The Real Estate Replacement Cost Gap: Where U.S. Housing Is Mispriced (v2)

Eight US metro areas meet a simple but demanding standard: their homes trade at a meaningful discount to what it would cost to build them new and their populations are still growing. These cities are clustered in Texas and select Midwest markets, and together they represent what we think is the most underappreciated opportunity in residential real estate today.

Key findings

■  54 of 100 MSAs trade below replacement cost — the average discount is 21%.

■  Oklahoma City (−35%), San Antonio (−29%), Indianapolis (−28%), Cincinnati (−26%), Houston (−25%) are the strongest combined signals: deep RC discount, growing population, and rapidly rising incomes.

■  The v2 model significantly reclassifies several markets previously obscured by coarse tier groupings — Nashville moves from a +22% premium to near breakeven; Seattle's premium falls from 55% to 22% as the model properly weights its labor costs.

■  Markets to avoid: San Jose (+91%), Honolulu (+85%), and Bridgeport-Stamford (+48%) carry the highest premiums over RC — and all three have flat or declining populations.

This post introduces the second version of our replacement cost model, rebuilt from the ground up with three independently-sourced factors: local labor costs indexed to MSA-level per capita income, state-level material costs from RS Means, and MSA-specific land estimates that replace the single national assumption that undermined our original model. The upgraded methodology surfaces sharper rankings and — in the case of Midwest and Texas markets — a more compelling fundamental case than the prior model was capturing.

Across the top 100 MSAs, 54 markets trade below replacement cost by an average of 21%. The most attractive are not simply the cheapest — Pittsburgh, at a 46% discount, reflects decades of population outflow as much as value. The more interesting set are markets like Indianapolis (−28%), San Antonio (−29%), Columbus (−21%), Kansas City (−22%), Houston (−25%), and Jacksonville (−14%): cities where homes are materially cheaper than building new, wages have risen sharply, and people are still arriving. That combination — structural undervaluation, income momentum, and positive migration — is rare. Right now it is concentrated in two regions of the country.

What Changed in Version 2

The v1 model applied a single blended city multiplier derived from RS Means regional tiers, then added a flat small-lot land cost assumption. Three things were fundamentally broken:

•       Labor was invisible. Construction wages are the most market-specific input in any building project. A general contractor in San Jose pays its framing crew and electricians at rates that are 40–60% higher than those in Memphis or Oklahoma City. The v1 model captured none of this granularity — it folded labor into a blunt regional average that told you almost nothing useful at the individual metro level.

•       Land was a national assumption. We used a single small-lot land cost regardless of market. In practice, land costs per square foot of building area range from $16 in Jackson, Mississippi to $180 in San Jose — a 10× spread that dwarfs the variation in construction costs (roughly 1.5×). A model that ignores this systematically misprices both ends of the market.

•       Tier granularity was too coarse. Five tiers across 100 MSAs means each tier contained ~20 cities treated identically. Cities at the top and bottom of a tier had materially different cost structures but landed at exactly the same replacement cost per square foot.The new model replaces all of this with three independently-sourced factors, each calibrated to MSA-level data:

Replacement Cost / sf  =

    NAHB Base ($162)  ×  [45% × Material Multiplier  +  40% × Labor Multiplier  +  15% overhead]

    +  MSA Land Cost / sf

Each factor has a defined source, weight, and logic:

Factor	Weight	Source	Calculation	Why it matters
Labor	40%	ACS 2024 per capita income	1 + 0.40 × (MSA PCI / $45k − 1)	Captures true local wage cost of skilled trades; moves continuously with local earnings
Materials	45%	RS Means state-level index	State multiplier (0.93–1.40×)	Regional variation in lumber, concrete, and steel driven by supply-chain geography
Overhead	15%	Industry standard	Applied at base rate	Contractor profit; treated as proportional to location
Land	Direct add	MSA-specific estimates	$16–$180 / sf of building area	Replaces single national average with observed local land values — the largest single driver of divergence

What the Model Reveals: A Summary of Key Findings

53 of 100 MSAs at a discount to replacement cost

−46% Deepest discount — Pittsburgh, PA — buying at less than half of replacement cost

+91% Highest premium — San Jose, CA — nearly double the cost to build new

10× Range in land cost — $16/sf (Jackson, MS) to $180/sf (San Jose, CA)

Across the top 100 MSAs, 53 markets trade at a discount to replacement cost and 45 trade at a premium. The average discount among discounted markets is 20%; the average premium among premium markets is 24%. But the distribution is highly skewed — a handful of coastal markets carry extreme premiums that pull the premium average up, while a deep cluster of Midwest and Southern cities sits well below replacement cost.

