By Paul Gomme and Peter Rupert
The BLS announced the CPI and PPI for June. The CPI rose 0.18% from May to June on a seasonally adjusted basis, 2.18% when annualized, and was up 3.1% year over year. Rather than strip out a lot of goods that comprise the CPI, we look at the CPI on a three month average basis. As shown in the graph, the year over year continues to decline while both the annualized monthly and the 3 month ticked up slightly. The CPI-X (ex food and energy) differs, however. The annualized monthly rate grew 1.91%, below the Fed’s 2% target.
Of course, the question the Fed faces comes down to: What has been happening to the trend in inflation? Given the 2% target, the Fed hopes to see progress toward that goal. But, as we have commented on in several previous posts, what measure of inflation is the best indicator? Headline CPI (PCE)? Core CPI (PCE)? The Fed’s new Supercore CPI (PCE) (the price of services excluding energy and housing services)? Paul Krugman (NYT, July 11) has his own measure of supercore: “My preferred measure these days is “supercore,” which excludes food, energy, used cars and shelter (because official measures of housing costs still reflect a rent surge that ended a year ago.)” And goes on to say he is not a fan of the Fed’s measure, “The Fed has a different measure of supercore — non-housing services — but when you look at the details of that indicator, it’s a dog’s breakfast of poorly measured components that I find hard to take seriously.” Here is a graph of the Fed’s supercore measure:
So what are we to make of all of this? The main problem is that all of these measures are purely arbitrary; there is no underlying theory as to how, or what to measure. The BLS states, “Inflation can be defined as the overall general upward price movement of goods and services in an economy. The U.S. Department of Labor’s BLS has various indexes that measure different aspects of inflation.” These new so-called supercore measures omit various categories of goods that are purchased in the market. The overall CPI is comprised of 243 commodities and services measured in 32 different geographic areas. To show just how big these omissions are, the following table recreates a subset of the BLS table where the relative importance is the share of total expenditure used to weight the respective change and the May 2023-June 2023 column gives the monthly change:
| Expenditure category | Relative importance | May 2023-June 2023 |
| All items | 100% | 0.2 |
| Food | 13.4% | 0.1 |
| Energy | 6.9% | 0.6 |
| All items less food and energy | 79.7% | 0.2 |
| Services less energy services | 58.3% | 0.3 |
| Shelter | 34.7% | 0.4 |
The table above shows that the core CPI includes about 79.7% of all expenditures. Removing shelter from the core means the price measure includes only 45% of the things we buy. If only services prices are included that would vastly reduce the set of prices used to calculate inflation.
The BLS announced that the Producer Price Index (PPI) for final demand increased 0.1% from May to June or 1.65% annualized. Year over year it increased 0.24%. Our 3 month measure fell -0.69%, the third consecutive month with a decline.
Trend Inflation
What we (the writers of this blog) mean by “trend inflation” is a series that tracks the movements of actual monthly inflation but somewhat smoothed (taking out the more extreme fluctuations). As we’ve discussed before, this is a signal-extraction problem. Every month, we get a new CPI number from which we want to make an inference concerning the current trend inflation rate. Trend inflation is the “signal”. However, the news (the new CPI number) is a noisy indicator of this signal, both because the CPI is reported in levels and because of transitory factors affecting monthly inflation. Further, the nature of the data is such that we cannot easily extract the signal (trend inflation) from the data. With all of this in mind, the reason for looking at core or supercore CPI inflation is that it has a higher signal-to-noise ratio than CPI inflation. Both seem promising in that they omit the more volatile price components of CPI: in the case of core CPI, food and energy; in the case of super core CPI, lots of other stuff. However, the charts above show that over the past couple of years, core CPI inflation has been running well above actual CPI inflation while the Fed’s supercore inflation has been well below actual.
An alternative to slicing and dicing CPI is to compute inflation over a horizon longer than a month. The idea in this case is that the monthly noise is an independent draw (in the jargon of economics, it’s independently and identically distributed) while trend inflation has memory and changes relatively slowly over time. Consider headline inflation. Typically, it’s calculated as the 12 month percentage change in the CPI. But a different way to think about this calculation is as the average of the 12 one month inflation rates. If there is no change in trend inflation over this period of time, the 12 noise terms will have a mean of roughly zero, and the year-over-year inflation rate will provide a good estimate of trend inflation. The rub is that there have been changes in trend inflation over the past couple of years. Consequently, the year-over-year inflation rate also takes an average of trend inflation over the past year. This is why, in the chart above, year-over-year inflation has consistently exceeded the monthly inflation rate for the last year and a half.
But there’s nothing magic about year-over-year inflation rates, and we’ve previously advocated for annualizing the three month inflation rate as a good compromise between capturing changes in trend in a timely fashion, and seeing through the noise in monthly inflation rates. By construction, the three month inflation rate tracks movements in monthly inflation while removing much of the noise – as can be seen in the chart above. One thing that the three month inflation rate lacks is a fancy name like supercore; oh, and the NYT soap box.
