Nobody knows the exact size of America's housing shortfall, and the credible published estimates do not agree with each other. Freddie Mac put the national deficit at 3.8 million units as of the fourth quarter of 2020, up from its own 2.5 million estimate in 2018. Other groups count differently and land lower or higher. The gap is real, the size is contested, and for an operator the number changes far less about your buying decision than your own unit-level math does.
Last updated: 2026-09-20
I have spent 18-plus years in real estate, focused on short-term rentals since 2016, and trained more than 10,000 operators through CashFlow Diary. In that time I have watched a lot of people buy a bad deal because a headline told them supply was scarce. Scarcity is a national average. You do not own a national average. You own one roof, in one submarket, with one rate structure and one set of carrying costs.
This piece does three things. It lays out what the published shortfall estimates actually measure. It flags the widely repeated figure I could not verify anywhere in the primary record. And it gives you the part that matters: what a structural supply gap does and does not change about your buy, hold, and rent decisions.
What the verified numbers actually say
Start with the source most often quoted. Freddie Mac's housing supply research estimated a national deficit of 3.8 million units as of Q4 2020, calculated against a target housing stock that assumes a 13 percent target vacancy rate and a "target headship rate" — the share of adults who would head their own household if housing costs were not suppressing formation. The same research put the 2018 figure at 2.5 million, a 52 percent increase in two years.
Inside that number is the detail most investors skip. Freddie Mac's work shows entry-level construction collapsing over decades: new entry-level single-family homes averaged 418,000 units a year between 1976 and 1979, fell to an average of 55,000 a year during the 2010s, and came in around 65,000 completions in 2020. The shortage is not spread evenly across price points. It is concentrated at the bottom, which is exactly where most first-time buyers and most workforce renters live.
Now the current production data. The Census Bureau and HUD report new residential construction monthly. For August 2026, building permits ran at a seasonally adjusted annual rate of 1,394,000, housing starts at 1,275,000, and completions at 1,128,000. Single-family permits were 878,000; units in buildings with five or more units were 467,000. Completions running below 1.2 million a year, while the accumulated gap is still measured in millions of units, is how a deficit gets maintained rather than closed.
And the demand side is not screaming scarcity right now. Census reported national vacancy rates for the second quarter of 2026 at 7.3 percent for rental housing and 1.2 percent for homeowner housing, with a homeownership rate of 65.0 percent. NAR reported existing-home sales down 2.0 percent in August 2026 with months' supply at 4.9 — its highest level in over ten years, per NAR chief economist Lawrence Yun.
Read those two paragraphs together. Long-run underbuilding at the entry level, alongside a resale market that is currently loosening. Both are true. An article that only tells you one of them is selling you something.
Why the estimates disagree
Different groups produce different shortfall numbers because they are answering different questions. None of them are lying. They just draw the boundary in different places.
Approach | What it counts | Why the number moves |
|---|
Household-based deficit (Freddie Mac) | Units needed to house actual plus suppressed households at a target vacancy rate | Very sensitive to the assumed target vacancy rate and target headship rate |
Underproduction (Up for Growth) | Homes not built relative to what growth and existing need would have required, measured metro by metro | Sensitive to the baseline year and to how "missing households" are modeled |
Production-gap arithmetic (Census + HUD data) | Annual permits, starts, and completions versus the prior year | Shows the flow, not the accumulated stock deficit |
Up for Growth runs the second method. Its 2025 report put national underproduction at 3.78 million homes in 2023, down from 3.85 million in 2022 — the second straight year the gap narrowed. Two independent methods landing near 4 million is worth more than either one alone.
Three mechanical levers swing any estimate by millions of units: the target vacancy rate you assume a healthy market needs, how many suppressed households you credit (adults doubled up who would form their own household at lower cost), and whether second homes, short-term rentals, and uninhabitable units count against the stock. Run Freddie Mac's own inputs — 126.2 million target households — at a 13 percent target vacancy rate and you get about 145 million units of required stock; at 11 percent you get about 142 million. Two points of assumption, about three million units of answer. That is arithmetic, not a scandal.
So when one group publishes a shortfall well under two million and another lands near four million, the honest answer is that both can be defended and neither is a fact in the way that "1,128,000 completions in August 2026" is a fact.
The 10 million number I could not verify
The earlier version of this article asserted that the White House published a report in April 2026 titled "Addressing America's Housing Shortage: A Path Forward" declaring a 10 million unit shortfall, attributed it to the Biden administration, and cited a March 2026 median price of $408,800 as the "33rd consecutive record high."
