2026 State of Home Construction Costs

1. Recovery Divergence: No State Is Back at Its Pre-Crash Peak

Using each state's own highest single-family permit year through 2007 (almost always 2005, the last full year before the housing bust) as its "pre-crash peak," we compared it to 2025 volume. Nationally, the U.S. built 911,903 single-family homes in 2025 versus 1,681,986 in 2005 — 54.2% of peak. As of 2025, not one of the 51 states (50 states + DC) is back at its own pre-crash volume. The closest the country came was 2021, at 1,115,360 permits — 66.3% of the 2005 peak — and volume has receded since.

Note on small-base states: Washington DC's pre-2007 "peak" (576 units in 2007) is itself a single-year spike within a typical range of roughly 100–350 units a year — at this scale, year-to-year swings are large relative to the base, so DC's 31.4% recovery figure should be read with that volatility in mind, not as a signal comparable to higher-volume states.

2. Lumber Normalized — Other Materials Didn't

Lumber prices were the public face of pandemic-era construction inflation. BLS Producer Price Index data through mid-2026 shows that story has flipped: lumber and wood products are up just 10.8% since 2020, and softwood lumber specifically is 2% below its 2020 level. The materials that actually kept climbing are ones that get far less attention.

Asphalt roofing and paving materials are up 144.2%, steel mill products 82.8%, and gypsum products (drywall) 49.2% — all far outpacing lumber. A cost narrative still anchored to lumber prices is looking at 2021, not 2026.

3. Wages Explain Some, Not Most, of the State Cost Spread

We correlated BLS OEWS state median construction-trade wages against this site's modeled cost per square foot across all 51 states. The relationship is real — r=0.80 (R²≈63.3%) — but wages alone leave more than a third of the state-to-state variation unexplained. Some states cost far more, or far less, than their wage levels alone would predict.

Alaska, District of Columbia, and Hawaii likely reflect shipping and logistics costs (Alaska, Hawaii) and land constraints (DC) layered on top of wages. Illinois and Minnesota have high measured trade wages but modeled costs that undershoot what wages alone would predict — a reminder that materials, land, and local competition move the number as much as labor does.

4. Where Land, Not Construction, Is the Real Barrier

FHFA's experimental land-price dataset splits a home's total value into land and structure. Nationally, land averages 28.3% of property value across the 51 states — but in a handful of states it's the majority. Land share also correlates with modeled construction cost even more tightly than wages do (r=0.85 vs. r=0.80 for wages), suggesting geography and land scarcity may move state cost differences more than labor costs alone.

FHFA land-price data is base-year-standardized to 2015 (pooled sample 2012–2022) — treat it as a dated structural reference on where land dominates, not a current-dollar figure. Source: FHFA Working Paper 19-01.

5. The March Filing Surge

Averaged across 2021–2025, single-family permit filings peak in March (9.59% of the year's total) and bottom out in December (6.75%) — 42.1% more permits filed at the peak than the trough. For builders and buyers, that means contractor calendars and material lead times are tightest heading into spring, and a permit filed in the fall trough faces less competition for inspectors and crews.

6. The Federal Benchmark vs. This Site's Tiers

The Census Bureau's Survey of Construction reports a median contract price of $171.57/sq ft (average $198.33/sq ft) for new contractor-built single-family houses started in 2025 — a figure that excludes land. That falls between this site's own basic ($139/sq ft) and mid-range ($188/sq ft) modeled tiers, and well under the premium tier ($256/sq ft).

We are not changing our modeled rates based on this single federal comparison — the Census figure covers a narrower universe (owner-contracted builds, excluding land) than our tiers — but a government figure placed openly next to our own is the honest way to present it. See our full breakdown of Census new-home data.

7. The Energy Code Patchwork

DOE's Building Energy Codes Program tracks the residential energy code edition each state has adopted. 11 states have no single statewide code at all ("home rule" — local jurisdictions set their own), and of those, 9 have no DOE-computed population-weighted default either. Among states that do have a statewide code, the most widely adopted edition is 2021 IECC (12 states) — a full code cycle behind the current 2024 IECC, which only 5 states have adopted.

What We Tested and Dropped

Three questions we set out to answer are not in the findings above, because the data did not support them. We are listing them anyway — a study that only reports what worked is not a study.