Between Headline and Core: Inflation Excluding Energy Goods

Federal Reserve Chairman Kevin Warsh has raised concern regarding how inflation can most effectively be measured. At his inaugural Federal Open Market Committee (FOMC) press conference (PDF), he announced the creation of various task forces, two of which (“data” and “inflation frameworks”) will address how to best inform policymakers on inflation.

The current monetary policy framework (PDF) judges that “inflation at the rate of 2 percent, as measured by the annual change in the price index for personal consumption expenditures, is most consistent over the longer run with the Federal Reserve’s statutory maximum employment and price stability mandates.” Yet, in communications, including the Summary of Economic Projections (SEP), Fed officials also refer to inflation “excluding the volatile food and energy categories.”For example, see Federal Reserve Chair Jerome Powell’s final FOMC press conference (PDF) on April 29, 2026.

In other words, when formulating its inflation objective over the medium and longer run, the Fed uses the personal consumption expenditures (PCE) price index (also known as headline PCE), while it often relies on excluding food and energy (also known as core PCE) to identify underlying trends and thus judge its performance in the short run.The concept of core inflation (PDF) was first introduced by economist Robert J. Gordon in 1975.

Purportedly, core PCE removes prices that are too volatile and may mask underlying inflation, at the cost of ignoring a substantial share of consumer spending. In this blog post, I will argue that we can construct a better measure of underlying inflation by removing only energy goods from headline PCE. This would preserve the desirable smoothness of core PCE while maintaining a more representative index, as it would remove a substantially lower share of expenditure.

Headline vs. Core PCE Inflation

The first figure shows headline and core PCE inflation from January 2001 until May 2026.The sample period is long enough to include three recessions (2001, 2007-09 and 2020) and the subsequent recoveries, but short enough to offer visual clarity. Similar qualitative results would be obtained if we started the analysis at an earlier date. Inflation is measured as the 12-month change in the corresponding price index and shown at a monthly frequency. As we can see, both measures of inflation track each other closely. Annual inflation averaged 2.2% and 2.1% for headline and core, respectively. However, headline inflation is clearly more volatile than core. Headline inflation spikes and dips more frequently and prominently than core, which is significantly smoother. In more-technical terms, the variance of core inflation is about half that of headline.

As mentioned above, the core price index is calculated by removing food and energy. On average, since 2001, food and energy accounted for 7.7% and 4.8% of PCE, respectively. In other words, core represents 87.6% of PCE. Over the sample period, the shares of food and energy in consumer spending have trended down somewhat. For example, in the 12 months ending in May 2026, the corresponding expenditure shares for food and energy were 7.2% and 3.7%, respectively.By contrast, back in 1960, the shares of food and energy in PCE were significantly larger: 18.9% and 7.3%, respectively.

The above figure illustrates the advantages of focusing on core inflation to understand inflation dynamics. Let’s focus on the recent high inflation episode associated with the COVID-19 pandemic and its aftermath. Core inflation rose with headline inflation during 2021, peaking in 2022. Peak core inflation was significantly lower than peak headline inflation, but the timing and significance of the increase remain. Afterward, core declined less abruptly and more steadily than headline, stabilizing closer to 3% than 2%, signaling that inflation has not yet been tamed.

In 2026, headline inflation rose dramatically, in large part driven by the rise in oil prices due to the conflict in Iran and the closure of the Strait of Hormuz. However, the concurrent, though more moderate, recent increase in core inflation suggests that there may be other inflationary pressures at play, which may not ease as oil markets return to normal.

Food and Energy Prices: How Volatile Are They?

Core PCE offers a less volatile measure of inflation at the cost of focusing on a narrower share of consumer spending. However, not all components removed from headline are equally volatile. Could we still get the benefits of core by limiting the ignored categories to only the most excessively volatile?

