Market Pulse
Mixed observation times, Oct 20–Oct 25, 2026 (5 days apart)
Real Estate · 7 min

CPI shelter ran 3.0% to August 2026, below the 3.4% headline

A third of the CPI is rent nobody pays in cash. Here is how BLS measures it, why the figure trails the market by design, and what Census vacancy data show.

◆ The Signal
30-second read
What happened
The shelter index of the CPI rose 3.0% in the 12 months to August 2026, released September 11, 2026, against 3.4% for all items. A year earlier, in the August 2025 report, shelter was running 3.6%. The September 2026 CPI is due October 14, 2026.
Why
BLS prices each sampled rental unit only once every six months and converts the six-month change into a monthly rate by taking its sixth root, so the index records a market shift gradually over about a year. Owners' equivalent rent, 25.918% of the index, is built from the same rental sample, not from house prices.
The numbers that matter
3.0%CPI shelter, 12 months to August 2026
35.343%Shelter's relative importance in the CPI-U, July 2026 weights
7.3%Rental vacancy rate, Q2 2026 (Census, released July 28, 2026)
$1,531Median asking rent, vacant units, Q2 2026
Why it matters
Shelter is the single largest line in the index that sets Social Security's COLA, the I bond rate and the real value of every wage. If the shelter line is a smoothed echo of rents from months ago, so is a third of your inflation figure.
What happens nextas we saw it on 13 Oct 2026
The September 2026 CPI is published Wednesday, October 14, 2026 at 8:30 a.m. ET. The Census Bureau's third-quarter Housing Vacancies and Homeownership release follows on October 28, 2026.
Go deeper ↓
CPI-U, 12-month changes
Shelter and its two parts against the headline, 12 months to August 2026

Shelter, 3.0% over the year, now runs below the 3.4% all-items figure and well below its 3.6% pace of a year earlier.

Shelter and its two parts against the headline, 12 months to August 2026
All items3.4%
Owners' equivalent rent3.1%
Shelter3.0%
Rent of primary residence2.7%
All items less food and energy2.4%
Shelter, year to Aug 20253.6%
% change, not seasonally adjusted, August 2025 to August 2026 (last bar: August 2024 to August 2025)Source: Bureau of Labor Statistics, Consumer Price Index August 2026, Table 1, released September 11, 2026; CPI August 2025, released September 11, 2025
Go deeper

The largest single component of US inflation is a rent that is never paid. Owners' equivalent rent of residences, the amount BLS estimates homeowners would charge themselves to live in their own houses, carried a relative importance of 25.918% in the Consumer Price Index as of the July 2026 weights. Add rent of primary residence at 7.735%, plus lodging away from home and tenants' insurance, and the shelter group as a whole is 35.343% of the index. That is more than food, energy, cars, medical care and clothing combined.

In the August 2026 report, published September 11, 2026, that 35% of the basket rose 3.0% over 12 months while the whole basket rose 3.4%. Shelter, for years the thing holding inflation up, is now holding it down. The reason has as much to do with how the number is built as with what landlords are charging, and the September reading arrives on Wednesday, October 14, 2026.

What the August report said

Data
Shelter rose 3.0% in the year to August 2026; rent 2.7%, owners' equivalent rent 3.1%
BLS Table 1 for August 2026 gives unadjusted 12-month changes of 3.0% for shelter, 2.7% for rent of primary residence and 3.1% for owners' equivalent rent of residences. All items rose 3.4%; all items less food and energy rose 2.4%. On the month, seasonally adjusted, shelter rose 0.3% in August after 0.1% in July, and the release states that “the index for owners' equivalent rent rose 0.2 percent in August as did the index for rent.”
Data
A year earlier shelter was running 3.6%
The August 2025 CPI, released September 11, 2025, said “the shelter index increased 3.6 percent over the last year” and that shelter “was the largest factor in the all items monthly increase.” In that report all items were up 2.9% and core 3.1%. Twelve months on, the ordering has flipped: headline 3.4%, core 2.4%, shelter 3.0%.
The shelter group in the CPI-U, August 2026 report
CategoryRelative importance, Jul 2026 (%)12 months to Aug 2026 (%)Jun to Jul 2026, SA (%)Jul to Aug 2026, SA (%)
All items100.0003.40.10.4
All items less food and energy79.1142.40.20.3
Shelter35.3433.00.10.3
Owners' equivalent rent of residences25.9183.10.30.2
Rent of primary residence7.7352.70.30.2
Services less energy services60.2723.00.20.3
Source: Bureau of Labor Statistics, Consumer Price Index, August 2026, Table 1, released September 11, 2026. SA = seasonally adjusted. 12-month changes are not seasonally adjusted.
Wealth Signal calculation
Take shelter out and the rest of the basket rose about 3.6%
A first-order split using the July 2026 weights: shelter contributes 0.35343 × 3.0% = 1.06 points of the 3.4% headline. The remaining 64.657% of the basket therefore accounts for 3.4 − 1.06 = 2.34 points, which is 2.34 ÷ 0.64657 = 3.6% on its own weight. The same arithmetic on core: (0.79114 × 2.4 − 0.35343 × 3.0) ÷ (0.79114 − 0.35343) = 0.838 ÷ 0.438 = 1.9% for core goods and services other than shelter. These are products of published weights and published changes, not BLS's own contribution figures, which chain monthly relatives and differ slightly.

