PCE, you complete me

Predicting Inflation from the Fed’s favorite Price Index

This week, just before Valentine’s Day, the Bureau of Labor Statistics released the January 2024 inflation rate based on the Consumer Price Index, which came in at 3.1 percent. That is close to the rates recorded for November (3.1 percent) and December (3.3 percent) of last year. Inflation is holding steady according to the CPI, but not yet showing signs of further decline. Hopefully the latter fact did not kill the mood for anyone’s Valentine’s day.

BE MINE

The good news, however, for anyone pining for interest rate cuts sometime soon, is that the rate of inflation as measured by the Fed’s favorite price index, the Personal Consumption Expenditure Price Index—or PCE for short—equaled 2.64 and 2.60 percent in November and December of 2023, respectively (measured as the percent change from a year ago; the estimate for January 2024 will not be out until the last week of February).

The Fed prefers the PCE index since it covers all consumer spending in the economy (whereas the CPI focuses on a basket of “regular” household items, as explained here by Macrosight). Figure 1 displays the PCE at the monthly frequency from 2019 through December 2023 (measured as the percent change from a year ago).

Like the case with the CPI over this time period (shown and discussed here), the PCE visual is dominated by the Beastflation of 2021 – 2023.¹ Yet, the PCE and CPI are not exactly the same, and it is the former to which the Fed primarily turns its gaze.

The PCE price index is also particularly intriguing, since, like consumer expenditure itself, the PCE price index is the sum of three sub-price indices—one each for consumer durables, nondurables and services. By examining each sub-price index, we can take a stab at predicting what the PCE rate of inflation will turn out to be for January, 2024.²

RED HOT

Figure 2 displays the rates of inflation for the three PCE sub-indices for consumer nondurables, durables and services (at the monthly frequency from 2020 through 2023 and measured as the percent change from a year ago).

Figure 2 reveals a couple of interesting things.

1. Beastflation was primarily driven by the two goods categories. The most aggressive increase occurred for nondurable goods. Nondurables inflation peaked at 13 percent in June of 2022, about four months later than the rate of inflation for durables (which peaked at 10.7 percent). The increase in prices for services was comparatively more gradual, peaking in early 2023 at 5.5 percent.

2. Since their respective peaks, the rates of inflation for each consumption category have declined. Durables has had persistent deflation the past seven months of 2023, while nondurables inflation steadily declined up until August 2023 (and has been up and down since). Services inflation, on the other hand, has declined gradually over the past year. Yet, overall the disinflation of services has been far more gradual than for nondurables and durables.  

How can the evolution of these three sub-price indices help us predict what’s going to happen with overall PCE inflation in 2024? Well, “predict” is a strong word. But, we can at least consider some “what-ifs” by making use of two consumption-related facts.

The Fed + PCE 4-Ever

FACT 1

The first fact relates to the historical averages of the sub-price indices, compared to what has happened recently. Table 1 displays the average rates of inflation for each consumption category from 2000 to 2019 versus 2020 to 2023 and the rates of inflation during Beastflation, as well as the peak rates during that period.

There are a number of interesting details in this table.

1. The average rates of inflation from 2000 to 2019 were relatively low. This is not news (as noted with CPI data in the Beastflation Macrosight post). However, what may be surprising is that the average rate of durables inflation over this period was negative. That’s not to say prices dropped every month for durable goods for two decades. Rather it means that on average the declines out-weighed the increases.

2. The average rates of inflation from 2020 to 2023 have been higher for each category by significant amounts. The rate of nondurables inflation, for example, is more than double the historical average. Nondurables includes food and beverage purchased for home, so the behavior of this series over this time period should come as no surprise to anyone that has been to the grocery store the past two years.

3. And, of course, the averages over the Beastflation period are remarkably higher than the historical averages for each category (as is the rate of inflation shown in the bottom row of the table for total consumption).

Given this historical juxtaposition, the disinflations for each category of consumer spending since their Beastflation peaks suggests the rate of inflation for each is converging back to those historical averages. At least, that is an optimistic reading of this information.

FACT 2

The second fact relates to the shares of the consumption categories relative to the total. Figure 3 displays the shares of each category to total consumption expenditure since 2020, along with the average for each over this time period (these shares are relatively stable going back further in time, as noted in this post).

