Ten Million Units Short, Still Not Building: Why the Math Doesn't Pencil on America's Housing Deficit
The headline number moves around. Up for Growth puts it at 3.9 million. Freddie Mac's 2021 study said 3.8 million. The National Low Income Housing Coalition says 7.2 million for affordable units alone. Zillow's 2024 estimate is 4.5 million. If you take cumulative U.S. household formation against cumulative housing completions since 2008 and compound the shortfall, you get to roughly ten million, which is the number that now appears in congressional testimony with predictable regularity.
The estimates disagree on definition. They agree on direction. Cumulative household formation has outpaced cumulative completions every year since 2010, and the gap has compounded into a structural deficit that every major methodology converges on at scale.
What the shortage estimates do not answer is why. We know the demand is there. We know rents have risen. We know starts are muted. The interesting question is the one almost no one in the public debate is answering: at current costs, current rates, current fees, and current entitlement friction, how does the pro forma pencil, and if it does not, what would have to change for it to.
This piece walks through the mechanics.
1. The Demand Side: How Ten Million Adds Up
The 2020 to 2024 window generated roughly 6.4 million net new households in the United States per Census ACS data. Over the same period, net new housing completions were approximately 5.7 million units. That is a 700,000 unit deficit accumulated in four years, on top of a pre-existing shortfall.
Three structural demand drivers deserve specific attention.
Delayed household formation. In 2020, roughly 26 million American adults aged 18 to 34 were living with their parents, the highest share since the 1940s. A meaningful fraction of those adults have formed households since, either through household formation proper or through doubled-up non-family arrangements. That release of pent-up household demand is already visible in the rental application data and is not finished.
Immigration. Net international migration to the United States was roughly 3.3 million in 2023 per CBO, the highest single-year figure on record. Immigrant households disproportionately rent for the first five to seven years after arrival, and they concentrate in metros with already-constrained supply.
Boomer aging-in-place. The population over 65 is retaining single-family housing stock at roughly double the rate of the prior cohort. AARP surveys show 77 percent of adults over 50 want to age in place, and the mortgage rate lock-in effect has amplified that behavior. Stock that would have historically rotated to younger buyers is not rotating.
The demand math is not mysterious. The question is the supply response.
2. The Supply Response Looks Broken
Annual housing starts ran at a long-run average of roughly 1.5 million units through most of the 1970s through 2007. They collapsed after the 2008 financial crisis and did not fully recover for more than a decade. The 2021 peak of 1.6 million starts was the first meaningful print above the long-run average since 2006.
Then rates moved, and starts fell back to roughly 1.35 million annualized by mid-2024. Multifamily starts specifically peaked at a 764,000 SAAR in Q4 2021 and fell to roughly 300,000 SAAR by mid-2024, a 60 percent decline in less than three years.
The completions we see in 2024 and 2025 are largely the tail of a cycle that started in 2020 and 2021, when construction debt was cheap and land was still available on long-dated options. The 2025 to 2027 completion window is going to be dramatically thinner because the starts that would have produced those completions are not happening.
3. Construction Cost Structure: Why the Pro Forma Breaks
The total development cost of a typical garden apartment has moved as follows over the five-year window 2019 to 2024, based on RSMeans, Turner Construction Cost Index, and aggregated sponsor survey data:
| Cost component | 2019 (per unit) | 2024 (per unit) | Change |
|---|---|---|---|
| Hard costs (materials + labor) | $165,000 | $225,000 | +36% |
| Soft costs (A&E, legal, insurance) | $42,000 | $62,000 | +48% |
| Land | $36,000 | $52,000 | +44% |
| Financing costs (construction period) | $22,000 | $38,000 | +73% |
| Reserves and contingency | $25,000 | $33,000 | +32% |
| Total development cost | $290,000 | $410,000 | +41% |
The aggregate 41 percent increase over five years is the number that breaks most pro formas. Over the same window, market rents for new Class A product increased roughly 25 to 30 percent depending on metro, meaning yield on cost compressed even before exit cap rates widened.
The input-level detail behind the hard cost line matters:
- Lumber. Spot prices peaked at $1,650 per thousand board feet in May 2021 versus a $350 pre-pandemic baseline. They have normalized to the $450 to $600 range but remain above pre-2020 levels.
- Steel. Up roughly 40 percent cumulatively versus 2019, with structural steel lead times extended from eight to sixteen weeks for mid-rise product.
- Concrete and cement. Up 25 to 30 percent, with regional variation driven by fuel costs and cement capacity.
- Insurance. Builders risk insurance in Florida, Texas, Louisiana, and coastal California is up 150 to 300 percent since 2020, and for some coastal zip codes is effectively unobtainable.
