Why I Published the True Cost Basket Index and the Shelter Index

If you ask the average federal inflation report what a household is paying this month, it answers at the country line. CPI national. Food at home national. Shelter component national. Then the press release rolls the same number into a single headline, and that single headline becomes the basis for a policy talking point, an employer cost-of-living adjustment, or a household relocation decision. The household standing in a grocery store in Bakersfield does not see the same prices as the household in Brooklyn, and the renter in rural Mississippi does not face the same monthly nut as the renter in Boston. The headline averages them anyway.

Aggregation at the national line is not malice. It is what most of the federal series support — sample sizes, survey design, release cadences. A careful analyst respects the limits. But the gap between what the household actually pays and what the headline reports is not small, and the federal data does in fact publish at finer grains than the headline uses. The work is putting those finer grains on a map a household can read without an econ degree.

That is the gap two new tools now fill on this site: the True Cost Basket Index (TCBI) and the True Shelter and Family Cost Index (TSCI). Both are interactive, both are free, both attribute every value to a federal source cell.

What the TCBI is

The TCBI is a five-domain weekly household basket. D1 Pump uses EIA Weekly Retail Gasoline at the state and PADD-region grain — refreshed to May 11, 2026 in this publication snapshot. D2 Pantry pulls BLS Average Price Data across the four Census regions. D3 Plate applies the USDA Thrifty Food Plan to a Mediterranean family-of-four week. D4 Restaurant uses BLS CPI Food Away From Home as the national baseline, scaled by state-level cost ratio. D5 is a composite that weights the housing-correlated domains against the gas series.

The map publishes at three resolutions. National choropleth across all 50 states and DC. County resolution across 3,142 counties. And tract resolution — about 17,500 census tracts — inside fifteen mega-counties covering roughly 17% of the U.S. population. New York City publishes 262 Neighborhood Tabulation Areas across all five boroughs, point-in-polygon aggregated from the underlying ACS5 tract values.

What the TSCI is

The TSCI is the fixed monthly counterpart. D1 Rent at the county grain from Census ACS5 median gross rent. D2 Mortgage built from ACS5 median home value and the owner-costs-with-mortgage table. D3 Insurance from the National Association of Insurance Commissioners HO-3 (homeowner) and HO-4 (renter) premium tables — state-level only, because that is what NAIC publishes. D4 Childcare from the Department of Labor National Database of Childcare Prices, at the county grain where the source publishes.

The whole point of TSCI is the monthly nut: the column on a household budget that does not move week to week. Coverage is all 3,222 counties plus Puerto Rico, with a 33,791-ZIP search index built from the Census ZCTA-to-county relationship file. A two-county comparator surfaces the rent, mortgage, insurance, and childcare deltas in one panel — the view a family weighing a real move actually needs.

How this got published

I should be direct about credit, as I was with the Veteran Livability Index. The underlying tools were built upstream as the TCBI v3.2.0 and TSCI v2.1.6 publication bundles — the data assembly, the SVG geometry, the composite formulas, the audit workbooks, the interactive layer. My work on this is the editorial wrap: deciding the tools belonged on the Foundation's surface, framing them for households and analysts rather than for econ researchers, writing the wrapper pages in this site's voice, integrating them cleanly with the rest of the analytics portfolio, and publishing them with the audit workbooks and the methodology one-pager so every number stays verifiable.

I am the publisher of record on this surface, not the original engineer of the bundles. That distinction matters because the work that is mine — the editorial framing, the integration discipline, the publishing posture — is the work I want to be judged on. The work that is not mine deserves its own attribution, which the methodology panel inside each tool, and the one-pager linked from each case-study page, provides.

What the tools do well

The first thing they do well is move the resolution of the conversation. A household weighing two counties no longer needs to assemble seven federal sources by hand. A renter comparing two ZIP codes can read the monthly nut in one panel. A planning office writing a county profile no longer translates national averages into local approximations.

The second is the provenance flag taxonomy. Every value carries a label — LIVE, PROXY, STATE-FLAT, COUNTY-LEVEL, DATA-SUPPRESSED. A reader knows immediately whether they are looking at a direct federal value, a documented inheritance, or a known suppression. The methodology one-pager prints to a single letter page and lists every flag class with its rule. If a user wants to verify a single cell, the audit workbook traces it back to source.

The third is the publishing posture. No signup. No tracking. No analytics on the standalone builds. The full audit workbooks download as single Excel files. The data is federal public domain. Anyone can verify, fork, or extend.

What the tools cannot do

Two structural limitations are worth stating plainly.

The first is the clamp. The tract value formula is state_baseline × tract_fc, where tract_fc is the tract median rent divided by the state average, clamped to [0.7, 1.5]. The clamp keeps extrapolation honest at the very-low and very-high rent extremes — rural counties and Manhattan — where the housing baseline starts to break down. The cost is a visually flat region in those extremes. The audit workbook's Calibration_Audit sheet documents every clamped tract.

The second is state inheritance. TSCI insurance is NAIC state-only. TCBI D4 Restaurant inherits county to tract because BLS CPI Food Away From Home does not publish at the tract grain. Eight states plus Puerto Rico carry a proxy flag for one or more DOL NDCP childcare slots. These are documented constraints, not modeling shortcuts. A rural Mississippi county will show the same HO-3 premium as Jackson because that is what the published source supports. The flag is on the tile; the rule is in the methodology panel.

What I want from you

Both tools are built to be used and critiqued, not admired. If you find a value that looks wrong, a state inheritance that distorts the read, a clamp that flattens a region you know is not flat, or a use case the layout does not support — tell me. The next vintage will be better for it. Decision tools earn their credibility by being correctable in public.

The TCBI case study is at patrickneilbradley.com/tcbi-case-study.html. The TSCI case study is at patrickneilbradley.com/tsci-case-study.html. The methodology one-pager is linked from both. Feedback goes to patrick@patrickneilbradley.com or through the contact form.