Methodology · Full Paper
This paper documents how the Reality Index is constructed, what it measures, how it differs methodologically from the Consumer Price Index, and where critics will find legitimate room to disagree. It is intended for the reader who wants the technical answer, not the journalistic summary.
The Reality Index v1 tracks the retail price of fifteen specific items in five categories: food, energy, housing, health care, and services. Each item is selected on three criteria: (a) the item is a fixed-specification good or service that households actually purchase, (b) an authoritative retail-price time series exists for it going back at least to 1980, and (c) the item is non-discretionary, in the sense that the typical household consumes some amount of it whether or not its price has risen.
The fifteen items are: eggs (Grade A Large, per dozen), ground beef (100% beef, per pound), chicken breast (boneless, per pound), whole milk (fortified, per gallon), white bread (per pound), gasoline (regular unleaded, per gallon), electricity (per kilowatt-hour), rent of primary residence (BLS subindex), home prices (FHFA House Price Index), first-class postage (per letter), family employer-sponsored health insurance (annual premium), and four additional health care measures (single premium, family worker contribution, single worker contribution, single deductible). Tuition appears as a separate item with multiple sub-series (public 4-year, public 2-year, private 4-year, BLS CPI subindex). Direct tax burden is tracked separately as a rate rather than a price.
This basket is partial. It explicitly excludes restaurants, vehicle purchases, durable goods, child care, entertainment, and discretionary spending. A more complete and synthesized index — the American Dream Index — extends the Reality Index methodology to measure the fixed-basket annual cost of an entire middle-class lifestyle, and is available now. We do not claim that the Reality Index is a complete measure of household cost of living; we claim that for the specific items it measures, it tracks them more honestly than any single line in the official CPI does.
A standing commitment to accuracy. Reality Index is built to be as accurate as possible, not as accurate as it is today. If we identify a more rigorous data source, a methodology improvement, or a measurement error — whether through our own continued research, reviewer feedback, or critique from readers — we will publish the update, announce the change in /news, and revise the affected pages with a timestamped methodology note. The goal is to be the most honest reference number for what inflation has actually cost American families, which requires being willing to revise the methodology when revision is warranted. Methodology suggestions and data corrections from any reader are welcome at info@realityindex.co.
Every data series in the Reality Index is sourced from an authoritative public time series. Where multiple sources exist for the same underlying measurement, we prefer the one whose methodology is most stable and most transparent. Where multiple methodologies exist (e.g., NCES vs. College Board for tuition), we publish both and discuss the difference.
| Item | Source | Series | Span |
|---|---|---|---|
| Eggs | BLS Average Price Data | APU0000708111 | 1980+ |
| Eggs (subindex) | BLS CPI | CUUR0000SEFH | 1935+ |
| Ground beef | BLS Average Price Data | APU0000703112 | 1984+ |
| Ground beef (subindex) | BLS CPI | CUUR0000SEFC01 | 1943+ |
| Chicken breast | BLS Average Price Data | APU0000706212 | 1980+ |
| White bread | BLS Average Price Data | APU0000702111 | 1980+ |
| Whole milk | BLS Average Price Data | APU0000709112 | 1995+ |
| Gasoline | EIA Monthly Energy Review | Table 9.4 | 1950+ |
| Electricity | BLS Average Price Data | APU000072610 | 1979+ |
| Rent of primary residence | BLS CPI | CUUR0000SEHA | 1940+ |
| Home prices | FHFA HPI All-Transactions | US national | 1975+ |
| First-class stamp | USPS rate history | regulatory record | 1971+ |
| Family health premium | KFF Employer Health Benefits Survey | Section 1, Fig. 1.12 | 1999+ |
| Public 4-year tuition | NCES Digest, Table 330.10 | in-state, tuition+fees | 1963+ |
| CPI tuition subindex | BLS CPI | CUUR0000SEEB01 | 1978+ |
| CPI-U All Items | BLS CPI | CUUR0000SA0 | 1947+ |
Three points worth flagging. First, the BLS Average Price Data series (the APU codes) are distinct from the BLS CPI subindex series (the CUUR codes) for the same category. The APU series report the average retail price at which the specified item was actually sold in the United States, with no adjustments. The CPI subindex applies BLS's substitution methodology and quality adjustments. The Reality Index uses the APU series as its primary line wherever it exists, and uses the CPI subindex as a backstop for items where APU coverage is shorter or unavailable.
