Trim Your Angst on Economic Measurement Tweaks

Good news! The Bureau of Economic Analysis (BEA) is about to fix some long-running bugs in its inflation measure! Data will get better! Only, it seems this isn’t good news to many. Instead, the news of forthcoming revisions to the Personal Consumption Expenditures (PCE) price index’s methodology rekindled fears of dirty data manipulated to meet the White House’s aims. Such griping is a bipartisan tradition, historically aimed at presidential administrations under both parties. It was a false fear before and remains so now, a sign stocks’ AI cheer has some skeptical ballast.

Our saga begins in March, when the BEA made a last-minute swap in the measure it used to deflate legal services in January’s PCE report. Traditionally, the BEA had used the legal service component from the Bureau of Labor Statistics’ (BLS’s) Consumer Price Index (CPI). The BLS discontinued that series in 2024 because its increased volatility rendered it unreliable, but it kept feeding the data to the BEA. They trundled along until discovering January’s “unpublished CPI showed an unprecedented increase that could not be substantiated by any other data,” as they explained in a later blog post.^[i] So they instead used the Producer Price Index (PPI) measure of legal services, which pulled down the headline inflation rate a smidge.

Economists agreed the move was necessary and sound. But the move’s timing and manner raised eyebrows, happening outside the normal annual methodology refinements and without a public announcement. It also followed last year’s kerfuffle over President Donald Trump firing the BLS’s head, which prompted concerns about data manipulation and politicization. To people inclined to be nervous about this, an unannounced, last-minute swap in the data was a bad look.

People quickly moved on, but now the BEA has announced its annual revisions, which will happen in September. There are three affecting PCE inflation, and the consensus is these will lower the rate. To some, this is uncannily convenient for new Fed head Kevin Warsh, whom some still see as eager to cut rates to please Trump. Several articles have married these upcoming changes with Warsh’s past fandom of the Dallas Fed’s “trimmed mean” PCE inflation measure, theorizing he is leading a data revolution to lower the inflation rate and please his appointer.

To us, this seems too reliant on cork boards and color-coded yarn. When you actually look at what the BEA is doing, it becomes clear the PCE price index’s problems ran deep. The three categories they are patching are the aforementioned legal services, plus portfolio management and investment advice and computer software and accessories. For legal services, the BEA is creating a bespoke composite of select legal services within the PPI report, which appears to be a finer-toothed comb than it used for January’s report. That tells you the goal here is accuracy, not inflation reduction. The means, not the end.

The computer software and accessories measure was similarly problematic. The Fed published a detailed report on the issues in May, noting this category made “unprecedented” inflation contributions that were “more than 9 standard deviations above their historical means” from November 2025 through March 2026. This is statistical jargon for “wackadoodle crazy volatility,” so they looked into it and found a measurement error. For convenience, the BEA used the computer software and accessories component of the CPI to deflate its measure of consumer spending on this category. Logical. But: “The BEA category contains no physical IT accessories, while the CPI category includes flash drives and blank media.”^[ii] A only partly overlapped with B, which meant they were using irrelevant hardware prices to deflate software.

The Fed also noted two conceptual problems. One, software providers had changed their pricing practices, which broke the BEA’s model and introduced bias because standard measurement systems weren’t equipped to deal with the new way of doing business. And two, the model wasn’t capturing software’s enhanced quality as AI tools got built into traditional programs like Microsoft Excel. The model was counting raw dollars spent on software without accounting for changes in what they were paying for.

So to account for all this, the BEA made a new composite blending CPI for computer software and accessories, the PPI for game software publishing and the PPI categories covering cloud services. It is all very finely tuned, replacing a flawed, blunt instrument.^[iii] The changes for portfolio management and investment costs are similar, better accounting for the amount of services consumers use.

This is all very technical, and you are probably bored despite our best efforts to be snappy, which kind of proves our point: None of these changes are an easy button. All amounted to more work, justified by a result that (hopefully) better matches reality. Today, it seems these changes will lower the inflation rate. In the future, they may lead to a higher rate. Either way, they should mean more accurate data for investors and policymakers to chew over. They are also the kind of thing the BEA does every year, regardless of who is in the White House and how they feel about data and interest rates. It just gets a spotlight now because that is where the zeitgeist is. That is sentiment for you.

As for the trimmed-mean inflation whining, we see a tempest in a teapot. A trimmed-mean measure tries to give a better look at underlying inflation by omitting the outliers on both ends. The logic: Big price changes up and down likely result from industry-specific one-offs, which don’t represent broad economic issues monetary policy could address. For example, a trimmed-mean measure would have excluded energy after the war broke out, which seems logical from a policymaking perspective because the Fed can’t drill for oil or magic natural gas export terminals into existence. The Bank of Canada has long looked at trimmed-mean CPI for this reason.

Yet, crucially, looking at doesn’t mean targeting. The Bank of Canada’s inflation target remains headline Canadian CPI, even though policymakers look to trimmed-mean and two other “core” measures for hints at underlying trends. We reckon Fed policymakers already look at the Dallas Fed’s trimmed-mean measure, just like they look at headline CPI, as well as PCE and CPI excluding food and energy. Not because any of these things dismiss headline PCE, which is their target, or render inflation there nonexistent, but because they give clues as to whether there is anything they can or should do to tame a faster-rising PCE.

Lastly, note Warsh never said anything about targeting trimmed-mean PCE. He just argued it is a better way to see underlying trends than the “core” measure that excludes food and energy. There is some merit to that, given one-off price changes happen in other categories, too. At the same time, there is no perfect measure that reduces all noise and gives perfect insight. Trimmed-mean CPI risks trimming out signal as well as noise. Still, a Fed head can like whichever measures he or she likes. But the Federal Open Market Committee has a big dashboard, and they will probably continue looking at—and arguing about—everything on it. So trim your angst and stay cool.