To be a good empirical economist, you must be prepared to make use of economic data without forgetting that the data is at best an imperfect guide to reality. I used to describe national income accounting — GDP and all that — as a peculiarly boring form of science fiction. That’s not to say that the statisticians just make things up; they try really hard, and their work is immensely valuable. It’s just that any close look at how the numbers are constructed reveals that data coverage is always incomplete and the gaps are filled in with estimates and imputations.

Lately, however, I’ve found myself drawn to another analogy: Economic measures, especially the measures we use to make sense of a rapidly changing situation, are like the shadows on the wall of Plato’s cave. That is, they’re imperfect images of an underlying reality that exists but that we can’t directly see. And sometimes it’s important, in interpreting the shadows, to think about the Platonic ideal we’re actually trying to discern.

What the heck am I talking about? Inflation, of course, which has been running high the past few months — although consumer prices rose a lot less in July than they did in June.

The big question about recent price increases has been: Are we looking at a transitory shock or a rise in the underlying rate of inflation?

I know a fair number of people, mainly Wall Street types, who get angry at anyone who even asks this question. Inflation is inflation, they insist, and attempts to define “core inflation” are just a way for the Fed to evade its responsibility to maintain price stability. But these critics generally don’t know why the concept was invented in the first place.

The truth is that back in the 1970s, economists noticed a sharp distinction in the behavior of some prices. The price of soybeans fluctuates a lot both up and down, whereas the prices of goods like new cars and the price of labor — that is, wages — seem to change reluctantly. The thing about these sluggishly moving prices is that once they do get moving — say, once they’ve been rising 6% or 8% a year several years in a row — it takes something big, like a severe recession, to stop them from just continuing to rise.


Why this distinction? That’s a fairly deep question, and economists are far from united in their answers. But the difference is real and important. A spike in inflation driven by goods without price inertia is easy come, easy go; inflation driven by goods with price inertia is very hard to get rid of — and to be avoided if possible.

How do we tell the difference? Back in 1975, Robert Gordon proposed that policymakers focus on an inflation measure that excluded food and energy — a rough cut at the distinction between inertial and noninertial prices that made sense at the time. (Remember, this was the era of wild swings in oil prices caused by wars and revolutions in the Middle East, and food prices were also a lot less stable in the 1970s than they have been since.)

Gordon’s suggestion proved so useful that “core inflation” — defined by excluding food and energy — became a standard measure and guide for Federal Reserve policy. And use of that measure has been a huge practical success. The Fed was able to keep its cool through several spikes in inflation driven mainly by oil prices, most recently in 2011, because its focus on the core told it that these were transitory shocks, that underlying inflation remained low — and the Fed was right.

But inflation excluding food and energy was always a quick-and-dirty approximation to the underlying concept — a shadow on the wall of the cave cast by the Platonic ideal of inflation in goods with inertial prices. And while this approximation worked well in an era of oil shocks, it’s not working well at all in an era of pandemics and vaccines, in which a remarkable amount of price action has been driven by used cars.

Nor are used-car prices the only price we didn’t used to think about much but that is having wild swings and should, conceptually, be excluded from core. In normal times, macroeconomists don’t pay much attention to shipping costs. But pandemic-related disruptions have created an incredible surge in the price of container shipping.

I’ve been trying to estimate how much shipping costs may have contributed to recent inflation, multiplying the reported change in the cost of shipping containers to the United States by the number of TEUs — 20-foot equivalent units — unloaded at U.S. ports. There’s quite a lot of uncertainty in these estimates, but as a rough guess, shipping may have added between one-quarter and one-half of 1% to inflation over the past year. This, too, should be excluded from the Platonic ideal of core.

Why does all this matter? As best I can tell, a fair number of people are still looking at the standard measure of core inflation — which has risen almost as much as headline inflation — and concluded that we really do have a fundamental problem. They could be right, and Team Transitory — economists who believe that this is a transitory blip, a group that includes the Biden Council of Economic Advisers — could be wrong. But you can’t settle that argument by looking at a number that, however well it worked in the past, is now a clearly inadequate measure of the underlying concept of inertial inflation.