Second, the BLS Average Price Data for some items has gaps in the early years. The eggs CPI subindex goes back to 1935 but the eggs APU series begins in 1980. We honestly report this: the eggs chart's solid red line begins in 1980, and the historical CPI subindex line extends behind it back to 1935.
Third, KFF's Annual Employer Health Benefits Survey is a private nonprofit source, not a government series. We use it because it is the most authoritative source for employer-sponsored insurance costs, its methodology has been stable since 1999, and its results are widely cited across the political spectrum. The Bureau of Labor Statistics has no comparable retail-dollar series for health insurance.
Each chart anchors at the earliest year for which its primary retail dollar series exists, indexing that year to 100. All other lines on the chart are scaled to converge with the primary line at the anchor year. This means we can read the chart as: "starting from this year and this base, how far has each measure of price grown?"
This choice has two consequences. First, it lets us compare items with different start dates on a common visual basis. Second, it means we are not claiming any line shows real-dollar value at the anchor year — we are showing relative growth from the anchor forward. For headline composite indexes, we choose 1980 as the common anchor because that is the earliest year all our primary data sources exist and because it roughly coincides with the beginning of the modern Federal Reserve monetary policy framework.
For every item where it is possible, we publish three lines: (1) the actual retail dollar price, (2) the BLS CPI subindex for that category, and (3) the BLS CPI-U All Items. The three-line treatment is methodologically important because it lets readers see two distinct gaps. The gap between (1) and (3) is the "Reality Index" gap — what families paid versus the average inflation rate the government reports. The gap between (1) and (2) is the methodological gap — the difference between BLS's own raw retail data and BLS's own published CPI subindex, after their substitution and quality adjustments. Where (2) is substantially different from (1), something inside the BLS methodology is producing a number that does not match its own underlying retail data.
CPI's modern methodology accommodates the fact that consumers respond to rising prices by buying less of expensive items and more of cheap substitutes. The Bureau of Labor Statistics calls this "geometric mean weighting" within categories, and it is one of the primary differences between the modern CPI-U and the older CPI-U-X1 and CPI-W series. The Reality Index does not do this. It tracks the price of the specified item year over year, regardless of whether consumer behavior has shifted.
This is a defensible choice, but it is a choice. We are measuring something different from what CPI measures. The right way to describe the difference is: CPI measures the cost of maintaining a constant standard of living, where standard of living is itself adjusting to what consumers can afford. The Reality Index measures the cost of maintaining a constant material specification of consumption. These produce different numbers because they answer different questions. Both questions are legitimate.
CPI applies hedonic quality adjustments to many items: when a new television model has more features than an old one, CPI treats some of the price increase as a quality improvement rather than inflation. This works reasonably well for technology goods. It works less well for food, where the same item specification can decline in quality over decades (smaller portions, lower meat content, more fillers) without the methodology adjusting prices downward to compensate. We call this reverse hedonic drift, and we do not adjust for it. The Reality Index tracks the price at which the BLS-specified item is sold. If a 2025 dozen Grade A Large eggs is in some way a different product than the 1980 specification — different flock conditions, different feed, different distribution chain — we hold the specification constant and let the price float.
The largest single divergence between the Reality Index and CPI is in housing. The Bureau of Labor Statistics measures the housing cost of homeowners through something called Owners' Equivalent Rent (OER), which estimates a hypothetical rental cost for owner-occupied homes. OER is the single largest component of headline CPI, weighing approximately 25 percent of the index. The Reality Index does not use OER. We use the FHFA House Price Index for homes (a repeat-sales index that controls for quality drift by tracking sales of the same homes over time) and the BLS Rent of Primary Residence subindex for renters. Both are real measures of real housing costs. Neither is OER.
We make this choice because, in our view, OER is the single largest source of the gap between official CPI and lived experience. Families do not pay an imputed rent on the home they own — they pay an actual mortgage on a home whose price is a function of the FHFA index. By using FHFA rather than OER, we capture the asset-price inflation in housing that CPI specifically excludes by design.
For the headline composite Reality Index, we currently use the BLS Consumer Expenditure Survey 2024 weights to determine how much each category contributes to the overall index. This is a transparency-and-defensibility choice rather than an endorsement of CES — we are using a published, citable weighting scheme so critics can replicate our composite. We are independently researching alternative weighting sources (Bank of America Institute, JPMorgan Chase Institute, Visa SpendingPulse, and others) that draw on anonymized credit card transaction data rather than household surveys. We intend to compare these in a future version.
