Charted: The US Authorized 0.94 Housing Units per Net New Household in 2024

Mortgage rates get the headlines. Permit tallies get the monthly press release. Household formation — the actual headcount of new dwelling demand — usually gets neither. That omission matters. In 2024, Census Building Permits Survey (BPS) authorized roughly 1.47 million privately owned housing units while the Housing Vacancy Survey path, as tabulated by Harvard’s Joint Center for Housing Studies, shows about 1.56 million net new households. The ratio is 0.94 authorized units per net household — still below one, and further below a replacement-aware benchmark near 1.2 once demolitions and other stock losses are acknowledged.

The dashboard above is built for that ratio, not for a single month’s seasonally adjusted annual rate. Toggle National ratio, Cumulative gap, Metro ladder, Metro scatter, Structure mix, and Region stack. Filter metros by Census region, sort by lowest ratio or largest unit gap, and stretch the national series back to 2015 when you want the pre-surge baseline. The question is simple: is the permit pipeline keeping up with the people who need roofs?

Why permits-per-household is the right scoreboard

Starts and completions lag permits. Existing-home sales recycle the same walls. Vacancy rates move for reasons that include second homes, short-term rentals, and measurement revisions. None of that replaces a stock-flow identity: over a multi-year window, net new households need net new housing units, plus replacement for units lost to demolition, disaster, and conversion out of residential use.

BPS counts authorized units in permit-issuing places — a forward signal, not a certificate of occupancy. HVS household totals are survey-based and revised when independent housing-unit controls update. ACS 1-year household estimates at the metro scale carry sampling error that can swamp a single year’s change in smaller markets. Treat the national ratio as a directional scoreboard and the metro ladder as a rank-order diagnostic, not as audited deed counts.

Still, directionally the scoreboard has been unkind. From 2019 through 2022, JCHS puts average annual household growth near 1.93 million. Permits rose too — peaking near 1.74 million authorized units in 2021 — but not enough to erase the gap. By 2023–2024, household growth cooled to 1.61 million and 1.56 million, while permits settled near 1.47–1.48 million. The ratio improved from the deep-surge lows near 0.7, yet never reclaimed a durable surplus.

The national path since the surge

Open the National ratio panel. Twin bars show authorized units and net household growth in thousands; the teal line is the ratio. Before the pandemic, the United States was already skating near or under 1.0 in several years. The surge years punched the denominator. 2020 and 2021 combined hot headship with strong immigration and remote-work reshuffling; permits recovered from the April 2020 trough but could not fully match the household pulse.

The Cumulative gap panel sums annual permit-minus-household residuals from the selected start year. On the 2019–2024 window used in the headline stack, the running gap sits near −920 thousand units — almost a million authorized units “short” of net household growth before any replacement add-on. That is not the same as a realtor.com-style cumulative shortage since 2012; it is a narrower statement about this six-year vintage of flow imbalance.

A replacement-adjusted reading is harsher. Housing analysts often budget 250–300 thousand units per year nationally for demolitions and other losses — roughly 0.2% of a ~145 million unit stock. Adding that to net household growth implies a permit (or completion) need closer to 1.8–1.9 million in a 1.56-million-household year. Against that ~1.2 benchmark, 2024’s 0.94 ratio is not a near miss; it is a structural shortfall that rents and prices eventually price in.

Which metros sit furthest below replacement?

National averages hide local extremes. The Metro ladder ranks a sixteen-metro sample by authorized units per net household over a recent multi-year window that pairs BPS CBSA annuals with ACS household change. San Francisco–Oakland–Berkeley prints near 0.42 — fewer than half a permitted unit per net new household. Boston, Los Angeles, New York, and San Diego cluster in the 0.49–0.59 band. These are not slow places in absolute permit volume; New York still authorizes tens of thousands of units. They are slow relative to the household flow they actually experience, once migration, headship, and international arrivals are netted.

Sun Belt metros tell a different story. Dallas, Houston, Austin, Charlotte, and Raleigh clear 1.0 on the sample ladder, with Raleigh near 1.18. That does not mean those markets feel “oversupplied” to renters — vacancy and absorption still matter — but it does mean the permit response is closer to measured household growth than the coastal constraint markets.

The Metro scatter makes the geometry explicit: household growth on the x-axis, permits on the y-axis, bubble size scaled to rental vacancy. Points below the 1:1 diagonal under-permit relative to net demand. Coastal metros hug the lower-right or mid-right with subdued y-values; Texas and Carolinas sit closer to or above the diagonal with larger absolute volumes.

