Data Methodology

Two geography types, never mixed

The dataset holds 4,764 HUD geographic entities: 3,161 counties and 1,603 county subdivisions. Each entity has one page. HUD publishes Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island and Vermont at town/city (county subdivision) grain, so those six states have no county-equivalent HUD record at all.

County figures for those states are never synthesised by aggregating their subdivisions. Inventing a county-level rent from town rows would produce a number HUD never published, so the site does not publish one.

Parent geography

Each county subdivision records the parent geography HUD assigns it. In five New England states that parent is a county. In Connecticut it is a Planning Region: Connecticut abolished county government in 2022, and HUD publishes the nine Planning Regions instead. The label follows HUD's own column, so a Connecticut page never calls its parent a county.

HUD-defined areas and the join

HUD does not publish income limits and rents per county or per town — it publishes them per HUD-defined area, which frequently spans several entities. Every entity is joined to its area on the 10-digit HUD FIPS code (state + county + county subdivision, where 99999 marks a county row); an area-code fallback is used only where the FIPS join is unavailable and the area's rows are unambiguous.

A consequence worth stating plainly: two entities in the same HUD area publish identical rents and limits. That is HUD's granularity, not a bug, and the site does not manufacture variation to disguise it. Each page names the HUD area it draws from.

Fair Market Rents (FMR)

The FMR is the 40th-percentile gross rent for standard-quality units in a local housing market. Each page shows every bedroom size HUD publishes for its area — studio, one, two, three and four bedroom — from the FY2026 tables. The two-bedroom value is used as the reference for comparisons because it is the size HUD itself uses most commonly for cross-area reference.

Income limits

Each page renders the full FY2026 matrix: three tiers — Extremely Low (30% of Area Median Income), Very Low (50% AMI) and Low (80% AMI) — across household sizes 1 through 8, for 24 published values in total. These are the thresholds HUD publishes for the area. They are not an eligibility determination: programs apply their own adjusted-income rules, deductions and allowances.

HUD Median Family Income

The income figure on every page is HUD's Median Family Income for the HUD area — the basis from which the income limits above are computed. It is not a Census Bureau median household income, and the two are not interchangeable: they differ in geography, in unit of analysis (family vs household) and in methodology. Do not mix these figures with ACS estimates without adjustment.

FMR-to-MFI benchmark

benchmark = (12 × FMR 2BR) ÷ HUD Median Family Income

A ratio between two HUD area benchmarks. It is deliberately not called a rent burden: no household in the area is observed paying it. A measured rent burden compares what real households actually spend against what they actually earn; this compares one published HUD figure with another. Treat it as a cross-area yardstick, not a statement about anybody's budget.

Within-state comparison and ranking

Each entity is compared against the other entities of its own geography type in the same state — counties against counties, county subdivisions against county subdivisions. Pooling them would rank a whole county against individual towns, so the two populations are kept strictly separate.

• The state median is the middle 2BR FMR of that population. It is unweighted — not weighted by households, population or land area.
• Rank 1 is the highest rent in the state. Ranking is competition-style: entities with identical rents share a rank, and the next distinct rent takes the rank after the whole tie block (1, 2, 2, 4).
• Because a HUD area can cover many entities, large ties are normal and expected — a single Boston-area rent can be shared by more than a hundred towns.

Purchasing-power module

A historical CPI-U adjustment of the HUD Median Family Income, not a forecast. The default window is an exact year-over-year pair: the most recent published month and the same calendar month twelve months earlier. If either endpoint was not published, the window steps back to the most recent month for which both endpoints exist — it is never approximated by counting list positions.

Adjusted = Nominal × CPI(to) ÷ CPI(from)

No interpolation is applied and missing months are not imputed. The module's “STABLE” threshold — a fixed +10% buffer over the local HUD Median Family Income — is a GovCalculator heuristic, chosen by us. It is not a HUD, BLS or IRS determination and carries no official standing.

Stored once, not per entity

National inputs are national: they do not vary by geography, so they are stored once and read per build rather than copied into entity records. An earlier version of this dataset replicated them into every county record, which made federal figures look as though they varied locally; that duplication has been removed.

Runtime CPI-U comes from the canonical full-history dataset data/inflation/cpi-u-all-items.json. Each geography page receives only a 36-published-observation presentation window derived from that same series — a display choice, not the extent of the data. The separate BLS Public Data API retrieval is kept solely to verify that series against an independent source; it supplies no displayed CPI value.

data/national_context.json now carries only the IRS Publication 15-T reference metadata — the publication’s identity and official URLs. Its withholding tables are not ingested, no nationwide withholding payload is stored or modelled, and GovCalculator does not compute federal withholding.

BLS CPI-U (CUUR0000SA0) — official series

The Consumer Price Index for All Urban Consumers, U.S. city average, all items, not seasonally adjusted. Unadjusted is the correct choice here because the purchasing-power ratio compares two specific named months; seasonal adjustment would distort a month-to-month comparison.

Each geography page presents a recent window of 36 published monthly observations covering 2023-07 – 2026-07, spanning 37 calendar months. That is a presentation window, not the extent of the data: the canonical CPI-U dataset behind the inflation calculator carries the full published history of the same series, monthly from 1913-01 to 2026-07 and annually from 1913 to 2025 — summarised on U.S. inflation.

Both come from one runtime source of truth. A second, independent retrieval of the same series through the BLS Public Data API is kept as a verification plane and is checked against the canonical dataset observation by observation on every build; it serves no value shown to a reader.

Published observations and calendar months differ because BLS did not publish 2025-10 (lapse in appropriations). A month the agency never published does not exist, so it is recorded as unpublished and never interpolated, carried forward or zero-filled. The year-over-year default window selects two months exactly twelve calendar months apart, so a gap between them does not affect the comparison.

IRS Publication 15-T — an official reference, not a dataset

GovCalculator cites Publication 15-T (2026), Federal Income Tax Withholding Methods, as the authoritative document for federal withholding.

Two things that are easy to conflate, stated separately:

• The publication does contain withholding tables — the percentage method and the wage bracket method, with their bracket rows.
• GovCalculator does not ingest or model them. No federal income tax withholding is calculated anywhere on this site, and no filing-status or pay-frequency table is served.

Purchasing-power figures are computed from BLS CPI-U and HUD Median Family Income only. IRS data is not an input to any calculation here. An earlier build shipped a simulated Publication 15-T structure with empty tables; it implied a capability that never existed and has been removed rather than left in place.