Harborwright Research · Methodology Brief

How we measure housing supply against demand

A county-level read on whether new construction is keeping pace with household formation across the United States. This brief covers the data sources, definitions, classification rules, and limitations behind the interactive report.

Coverage 3,144 US counties Cadence Annual, refreshed quarterly Vintage Latest available Census BPS & ACS releases

The question we are answering

For any given county, are builders permitting enough new housing to absorb the households being formed there each year? Persistent gaps compound — undersupplied counties see prices and rents outpace incomes; oversupplied counties accumulate vacancy. Both signals matter for capital allocation, public policy, and economic development.

Our framework reduces that question to a single ratio per county, computed on a rolling three-year basis to smooth permitting volatility and one-off ACS sampling noise.

Core metric: the supply ratio

supply_ratio = mean(permits, last 3 years) / mean(household_formation, last 3 years)

A ratio of 1.0 means permitting roughly matches household formation. Ratios well below 1.0 indicate supply is failing to keep up; ratios well above 1.0 indicate construction is running ahead of demand.

A three-year average is a deliberate trade-off. It dampens the noise of single-year permit swings (which often reflect approvals timing rather than fundamentals) at the cost of slower response to genuine trend breaks. We surface single-year values in the county detail view so analysts can see the underlying volatility.

Classification thresholds

The ratio is bucketed into six categories. Thresholds are calibrated to the empirical national distribution and align with the bands typically used in housing market reporting.

Category	Ratio band	Reading
Severely undersupplied	< 0.6	Permits at less than 60% of household formation. Persistent affordability pressure.
Undersupplied	0.6 – 0.9	Permits trail demand. Tight inventory, upward price pressure.
Balanced	0.9 – 1.3	Construction roughly tracking household growth.
Oversupplied	1.3 – 2.0	Permits running ahead of demand. Watch absorption and vacancy.
Severely oversupplied	> 2.0	Permits at more than double demand. Elevated risk of vacancy build-up.
Declining households	n/a	Net household formation is negative. Ratio undefined; reported separately.

Inputs and definitions

Supply: residential building permits

From the U.S. Census Bureau's Building Permits Survey (BPS), county-level annual file. We sum permitted units across all four product categories — single-family (1-unit), 2-unit, 3-to-4-unit, and 5-or-more-unit structures — to get total annual permits. Product mix is preserved separately for the county detail view.

BPS captures permitted units, not starts or completions. In most counties the gap between permits and starts is small and stable; in jurisdictions with long permit-to-start lags (e.g., parts of California and the Northeast), permits will overstate near-term delivery.

Demand: net household formation

Derived from the Census American Community Survey (ACS), table B11001 (total households). Annual household formation is the year-over-year change in household count:

household_formation_t = households_t − households_{t-1}

We then average across the most recent three years. For counties above the ACS 1-year publication threshold (population ≥ 65,000, roughly 830 of 3,144 counties), we use the ACS 1-year series — the appropriate vintage for annual cadence. Smaller counties fall back to ACS 5-year estimates, which cover every US county but report a five-year rolling average. Year-over-year formation deltas are correspondingly smoother for the small-county tail; this is a deliberate trade-off to maintain full national coverage rather than dropping rural counties from the panel.

Geography

FIPS-coded counties and county equivalents (parishes in Louisiana, boroughs and census areas in Alaska, independent cities in Virginia and elsewhere). State and national rollups are simple sums of constituent county values, computed from the same underlying inputs to ensure consistency across geographic scales.

Product mix

Within the supply total, we report the share of permits going to single-family vs. small multifamily (2–4 units) vs. larger multifamily (5+ units). Product mix is informative on its own — a county whose permits are 90% single-family is structurally constrained from delivering rental supply quickly, even if the headline ratio looks balanced.

Refresh cadence

The underlying API caches results for 30 days and rebuilds on demand after each Census release (BPS annual file in spring; ACS 1-year in late summer). The interactive report fetches live data on each load, so users always see the most recent vintage without any client-side action.

Known limitations

Permits ≠ deliveries. A permitted unit may be built late, scaled back, or never built at all. Drop-off rates vary by region and cycle. The ratio is a leading indicator, not a measure of completed housing.
ACS sampling error. Year-over-year changes in household count for small counties carry meaningful uncertainty. We mitigate with the three-year smoothing window but cannot eliminate it.
Cross-border commuting. County boundaries do not align with housing markets. A bedroom community may look severely undersupplied in isolation when the metro it feeds is balanced. Read counties in their metro context.
Demand from in-migration is partially captured. ACS counts households at the point in time of the survey, so household formation reflects both natural family formation and net migration. Counties with rapid in-migration will show high formation; counties losing population will show low or negative formation.
Unpermitted construction. BPS misses informal, accessory, and unpermitted units. The undercount is small in most jurisdictions but can be material in rural counties and in markets with significant ADU activity that is not separately permitted.

Methodology updates

We publish methodology changes alongside data refreshes. Material changes to thresholds, smoothing windows, or input sources are versioned and noted in the API metadata (meta.version) and at the top of this page.

Citation

Harborwright Research. National Housing Supply vs. Demand, county-level. Source data: U.S. Census Bureau, Building Permits Survey and American Community Survey. Accessed via the Harborwright public data API.