← gpu-pilot · json · source: tlccosts/RESULTS.md

True Lifestyle Cost Index v1 — NYC Pilot Results

11 counties × 3 income bases × 2 U scopes — HUD-vs-median gap quantified (+47% avg), LEAD county energy live, CNT r=0.90 holds. · updated 2026-06-12 04:42

TL;DR

  • At HUD AMI (the product default), TLC% (u_full) runs 23.2 (Middlesex) → 30.8 (South Central CT) and every pilot county clears the 48% line — because HUD FY2026 AMI runs +37% to +66% above the CBSA median household income v0 used. At each county's OWN median the picture flips: New Haven 43.3 / Bridgeport 43.1 press the line. The income basis is now the story — hence the radio-button picker.
  • U is real now: LEAD county energy $180/mo (Hudson, urban gas) to $508/mo (Western CT, oil heat) at the AMI bracket vs v0's flat $221–243 state placeholder; EIA per-fuel trending (NJ elec ×1.37 since 2022, oil ×0.85). u_full adds $267–282/mo (water/sewer $69 + broadband $85–100 + cell $113 + trash $0-toggle).
  • HUD gap table (decision #2 confirmed): Nassau/Suffolk/Westchester HUD MFI $164.3K vs $99.2K CBSA median (+66%); Jersey City HMFA's published 80% limit ($117,900) EXCEEDS its own MFI ($110,100) — HUD high-housing-cost adjustment, verified not a bug.
  • Validation holds: comparable construct vs CNT ht_ami r=0.90 / ρ=0.87 (v0: 0.90/0.89) — the workers-per-HH reweight (C up 17–38%) did not degrade the benchmark. Tiers shipped: metro quintiles + CNT national percentile context (Hudson sits at the 1.8th percentile of CNT's national distribution; Nassau 82.9th).
  • All five open decisions RESOLVED 2026-06-12 (LEAD+u_full, HUD income picker, spelled-out name, 48%+tiers, household basis); v2 queue: remote-share discount, LEHD destinations, metro water rates, tract pages (LEAD tract energy already staged).

True Lifestyle Cost Index — NYC Pilot Results (v1)

True Lifestyle Cost Index — Housing + Utilities + Commute. Computed: v1 2026-06-12 · compute_tlccosts_v1.py · ACS 2020–2024 · DOE LEAD 2022 EIA-trended · FCC URS 2026 · BLS CEX 2024 · HUD FY2026 income limits · C from the 30-town GPU commute matrix × ACS B08202 workers/HH · benchmark = CNT H+T® 2022 (hta_index, © CNT, comparison only) Output: tlccosts/output/tlccosts_v1_nyc.csv (11 counties × 3 income bases × 2 U scopes) · v0 preserved at output/tlccosts_v0_nyc.csv

Headline (v1)

  • At the housing-industry income standard (HUD AMI, the product default), the True Lifestyle Cost of a Manhattan commute runs 23–31% of income across the 11 pilot counties (u_full) — every county clears the 48% affordability line, because HUD FY2026 area median family income runs +37% to +66% above the CBSA median household income v0 divided by. Affordability claims are extremely sensitive to whose "median" you use — which is why the income basis is now a user choice (see decision #2).
  • At each county's own median household income, the picture flips: New Haven 43.3% and Bridgeport 43.1% (u_full) press against the 48% line — metro-priced commutes on city-median incomes.
  • U is now real (county, income-bracketed): LEAD county energy spans $180/mo (Hudson) to $508/mo (Western CT) at the AMI bracket — the v0 state placeholder was flat $221–243. EIA per-fuel trending matters: NJ residential electricity is ×1.37 vs 2022, while heating oil is ×0.85.
  • u_full (the product number) adds $267–282/mo over u_energy: water/ sewer $68.83 + broadband $85–100 + cell $113.25 + trash $0 (NJ norm).
  • Validation vs CNT H+T® holds: r = 0.90 / ρ = 0.87 on the comparable construct (gross housing + household commute over CBSA median income) — unchanged from v0's r = 0.90 despite the workers-per-household reweight.

v1 county table (HUD-AMI basis, sorted by TLC% u_full; monthly $)

county st HUD MFI H shelter u_energy u_full C hh/mo TLC% med TLC% AMI TLC% 80 TLC% AMI (u_energy) metro q CNT ht_ami CNT nat %ile
South Central CT (New Haven) CT 123,200 1,409 266 533 1,217 43.3 30.8 38.0 28.2 5 46 22.6
Hudson NJ 110,100 1,876 180 458 369 35.1 29.5 27.5 26.4 5 37 1.8
Suffolk NY 164,300 2,018 421 703 1,238 36.5 28.9 37.0 26.9 4 55 75.8
Nassau NY 164,300 2,357 412 694 898 32.6 28.8 36.9 26.8 4 57 82.9
Western CT (Stamford) CT 156,800 2,037 508 775 905 34.4 28.5 36.6 26.4 3 46 22.6
Morris NJ 138,400 2,044 237 514 718 29.1 28.4 36.9 26.0 3 53 65.3
Union NJ 138,400 1,953 205 482 743 37.2 27.6 35.7 25.2 2 45 17.8
Mercer NJ 139,800 1,552 236 513 1,013 36.9 26.4 34.5 24.0 2 43 10.0
Westchester NY 164,200 1,989 407 689 670 34.8 24.5 32.0 22.4 1 48 33.4
Greater Bridgeport CT 156,800 1,568 296 563 1,063 43.1 24.4 34.1 22.4 1 39 3.1
Middlesex NJ 154,800 1,795 206 483 715 32.4 23.2 31.3 21.1 1 46 22.6

