Sovereign
All inquiries
SystemsINQ·17

Inflation Is Not One Number

Who is actually losing purchasing power? Build a separate inflation rate for each tenth of U.S. households, set it against that tenth’s own income from 2019 to 2025, price one paycheck against rent, food, and gasoline, then measure the cost the consumer index never sees: the price of becoming an owner.

INTERACTIVE · MODEL·10 income deciles · income 2019–2025·BLS CPI-U through Sep 2026

A national inflation rate is an average over households that do not exist. Nobody buys the CPI basket. In 2025 a family in the bottom tenth spent 55.6% of its budget on food and housing; the tenth just above it spent 56.9%; a family in the top tenth spent 40.5%. When prices move unevenly, the same month is a different month depending on where you stand — and a single number cannot say who is falling behind.

Data window · BLS Consumer Expenditure Survey 2025 decile shares and income before taxes · BLS CPI-U monthly index levels Jan 2018 – Sep 2026 · FHFA HPI Aug 2026 · Freddie Mac PMMS Oct 29 2026 · paycheck: Census household income 2025, BLS usual weekly earnings Q3 2026, SSA wages 2023, IRS Rev. Proc. 2025-32, USDA food plans Dec 2025, HUD FY2026 fair market rents, EIA gasoline Nov 2 2026, FHWA Highway Statistics 2024 · reconciled 2026·11·04

01Three measurements, not one

The inquiry starts by refusing to treat the published figure as everyone’s experience. It builds three measurements and reconciles them against each other.

Official
CPI-U
The Bureau of Labor Statistics all-urban index: one basket, one rate, the number in the headline. 3.7% for the twelve months to September 2026.
Independently observed
Prices actually paid
Receipts, invoices, rents, utility bills, insurance premiums — matched item for item, a year apart. The audit table in the lab is built for this.
Economic-group
Inflation per decile
The price change each income group faces given what it actually buys. This is the measurement this version computes.

02Two different ways a group can face more inflation

Basket composition. Lower-income households devote more of their budgets to necessities; higher-income households more to discretionary spending. Identical price changes then produce different rates. Prices actually paid. Households also shop at different stores, buy different brands, live in different housing markets, and borrow on different terms. BLS states that its experimental income-quintile indexes capture the first effect far more fully than the second. Neither does this model — yet.

The first effect is pure arithmetic. Suppose food rises 10% and everything else 2%:

Same prices · two budgets (illustrative)
Food 50% · other 50%
0.0%
Food 10% · other 90%
0.0%
Gold = food’s contribution, slate = everything else. One household experiences more than twice the inflation of the other from identical price tags.
0.5 times 10 percent plus 0.5 times 2 percent equals 6.0 percent, versus 0.1 times 10 percent plus 0.9 times 2 percent equals 2.8 percent
the basket-composition effect, isolated

03The model

Split households into ten income deciles. For each group g, hold its baseline basket fixed and price it forward:

I sub g of t equals 100 times the sum over items i of w sub g,i times P sub g,i,t over P sub g,i,0
wg,i = group g’s baseline spending share on item i · Pg,i,t = price that group pays at time t

A fixed basket isolates price change from changes in quantity: it answers how fast is the cost of keeping this group’s existing standard of living rising? A changing-basket version can be added later to measure substitution. In this version, the price term is the national CPI category index — the same for every group — so every difference between deciles comes from the weights.

The weights are BLS Consumer Expenditure Survey shares for 2025 across seven categories: food, housing, transportation, healthcare, entertainment, personal care, and miscellaneous. These cover 88.4% of the bottom decile’s outlays and 70.2% of the top decile’s. The remainder — insurance and pensions, education, apparel, cash contributions — is charged the headline rate. The 2022 shares, which the first version of this model used, stay in the lab as a vintage:

pi sub g equals one hundredth of the sum over seven categories of w sub g,i times pi sub i plus w sub g,res times pi sub res
the full-basket proxy · πres = headline CPI-U by default · the lab also offers a mapped-only variant that excludes the residual
One rule governs every result below. A gap between a decile’s rate and the published CPI measures that decile’s exposure to uneven prices. It is not, by itself, evidence that the published prices are wrong. Testing that requires matched transaction prices — the independent measurement this inquiry is built to collect next.

04September 2026: the fuel year

Twelve months to September 2026, CPI-U rose 3.7%. Underneath, energy rose 20.4% and motor fuel 34.9%, pushing transportation to +7.9%. Food (+2.5%) and housing (+3.3%) — the categories that dominate low-income budgets — ran below the headline.

