The Inflation Number Lands at 8:30. A Quarter of It Is Rent Nobody Pays.

At 8:30 this morning the Bureau of Labor Statistics publishes the Consumer Price Index for July. By 8:34 you will know what inflation "is," to one decimal place, and a great many people will tell you what it means for interest rates, for the election, and for you.

One of those three is a lie by omission, and it's the one about you.

Not because the number is wrong — BLS produces it carefully and documents its methodology in exhausting public detail, which is more than most numbers you'll encounter this year can claim. It's because the CPI is a specific measurement of a specific basket, and that basket is a statistical American household. It is not your cost structure. It was never built to be, and BLS has never claimed otherwise.

Here's what's actually in it.

The arithmetic nobody shows you¶

Every component of the CPI carries a weight — its share of the notional basket, published by BLS as "relative importance." As of December 2025, using 2024 expenditure weights, the big ones look like this.

Shelter35.6%

Transportation16.3%

Food13.7%

Medical care8.4%

Energy6.4%

Apparel2.4%

Now take the June figures and do the multiplication, because this is the part that almost never appears in the coverage and it takes ten seconds.

Core inflation — everything except food and energy, 79.919% of the index — ran 2.6% over the year. Multiply: 2.08 percentage points.

Food, 13.698% of the index, ran 3.0%. That's 0.41 points.

Energy, 6.383% of the index, ran 15.7%. That's 1.00 point.

Add them: 2.08 + 0.41 + 1.00 = 3.49, which rounds to the 3.5% headline BLS published. The decomposition closes.

The quarter of the index that is a survey answer¶

Now the part that surprises people who have been quoting this number for years.

Shelter is 35.625% of the CPI. But the CPI doesn't measure house prices, and it never has, for a defensible reason: a house is largely an investment asset, and a consumption price index that included asset purchases would be measuring something other than the cost of living.

So BLS measures the service a home provides. For renters, that's straightforward — it's the rent. For homeowners, there's no transaction to observe at all, so BLS estimates one. It's called owners' equivalent rent, it is 26.204% of the entire index, and here is the actual question the Consumer Expenditure Survey puts to homeowners:

“

If someone were to rent your home today, how much do you think it would rent for monthly, unfurnished and without utilities?

”

That's it. That's the origin of roughly a quarter of the American inflation statistic: homeowners estimating a hypothetical rent for a house they are not renting, to a tenant who does not exist.

To be completely fair to BLS — and it deserves the fairness, because it publishes all of this openly rather than burying it — the price changes applied to that weight come from the CPI Housing Survey, which collects actual rents from sampled units every six months, adjusted for quality and the aging of the housing stock, with the monthly change derived as the sixth root of the six-month change. It is not vibes. It's a real measurement of real rents, applied to an imputed expenditure. BLS is also explicit that where units aren't collected, it "imputes price change for the non-collected units."

Shelter rose 3.3% over the year to June. Against a 35.625% weight, that's 1.18 percentage points — about a third of the entire 3.5% headline, and roughly three-quarters of that weight is the imputed portion.

Same series, opposite signs¶

One more thing to hold in your head when the number lands this morning.

In June, the energy index fell 5.7% on the month. The same energy index was up 15.7% over the twelve months ending that June.

Both are true. Both are correctly computed from the same underlying series. They point in opposite directions because they answer different questions — "what happened last month" and "where are we versus a year ago" — and whichever one gets quoted at you is a choice somebody made, sometimes innocently and sometimes not.

The general form of the trick: when a number can be framed over multiple windows, the person choosing the window is making the argument. You'll see both versions of the July figure before lunch today. Neither is a lie. Notice which one the speaker picked, and ask what the other one says.

Build the only index that matters to you¶

The useful response to all of this isn't cynicism about official statistics. It's ten minutes with a spreadsheet, once, and then five minutes a month forever.

1. 1

   Write down your actual basket — before you look at any prices

   List your real cost lines and their share of total spend: rent or mortgage on your space, labour, materials, fuel, software, insurance, payment processing. The order matters enormously and almost nobody has ever written it down. If shelter is 4% of your costs and labour is 60%, a 35.6%-shelter index is telling you almost nothing about your year.

2. 2

   Fix the weights first, then go get the prices

   This is the discipline that separates an index from cherry-picking, and it's exactly what BLS does — the weights come from expenditure surveys conducted before the prices are observed. Set your weights from last year's P&L, then look up what happened. Setting weights after you've seen the data is how you prove whatever you already believed.

3. 3

   Track the three lines that actually move you

   For most small operations it's labour, occupancy, and one input category. Those three usually explain the large majority of cost movement. Everything else is rounding, and tracking it is a hobby rather than management. Your overhead structure decides which three you get.

4. 4

   Price against your index, not the national one

   "CPI was 3.5%, so I'll raise 3.5%" is a decision made with someone else's basket. If your true input inflation was 6%, you just took a real margin cut and called it a price increase. If it was 1%, you left money on the table and annoyed customers for nothing. Run the actual numbers — properly — and watch what it does to gross margin.

5. 5

   Check it against the room

   The data has a lag and your customers don't. This desk has spent the month on the divergence between small business sales and transaction counts that have now fallen for nine straight months. If your index says costs are fine and the room says otherwise, the room is usually earlier.

Frequently asked questions¶

The number is fine. The inference is the problem.¶

At 8:30 a figure arrives, and within minutes it will have been converted into a claim about your life. Some of that conversion is careful and some of it is people who have never once looked at the relative importance table telling you what your costs did.

The number itself is one of the more honest artifacts the American government produces. BLS publishes its weights, its methodology, the exact wording of its survey questions, and its own historical error rates — which is precisely how I was able to write this. Every fact above came from BLS documenting its own limitations in public.

That's the standard, and it's worth stealing. Not "our numbers are perfect," but here is exactly how we got them and here is where they'll fail you. It's the same reason a jobs figure that gets revised is more trustworthy than one that doesn't.

Know what's in the basket before you let it tell you what things cost.

It's just business — and the basket isn't yours.

Sources

1. U.S. Bureau of Labor Statistics — Consumer Price Index Summary, June 2026 (released July 14, 2026)
2. U.S. Bureau of Labor Statistics — CPI relative importance of components, December 2025
3. U.S. Bureau of Labor Statistics — Measuring Price Change in the CPI: Rent and Rental Equivalence
4. U.S. Bureau of Labor Statistics — CPI release schedule