TLC% med/AMI/80 columns are the u_full variant at the three income bases; the u_energy variant at AMI is shown for the CNT-comparable scope. metro q = quintile among the 11 pilot counties at the product default (HUD AMI × u_full), 1 = most affordable. CNT nat %ile = where the county's CNT ht_ami sits in CNT's national county distribution (CNT-vs-CNT, no construct mixing). H+T® Index data © Center for Neighborhood Technology (htaindex.cnt.org), used for comparison; the True Lifestyle Cost Index is independent and not affiliated with or endorsed by CNT.

u_full component build (the decision-#1 expansion)

component $/mo source (vintage) geography used quality tier
energy 180–508 by county/bracket DOE LEAD 2022, EIA-trended per fuel (elec ×1.18–1.37, gas ×0.92–1.14, oil ×0.83–0.87) county (tract staged) A
water/sewer 68.83 (hh-size 44–98 staged) BLS CEX 2024 Table 1400 "water and other public services" national B
broadband NJ 95 / NY 100 / CT 85 FCC Urban Rate Survey 2026 (DA 25-1088), median ≥100/20 Mbps standalone state (natl 95.00 wtd, tier A) B state
cell phone 113.25 (hh-size 61–179 staged) BLS CEX 2024 "cellular phone service" national A
trash 0.00 (toggle) NJ norm: municipal collection inside property tax; private-hauler townships ~$25–45/mo state default C

Tier A = federal source at the target construct/geography · B = federal source applied at coarser geography (local water rates vary 2–3×; AWWA paywalled, Circle of Blue 30-city survey ended 2019) · C = tariff/municipal anchor. NJ cross-check anchor: NJ American Water (BPU No. 8 tariff): 5/8" meter $19.85/mo + $7.7752/kgal → $50.95/mo at 4 kgal, water only — consistent with the CEX water+sewer national mean. All rows live in the utility_baselines table (82,833 rows: 81,672 tract + 1,113 county + 28 state energy; broadband/cell/water-sewer/anchors).

HUD AMI vs the medians (decision #2 — Krishna's read confirmed)

HUD FY2026 income limits (huduser.gov, file dated 2026-04-16) vs ACS 2020–2024 medians. The industry "AMI" is median family income for a 4-person benchmark, trended forward — it reads far above the median household (households include singles and seniors):

county HUD area (FY26) HUD MFI (100% AMI) vs CBSA median HH (+%) vs county median HH (+%) HUD 80% (4p)
Nassau Nassau-Suffolk, NY HMFA 164,300 +65.7% +12.4% 128,100
Suffolk Nassau-Suffolk, NY HMFA 164,300 +65.7% +25.7% 128,100
Westchester Westchester County, NY Statutory Exception Area 164,200 +65.6% +38.0% 125,850
Middlesex Middlesex-Somerset-Hunterdon, NJ HMFA 154,800 +56.1% +38.8% 114,600
Mercer Trenton-Princeton, NJ MSA 139,800 +41.0% +38.9% 106,800
South Central CT New Haven, CT MSA 123,200 +39.7% +39.7% 99,600
Morris Newark, NJ HMFA 138,400 +39.6% +0.8% 106,800
Union Newark, NJ HMFA 138,400 +39.6% +34.1% 106,800
Greater Bridgeport Bridgeport town, CT Exception Area 156,800 +36.9% +74.9% 112,500
Western CT Stamford town, CT Exception Area 156,800 +36.9% +22.3% 117,100
Hudson Jersey City, NJ HMFA 110,100 +11.0% +19.9% 117,900

CT rows anchor on the planning region's principal commute town (HUD FY26 publishes CT as town-level rows under planning-region FIPS; Bridgeport and Stamford towns are HUD "exception areas" within the Bridgeport-Stamford region).

  • Average gap: HUD AMI ≈ +47% over the CBSA median household income (range +11% to +66%) and +32% over county medians. Krishna's expectation ("our CBSA-median proxy reads low vs industry AMI") is confirmed and quantified — hence the radio-button income picker with HUD AMI default.
  • Jersey City HMFA anomaly (verified, not a bug): the published 80% limit ($117,900) exceeds the area's own MFI ($110,100) — HUD's high-housing-cost adjustment. Where that happens, TLC% at "HUD 80%" is lower than at "HUD AMI". Published limits ≠ 0.8 × MFI.

