Market-Based PCE Excludes Imputed Prices Except One Line Worth $2.5 Trillion

The U.S. Bureau of Economic Analysis, Personal Income and Outlays, July 2026, released August 26 as BEA 26-39, carries the sentence every rate desk reads first: "From the same month one year ago, the PCE price index for July increased 3.7 percent. Excluding food and energy, the PCE price index increased 3.3 percent from one year ago."

A reasonable follow-up is how much of that 3.7 percent came from a price someone actually quoted to somebody. BEA publishes an answer and it has a name. The market-based PCE price index is built, in the agency's words, from components "based primarily on observed market transactions for which there are corresponding price measures."

Read that phrase alone and you will conclude that the market-based index is the headline index with the imputations taken out. It is not. One imputation stays in, deliberately, and in 2025 it was $2,500.256 billion of consumer spending. The line is drawn in a sentence whose meaning inverts if you stop reading two words early.

The Clause That Reverses the Definition

The U.S. Bureau of Economic Analysis, FAQ 83: What is the 'market-based' PCE price index? states the exclusion and then immediately overrides part of it. The first two sentences have to travel together, and the rest of the paragraph is quoted further down: "The PCE market-based price index excludes most imputed expenditures, such as 'financial services furnished without payment,' for which deflators are implicit, based on independent current dollar and chained-dollar estimates. However, it includes 'imputed rental of owner-occupied nonfarm housing,' which is deflated by the CPI for owner's equivalent rent, so that the index will be comparable with the overall CPI."

Stop at "implicit" and you have an index of observed transactions. Read one sentence further and you have an index whose largest housing component is a rent nobody was ever billed for, kept in on purpose so that the series lines up against the CPI.

BEA said the same thing more bluntly eighteen years ago. The U.S. Bureau of Economic Analysis, Survey of Current Business, A Reclassification of Personal Consumption Expenditures (May 2008) puts the whole boundary in one line: "Under the current structure, the market-based PCE price index excludes the expenses of NPISHs and imputed expenditures except for owner-occupied space rent."

Two documents, two vintages, and they do not say quite the same thing. The 2008 article names owner-occupied space rent as the single exception. FAQ 83 names a second inclusion: "Household insurance premiums and normal losses, which are deflated separately by CPIs, are included in the market-based PCE index, but medical and hospitalization insurance, income-loss insurance, life insurance, motor vehicle insurance, and workers' compensation insurance are excluded." Both branches are load-bearing, and a reader who keeps only the first half will expect the wrong behaviour from insurance inside the gap.

Reconciling 3.7 Percent With the Index Levels

FAQ 83 also tells you where the market-based series lives: "These indexes are shown in the GDP and personal income news releases, in NIPA tables 2.3.4, 2.3.7, 2.8.4, and 2.8.7, and as addenda items in the underlying detail PCE tables." The narrative sentences of the July release quote the PCE price index and the index excluding food and energy; they do not quote a twelve-month market-based change. So the comparison has to be assembled from published index levels, and labelled as arithmetic rather than passed off as a BEA print, a distinction this blog has walked into before in Power Construction Put in Place: The Monthly Number Comes From an Annual Forecast.

Start with the check. The Federal Reserve Bank of St. Louis, FRED series DPCMRG3M086SBEA, Personal consumption expenditures: Market-based (chain-type price index) and its headline companion are both index 2017 = 100, monthly, seasonally adjusted, last updated 2026-08-26. The headline index reads 126.960 for July 2025 and 131.659 for July 2026, a ratio of 3.7012 percent against a printed 3.7 percent. The index excluding food and energy reads 126.430 and 130.658, or 3.3441 percent against a printed 3.3 percent. The levels reproduce the release, so the same operation on the market-based levels is doing the same thing.

Market-based PCE reads 124.656 for July 2025 and 128.994 for July 2026. That is a rise of 4.338 index points, or 3.4800 percent. Market-based excluding food and energy reads 123.697 and 127.439, or 3.0251 percent. Set them against the printed figures and the headline gap is 0.2212 percentage point and the core gap is 0.3190 percentage point. None of those four decimal figures is a BEA publication. They are ratios of two published levels, and the rounding BEA applies to its own print is why 3.7012 becomes 3.7.