Core removes three broad categories:

• First, “food and beverages purchased for off-premises consumption” or “food” for short. (Note that this does not include food consumed in restaurants, which is included in “food services” and in the core price index.)
• Second, “gasoline and other energy goods,” which consists of gasoline and other fuels, as well as lubricants and fluids.
• Third, “electricity and gas,” which is considered a household utility (i.e., a service).

The second category comprises energy goods, and the third category comprises energy services; collectively, the two are referred to as “energy.”

The second figure shows the price indexes for food, energy goods and energy services. Currently, price indexes in the national accounts use 2017 as the base year, meaning all prices average 100 over 2017. Prices in the figure below are expressed as natural logarithms, which is a common way to express macroeconomic variables with a trend.

All three price indexes show significant ups and downs, but the volatility in energy goods clearly stands out. Why are prices for energy goods so volatile? Their prices closely track the evolution of oil prices. The third figure shows the clear comovement between energy goods and oil prices. The correlation between these two prices is 92%. This high correlation is preserved for the 12-month change in prices (i.e., inflation): 86%.

The behavior of the prices of food, energy goods and energy services during the COVID-19 pandemic and its aftermath was also significantly different. The price of energy goods fell at the onset of the pandemic and then rose sharply, peaking in June 2022. Then it declined, with some hiccups, through the beginning of 2026 when it rose sharply again during the conflict in Iran. As shown in the third figure, this behavior mirrors what happened to oil prices.

In contrast, the prices of food and energy services (electricity and natural gas) exhibited much more gradual and persistent increases. This pattern is more closely associated with the behavior of aggregate prices: a significant increase in 2021 through 2022, followed by a rate of growth that is somewhat above the prepandemic period.See this recent blog post for more analysis on the behavior of aggregate prices.

Given that oil prices are determined in international markets—largely independent of U.S. monetary policy and other domestic factors—and experience large and short-lived swings, it seems reasonable to exclude them for an inflation index designed to ignore short-lived fluctuations and understand underlying trends. However, the much lower volatility and more persistent behavior exhibited by the prices of food and energy services open up the question of whether it is appropriate to exclude them as well.

Should Food and Energy Service Prices Also Be Excluded?

To answer this last question, we can compare the volatility of inflation in food and energy services with that of other prominent product categories included in core PCE. The table below shows the relative variance of inflation and the corresponding expenditure shares for various consumption categories.

As we can see, inflation in durable goods, clothing and footwear, transportation and financial services all exhibit volatility similar or higher than food. Yet they are all prominent categories that are not excluded from core PCE. So, does it make sense to exclude food? Food is not excessively volatile and, in fact, is a good predictor of headline inflation, as my colleague Michael McCracken noted in 2023. And inflation in energy services is indeed much more volatile than these other categories, but much less so than energy goods, which appear to be a significant outlier.

Could PCE Excluding Energy Goods Offer a Better Measure of Underlying Inflation?

Given the findings above, consider constructing a price index that sits in between headline and core PCE. The objective is to remove the most volatile components that could mask underlying inflation trends and potentially lead to misdiagnosing a situation or overreacting to changes in inflation.

We could, for example, exclude all energy categories, which represents 4.8% of PCE (3.7% in the last 12 months), as I have done in previous blog posts using the PCE price index excluding energy published by the Bureau of Economic Analysis (BEA). However, we can obtain the benefits of core (lower volatility of prices and inflation) while preserving a larger share of consumption by removing only energy goods, which, as shown above, basically track oil prices. A price index for PCE excluding energy goods is not available from the BEA and needs to be constructed from the individual PCE product categories, using the same chain-weighting methodology as for headline, ex energy and core.

PCE excluding energy goods can be built from the ground up by combining the price and spending data for every non-energy category, or by taking energy-goods spending out of total PCE and using the official price indexes to construct the chain-weighted index for the remainder. The residual Fisher method is preferable because it preserves the BEA’s own aggregate calculations and matches the official core and ex-energy indexes much more closely during the unusual 2020-21 period. Outside that episode, the two methods produce almost identical results.