v0 → v1: what changed and why

change v0 v1 effect
U source state EIA placeholder ($221–243 flat) LEAD 2022 county × income bracket, EIA-trended per fuel county energy now $180–508; Hudson −19% vs placeholder, Nassau +87% — urban-gas vs oil-heat reality
U scope energy only u_energy + u_full (water/sewer, broadband, cell, trash) u_full adds $267–282/mo; the product number
income basis CBSA median HH ("ami" proxy) median / HUD AMI / HUD 80% / custom (default HUD AMI) headline TLC% drops ~9–13 pts at AMI vs v0 — denominator, not costs
C basis 1 commuter/HH flat × B08202 workers/HH (1.17–1.38 by county) C up 17–38%; "the household's index" (CNT's Regional Typical Household does the same — Methods Nov 2022, p.25)
H decomposition gross − U(state) gross − LEAD ALL-bracket energy − CEX water/sewer (basis-independent shelter) H+embedded-utilities still ACS-consistent; shelter no longer moves with the income basis
threshold 48% line 48% line and metro quintiles + CNT national percentile context tiers survive denominator debates

v1 TLC% values are not directly comparable to v0's (income basis and U scope both changed); the benchmark-construct columns (tlc_pct_benchmark_cbsa*) carry the v0-comparable series.

CNT H+T® benchmark recheck (© CNT, comparison only)

  • Comparable construct (12×(gross housing + household commute)/CBSA median HH income — CNT's own income basis is "regional area median income", Methods Nov 2022) vs ht_ami: r = 0.90, ρ = 0.87 (n=11).
  • Same construct with v0's 1-commuter C: r = 0.90 — the household reweight did not degrade the benchmark (v0 reported r = 0.90 / ρ = 0.89).
  • The construct gap stays interpretable: CNT T = all household travel incl. auto ownership; our C = the priced commute, chosen mode.

v1 caveats (all flagged in the CSV flags column)

  1. C = representative commute towns, chosen mode, × county workers/HH, 22 days — every worker is treated as a CBD commuter; no remote-share discount yet (B08303/B08301 staged), no LEHD destination mix until the national hex run.
  2. LEAD bracket boundary convention: a household earning the basis income is the top of its LEAD bracket (AMI → 80–100%, HUD-80 → 60–80%; median → ALL). Bracket-to-bracket energy spread is ±15%, second order.
  3. water/sewer + cell are national CEX means (hh-size cohorts staged); NE-region CEX cut and metro water-rate scraping are v2 items. Broadband state medians are thin-sample (URS weights are national-urban).
  4. LEAD is 2022 (ACS 2018–2022) trended by state EIA price indices — it captures 2022 consumption patterns at current prices, not post-2022 electrification/efficiency shifts.
  5. CT "counties" are planning regions on every side of the join now (ACS 2024, CNT 2022, LEAD 2022, HUD FY26 towns) — verified direct FIPS joins.
  6. CNT comparison crosses vintages (2022 inputs vs our 2024/2026 stack).

Decisions — RESOLVED 2026-06-12 (Krishna)

  1. ~~U source~~ RESOLVED: DOE LEAD blessed for v1 energy-U, EIA-price-trended; U expanded beyond energy with two displayed figures — u_energy (LEAD-comparable, used in CNT benchmarks) and u_full (+ water/sewer, broadband, cell, trash where billed — the product number). Implemented above; staged in utility_baselines.
  2. ~~Income variants~~ RESOLVED: income basis is a user choice (radio: median HH / HUD AMI 100% / HUD 80% / enter-your-income), default HUD AMI. HUD FY2026 limits staged in hud_income_limits; gap vs medians quantified above (avg +47% vs CBSA median HH).
  3. ~~Naming~~ RESOLVED: fully spelled out — "True Lifestyle Cost Index" (13 years of public use by Krishna/TLCengine). Subtitle pattern: True Lifestyle Cost Index — Housing + Utilities + Commute. Repo slug stays tlccosts. Applied across docs + page spec.
  4. ~~Threshold~~ RESOLVED: BOTH — the 48% affordability line (30 H + 15 C + 3 U) and percentile tiers: metro quintiles shipped in v1; national tiers on our construct await the national run (CNT national percentile shown as context meanwhile).
  5. ~~Commuters per household~~ RESOLVED: household basis ("Household Index" framing), C weighted by ACS B08202 workers/HH. Verified CNT precedent: the H+T "Regional Typical Household" fixes income, household size and commuters per household at regional values (H+T Index Methods, CNT, Nov 2022, p.25 — data/cnt/method-2022.pdf). A public, press-friendly methodology now ships separately: PUBLIC_METHODOLOGY.md (framework + sources + validation, no coefficients).
  6. ~~Census API key~~ RESOLVED 2026-06-12 (earlier): key at ~/.census_api_key (mode 600), auto-read; verified against ACS 2024.

Open items (v2)

  • Remote-work discount on C (B08303/B08301) + LEHD destination mix.
  • NE-region CEX cut for water/cell; metro-level water rates (NJAW anchor pattern → more utilities); trash toggle UI.
  • Tract/blkgrp pages: LEAD tract energy is already staged (81,672 rows).
  • National tiers on our construct (CONTINUE.md Phase 2 hex maps).
  • "Enter your income" custom basis in the page spec (BRAND.md §3).

Comments