DecileFood + housingTransportCoverageInflationvs CPI
Bottom 10%55.6%15.9%88.4%3.80%+0.10
10–20%56.9%14.7%90.3%3.70%0.00
20–30%55.0%15.2%87.0%3.75%+0.05
30–40%52.9%16.1%85.8%3.81%+0.11
40–50%50.9%16.9%83.9%3.86%+0.16
50–60%48.2%18.6%82.0%3.96%+0.26
60–70%46.7%18.5%81.4%3.96%+0.26
70–80%45.2%16.7%78.1%3.90%+0.20
80–90%44.0%17.4%75.8%3.96%+0.26
Top 10%40.5%14.8%70.2%3.90%+0.20

The modeled range is 3.70% to 3.96%. Three groups tie for the hardest hit — the 50th–60th, 60th–70th, and 80th–90th percentiles — all at 3.96%. The lightest is the 10–20% group, at the headline, because it puts the most of its budget into food and housing and the least into transport. Hold transport at the headline and the order flips: the bottom decile falls to 3.13% and the top sits at 3.27%.

05Seven years of real prices

One month proves little. The lab replays every month from January 2019 to September 2026 using the actual BLS index levels for all eight series, applied to the same baskets. Track one number — the bottom decile’s rate minus the top decile’s:

Delta sub t equals pi of the bottom decile at t minus pi of the top decile at t
positive Δ: the bottom tenth faces more inflation · negative Δ: the top tenth does
PeriodWhat led pricesAverage ΔRange
2019Calm; shelter steady+0.16+0.10 … +0.21
2020Gasoline collapses, food rises+0.23+0.11 … +0.38
2021Used cars, fuel rebound−0.16−0.31 … +0.15
2022Fuel peak, then food and shelter−0.03−0.23 … +0.23
2023Food and shelter, fuel deflating+0.18+0.04 … +0.30
2024Shelter leads+0.16+0.10 … +0.23
2025 (11 months)Shelter leads+0.15+0.10 … +0.21
2026 (Jan–Sep)Energy shock0.00−0.10 … +0.15

In 70 of 92 published months the bottom tenth faced more inflation than the top tenth. The gap peaked at +0.38 percentage points in May 2020 and at −0.31 in June 2021. On the 2022 baskets the same prices still reach 0.61 points, in May 2020: the bottom decile’s food share has since fallen from 17.2% to 14.0%, so the same grocery shock moves it less. At the June 2022 peak — 9.1% headline — the hardest-hit groups were the 50th–70th percentiles at 10.0%, and the bottom decile ran 9.5%. The bottom decile’s worst relative stretch came after the peak, in 2023, when fuel deflated and groceries and rent kept climbing. Compounded from January 2019 to September 2026, prices rose 33.9% on the headline index and between 35.6% (top decile) and 36.3% (50th–60th percentile) on the 2025 baskets. The bottom decile compounded 36.2%.

06Purchasing power: who falls behind

A group loses purchasing power when its income grows more slowly than its own inflation rate:

real income growth for group g equals one plus nominal income growth over one plus that group's inflation, minus one
a group facing 7% inflation with 3% income growth loses 3.74% of its real purchasing power

The model gives each decile a break-even for whatever price scenario is on screen. In September 2026 the bottom decile needed at least 3.80% nominal income growth to stand still, and the three groups at 3.96% needed that. One month is a stress test. The matched record is six years of income for the same ten deciles.

That record is the Consumer Expenditure Survey’s mean income before taxes, calendar 2019 through calendar 2025, the same brackets as the baskets. After-tax income is the closer measure of what a household can spend, and BLS publishes it through 2023. For 2024 and 2025 the after-tax series is absent, because the tax model was not updated. The full window uses before-tax income. Price change over the window is the basket in the figure, which opens on the 2025 weights, priced with annual-average CPI. October 2025 was never published, so the 2025 average uses the other eleven months.

Real before-tax income · calendar 2019–2025

Each bar is one income decile. Real income divides the income BLS recorded by that decile’s own basket. Dollar change is the same recorded income in dollars, so it stays put when the basket changes.

Real income, percent · 2025 baskets
Dollar change in mean income before taxes
Each year on the fixed basket. A line below zero is a year when that bracket’s recorded income grew more slowly than its prices. Gold is the bottom tenth. Cyan is the top tenth.

The six-year window

DecileIncome 2019Income 2025NominalBasket inflationReal incomeDollar change

Each year · real before-tax income

Every decile’s real before-tax income rose. The bottom tenth rose 28.6%. The top tenth rose 2.5%. Basket inflation sat between 27.3% and 27.9% — a 0.6 point spread — while nominal income growth ran from 30.4% at the top to 64.5% at the bottom. The spread in basket inflation is 0.62 points. Over this window, income growth is what separated the brackets.