Where the Opportunities Are

A replacement cost discount alone is a necessary but not sufficient condition for an attractive market. A city can trade below replacement cost because it has been in structural decline for decades — Cleveland and Detroit are examples. The more interesting set is markets that combine a meaningful discount to replacement cost with positive population dynamics and rising incomes, suggesting that demand is present and the discount reflects a pricing inefficiency rather than fundamental weakness.

The following 8 markets meet all three criteria: trading at a discount to v2 replacement cost, and rapidly growing in population in 2024–25, and showing 25%+ income growth since 2020.

MSA	State	Exist. $/sf	v2 RC	Discount to RC	2024–25 pop. growth	5yr ann. pop. growth
Oklahoma City	OK	$114	$176	−35%	+0.69%	+0.99%
Indianapolis	IN	$136	$189	−28%	+0.80%	+1.50%
San Antonio	TX	$131	$185	−29%	+0.60%	+1.39%
Houston	TX	$144	$191	−25%	+0.80%	+1.61%
Columbus	OH	$152	$193	−21%	+0.87%	+1.52%
Kansas City	MO	$149	$192	−22%	+0.73%	+0.96%
Cincinnati	OH	$142	$191	−26%	+0.52%	+0.76%
Dallas	TX	$171	$203	−16%	+0.69%	+1.56%

The Texas cluster

Texas warrants specific attention. Houston, Dallas, and San Antonio all appear in the attractive markets table, and all four share a similar profile: below-average material costs (0.98× state multiplier), moderate land costs ($18–$40/sf), and strong population inflows. The state's business-friendly regulatory environment and relative housing supply responsiveness mean that discount-to-RC compression has historically been slower there than in supply-constrained markets — but that same supply responsiveness also means the discount provides genuine protection against downside.

The Midwest cluster

The Midwest tells a structurally distinct story from Texas. Cities like Cleveland, Cincinnati, Columbus, Indianapolis, Kansas City, and Chicago all trade at 22–39% discounts to replacement cost — some of the deepest in the entire dataset. The population picture is more mixed than Texas — Cleveland, Toledo, Dayton, and Akron are still losing residents, while Columbus, Indianapolis, Cincinnati, and Kansas City are growing steadily. That split matters: the growing Midwest cities offer the discount of a distressed market without the demand weakness of one, and replacement cost floors that are rising as wages catch up.

Limitations of the Model

No replacement cost model is complete. The v2 model improves significantly on v1, but several dimensions remain outside its scope:

•       Permitting and soft costs. In some markets — particularly coastal California and New York — permitting timelines, impact fees, and related soft costs can add 15–25% on top of hard construction costs. Our model does not capture these, which means it likely understates true replacement cost in heavily regulated markets. This makes the premiums in those markets look smaller than they actually are.

•       Product type. The model reflects single-family construction economics. In dense urban cores where new supply is predominantly multi-family, cost structures differ: higher per-unit land costs, shared infrastructure, elevator requirements, structured parking. The model is not calibrated for vertical product.

•       Material price timing. RS Means indices reflect conditions at the data vintage. Lumber, steel, and copper prices move significantly year to year. The model should be treated as current as of Q1 2026 and recalibrated at least annually.

•       Vintage and condition. Existing homes are not new homes. A 1960s ranch in Pittsburgh is not worth its replacement cost even at $177/sf — condition, layout, energy efficiency, and functional obsolescence all affect realized value. The replacement cost figure is best understood as a structural floor, not a precise appraisal.

The Three Factors in Detail

Factor 1: Labor, indexed to local per capita income

Labor represents 40% of construction cost and is the most market-specific component. We use ACS 2024 per capita income as a proxy for local skilled trades wages, benchmarked against a national median of $45,000. The formula is: Labor Multiplier = 1 + 0.40 × (MSA PCI / $45,000 − 1).

A market at the national PCI median gets a labor multiplier of exactly 1.0. San Jose, with a PCI of $93k, gets approximately 1.38. McAllen, Texas at $23k gets 0.86. Rather than snapping between five tier values, the labor component now varies continuously across all 100 markets.

Why per capita income rather than construction-specific wage surveys? MSA-level construction wage data is inconsistently reported and often conflates residential and commercial work. Per capita income correlates strongly with skilled trades wages in practice and is drawn from a consistent, Census-quality source updated annually.