Background: how CPI methodology itself has changed since 1983. The Consumer Price Index that BLS publishes today is the product of a series of methodological changes — Owner's Equivalent Rent (1983), the Boskin Commission recommendations (1996), hedonic quality adjustment (expanded 1998-onward), the geometric mean formula (1999), Chained CPI (introduced 2002, applied to federal tax brackets in 2017), and continuing hedonic expansion (2018-onward). Each change has a stated rationale and a critique. Each has the arithmetic effect of producing a lower reported inflation rate than the methodology it replaced. We have published a plain-language timeline of these changes at CPI Changes Timeline. Understanding what has changed in the CPI itself is essential context for understanding why a fixed-basket alternative like the Reality Index produces a different number.
"CPI is the most scrutinized economic statistic in the world. Why should anyone trust your alternative?"
We do not ask anyone to trust the Reality Index as a substitute for CPI. We publish a different measurement that answers a different question. CPI tells you how much the cost of a substitutable, quality-adjusted basket has grown. The Reality Index tells you how much the cost of a fixed specification of common purchases has grown. Both are useful. They are not in competition; they are complementary. The Federal Reserve and the federal government will continue to use CPI for monetary policy and indexing. Families wondering why their grocery bill feels higher than the news reports will find the Reality Index more informative.
"You're cherry-picking items that ran faster than CPI."
We are not. The Reality Index includes five items that ran slower than CPI over 1980–2025: chicken breast, whole milk, white bread, electricity, and gasoline. The pattern is clear: items where industrial scale and global supply chains drove costs down (mass-produced food, energy) actually got cheaper in real terms. Items where supply is constrained and substitution is impossible (housing, health care, education) ran dramatically faster than CPI. We report both halves of this pattern honestly. Cherry-picking would mean including only the hot items. We include the cool ones too — and we publish them with the same prominence.
"Quality adjustment is a real economic concept. Cars in 2025 really are better than cars in 1980."
Agreed. Hedonic adjustment is methodologically sound for many goods, particularly technology and durable goods where measurable feature improvements clearly compound over time. We do not dispute the principle. We dispute the consistency of application: BLS adjusts prices downward when product quality rises, but does not symmetrically adjust upward when product quality falls. The result is asymmetric methodology that systematically biases reported inflation downward in categories where quality is declining. We hold the specification constant and let the price float. This is a deliberately conservative methodological choice — the opposite of the BLS approach, but transparent about being so.
"Your eggs chart actually shows that BLS understates inflation by only 12% over 45 years. That's hardly chicanery."
Correct. The eggs gap is real but modest, and concentrated almost entirely in the 2022–2025 period when avian flu shocks hit retail prices harder than the smoothed CPI subindex captured. We do not claim eggs are an example of dramatic BLS chicanery. Eggs are an example of how the Reality Index works: we show the gap where it exists and report it at its actual size. The dramatic gaps are in housing (1.93x), health care (2.41x), and tuition (3.42x). Eggs are not the story. They are a fixed-specification staple that demonstrates the methodology.
"Your basket is too small to draw conclusions about overall inflation."
Agreed, and we say so on the homepage. The fifteen items in the Reality Index cover only a subset of household spending. We make no claim that the index represents the average American family's full budget. We claim that for the items it does measure, it measures them honestly. The American Dream Index — now live — extends the methodology to a full fixed-basket lifestyle cost, including restaurants, vehicles, child care, and discretionary spending.
"You're using government data while claiming to be independent of government data."
A fair point. The Reality Index is built primarily on BLS, Census, FHFA, EIA, and NCES data — all government sources. We are not anti-government-data; we are anti-particular-government-methodology in the construction of CPI. The raw retail prices BLS publishes in its Average Price Data series are entirely usable for our purposes, because they are not subject to the substitution and hedonic adjustments that we critique in CPI. We use BLS's own underlying retail data to build a different number than BLS's own headline CPI. This is internally consistent: we trust the data, we are publishing a different methodology that produces a different number from the same data.
All Reality Index calculations are reproducible from the citations on each chart page. The underlying time series can be downloaded directly from BLS, Census, FHFA, EIA, NCES, and KFF. We publish the Python source code that constructs each chart from these inputs on request.
When source data is revised by the underlying agency (BLS, KFF, NCES periodically revise prior-year estimates), we update the Reality Index to match. We do not retroactively change methodology — only data. Methodology changes are versioned (v1, v1.5, v2) and accompanied by an explanation of what changed and why. Prior versions of the index remain accessible.
If you find an error in our data or our methodology, please email info@realityindex.co. Errors will be corrected and credited. Corrections will be logged on this page.