How old are America’s homes?

Key findings

• About 47.7% of primary occupied U.S. homes were built before 1980 in the preliminary 2024 RECS. The Room Study calculated the share from survey weights. Source

  Link to finding

• The pre-1980 share was 66.9% in the Northeast and 35.5% in the South: a 31.3-percentage-point difference calculated before rounding. Source

  Link to finding

• Among owner-occupied homes, 48.0% were built before 1980; our approximate 95% jackknife interval is 47.3%–48.7%. Source

  Link to finding

• Among rented homes, including homes occupied without rent, 47.1% were built before 1980; our approximate 95% jackknife interval is 45.6%–48.5%. Source

  Link to finding

Which homes are counted?

This is the age of occupied homes, including apartments and mobile homes as well as houses. It excludes vacant properties, second homes and seasonal homes. “Before 1980” groups four published construction periods; it is a useful dividing line in the available data, not a definition of good or bad housing.

A home’s construction period tells us when it began life. It cannot tell us when the wiring was replaced, whether a bathroom was adapted, or how well the building has been maintained. We do not turn age into a repair diagnosis.

Where are older homes concentrated?

The Northeast has the largest pre-1980 share of the four regions. In the South, roughly one home in three belongs to that group. Both owners and renters follow this broad regional pattern.

The national owner and renter estimates are close. Their uncertainty intervals overlap, and this page makes no claim that tenure itself explains housing age. The table supplies the regional detail instead of reducing the country to an “oldest homes” ranking.

The Room Study analysis of EIA 2024 RECS public file v1 (May 2026). All-tenure margins are benchmark-calibrated. Method and limits

Download chart (SVG)

Why are some uncertainty figures left blank?

EIA aligns construction-period totals with Census benchmarks. For all tenures together, the replicate estimates barely move because those margins are fixed by the weighting process. Publishing that tiny variation as a near-perfect confidence interval would be misleading. The table labels those rows “benchmark-calibrated” instead.

Owner/renter intersections can vary across the replicate samples, so we show approximate 95% intervals for them. These describe sampling uncertainty conditional on the calibration. They do not include every source of error in the underlying Census benchmarks, respondent answers or imputation.

What does this mean for a room plan?

For our design studies, the useful question is what the existing building asks us to work with: the position of windows, the depth of a room, the route to a bathroom, or a door that takes up the only clear wall. An older construction date is a reason to investigate those particulars.

The companion repair-needs page measures reported problems using a separate survey and cost model. Read it alongside this analysis for context, but do not combine the two datasets as though they describe the same households.

Methodology & sources

The Room Study analyzed all 16,676 records in recs2024_public_v1.csv. We sum NWEIGHT where YEARMADERANGE is 1, 2, 3 or 4, divide by all NWEIGHT in the relevant group, and multiply by 100. REGIONC supplies the four Census regions. KOWNRENT = 1 means owned; 2 and 3 are combined as rented, including occupied without payment. No records are dropped for these variables.

For each region–tenure estimate, we recalculate the numerator and denominator with NWEIGHT1 through NWEIGHT60. Variance = (59/60) × the sum of squared deviations of those 60 estimates from the full-weight estimate. Intervals use estimate ± 2 standard errors, following the approximate 95% convention in EIA’s guide with 59 degrees of freedom. All-tenure construction margins are calibrated: their near-zero replicate errors are suppressed, not interpreted as certainty.

A separate row-by-row JavaScript implementation independently reproduced the Python weighted sums and jackknife calculations from the original CSV. Weighted home totals and mean floor areas also reconcile to the official rounded tables. The downloadable aggregate CSV retains denominators, sample counts, standard errors and interval flags. The linked Python script reproduces the age analysis from the public source file.

All calculations use unrounded weights. Display percentages are rounded to one decimal, so subtracting two displayed shares may differ by 0.1 point from a difference calculated before rounding. Preliminary 2024 data were released in 2026; this is not a 2026 stock estimate. No median year built, state ranking, causal effect or time trend is inferred from the grouped years.

We review these resources when the underlying releases change. A review date does not make a figure expire: its measurement period stays visible until a checked revision replaces it.