Core Inflation: What Is It Good For?

Lots of discussion now about whether the Federal Reserve should be looking at core inflation.  Core, in the typical definition, is total inflation excluding food and energy prices.  On the surface, this seems silly:  food and energy are significant parts of our lives.  That’s certainly true of me.  I eat food most every day, and drive like crazy.

But the Fed doesn’t really care too much about my daily life.  Oh sure, Ben checks in on how I’m doing every now and then, but I’m not sure how much he really cares.  What the Fed is trying to do is keep overall inflation under control.  But food and energy contribute a lot of noise to the data.  When inflation spikes up because of oil, the Fed’s board members rub their chins and wonder, are we keeping inflation under control or not?  Maybe the oil price impact on inflation is transitory.  (Note that oil prices don’t have to come down for the spike to be transitory; so long as they stop rising eventually, the effect of oil on inflation will be temporary.)

Over the long haul, total and core inflation tell the same story:


But over the short-run, they are very different:


One of the hardest things to do in economics figure out transitory changes from underlying trends. Core inflation helps do that.  Here’s the simple experiment that motivates the use of core inflation.  Let’s say that I want to predict future total inflation based on current or past inflation.  Which measure of current and past inflation should I use?  This is a simple empirical question.  The answer is that core inflation does a better job of measuring future total inflation than does current total inflation.  So when the Fed wonders if they are keeping inflation under control, core inflation is a better gauge than total inflation.

Is core inflation the best measure?  It turns out not to be.  The best seems to be the "16% trimmed mean core inflation" measure.  What’s that?  Imagine laying out in a table all the individual components of inflation.  Corn is on one line, motor oil on another, haircuts on a third, etc.  Now, pull out 16% of the items (based on their weight in the total index) that are farthest away from average.  That is, compute an average after trimming out the outliers.  This measure does better than the regular core inflation at predicting future inflation, and thus telling the Fed how it’s doing on keeping inflation under control.

The Federal Reserve Bank of Cleveland computes the trimmed mean CPI; here’s the recent data:

David Altig over at MacroBlog has a good discussion of this issue.