So the picture inside the August report is the opposite of 2025: inflation outside housing is running hotter than inflation inside it. That makes the construction of the shelter line worth understanding, because a lagging index can flatter or flatten the headline for a year at a time.

How BLS prices a rent

The method is documented in the BLS fact sheet on owners' equivalent rent and rent and in the Handbook of Methods. We cover the broader index in a separate piece on how the CPI is built; this section is only about the shelter lines.

Fact
Each sampled rental unit is priced once every six months
BLS states: “The CPI program collects rent data from each sampled unit every 6 months.” The sample in each area is split into “six subsamples called panels”; panel 1 is collected in January and July, panel 2 in February and August, and so on. BLS gives its reason directly: “Many rents change infrequently, being locked in place for a given lease term, and collecting rent data less frequently allows for a larger sample.” Data collectors visit “a small number (usually 5) of renter-occupied housing units selected to represent each segment,” the segments being small areas defined from Census data.
Fact
The monthly change is the sixth root of a six-month change
For each area, the weighted average rent in month t is divided by the weighted average rent in month t−6. In BLS's words, “the sixth root of this relative is the 1-month change in OER for index area.” The Handbook of Methods describes the index as built by “applying the sixth root of the 6-month rent change to the index for the previous month.” The monthly figure the newspapers report is therefore one-sixth of a half-year change, spread evenly, from a sixth of the sample.
Interpretation
This is why the shelter line trails the market by design
Suppose asking rents jump in one month and then go flat. The units repriced that month show the jump in their six-month comparison; so do the units repriced in each of the next five months, because their comparison window still straddles the jump. Each contributes a sixth of its change per month. The index therefore absorbs a one-off market move over roughly a year, and a turn in the market shows up in the CPI gradually rather than at once. None of this is an error. It is a trade-off BLS has chosen, stated on its own pages, between timeliness and sample size.
Fact
Most rents in the sample are not new leases
The official rent index, BLS says, “measures the change in all rents, including new leases, renewals, and rents in the middle of a lease.” On the BLS research page for its new-tenant rent work, “tenure lengths for tenants average about three years.” A tenant in year two of a lease sees no change in their rent, and the index records that correctly. Asking rents on listing sites describe only the units that are empty and on the market.

Owners' equivalent rent is not a house price

Fact
OER is built from rents, and house prices play no part
BLS states that “owned housing units themselves are not priced in the CPI Housing Survey.” The CPI treats an owned home as a capital good and measures only the shelter service it provides, using rents of comparable rental units as the proxy. Homeowners in the Consumer Expenditure Survey are asked: “If someone were to rent your home today, how much do you think it would rent for monthly, unfurnished and without utilities?” That answer sets the weight only. BLS is explicit: “these responses are not used in estimating price change for the shelter categories, only the weight.”
Fact
Utilities are stripped out of the rent before it enters OER
Because owners pay their own electricity, gas, oil, water and sewer, BLS removes any utilities bundled into a sampled unit's contract rent before using it in OER: “the pure rent excludes the cost of any utilities included in the contract rent.” Energy use is modelled from the Residential Energy Consumption Survey and valued at average prices lagged 12 months. The rent index, by contrast, uses “economic rent,” which keeps the utilities a landlord includes.

Two consequences follow. First, a rise in mortgage rates or house prices does not enter the CPI directly; it enters only if and when it moves rents. Second, because 25.918 points of the index rest on a rental sample weighted to look like owner-occupied housing, the rent of single-family houses in suburbs matters far more to OER than the rent of city-centre apartments, even though the latter dominate the asking-rent data most people see. See renting versus buying for the household side of the same arithmetic.

What the Census vacancy survey shows

The closest official counterpart to the CPI shelter line is the Census Bureau's quarterly Housing Vacancies and Homeownership release, a supplement to the Current Population Survey. It does not measure rent inflation. It measures how much of the rental stock is empty and what empty units are asking, which is roughly the thing the CPI lags.