Dollars run the other way. The bottom tenth’s mean rose by $4,043. The top tenth’s rose by $89,562. A larger percentage of a much smaller income is still a smaller amount of money.

Each year’s decile is whoever was in it that year. A family that earned its way out of the bottom tenth is replaced by another. The 28.6% is the change in the income BLS records for that bracket. In 2025 the bracket reports $10,311 of income before taxes and $35,986 of spending, so the recorded income is a fraction of the budget the basket is built on.

The six-year gain hides 2021. That year every decile lost real income. The bottom lost 7.6% and the top lost 2.4%. In 2022 the bottom’s income caught up, +3.5% real, while nine deciles were still behind. In 2024 the top tenth’s mean before-tax income fell and its real income fell 4.6%. It reversed in 2025, +7.8%.

Where after-tax income exists, 2019 through 2023, the direction matches. The bottom tenth’s real after-tax income rose 17.6%. The top tenth’s was −0.4%. The bottom’s after-tax mean sits above its before-tax mean, which is the CE tax estimate picking up refundable credits.

That chart prices the bracket BLS records. The next section prices one wage against a city rent, food for four, and gasoline.

07A paycheck, priced

In 2023 the Social Security Administration counted 173,670,935 wage earners. Through the interval that ends at $69,999.99, the cumulative share is 71.47%. The median net compensation that year was $43,222.81. Full-time wage and salary workers are a smaller group with higher usual earnings: in the third quarter of 2026 the median was $1,232 a week, which is $64,064 over 52 weeks. A household can hold more than one of those earners. Census Table 6 for 2025, household money income in 2025 dollars, puts the 40th percentile at $68,600, the median at $87,460, and the 60th percentile at $110,300.

A $70,000 wage, single filer, tax year 2026, owes $6,570 of federal income tax after the $16,100 standard deduction and $5,355 of employee Social Security and Medicare tax. The paycheck is $58,075. The same wage on a joint return with one earner owes $4,040 of federal income tax and the same $5,355, and the paycheck is $60,605. The employer owes a matching $5,355 on top of the wage. On the $64,064 median the single paycheck is $53,655 and the joint one-earner paycheck is $55,835. The distance from the $58,075 paycheck to $50,000 of cash is $8,075. That is the amount state income tax and sales tax would have to collect to land there, and those taxes differ by state. This page leaves them uncomputed.

HUD’s fiscal 2026 fair market rent is the 40th-percentile gross rent, utilities included. The two-bedroom figure is $2,910 in the New York HUD metro, $2,903 in Los Angeles County, $1,931 in Dallas, and $1,573 in Houston. A $2,700 two-bedroom sits above Dallas and Houston and below Los Angeles and New York. Inside a metro the zip-code rents run on either side of the metro figure. Houston’s own release publishes zip-code two-bedrooms of $2,360.

USDA’s December 2025 moderate-cost food plan, all meals and snacks prepared at home, prices a four-person household — a man and a woman age 19–50, a child 6–8, and a child 9–11 — at $1,347.20 a month. That is $44.29 a day and $16,166 a year. The liberal plan for the same household is $1,628.40 a month, $53.54 a day. Fifty dollars a day is $18,250 a year and sits between those two plans. Restaurant meals sit outside both.

FHWA Highway Statistics 2024 records 3,294,031 million vehicle-miles and 239,824,944 licensed drivers. Dividing the first by the second and rounding to the nearest mile gives 13,735 miles per licensed driver, and that quotient hands truck and bus miles to drivers along with cars. A short-wheelbase light-duty vehicle averaged 10,812 miles and 25.6 miles per gallon. At the November 2, 2026 U.S. regular gasoline price of $4.153 a gallon, 13,735 miles costs $2,228. At $4.50 a gallon the same miles cost $2,414, and two such drivers cost $4,829. Four thousand five hundred dollars at $4.50 a gallon is 1,000 gallons. At 25.6 miles per gallon that is 25,600 miles.

That $2,228 is the household’s yearly cost of oil, paid by one driver to the seller of the gasoline. On the world record, a 2024 petro-dollar of oil bought 70.73 MJ of oil that was consumed in every observed year from 1920 through 2025, and gas bought 116.31 MJ over 1965–2024. What $100 Pays For is that comparison. It does not draw this pump price into a higher future year.

