# NYC electric vehicle denominator methods

## Scope and source

HousingStabilityIQ uses 2025 ACS one-year table B25148 for New York city, New York, GEO_ID 1600000US3651000. The year 2025 is the survey period; the release date is October 29, 2026. The geography is the five-borough city, not the New York metro or state. Occupied housing units are households; these estimated counts are not survey sample sizes.

The official DAT and JSON are retained in the batch at ev-housing/raw, with the parent's acquisition manifest ev-housing/source-manifest.json. The local reproduction script reads that batch-relative location, plus housing-new/public/evidence/raw/Geos20251YR.txt for geographic identity. Keep both branches when extracting the public-clean package. The script does not use network access or credentials.

## Six measures

DAT columns are E005/E002 for owner households with an EV among all owner households, E010/E007 for renters among all renter households, E005/E004 for owners among owner households with a vehicle available, E010/E009 for renters among renter households with a vehicle available, E003/E002 for owners with no vehicle available, and E008/E007 for renters with no vehicle available. Multiply each by 100. The corresponding API variable names place the three-digit line before E/M (B25148_005E versus DAT B25148_E005).

The vehicle-availability denominator is not synonymous with registered vehicles, licensed drivers or personally owned cars. Electric vehicle includes all-electric and plug-in hybrid vehicles; conventional non-plug-in hybrids are excluded. Definitions and question wording are retained in the official 2025 subject definitions, printed page 18, and questionnaire. The EV question is asked of occupied households keeping at least one vehicle at home. It is new to 2025; no earlier ACS EV trend is asserted.

## MOEs and inference

All source MOEs are 90%. For p=N/D use M_p=100×sqrt(M_N²−p²M_D²)/D; use the Census plus-sign ratio fallback if the radical is negative. No selected metric requires fallback. Difference MOE is sqrt(M_A²+M_B²), an approximation without covariance. All selected cells are valid finite nonnegative values, and five table partitions reconcile. Missing, negative jam values or annotations are never zeroed.

Owner–renter difference among all households is 3.4582239905pp, approximate MOE 0.4522832286pp. Among vehicle-available households it is 0.9961015442pp, approximate MOE 1.0044741181pp, giving 90% interval [−0.0083725739,2.0005756623]pp. The latter is not statistically distinguishable from zero under this approximation. The test uses full precision, not rounded 1.0 versus 1.0. It is not an equivalence test and does not establish equal underlying rates.

Changing denominators produces different descriptive gaps. We do not test the difference between those two gaps, because the underlying observations overlap. No causal claim, income adjustment, national rank or multiple-comparison-controlled discovery is made. The NYC cut was an exploratory, locally motivated selection, not a preregistered confirmatory result. No claim is made that tenure has no effect or that adjusting for vehicle availability resolves a broader equity question.

## Allocation and nonsampling limitations

B992527's universe is occupied housing units with at least one vehicle. In NYC, 139,846 of 1,402,085 were allocated for EV information, 9.9741456474%. The denominator exactly equals the vehicle-available owner and renter totals 748,495+653,590. The table contains no published MOEs; none are invented. This citywide quality indicator does not reveal owner/renter-specific allocation. Allocation is not a quantified estimate of bias. Sampling margins do not cover all nonsampling error.

The data do not measure charger access, home charging, building wiring, off-street parking, battery size, driver profession, vehicle miles, income, costs or reasons for having no vehicle. Housing tenure is not vehicle tenure. The city's 2026 curbside program is contextual, not joined at household level or treated as an explanation of 2025 outcomes. A low EV share is not itself a housing hardship or transport deprivation measure.

## Reproduction and chart

Run `python housing-new/public/evidence/reproduce-nyc-ev.py` from an extracted batch, retaining the original folder structure. Add `--charts` with matplotlib installed for SVG/PNG output. The script regenerates nyc-ev-estimates.csv/json, nyc-ev-chart-data.csv, nyc-ev-cell-audit.csv, nyc-ev-allocation.csv and nyc-ev-validation.json. The four-bar chart uses each bar's named household denominator and approximate 90% MOEs. PNG visually inspected; labels, source, units, denominator note and confidence limitation are legible and unclipped.