7.3%
Rental vacancy rate, Q2 2026
Q2 2025: 7.0%. Released July 28, 2026
$1,531
Median asking rent, vacant for-rent units, Q2 2026
Asking price of empty units, not rent paid by tenants
65.0%
Homeownership rate, Q2 2026
“Virtually the same” as Q2 2025, per Census
Data
Rental vacancy was 7.3% in the second quarter of 2026, up from 5.6% four years earlier
Census states: “National vacancy rates in the second quarter 2026 were 7.3 percent for rental housing and 1.2 percent for homeowner housing.” The release's Table 1 gives second-quarter rental vacancy rates of 5.6% in 2022, 6.3% in 2023, 6.6% in 2024, 7.0% in 2025 and 7.3% in 2026. Census notes the 2026 figure “was not statistically different from the rate in the second quarter 2025 (7.0 percent)”; the 90% confidence interval it publishes is 7.1% to 7.5%.
Data
Vacancy is highest in the South and rising in the Northeast
By region, the Q2 2026 rental vacancy rate was 9.5% in the South (9.0% a year earlier), 6.9% in the Midwest (6.6%), 5.9% in the Northeast (5.2%) and 5.3% in the West (5.7%). Census says only the Northeast change is statistically significant; the Northeast rate “was higher than the rate in the second quarter 2025.”
Interpretation
A loosening rental market in 2022 to 2026 is consistent with shelter CPI slowing in 2025 to 2026
A vacancy rate that has climbed for four straight second quarters means more empty units competing for tenants. That tends to show up first in asking rents, then in renewal rents as leases roll, and only then, after the six-month panel and the sixth root, in the CPI. The CPI shelter line moved from 3.6% in the year to August 2025 to 3.0% in the year to August 2026. We cannot measure the lag from these two surveys alone, and the Census definition of asking rent (“the amount asked for the unit at the time of interview”) is not comparable with a rent index in level or construction. The direction, however, is the same.
Fact
BLS's own leading indicator is paused
BLS publishes a research series, the New Tenant Rent Index, built from “the first survey observations after new tenants move into their sampled housing units.” It is “perpetually revised, with recent periods being prone to large revisions” and calculated “outside of the official CPI production system.” On a page last modified April 24, 2026, BLS states that “due to a lapse in appropriations resulting in uncollected CPI Housing Survey data for October 2025 and competing priorities, BLS paused publication of the R-CPI-NTR and R-CPI-ATR data in April 2026.” The series that would most directly show where the official rent line is heading is, as of October 12, 2026, not being published.

What it means for your money

For a renter, the CPI rent line is a weighted average of everyone's rent, most of them mid-lease. If your own lease renews in 2026, your increase is set by your landlord and your local market, not by the 2.7% national figure, and could be far above or below it. For an owner with a fixed mortgage, the 3.1% owners' equivalent rent figure is a statistical construct: your housing cost did not rise 3.1%, and the index does not say it did.

Wealth Signal calculation
Shelter's share of a 3.4% year, in dollars
A household spending $60,000 a year at the CPI-U weights allocates 0.35343 × $60,000 = $21,206 to shelter as the index defines it. A 3.0% rise on that is $636 a year; a 3.4% rise on the whole $60,000 is $2,040. The $636 is 31% of the $2,040, below shelter's 35% weight, which is the arithmetic of a component rising more slowly than the whole.

Where the lag matters most is in the things indexed to the CPI. The Social Security cost-of-living adjustment, the I bond rate and the tax bracket thresholds all take the shelter line as published, lag and all. Our piece on the COLA formula and the I bond formula walk through that indexing. A reader checking whether an inflation figure is current can use the method in this article; for the shelter line, the honest answer is that the figure published on September 11, 2026 describes rents sampled across the six months to August and priced against the six months before that.

The September 2026 CPI is on the BLS calendar for Wednesday, October 14, 2026 at 8:30 a.m. ET, and the third-quarter Housing Vacancies release for October 28, 2026. Both dates are on the release calendar.

This article does not claim that CPI shelter is wrong, that it overstates or understates true rent inflation by any specific amount, or where the shelter line goes next. It does not estimate the length of the lag between asking rents and the CPI; the only lag described is the six-month panel and sixth-root mechanism as documented by BLS. Census asking rents and vacancy rates are survey estimates with published margins of error, and the vacancy comparison Census itself labels as not statistically different from a year earlier is reported as such. No figure here comes from a private rent index.

Sources

Wealth Signal prefers primary sources — regulators, statistical agencies and company filings. Named secondary sources are used where a primary document does not exist or is not public. Our source standards.

Editor and publisher · Markets, economics, personal finance

Wealth Signal is researched, written, edited and published by one person, under the Wealth Signal name. Every figure links to the document it came from, and errors are corrected on the page and logged publicly.

Related analysis

Cite this: Wealth Signal NYC (October 13, 2026). “CPI shelter ran 3.0% to August 2026, below the 3.4% headline.” https://www.wealthsignalnyc.com/stories/rent-vs-cpi-shelter-august-2026/ — the figures are the issuing agency’s, listed with their dates in the sources above; quote the agency, link the explainer.

This article is general information and analysis, not investment, tax or legal advice; read the financial disclaimer. Spotted an error? Corrections are made on the page and logged on the corrections page — tell the editor.