We can make the same points about nonfarm payroll data as we have about the unemployment rate. And those criticisms may be more familiar. Economists (some) have been long complaining of the quantity of low skill jobs being created. We have a service sector economy and it churns out more low low-skill low-paid jobs. Indeed, this is one of the reasons that productivity growth is so poor. Low productivity growth is not an exogenous development. Productivity is low also because investment is low. Investment is low because the US skill pool seems to be lacking some of the more specific skills needed despite the proliferation of highly educated workers. Moreover, the strong dollar makes the US uncompetitive (makes US wages higher in foreign currency terms) and this scares off investment. We have many interrelated problems here. But it comes down to trying to understand the micro-economic foundations of the macro-economic failures. Not all failures will simply reverse themselves. Economies do not necessarily grow out of them into success stories.
The Fed simply does not seem to have spent enough time or effort trying to understand why its pet macroeconomic statistically sophisticated metrics had stopped working. The answer is always that the devil is in the details. It was the same answer this time around.