E-piphany

I have neglected to post here the links to recent TV appearances. Here’s one on TD Ameritrade network from February 12th; here’s one from Bloomberg TV on February 14th, the day of the shocking CPI.

I get asked frequently about Core PCE inflation. Because the Fed obsesses over Core PCE, as opposed to one of the many flavors of CPI (core, median, trimmed-mean, sticky-price), investors therefore obsess over it as well.

My usual response is that I don’t pay much attention to Core PCE, for several reasons. First, there are no market instruments that are remotely tied to PCE, so you can’t trade it (and, for the conspiracy-minded among you, that means there is no instrument whose market price can call shenanigans if the government figure is ‘massaged’). Second, while PCE is interesting and useful for some uses – it measures prices from a different perspective, mainly from the supplier-side of the equation so that, for example, it captures what Medicare pays for care as opposed to just what consumers pay – those aren’t my uses. Markets respond to inflation, and to perceptions of inflation, but what the government pays for healthcare isn’t something we perceive directly.

So, I care about PCE more than, say, PPI, but only just. The only reason I care about PCE is that the Fed cares about it.

Now,  PCE differs from CPI in a couple of key ways – apart from the philosophical way mentioned above, that one measures the price of things businesses sell and one measures the price of things people buy. But those key ways are mostly interesting to pointy-head economists who are interested in calculating the third decimal point. Me, I’m just trying to get “higher” or “lower” correct. (Ironically, those folks who are interested in the third decimal point are the same folks who miss the big figure in front). So they wail at the following chart (source: Bloomberg), and moan about how the Fed has been unable to get inflation higher because of this persistent shortfall of PCE compared to CPI. Try harder!

The pocket-protector set can keep their ‘formula differences.’ There’s really only one important difference that has caused the gap between PCE and CPI over the last half-dozen years: the difference in the weight allocated to housing. In the PCE, “Rental of tenant-occupied nonfarm housing” and “Imputed rental of owner-occupied nonfarm housing” add up to about 16% of the overall PCE. In the CPI, “Rent of primary residence” and “Owners’ equivalent rent of residences,” the corresponding categories, sum to about 32% of overall CPI.

Housing is both the largest weight in CPI, as well as one of the most stable parts of CPI. So, when shelter costs are running ahead of “the rest of CPI”, then Core CPI tends to be above core PCE. I’m actually soft-pedaling that. Given how shelter inflation has been relatively elevated now for some time, it is far and away the most important difference in CPI and PCE. So much so that I can create the following chart in a couple of minutes: I merely took Core CPI and backed out half of the weight of housing inflation, and compare that to Core PCE.

Remarkably, making that simple, back-of-the-envelope adjustment puts core CPI right smack on top of core PCE.

The implication is that in choosing to focus on Core PCE, rather than Core CPI[1], the Fed is saying that housing is just not as important as it seems to consumers. Although you spend about a third of your money on shelter (about 40%, after we take out food and energy), the Fed is behaving as if that inflation only matters about half that much. I don’t think the central bankers are doing this on purpose; I think they believe that the PCE is a more-pure economic statistic that perhaps gives them a better read on ‘overall’ inflation. But in this case, they’re effectively biasing their monetary policy looser for one reason: because housing costs are already going up faster than overall inflation.

Now, to me that sounds like they’re biased the wrong way. But the academics will write a paper in five or ten years and explain why that’s wrong, even though it turned out badly.

[1] Median CPI is better yet, but let’s take baby steps.