Vacancy is not a perfect inverse of the ratio — Phoenix and Austin can clear higher vacancies while still permitting near household growth because prior boom years overshot and because single-family greenfield pipelines behave differently from urban infill. But the coastal cluster’s combination of sub-0.7 ratios and mid-single-digit rental vacancies is the textbook underbuilding signature.

Structure mix: apartments helped, then cooled

The Structure mix panel tracks the share of authorized units in 1-unit, 2–4, and 5+ buildings. Multifamily’s share climbed into the mid-30%s around 2022–2023 as apartment developers raced to meet pandemic-era demand and cheap pre-rate-hike financing. By 2024, the mix eased as multifamily starts rolled over — JCHS notes multifamily starts down sharply even while rents remained elevated. Single-family regained share, but not enough volume to close the national household gap.

That composition shift matters for the metro ladder. Coastal underbuilders often rely on 5+ product because land and entitlement favor vertical delivery; when multifamily capital stalls, their ratios worsen first. Sun Belt metros can substitute 1-unit subdivisions on the fringe. The national pie for 2024 still shows roughly three-fifths single-family and about one-third 5+ — a reminder that “housing supply” is not one industrial process.

The South still writes most of the permits

Open Region stack. The South’s share of authorized units sits near 55% of the 2024 national total in this dataset — a continuation of a long geographic reallocation of construction. Northeast and Midwest volumes are smaller in absolute terms and often serve slower household growth; West permits are larger than the Northeast but still insufficient for several Pacific metros’ household tallies.

Geography therefore splits into two policy problems that look alike in national media and diverge in data:

1. Constraint metros (coastal CA, Boston, parts of NY)ratios far below 1.0, limited lots, long entitlement clocks, and political resistance to density where jobs already are.
2. Throughput metros (Texas triangle, Carolinas, parts of Arizona)ratios near or above 1.0, but infrastructure, insurance, and water constraints can still bind, and cumulative national shortages do not vanish just because Dallas clears parity.

Sorting the ladder by largest unit gap rather than lowest ratio elevates the biggest absolute shortfalls — New York and Los Angeles can outrank a smaller metro with a worse ratio simply because tens of thousands of households are involved.

Caveats, measurement traps, and what the ratio cannot say

Several traps deserve equal billing with the headline.

Permits are not homes. Cancellation, multi-year build times, and manufactured housing outside the BPS frame all wedge authorization from occupancy. Completions series and Master Address File-based unit counts (where available) are better for stock accounting; permits remain the best high-frequency intent signal.

Household surveys disagree. ACS, HVS, and AHS can differ by hundreds of thousands of households in a given year. JCHS emphasizes trailing averages for good reason. This post leans on the HVS/JCHS national path for the headline and ACS for metro change; swapping survey vintage would move point estimates without erasing the coastal-vs-Sun-Belt rank order in most published comparisons.

Metro definitions drift. CBSA boundary updates and ACS margins of error mean a 0.42 vs 0.49 gap between San Francisco and Boston should be read as both badly under 1.0, not as a precise horse race.

Replacement is local. Fire, flood, and teardown intensity vary. A national 1.2 benchmark overstates need in slow-teardown metros and understates it after disasters.

Demand is not destiny. Some household growth is itself a function of rents and vacancies — people double up when units are scarce. Underbuilding can suppress measured household formation, which improves the apparent ratio while worsening welfare. The 2019–2022 surge showed the opposite: when incomes and preferences allowed, headship jumped and the ratio collapsed.

What to watch next

Three prints will tell whether 2024’s 0.94 was a floor or a plateau. First, BPS annual 2025 — whether authorized units hold near 1.45–1.50 million or slip as multifamily financing stays expensive. Second, HVS household change — JCHS already flags a further slowdown toward a ~1.26 million annual pace into early 2025 as immigration and demographics cool; a softer denominator can lift the ratio even without a construction boom. Third, metro ACS releases — watch whether coastal ratios stay pinned below 0.6 while Sun Belt vacancies normalize from their post-boom highs.

For desks that usually frame housing as a mortgage-credit story, the permits-per-household scoreboard is a useful discipline. Credit conditions explain who can buy the marginal unit. Authorization versus household formation explains whether the unit exists at all — and which metros remain furthest below replacement when the national average looks almost fine.