What are trimmed mean and median inflation rates? And why does Kevin Warsh prefer them?

During his Senate confirmation hearing, Kevin Warsh, President Trump’s nominee for Federal Reserve chair, said that “the data that’s being used to judge inflation is quite imperfect.” In place of traditional inflation metrics, Warsh argued, the Fed should use “trimmed averages” to better understand the “underlying, generalized change in prices in the economy.”

What are the “trimmed averages” to which Warsh refers?

What is core inflation? And why does the Fed watch it?

To measure progress towards its congressionally mandated goal of price stability, the Federal Reserve sets an inflation target: 2%, measured by changes in the price index for Personal Consumption Expenditures (PCE), an index the Fed prefers to the more widely used Consumer Price Index (CPI). Both price indices, though, reflect a combination of transitory price changes—often referred to as “noise”—and permanent price changes, which represent departures from long-run inflation trends. These transitory shocks often have little to do with cyclical inflation pressures.

Policymakers therefore supplement overall, or “headline,” inflation with estimates of “core” inflation, which they regard as the better predictor of where inflation is headed. Using various techniques, core inflation removes the most volatile elements of the price index—those items whose price change the most in a given month. What remains is often a better measure of underlying inflation, and thus a more useful tool for policymakers when gauging cyclical inflation pressures, forecasting future inflation, and setting interest rates.

Warsh has highlighted the importance of distinguishing trend from noise: “What I’m most interested in is what’s the underlying inflation rate,” he said during his confirmation hearing, “not what’s the one-time change in prices because of a change in geopolitics or a change in beef.” Current Fed Chair Jerome Powell has said much the same thing: “Core inflation is something we think about because it is a better predictor of future inflation.” 

What is the most common measure of core inflation?

The most often cited measure of core inflation removes food and energy components from the CPI or PCE price indices. Food and energy prices tend to be volatile and are heavily influenced by the weather and geopolitics, potentially disguising the underlying inflation trend.

What is trimmed mean inflation?

The trimmed mean is an alternative measure of core inflation. It removes outliers by taking the price change of all items in an index, sorting them from smallest to largest, and then cutting off the most extreme observations from the bottom and top of the distribution. The inflation rate is the weighted average of all the remaining items.

The threshold above and below which items are dropped, or trimmed, varies across estimators. The Federal Reserve Bank of Dallas’ trimmed mean PCE inflation rate excludes components whose share of total expenditures falls below the 24^th percentile or above the 69^th percentile of the price change distribution. In February, for example, the Dallas Fed dropped telephone and related communication equipment (with an annualized one-month price change of -50.8%) and moving, storage, and freight services (with an annualized price change of +384.6%). So, while the headline PCE measure rose by 2.8% for the 12 months ending in February 2026, and PCE excluding food and energy rose 3.0%, the trimmed mean rose 2.3%.

The Federal Reserve Bank of Cleveland’s trimmed mean inflation rate follows a similar logic  with two key differences. First, the Cleveland Fed uses the Consumer Price Index (CPI), rather than the price index for PCE. (For more on the differences between CPI and PCE, see this Hutchins Center explainer.)  

Second, the Cleveland Fed drops items below the 8^th percentile and above the 92^nd percentile of the price change distribution—a symmetrical 8% off the top and the bottom. In contrast, the Dallas Fed cuts more off the top of the distribution than it does off the bottom. The Dallas Fed’s thresholds account for the fact that prices at the top of the distribution have historically increased less than prices at the bottom of the distribution have fallen.

For the year ending in March 2026, the headline CPI measure rose 3.26%, while CPI excluding food and energy rose 2.60% and trimmed mean CPI rose 2.64%.

What about median inflation?

The Cleveland Fed also publishes median inflation for both the CPI and the PCE price index. In this measure, the monthly price changes for each item in the index are ordered from most negative to most positive, and the item in the middle of the distribution—that is, the item whose share of overall expenditures falls in the 50^th percentile of price changes—is selected. The price change of that 50^th percentile item is the median inflation rate.

In March, that median item was rent. Prices of half the items in the index rose more than rent, and prices of half the items rose less. Rent increased by 2.31% at an annualized rate in March, so the median CPI inflation rate was 2.31%. Core CPI increased by an annualized rate of 2.38% in March.

How do trimmed mean and median inflation compare to other inflation measures?