The fourth figure compares inflation rates for headline PCE with inflation rates for PCE excluding energy goods. As we can see, inflation ex energy goods is substantially less volatile than headline but preserves its broad movements and trends while removing only 2.8% of expenditure (2.1% in the last 12 months). Relative to core, inflation ex energy goods is somewhat more volatile, experiencing a sharper decline during the Great Recession and a more pronounced rise during the COVID-19 pandemic. But like core, it captures the stall in inflation following the pandemic, which was somewhat masked in headline because of the behavior of oil prices, as well as the more recent increase in inflation not directly attributable to oil prices.

Key Advantages of PCE Ex Energy Goods over Other Measures

The second table presents formal measures of correlation, volatility and expenditure shares of alternative aggregate price indexes, or aggregates. All three aggregates discussed here—PCE ex energy goods, PCE ex energy goods and services, and core PCE—are highly correlated with headline inflation, with PCE ex energy goods exhibiting the highest correlation (0.92). We can also see that PCE ex energy goods exhibits the highest volatility of the three, but it is still substantially below that of headline and only marginally higher than that of the other aggregates. In addition, the share of PCE is substantially higher when we remove only energy goods (97.2%) as opposed to energy goods and services (95.2%) or food and energy (87.6%).

The table also compares these aggregates with an alternative approach: the Dallas Fed trimmed mean inflation. This measure removes “24 percent of the weight from the lower tail and 31 percent of the weight in the upper tail”, meaning it represents about 45% of PCE.

Compared with the other aggregates presented here, trimmed mean inflation has a substantially lower correlation with headline inflation (0.74) and significantly lower volatility (about a third of the variance of headline inflation). The low correlation between trimmed mean and headline inflation is evident when charting both series together: Significant dips and peaks (e.g., during the Great Recession and the COVID-19 pandemic) occur with a significant lag. And in the last 12 months, trimmed mean inflation has been trending down rather than up. Altogether, it is a series that appears to struggle more than the other aggregates presented here in tracking the underlying trends in inflation.

PCE Ex Energy Goods Offers Improved Measure to Track Underlying Inflation

In this post, I have proposed PCE ex energy goods as an alternative aggregate to core PCE to track underlying inflation. PCE ex energy goods inflation retains the relative smoothness of core PCE inflation and achieves this by removing a significantly smaller share of PCE.

Importantly, this post does not suggest replacing headline PCE as the measure of inflation over the medium and longer runs, as stated in the current Fed’s monetary policy framework. Rather, it proposes a measure that improves on core to interpret and judge short-term fluctuations in prices and inflation.

As I have shown, food is not particularly volatile relative to other categories and exhibits behavior more in line with headline than do energy prices. Electricity and natural gas prices, though very volatile on their own, are still more persistent and significantly less volatile than the price for energy goods.

In essence, by just excluding energy goods, we are removing categories that correlate highly with oil prices, which exhibit large but short-lived fluctuations. This measure arguably better captures underlying trends in inflation that do not get masked by short-lived fluctuations in oil prices, as may occur with headline inflation. This is particularly important for policymakers evaluating when to adjust monetary policy to react to specific shocks rather than “looking through” them.

Notes

1. For example, see Federal Reserve Chair Jerome Powell’s final FOMC press conference (PDF) on April 29, 2026.
2. The concept of core inflation (PDF) was first introduced by economist Robert J. Gordon in 1975.
3. The sample period is long enough to include three recessions (2001, 2007-09 and 2020) and the subsequent recoveries, but short enough to offer visual clarity. Similar qualitative results would be obtained if we started the analysis at an earlier date.
4. By contrast, back in 1960, the shares of food and energy in PCE were significantly larger: 18.9% and 7.3%, respectively.
5. See this recent blog post for more analysis on the behavior of aggregate prices.