Surely we’re behind some curve!

Let us take a moment explicitly to place the CFE in the pro-facts camp. We are honored to work at Johns Hopkins University, the first modern research university in the U.S., founded on a hard-nosed commitment to facts and with a mission to discover facts and to teach how to deploy them constructively. Ok, got that out of our systems.

The economic policy debate is currently focused in part on just how close we are to full employment, on how much slack may remain in the labor market, and on how much risk we face of economic overheating due to excessive job growth. These are contentious political topics, and so we probably should not be surprised that we’ve recently seen the full range of crimes against economic facts. There has been a complete disregard for the meaning of data by the Trump administration, a failure to pay attention to simple details of data by some analysts, and perhaps most insidious, a persistent clinging to folk wisdom by conventional macroeconomists in the face of facts.

Bigger than Bowling Green?

At his Jan. 11 press conference, the soon to be soon-to-be-president declared,

…millions more workers right now in the United States — 96 million really wanting a job and they can’t get. You know that story. The real number – that’s the real number.

If true, this would be a Great-Depression-sized tragedy, but it is actually more of a Bowling-Green-Massacre-style tragedy.

The 96 million number is the estimated number of working-age folks who report themselves as not looking for work, and is taken from the BLS’s household employment survey. The vast majority of these folks are not “really wanting a job.” For example, nearly 30 million of these folks are over 70 years old, 20 million or so are disabled, approximately 10 million are married and living with their spouse (think Melania—until she starts up that profitable brand), and millions more are high school and college students.

Many others have already explained why Trump’s claim is ridiculous. We bring this up because of a feature that particularly grated on our fact-oriented sensibilities: one group omitted from the 96 million is the roughly 7.5 million people who are actually unemployed and looking for work: the president-to-be overlooked the main group he was ostensibly promising to help. (Is there a metaphor in there somewhere?)

Was the January household survey strong or weak?

Staying on the theme of the BLS household survey, and in the “black means white” world we seem to be living in, we come to last Friday’s employment report.

The labor report includes two sources of information on jobs. The establishment survey tallies payrolls reported by businesses while the household survey reports a survey of households regarding the employment status of members of the household. The establishment survey reported job gains of 227,000 for January 2017, while a naïve reading of the household survey suggested a net loss of 30,000 jobs. This apparent contradiction led a host of analysts to conclude that the “jobs data didn’t add up… The household survey of employment was significantly weaker than the payroll survey, showing a net loss of 30,000 jobs in January.”

Perhaps the jobs data don’t add up, but, if anything, the household survey was surprisingly strong, not weak. The false naïve reading comes from ignoring technical details regarding an adjustment in the data that occurs each January. The short explanation is that, every year the jobs data for January, which are released in February, include a population adjustment. But BLS does not adjust the prior data. Thus, subtracting the December jobs total from the January total does not give an apples-to-apples comparison. Each year, the BLS report explains these facts and gives the appropriate job gains number. This year, the January job gains, as gauged by the household survey, were a whopping 457,000.


It is easy to pick on outrageous and merely sloppy misrepresentations of labor market facts. We are still left with the substantive policy question: Is employment currently surpassing its maximum sustainable level? Have we reached the point at which more people getting jobs is a bad thing? More carefully,

Would it be better for the economy overall if the folks who are really wanting jobs found them more slowly?

Keeping people out of work longer for the greater good is the sort of difficult judgment call that national policymaking sometimes requires, and it would be foolish to shy away from facing this fact. But it would be even more foolish, in our view, to base the decision on a folk wisdom of macroeconomics that has never been supported by facts.

Conventional macroeconomists speak confidently as if there is some ‘natural’ value for the unemployment rate, and that as the unemployment rate falls beyond this cutoff, rising inflation and other potentially destabilizing excesses reliably follow. Let’s agree that some version of this ‘Phillips curve’ mechanism operates in reality, along with the myriad other forces linking employment and inflation. The folk wisdom is that macroeconomists understand this mechanism well enough, and can measure the natural rate well enough, to predict when we’re falling behind the curve. Unfortunately, the economy throws lots of curves, and as we’ve argued in a number of posts,[1] nobody has a good record calling when we are behind the inflation curve.

Specifically, history makes pretty clear that when the Fed acts to stem rising or falling inflation based on an assessment of labor market slack, but without confirming evidence from inflation itself, the Fed has fairly generally been making a mistake.

The great inflation of the 1970s is one such mistake, and policy during the deflation scare in the early 2000s is another. Greenspan overruled the folk wisdom in the late 1990s or that period would provide another example of tightening policy to fight inflation that was not, in fact, on the way.[2]

We suspect that the folk wisdom regarding the Phillips curve persists because, while the mechanism truly does operate, macroeconomists are loath to accept the fact that we see its effects clearly only in the rearview mirror. As the unemployment rate falls, one of two things happens: either inflation begins to rise, or mainstream analysts figure that the natural rate must be lower and mark down their estimate. When inflation does begin to rise, the mainstream analysts decide more confidently that we have, indeed, reached the natural rate.

This process is illustrated by the last 4 years of FOMC estimates of the longer-run normal rate of unemployment as reported in the Survey of Economic Projections.[3] From Dec. 2012 to Dec. 2016, as the unemployment rate fell but inflation did not rise, the FOMC steadily lowered its estimate of what must be normal. The upper bound of the central tendency for the normal rate fell on average more than 2 tenths of a percentage point per year, the lower bound fell a bit more slowly.[4] If tradition holds, this process of downward adjustment will continue until the Phillips curve mechanism finally comes to dominate the other effects on inflation. The downward-drifting estimates over the past 4 years were presumably part of the consistently-erroneous FOMC forecasts of rising inflation over this period.

Defenders of the Phillips curve folk wisdom often appeal to metaphor rather than fact: behind the curve, overheating, genie out of the bottle, can’t wait until we see the whites of their eyes. We ask this: Where is the episode in which inflation jumped from persistently too low to painfully out of control without allowing ample time for a sensible, moderate, and adequate policy response? Since WWII, we’ve had one main episode of losing control of inflation, and inflation in that episode ramped up slowly over more than a decade.

Bottom line

We agree with most folks that at some point as the labor market tightens, inflation will move upward. For what it’s worth, we suspect that the economy is very near that point; the inflation data now seem to be starting to confirm that suspicion. Thus, the Fed’s cautious and gradual upward moves in the policy interest rate seem justifiable to us. And the Fed shows every sign of remaining on this path unless conditions change markedly.

Absent clear change in the inflation data, however, arguments that the low unemployment rate calls for a sharp policy change fly in the face of hard facts. Similarly, we think that moving preemptively to offset policies that might or might not be coming from the new administration would not be prudent. But that is whole different kettle of, well, something smelly.


1. For example, Six degrees… and Six degrees … in practice and the research cited therein. [back]

2. Faust and Leeper describe these examples in more detail. In the 1970s, the Fed arguably ignored rising inflation, in part, due to its view that there was significant slack in the labor market. In the early 2000s, despite little change in inflation and a rapidly growing economy, the Fed saw high deflation risk because of its erroneous estimate of slack and its effects. In the late 1990s, the FOMC saw an overheating economy with rising inflation on the way, but Greenspan famously overruled the committee, believing instead that a productivity boom was underway. Greenspan’s handling of policy in this period may or may not have been sound, but he was right that the tight labor market was not reliably signaling rising inflation. [back]

3. Links to the quarterly surveys can be found here[back]

4. The upper bound fell from 6.0 to 5.0. At the beginning of this period, no median was reported, so we are reporting the central tendency of the estimates instead. [back]