- Appliances and finishes. Up 20 to 35 percent, with continuing pressure from supply chain normalization that is still incomplete.
4. Labor: The Invisible Constraint
The Associated Builders and Contractors estimate the U.S. construction industry is short roughly 500,000 workers against current demand. The gap is larger in the skilled trades: electricians, HVAC technicians, licensed plumbers, commercial framers. Construction labor wages have risen 25 to 35 percent over the 2019 to 2024 window depending on trade, with skilled specialty trades at the top of that range.
Construction productivity has been roughly flat in real terms for forty years, a pattern McKinsey documented in 2017 and which has not meaningfully changed. Of the major U.S. industries, construction and healthcare are the two where unit productivity has not grown. The implication is that cost increases flow directly into end product prices because there is no productivity gain absorbing them.
The labor gap is not going to close on its own. Immigration reform, vocational education at scale, and factory-built construction are the three levers that could move it. None of them are moving fast.
5. Land and Entitlement: The Binding Constraint
Raw hard cost inflation is the visible part of the problem. The less visible part, and in many markets the larger part, is land and entitlement.
Land basis. Infill land prices in Los Angeles, Miami, Boston, Seattle, Denver, Austin, and Nashville are up 200 to 400 percent since 2015. The tension is that the metros where demand is strongest are the metros where supply elastic land is scarce. In most Sun Belt secondary markets land basis remains 10 to 15 percent of total development cost. In California Coastal, Boston, and New York it can reach 30 to 50 percent, at which point the deal is effectively a bet on the land reversion, not on the operating asset.
Entitlement timelines. Typical multifamily entitlement runs 18 to 36 months in California, New York, Massachusetts, and Washington. It runs 6 to 12 months in most of Texas, Florida, and Georgia. The two-year gap is worth real money. At 7 percent cost of capital, carrying pre-development land and soft costs for an extra 24 months adds 14 to 18 percent to total development cost before a shovel hits dirt.
Discretionary review. The single most underappreciated variable in housing economics is whether a jurisdiction is by-right or discretionary. A by-right jurisdiction permits a project that complies with the code. A discretionary jurisdiction subjects every project to a commission hearing where approval is optional. The same project is a fundamentally different risk in the two regimes, and capital prices that risk accordingly.
6. Impact Fees and the Regulatory Tax
The National Association of Home Builders estimates that 23.8 percent of the final price of a new single-family home is attributable to regulatory costs, including impact fees, land use regulation, and permit requirements. On multifamily, the comparable share varies by jurisdiction but regularly exceeds 20 percent in coastal California, the Northeast, and portions of the Pacific Northwest.
Representative per-unit impact fees for new multifamily construction:
| Jurisdiction | Typical per-unit impact fees | Notes |
|---|---|---|
| Fremont, CA (Bay Area) | $60,000 to $80,000 | School, park, traffic, affordable in-lieu |
| San Diego, CA | $40,000 to $55,000 | Includes public facilities fee |
| Austin, TX | $10,000 to $15,000 | Lower, but rising |
| Nashville, TN | $5,000 to $9,000 | Minimal impact fee structure |
| Miami-Dade, FL | $12,000 to $18,000 | School concurrency fees dominant |
| Dallas, TX | $8,000 to $12,000 | Minimal regulatory tax |
A project in San Jose pays roughly 60 to 70 thousand dollars per unit more in impact fees than an identical project in Dallas. That differential is not a function of construction quality or market demand. It is a pure regulatory transfer, and it flows directly into required rent to cover the incremental cost.
Parking requirements are the other line item that quietly distorts supply. Urban infill parking construction costs $30,000 to $75,000 per structured stall. A minimum parking requirement of 1.5 stalls per unit on a 200-unit building can add $9 to $22 million in non-revenue-producing cost to a pro forma. Jurisdictions that have eliminated or reduced parking minimums (Minneapolis, Buffalo, most of Oregon post SB 8) have seen meaningful increases in apartment starts specifically in the product segment most sensitive to per-unit cost.
7. NIMBYism and the Political Economy of Zoning
The NIMBY problem is not just local opposition. It is an institutional structure that gives local opposition legal leverage. Three mechanisms do most of the work.
Environmental review. The California Environmental Quality Act (CEQA) and its state-level analogs in Washington, New York, and Massachusetts were written for environmental protection. In practice, they are now used primarily to delay housing. Cal YIMBY estimates that CEQA review adds two years and approximately $250,000 in compliance cost to the average housing project in California. The lawsuit rate on housing CEQA filings is approximately 40 percent.
Discretionary review boards. Zoning boards of appeals, design review commissions, planning commissions, and local historic commissions each carry veto power in most jurisdictions. A project that requires five separate discretionary approvals has five separate opportunities to fail. The compound probability math is unforgiving.