Factor 2: Materials, via RS Means state-level index

Materials account for 45% of construction cost. Unlike labor, material costs follow regional supply-chain patterns — proximity to manufacturing, transportation infrastructure, and local building code requirements — that are better captured at the state level than the metro level.

We apply a state-level material cost index derived from RS Means regional data. The index ranges from 0.93× in Arkansas and Mississippi to 1.40× in Hawaii. California carries 1.25×; New York 1.20×; Texas 0.98×. Applied uniformly across all MSAs within a state, this is the appropriate granularity for how material supply chains actually operate.

Factor 3: Land, MSA-specific estimates

Land is where v2 diverges most dramatically from v1. Rather than a single national assumption, we estimate small-lot land cost per square foot of building area for each of the 100 MSAs individually. The methodology uses observed median land values for a typical 0.15-acre starter home lot, divided by a 1,500 sf building footprint.

The resulting spread is the single most important change in the model. San Jose sits at $180/sf, Honolulu at $160/sf, San Francisco at $165/sf. At the other end, Jackson Mississippi is $16/sf, McAllen $18/sf, Birmingham and Memphis in the low $20s. This spread — 10× between cheapest and most expensive — is far larger than any other input in the model.

Crucially, land cost is added as a direct, transparent component rather than embedded in a multiplier. This means it can be updated independently as better local data becomes available, without touching the construction cost calculation.

How the model changed from v1

The shift from v1 to v2 is not cosmetic. The table below shows how replacement costs and discount/premium classifications changed for a representative set of markets:

MSA	Exist. $/sf	Construction build-up	Land $/sf	v2 RC	v1 RC	v2 disc/ prem	v1 disc/ prem
Pittsburgh, PA	$104	$162 × 1.08 mat × 0.99 lab = $167	$25	$192	$209	−46%	−50%
Oklahoma City, OK	$114	$162 × 0.94 mat × 0.95 lab = $154	$22	$176	$168	−35%	−32%
Indianapolis, IN	$136	$162 × 1.00 mat × 0.98 lab = $161	$28	$189	$174	−28%	−22%
Houston, TX	$144	$162 × 0.98 mat × 0.98 lab = $159	$32	$191	$170	−25%	−15%
Nashville, TN	$213	$162 × 0.96 mat × 1.08 lab = $164	$50	$214	$174	−1%	+22%
Denver, CO	$266	$162 × 1.06 mat × 1.14 lab = $176	$75	$251	$187	+6%	+42%
Seattle, WA	$348	$162 × 1.12 mat × 1.24 lab = $186	$100	$286	$225	+22%	+55%
San Francisco, CA	$525	$162 × 1.25 mat × 1.40 lab = $206	$165	$371	$251	+42%	+109%
San Jose, CA	$742	$162 × 1.25 mat × 1.43 lab = $208	$180	$388	$251	+91%	+196%

Several patterns emerge. High-income coastal markets saw their replacement costs rise sharply as the model properly captures the weight of local wages and land. San Jose's implied replacement cost rises from $251 to $388/sf under v2 — not because construction has gotten more expensive, but because the v1 model was drastically underweighting both its labor premium and its land costs. The market still trades at a premium to that higher figure, but the premium compresses from a misleading +196% to +91% — still high, but now reflecting genuine cost reality rather than a modeling artifact.

Mid-tier Sun Belt markets showed more nuanced shifts. Nashville, for example, moved from a +22% premium under v1 to a near-breakeven −1% under v2 — a significant reclassification driven by the model now correctly reflecting that Tennessee's relatively modest material costs and land values make Nashville cheaper to build in than its rising price point suggests.

Rust Belt markets deepened their discounts slightly in most cases, as the labor multiplier correctly reflects wages that are at or below the national median. Pittsburgh moves from −50% to −46%. The ranking at the top of the discount list is unchanged, but the magnitude is more accurate.

Data Sources

Input	Source	Vintage
NAHB construction baseline	NAHB Cost of Constructing a Home Survey	2024
Existing home $/sf	Realtor.com median listing $/sf via FRED	Feb 2026
Per capita income (PCI)	US Census Bureau ACS 1-Year Estimates	2024
Population growth	US Census Bureau Vintage 2025 Estimates	April 2020 – July 2025
Material cost index	RS Means regional construction cost data	2024
Land cost estimates	Internal — small-lot median land value ÷ 1,500 sf	2025

 This blog post reflects data available as of March 2026. Replacement cost inputs — particularly labor proxies and land estimates — are subject to revision as new Census and market data become available. Model assumptions are available on request.