On the $53,655 single paycheck, the Houston rent, the moderate food plan, and the $2,228 of gasoline leave $16,385 for the year. The same food and gasoline with the New York rent leave $341. Los Angeles leaves $425. Dallas leaves $12,089. A $2,700 rent, food at $50 a day, and 13,735 miles at $4.50 leave $591. The same rent and food with 25,600 miles, which is $4,500 of fuel, spends $55,150 and leaves the paycheck $1,495 short. State income tax, health insurance, a car payment, and childcare are still ahead of whatever remains.

The 28.6% in the chart above is the change in a bracket whose 2025 mean income before taxes is $10,311. The $64,064 is one full-time worker’s usual weekly earnings, annualized. Both figures are official. They describe different people.
One paycheck · rent, food, fuel

The paycheck is the wage minus federal income tax and employee FICA. Left is that paycheck minus a year of rent, food, and gasoline. Employer FICA is shown on top of the wage and is kept out of the subtraction.

Federal income tax is the annual liability after the 2026 standard deduction. Employee FICA is 6.2% up to the Social Security wage base of $184,500 plus 1.45% Medicare on all wages. The 0.9% Additional Medicare Tax starts at $200,000 of wages for a single filer and $250,000 on a joint return. Withholding can move a few dollars between paychecks. The annual liability is the figure in the readout. Food uses a 365-day year. The moderate day, $1,347.20 × 12 / 365, is $44.29.

Federal paycheck · tax year 2026

WageFilingFederal income taxEmployee FICAPaycheckEmployer FICA, on top
$64,064Single$5,508$4,901$53,655$4,901
$64,064Joint, one earner$3,328$4,901$55,835$4,901
$70,000Single$6,570$5,355$58,075$5,355
$70,000Joint, one earner$4,040$5,355$60,605$5,355

What $53,655 covers

Case, single filerRentFoodFuelLeft
Houston · moderate food · $4.153 · 13,735 mi$18,876$16,166$2,228$16,385
Dallas · same food and fuel$23,172$16,166$2,228$12,089
Los Angeles · same food and fuel$34,836$16,166$2,228$425
New York · same food and fuel$34,920$16,166$2,228$341
$2,700 · $50/day · $4.50 · 13,735 mi$32,400$18,250$2,414$591
$2,700 · $50/day · $4.50 · 25,600 mi$32,400$18,250$4,500−$1,495

08The cost of becoming an owner

Consumer inflation does not measure how much harder it becomes to acquire a home, a business, farmland, or financial assets. Those acquisition costs govern wealth accumulation and mobility, so they get their own index rather than being mixed into the consumer basket. The first is the monthly cost of buying a house:

M equals L times r over one minus one plus r to the minus 360, where L equals P times one minus d and r equals R over 12
principal and interest on a 30-year fixed loan · P price · d down-payment share · R annual rate
Illustrative $400k house, 20% downA year earlierNow
Home price (FHFA +2.4%)$400,000$409,600
30-year fixed rate (Freddie Mac)6.17%7.24%
Monthly principal & interest$1,954$2,233
Income needed at 28% of gross$83,729$95,706
Change in monthly entry cost+14.3%
Twelve-month change · Sep–Oct 2026
Consumer prices
0.0%
House price
0.0%
Monthly payment
0.0%
The house itself rose less than consumer prices. Financing turned it into a 14.3% increase in the monthly cost of entry — 3.9× the CPI. About 82% of that increase comes from the rate move alone.

The FHFA price window (August to August) and the mortgage-rate window (October to October) are not matched, so this is a financing stress test rather than a matched-time housing index. It excludes taxes, insurance, HOA dues, and closing costs. The direction is unambiguous anyway: the income required to enter rose by $11,977 in a year in which CPI rose 3.7%.

09Run the model

The purchasing-power figure draws the 2019–2025 income record and recomputes real income when the basket vintage changes. The paycheck figure prices one wage against rent, food, and fuel. The lab sets category price changes or a scenario, switches the same 2025 and 2022 baskets, and rebuilds the decile bars from their category contributions. Replay the 92 published months. A separate slider still tests a hypothetical growth rate against the scenario on screen. Stress the home-purchase inputs, or start logging matched prices of your own.

INTERACTIVE · 2025 baskets × BLS prices · income 2019–2025 · home access · audit open full screen ↗

10What the arithmetic insists on

First: who faces the higher rate depends on what is rising. When groceries and rent lead, the lower deciles face the higher rate. When fuel leads, the car-dependent upper-middle faces it. From 2019 to September 2026 the bottom-minus-top gap changed sign three times, and the single hardest-hit decile changed 17 times. With the 2025 baskets, the 10–20% group — the one with the heaviest food-and-housing share — was the hardest hit in 51 of the 92 months.