The Gap Has an Address

A 0.22 point wedge could be diffuse. It is not. The Federal Reserve Bank of Richmond, Macro Minute: A Tale of Two PCEs: Market- and Nonmarket-Based Prices, published February 3, 2026, set the two indexes side by side category by category in a table titled "Year-Over-year Changes in PCE Category Price Indexes". For the November reading current at publication, PCE inflation was 2.8 percent and market-based PCE inflation was 2.5 percent, and the note says where the wedge sits: "The largest disparity between the standard and the market-based price indexes is attributable to the financial services and insurance category." That row reads 6.12 percent against 0.82 percent.

The instructive part of the table is how many rows do not move at all. Health care prints 2.82 against 2.82, household utilities 6.41 against 6.41, transportation services 3.33 against 3.33, gasoline and other energy goods 1.52 against 1.52, clothing and footwear negative 0.44 against negative 0.44. Where a category is priced by a collected CPI or PPI, the two indexes are the same number. The wedge is not a general quality gap between the series. It is one category with an implicit deflator, plus the insurance split quoted above, whose second branch is the boundary this blog traced in Which Portion of a Homeowners Premium Enters the CPI and Which Enters GDP.

The Imputation That Stays, and Why

Now the exception. The U.S. Bureau of Economic Analysis, NIPA Handbook Chapter 5: Personal Consumption Expenditures lists what is invented outright, and the qualifier at the end of the financial-intermediary item belongs to BEA: "Estimates for the following PCE components are entirely imputed: the space rent of nonfarm owner-occupied housing, farm products consumed on farms, wages and salaries paid in kind, private workers' compensation, services furnished without payment by financial intermediaries except life insurance carriers, and the expenses associated with life insurance and pension plans." The chapter gives the reason in Table 5.1, Content of PCE, whose row label runs five branches: "Purchases imputed to keep PCE invariant to whether: Housing and institutional structures and equipment are rented or owned; Employees are paid in cash or in kind; Farm products are sold or consumed on farms; Saving, lending, and borrowing are direct or are intermediated; Financial and insurance service charges are explicit or implicit". Housing is the first of five, and the sentence quoted above sits in the Comments column of that same table.

Size it. The Federal Reserve Bank of St. Louis, FRED series DOWNRC1A027NBEA, Personal consumption expenditures: Services: Housing: Imputed rental of owner-occupied nonfarm housing carries the line BEA publishes in NIPA tables 2.4.5 and 2.5.5, annual, current dollars, not seasonally adjusted, and it reads $2,500.256 billion for 2025 against total PCE of $20,954.857 billion. Dividing gives 11.9316 percent, which is arithmetic on two published levels and not a BEA statistic; subtracting leaves $18,454.601 billion of PCE that is not owner-occupied space rent. FAQ 83 names this component as included, in those words. It does not rule on the other five that Chapter 5 lists: it says only that the index "excludes most imputed expenditures", and the one example it gives is financial services furnished without payment. Treating "most" as "the other five are out" is this article's inference, not a BEA classification, and neither document retrieved here assigns farm products consumed on farms or wages and salaries paid in kind to either side of the line. BEA does not rank the six by size and this article has not priced the other five.

The reason for retaining it is stated, not inferred. FAQ 83 says the item is deflated by the CPI for owner's equivalent rent "so that the index will be comparable with the overall CPI", and adds the mechanism in a parenthesis: "(Owner's equivalent rent within the CPI is measured by reweighting the renter sample of market transactions.)" That parenthesis is the hinge. The imputed expenditure has an observed price behind it, because the price comes from rents that renters actually pay.

The U.S. Bureau of Labor Statistics, CPI Fact Sheet: Owners' Equivalent Rent and Rent defines the concept with its own restrictive clause attached: "the implicit rent that owner occupants would have to pay if they were renting their homes, without furnishings or utilities". It carries a second qualifier that gets dropped constantly: "Pure rents are derived in a manner similar to that of economic rents, with some exceptions. Because owners pay for utilities separately, the pure rent excludes the cost of any utilities included in the contract rent." And it rules out the costs most homeowners think of first: "Interest costs (such as mortgage interest), property taxes, real estate fees, most maintenance, and all improvement costs are part of the cost of the capital good and are also not treated as consumption items."

Concrete Framework

Five rules that survive contact with the release day.

One. Never describe market-based PCE as the index without imputations. It is the index without imputations except owner-occupied space rent, and the exception is 11.9316 percent of 2025 PCE by arithmetic on published levels. The word "except" is the definition.