Appeals and referenda. In most jurisdictions, an approved project can be appealed to the city council, to the state planning board, and in some cases, via local ballot referendum. Each layer adds 6 to 18 months of delay, regardless of eventual outcome.
There has been meaningful reform motion. California SB 9 and SB 35 created by-right pathways for certain housing types. Oregon's 2019 HB 2001 abolished single-family-only zoning statewide. Minneapolis similarly ended single-family zoning in 2019. The Montana Land Use Planning Act of 2023 requires cities to allow by-right multifamily in most zones. These reforms are the right direction. They are also much slower in implementation than headlines suggest, and the California data specifically shows that SB 9's first three years produced fewer than 100 additional units statewide because of implementation barriers.
8. Capital Markets: The Construction Finance Reality
Construction debt is where the cycle change is most visible.
- 2019 to 2021. Construction loans from regional banks at SOFR plus 225 to 300 basis points, all-in cost 3.5 to 4.5 percent. Non-recourse at the sponsor GP level with customary carve-outs. 65 to 75 percent loan-to-cost.
- 2024 to 2025. Same project, same sponsor, construction loan at SOFR plus 300 to 450 basis points, all-in cost 8.5 to 10 percent. Full recourse or significant completion guarantees. 55 to 62 percent loan-to-cost. Many regional banks out of the market entirely post-SVB.
Private credit filled part of the gap but at 200 to 400 basis points above bank pricing, with mezzanine layered on top at 12 to 15 percent preferred returns, blending the total cost of capital up further.
Exit cap rate assumptions used by underwriters have widened 150 basis points or more from the 2021 trough. A deal that penciled at a 4.25 percent exit cap does not pencil at a 5.5 percent exit cap, and most sponsors are now underwriting 5.5 to 6.0 percent exits on new construction.
9. A Worked Example: Why the Pro Forma Breaks
Consider a typical 200-unit Class A Sun Belt garden apartment deal underwritten in Q1 2026.
- Total development cost: $82 million ($410,000 per unit)
- Market rent at stabilization: $2,100 per month, $25,200 per unit per year
- Operating expense ratio: 38 percent
- Stabilized NOI: $15,600 per unit, $3.12 million total
- Yield on cost: 3.80%
- Required yield on cost for an institutional GP to take the trade: 6.00 to 6.50%
The gap between the 3.8 percent the asset produces and the 6.0 percent the capital requires is the reason the deal does not happen. There are three ways to close the gap:
- Reduce TDC by 30 percent. Implausible without structural change in construction cost, labor, or fees.
- Increase stabilized rent by 45 percent. Not happening organically absent a major supply shock or rapid wage inflation.
- Compress exit cap by 150 basis points. Requires a significant Fed pivot and a risk-on reversal in the capital markets.
None of the three is inevitable. None is impossible. The underwriter's job is to pick the combination of inputs that is most plausible, size the bet accordingly, and be transparent about what has to be true for the trade to work.
10. What Actually Moves the Needle
By-right zoning reform. The single highest-leverage policy intervention, and the most politically difficult. State preemption of local discretionary review (CA, OR, MT) is the current template. Federal preemption is legally possible but untested.
Factory-built and modular at scale. The technology is proven. The bottleneck is volume. A modular factory producing 1,000 units per year at fifteen to twenty percent cost savings versus stick-built does not move the national number. A hundred such factories does. We do not yet have the capital stack that funds that expansion.
Impact fee reform. Converting fixed per-unit fees into value-capture or tax-increment mechanisms would reduce the barrier for small units while still funding infrastructure. This is mechanically straightforward and politically painful.
Federal construction lending backstop. A Ginnie Mae or FHA-style program for construction lending on multifamily would be the multifamily analog to what Fannie and Freddie do for stabilized debt. It is politically sellable given the housing shortage framing. It has not been done.
Rate normalization. The single largest near-term lever. A Fed funds rate 200 basis points below current levels would restart construction finance at scale without any other change. Whether it happens is not in the developer's control.
Closing
The ten million unit shortage is real, the estimates of its size all point in the same direction, and yet starts remain muted because the unit-level economics of building are misaligned with the macro-level need for units. A shortage is a price signal. A supply response is a different problem, and it lives in the intersection of construction cost, labor, land, fees, zoning, and capital markets.
Nobody in the public debate is going to solve the shortage by publishing another estimate of the shortage. It gets solved by fixing the specific line items that broke the pro forma, one by one, in the jurisdictions where the gap is largest and the political will exists. That is slower, less photogenic, and more durable than any of the headline proposals that dominate the housing press.
The math has to pencil before the buildings go up. The shortage is proof that the math is not penciling. Everything else is commentary.