Second: basket composition is a small effect, and income growth is the large one. On the 2025 baskets the bottom-minus-top gap never exceeded 0.38 percentage points in a year. Compounded from January 2019 to September 2026 the ten baskets span 35.6% to 36.3%, against a 33.9% headline. Over the income window, 2019 to 2025, basket inflation spans 27.3% to 27.9% while nominal before-tax income growth spans 30.4% to 64.5%. Real before-tax income of the Consumer Expenditure brackets rose in every decile: 28.6% at the bottom, 2.5% at the top. The dollar changes run the other way, +$4,043 against +$89,562.

Third: the spike still took purchasing power, and it took it from the bottom first. In 2021 every decile’s real before-tax income fell. The bottom fell 7.6% and the top fell 2.4%. In 2022 nine deciles were still behind. The six-year gain is the recovery after that, plus income growth that outran a 0.62-point basket spread. The households inside a decile change from year to year, so these percents describe the bracket BLS records.

Fourth: the same percentage is a different burden. In September 2026 the bottom decile’s 3.80% is close to the headline. 55.6% of that budget is food and shelter, spending with little room to postpone. A fixed-basket index measures the cost of keeping a standard of living. The income table is what measures whether the bracket’s recorded income kept up. In 2025 the bottom tenth’s spending was $35,986 against $10,311 of reported income before taxes. The recorded income is a fraction of the budget the basket is built on.

Fifth: every decile basket ran ahead of the headline. From January 2019 to September 2026, all ten 2025 baskets compounded 1.8 to 2.4 points above the published 33.9%. Households concentrate measured spending in food, shelter, and transportation, the categories that rose fastest. That is a property of the weights. It also understates healthcare: the CE healthcare share is mostly insurance premiums, and the medical-care CPI used as its price is not a premium index.

Sixth: the cost of entering ownership is diverging from the cost of living. In a single year, the monthly cost of buying a median-type house rose 3.9× faster than consumer prices, and four-fifths of that came from financing, not the house. Owners with fixed-rate mortgages were insulated and gained 2.4% in equity; aspiring owners paid a $11,977 rise in the income bar. That is a transfer between those who already hold the asset and those trying to acquire it — the channel this inquiry shares with Control the Flow, Control the System, where monetary expansion and credit pricing redistribute wealth.

Seventh: a paycheck in a city is a tighter object than a decile mean. In 2023, 71.47% of wage earners had net compensation under $70,000, and the median was $43,222.81. The third-quarter 2026 median for full-time workers was $1,232 a week, $64,064 over 52 weeks. The 2026 federal paycheck on that wage is $53,655 for a single filer. HUD’s fiscal 2026 two-bedroom fair market rent is $1,573 in Houston, $1,931 in Dallas, $2,903 in Los Angeles County, and $2,910 in the New York metro. With the USDA moderate food plan and one licensed driver’s gasoline at the November 2 pump price, the Houston rent leaves $16,385 of that paycheck and the New York rent leaves $341. A $2,700 rent, $50 a day of food, and $4,500 of fuel leave it $1,495 short for the year, before state income tax, health insurance, a car payment, and childcare.

11What makes it independent

The 2025 baskets and the income series are now in the model. What remains is what would make the rates independent of official category prices:

UpgradeWhy it mattersStatus
2025 decile weights2022 baskets predate the later shifts in food and shelter shares. The bottom decile’s food share fell from 17.2% to 14.0%.Loaded · 2022 still selectable
Matched-price databaseReceipts, utility bills, rent and insurance statements, retailer prices — the way to measure prices paid by each group and test the official category rates.Audit structure ready · no receipts invented
Income by decile, same periodTurns break-evens into real purchasing-power gains and losses per group.Before-tax, 2016–2025 · after-tax through 2023
More asset-access indexesSmall-business entry, farmland, and a financial-asset yield-per-dollar index alongside housing.Housing only
SubgroupsRenter vs owner, age, region, household size within each decile — only with measured price premiums.Overlay ready

What this version now does is price each income decile’s own basket and set that price against the income BLS records for the same decile. It also prices one wage against a city rent, a four-person food plan, and gasoline. Replacing the national category price with the price each group actually pays waits on matched receipts in the audit table. Until those receipts exist, every inflation rate here is official prices reweighted. The paycheck section uses the rent, the food plan, and the pump price directly. That is enough to show that basket composition is the small term in the decile rates, and that a two-bedroom in New York consumes a median full-time paycheck. Auditing the official prices takes the receipts.

Download the v1 workbook (.xlsx) Open the lab ↗

Primary sources (reconciled Nov 2026)