Two. When you quote a twelve-month market-based figure, say whether you computed it. A change assembled from two index levels is your arithmetic, and the four-decimal precision is a property of your division, not of BEA's estimate.

Three. Reconcile before you compare. Run the ratio on the headline levels first. If 131.659 over 126.960 does not give you the 3.7 percent BEA printed, the problem is your levels or your vintage, not the market-based series.

Four. Carry both insurance branches. Household premiums and normal losses are inside the market-based index. Medical and hospitalization, income-loss, life, motor vehicle and workers' compensation insurance are outside it. A gap attributed to "insurance" without that split is not an attribution.

Five. Check which document you are quoting. The 2008 article names owner-occupied space rent as the sole exception. FAQ 83 adds household insurance premiums and normal losses, plus nonprofit primary services from the 2009 comprehensive revision. Both are BEA, describing the boundary at different dates.

What to Watch Next Week

The July release states the next date in plain text: "Next release: September 30, 2026, at 8:30 a.m. EDT Personal Income and Outlays, August 2026". Its own footnote says that date carries more than one month of data: "Revised wage and salary data from the U.S. Bureau of Labor Statistics Quarterly Census of Employment and Wages program for January through March will be incorporated with the annual update of the National Economic Accounts on September 30. Updated monthly estimates of personal income and outlays will be released on September 30, along with the estimate for August 2026."

The U.S. Bureau of Economic Analysis, Preview of the 2026 Annual Update of the National Economic Accounts gives the scope. The update begins "on September 30, 2026" and covers "the first quarter of 2021 through the first quarter of 2026." Three deflator changes are announced, and the first sits in the category the Richmond Fed identified as the source of the wedge: "BEA will improve the deflation methodology for consumer spending on portfolio management and investment advice services to better reflect the timing and quantity of services consumed." The preview also promises a better deflator for legal services and a second one whose purpose clause matters: "BEA will improve the deflator for consumer spending on computer software and accessories to better reflect the composition of products included in this category." One thing is fixed: "The reference year for index numbers and chained-dollar estimates will remain 2017."

So the thing to do on September 30 is not to read the August number first. It is to re-run the reconciliation above on the revised July levels. The market-based levels for July 2025 and July 2026 are both downstream of deflator methodology being changed that day.

What Would Invalidate This

Four ways this reading fails, in descending order of how much they would cost.

First, the deflator change may not move the gap at all. Portfolio management and investment advice services appear in the 2008 article's list of financial services, but that article does not assign them to the market-based or the other-than-market-based side, and neither document retrieved here does. If those services sit inside the market-based set, the September 30 change moves both indexes together and the 0.2212 point wedge need not narrow at all. BEA publishes category lists that settle it. This article has not opened them, and claims only that the change lands in the category the Richmond Fed identified, not that it lands on the excluded side of the line.

Second, an expenditure share is not an index weight. Eleven point nine percent of current-dollar PCE is a scope figure: it says how much spending the exception covers, not how much of a twelve-month price change owners' equivalent rent contributed. If that rent were rising at the market-based aggregate's own rate, keeping it in would barely move the index, and the argument here would collapse into a statement about definitions.

Third, the category attribution is borrowed from a different month. The Richmond Fed table describes a November reference month as of a February 3, 2026 publication. Nothing retrieved here shows that the financial services and insurance row is still carrying the July 2026 gap of 0.2212 point on the headline and 0.3190 point on the core. The mechanism travels; the decomposition does not travel with it automatically.

Fourth, none of this changes the mandate number. The Board of Governors of the Federal Reserve System, Statement on Longer-Run Goals and Monetary Policy Strategy, adopted effective January 24, 2012 and as reaffirmed effective January 27, 2026, is specific: "The Committee reaffirms its judgment 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." That paragraph names the price index for personal consumption expenditures. It does not name the market-based variant and it does not name the index excluding food and energy. An analyst who prefers market-based PCE on measurement grounds is still not looking at the series the statement points to.

The pattern is familiar from the derived series this blog took apart in Nondurable Orders and Nondurable Shipments: The Same $324,224 Million Appears in Both Tables. A statistic with a clean name turns out to be the output of a rule, the rule carries an exception, and the exception is where the money is. Here the rule excludes imputed expenditures, and the exception is $2,500.256 billion of rent no owner was invoiced for, deflated by the rents that renters were.