Averaged over 2005 to 2018, Dallas Fed trimmed mean inflation ran higher than core PCE inflation, and was sometimes higher and sometime lower than the all-items headline inflation.

But today, trimmed mean inflation is lower than core PCE inflation. Why? Because the distribution of price changes has shifted over time. Historically, prices in the bottom of the distribution fell more than prices in the top of the distribution rose. To account for this fact, the Dallas Fed cuts more off the top of the distribution than the bottom. Recently, however, prices of items at the top have risen as much as or more than prices at the bottom have fallen, which could lead the Dallas Fed trimmed mean to understate inflation. Such changes occurred during other inflation surges, such as the 1974 oil embargo and the COVID pandemic, when trimmed mean inflation also lagged behind core and headline inflation measures. During these periods, researchers at the Dallas Fed conclude, core PCE inflation (that is, excluding food and energy) is a better predictor of where inflation is headed.

One reason for the recent shift in the distribution of price changes is tariffs. In the 12 months ending in February 2026, there were more large increases in prices of core goods (goods other than food and energy) and fewer large decreases. Because trimmed mean inflation measures cut off the upper end of the price change distribution, though, they exclude much of this tariff shock. If these trends continue, “the currently lower Trimmed Mean PCE inflation rate relative to the core PCE inflation rate may not be strong evidence of disinflation ahead,” the Dallas Fed researchers conclude. However, if the recent increase in core goods prices is “primarily tariff-driven and a one-off price level change,” they argue that trimmed mean inflation would be a better predictor of where inflation is headed in the medium run.

Why do some economists favor trimmed mean and median inflation?

The economists who developed trimmed mean inflation cite three main advantages over core PCE inflation. First, research from the Dallas Fed suggests that first-release trimmed mean inflation rates are a better measure of inflation trends than first-release core PCE inflation rates, both of which are revised after the initial release. Between 2005 and 2018—a period long enough to highlight long-term trends—first-release trimmed mean inflation averaged 1.78%, only 0.1 percentage point off the average headline PCE inflation rate of 1.77%. First-release core PCE inflation, meanwhile, averaged 1.61%. Related research from the Dallas Fed also suggests that the trimmed mean is generally a better predictor of where inflation is headed. And because monetary policy affects inflation rates with a lag, successful policymaking requires both accurate real-time signals of inflation and accurate forecasts of where inflation will be in the future.

Second, the most volatile elements in both the CPI and the PCE price index are not fixed, nor are they exclusively food or energy. Trimmed mean inflation’s statistical approach to cutting off outliers—as compared to core PCE’s categorical approach—gives it the flexibility to remove the most volatile elements on a monthly basis.

There is also some evidence that trimmed mean inflation better reflects how much slack, or underutilization of resources, there is in the economy. Judgments about the amount of slack in the economy, often inferred from the inflation rate, help guide policymakers’ decisions on whether to raise or cut rates. If trimmed mean inflation is more tightly related to slack than core PCE inflation is, as some argue, then it could be a more useful tool for policymakers.

What are the downsides of trimmed mean and median inflation?

Trimmed mean and median inflation measures are unbiased only if the underlying price change distribution remains constant—for example, if prices at the bottom end of the distribution always fall more than prices at the top end of the distribution rise. But that’s not always true. Observing this, the Cleveland Fed notes that “median and trimmed-mean inflation rates tend to be useful estimates of trend inflation over long periods, but they can exhibit persistent departures from underlying trend over shorter horizons.”

In addition, trimmed mean and median measures, by design, exclude much of the impact of President Trump’s tariffs and much of the energy shock from the war in Iran. These shocks could be viewed as one-time increases in the price level—not the inflation rate—and thus should be excluded from any measure of the underlying inflation trajectory. But these shocks can influence market, business, and consumer inflation expectations, and containing those expectations is a key Fed objective. Moreover, by disregarding price increases due to tariffs or higher oil-related input costs, trimmed mean and median inflation measures “could systematically and substantially understate the inflation experience for a period,” says Krishna Guha of Evercore in a report to clients. Dallas Fed researchers note that trimmed mean inflation might not pick up on “the beginning of broadening inflationary pressures.”

A final reason to be skeptical of trimmed mean and median inflation: Switching to a new metric when inflation has been above the Federal Reserve’s 2% PCE target for five years could be seen as moving the goal posts, potentially threatening the central bank’s credibility.