These shares are important since the overall PCE price index is a weighted average of durables, nondurables and services. That means that price changes for services has the largest effect on the overall PCE index, nondurables has the second largest effect, and durables has the smallest effect.

For example, the PCE rate of inflation for December 2023 equaled 2.6 percent. The rates of inflation for each category, and their relative shares of the total, reveal how that number came to be. This relationship is expressed as follows:

PCE Inflation = (Services Share × Services Inflation )

+ ("Nondurables Share × Nondurables Inflation )

+ (Durables Share × Durables Inflation )

With numbers, we get:

PCE Inflation = (0.66 × 3.9%) + (0.22 × 1.3% ) + (0.12 × (-2.3) )

PCE Inflation December 2023 = 2.6%

Okay, at this point, I suspect, like my students watching the clock as I drone on about macro topics, you are wondering what my point is with this demonstration. The point is the breakdown of consumer spending, as defined by the shares of each category and the historical rates of inflation for each, provides insight into what to expect for PCE inflation in the coming months.

To predict what may lie ahead, let’s use the average rates of inflation for each consumption category the past six months of 2023 as our historical inputs. These values equal 4.5, 1.5, and -2.0 percent for services, nondurables, and durables, respectively. Assuming the same shares as above, we input the inflation averages from the past six months to generate the following “prediction” for the PCE rate of inflation for January 2024:³

PCE Inflation = (0.66 × 4.5% ) + (0.22 × 1.5% ) + (0.12 × (-2.0))

PCE Inflation January 2024 (predicted) = 3.1%

That “predicted” number of 3.1 percent is not altogether surprising since PCE inflation averaged 3 percent the past six months.

How about if the rates of inflation unexpectedly converge back to their 2000 - 2019 averages (shown in Table 1)? In that case, the “prediction” for the PCE rate of inflation for January 2024 is as follows:

PCE Inflation = (0.66 × 2.6% ) + (0.22 × 1.7% ) + (0.12 × (-1.9))

PCE Inflation January 2024 (predicted) = 1.9%

That value, 1.9 percent, is pretty close to the historical average of PCE inflation (equal to 1.8 percent, as shown in Table 1).

XOXO

Of course, it is unlikely that the PCE rate of inflation will suddenly “disinflate” back to its historical average in one month, from 2.6 percent in December 2023 to 1.9 percent in January 2024. But, believe it or not, the potential for that outcome is there. The average rates of inflation for both durables and nondurables the past six months of 2023 were pretty darn close to their pre-Covid-era averages—1.5 vs. 1.7 percent for nondurables, and -2.0 vs. -1.9 percent for durables.

It is the services category that is currently keeping the PCE rate of inflation above its historical average. While during the Beastflation era services never spiked as much as nondurables and durables, the rate of inflation for this series has proven the most stubborn. The rate of services inflation is still double what it averaged from 2000 to 2019.

As we look ahead to the coming months—and wait with bated breath the Fed’s first interest rate cut—the stubborn disinflation of consumer services is the category to watch with the most scrutiny.⁴

That is not to say, of course, that factors affecting nondurables inflation and durables inflation could not flare up and change the recent trends within those categories. But, at least over recent months, those two categories are near their historical averages. As of now, it is the services category within the Fed’s favored PCE price index keeping those yearned-for-interest rate cuts at bay.

1. Macrosight identified Beastflation as lasting from April 2021 through May 2023.

2. Dear reader, I apologize in advance that this post is longer than normal for Macrosight. Truth-be-told, I could have edited it down. But I didn’t. The heart wants what the heart wants.

3. I am using the word “predict” in a colloquial way in this post. The equation showing PCE inflation as a function of the sub-indices’ inflation rates times their respective shares is not a “behavioral equation” one would use as the foundation of a forecasting model. Macrosight explains that type of model here and here. Rather the PCE equation is an accounting identity. It is true by definition. Here I am using this identity to assess which of the consumption categories appears to be the most important for guessing future inflation.

4. This category, as defined by the BEA, includes Housing and Utilities, Health Care, Transportation services, Recreation services, Food Services and accommodations, Financial services and insurance, and other services. See these Macrosight posts for related discussion.