I went looking for that report on whitehouse.gov and across federal housing data publishers and could not find that title, that report, or that 10 million unit figure anywhere in the primary record. I also could not tie the $408,800 median or the 33-month record streak to NAR, Census, or HUD published data. Rather than restate it with a hedge, I am removing it. If a federal report with that title and that number is published and I can pull it from a .gov domain, I will update this page and say so.
That matters beyond housekeeping. A 10 million figure is roughly triple the best-documented household-based estimate. If you underwrote a purchase on the assumption that federal policy had just certified a shortage three times larger than the research says, you priced in appreciation you have no evidence for. I would rather hand you a smaller number I can source than a bigger one I cannot.
This is educational content, not tax, legal, or investment advice. Run your own numbers with your own professionals.
What a supply gap does and does not change for an operator
Here is the part the headlines never get to. A national supply deficit is a slow variable. Your deal is a fast one. The deficit does not pay your mortgage in February.
Decision | What a structural supply gap supports | What it does not support |
|---|
Buy | Confidence that entry-level and workforce housing stays in demand over a long hold | Paying above your underwritten price because "they are not making more land" |
Hold | Patience through a soft resale window instead of a forced sale | Holding a property whose cash flow is negative every month with no path to fixing it |
Rent pricing | A floor under long-run rent demand in constrained submarkets | Assuming you can push nightly or monthly rates without comps supporting them |
Financing | A case for fixed-rate, longer-term debt on a long hold | Stretching debt service coverage on the theory prices only go up |
Market selection | Weighting metros where permits lag household growth | Chasing a metro on migration headlines with no live rate or occupancy data |
Exit | A reasonable expectation of buyer depth at the entry level | A specific appreciation rate baked into your model |
The failure mode I see most often: an operator reads a scarcity story, raises their maximum purchase price by ten percent, and buys a property whose numbers only work if rents grow. Then August arrives, occupancy sits below what the pro forma assumed, and the shortage does not send a check. The fix is boring. Cap your purchase price at what today's verifiable revenue supports, and treat every macro tailwind as upside you did not pay for.
Here is a worked illustration. The assumptions are mine, not a projection from any data source, and your inputs will differ. Say a property underwrites to $52,000 of trailing-twelve-month comparable revenue, with $19,000 of operating cost and $24,000 of annual debt service at a $400,000 purchase price. That leaves roughly $9,000 of margin. Raise your offer ten percent on the same financing terms and debt service climbs by about $2,400 a year, cutting that margin by more than a quarter — before a single soft month. Scarcity headlines do not fund that gap. Your pricing discipline does. Run the same three lines on your own deal before you decide what a national estimate is worth to you.
The second failure mode is the mirror image. An operator reads that months' supply hit 4.9 and the resale market is loosening, panics, and dumps a property that cash flows fine. Resale liquidity and rental demand are different markets. A slower buyer market often means more motivated sellers and better acquisition terms for you, not worse operations.
How to use supply data without letting it underwrite for you: a five-step process
Use macro data to choose where to look. Use unit-level data to decide what to pay. Never mix the two up.
Step 1 — Screen markets with production versus formation
Pull permits and completions for your candidate metros from Census new residential construction data and compare them against household growth. A metro permitting well below its household growth for several years has a structural tailwind. That gets a market onto your list. It does not price a single deal.
Step 2 — Check the rental side separately
Rental vacancy nationally was 7.3 percent in Q2 2026. Your submarket may be far tighter or far looser. Pull local vacancy and active listing counts before assuming a national deficit story applies to your street. If competing supply is rising in your exact class of property, the national number is irrelevant to you.
Step 3 — Underwrite from live revenue, not a forecast
Build revenue from actual comparable performance over the trailing twelve months, segmented by season. For short-term rentals, that means real booked nights and real average daily rates for properties of your bedroom count and amenity set, not a platform projection. Cut the top decile of comps out of your average. They are usually better operated than a first-year listing will be.
Step 4 — Stress the model before you sign
Run three cases. Base at your comp-derived revenue. Downside at 20 percent lower revenue and 10 percent higher operating cost. Failure case at long-term-rental rent only. If the failure case still covers debt service plus taxes and insurance, you own a property with an exit that does not depend on any macro story. If it does not, you are betting on the shortage.
Step 5 — Set a review date, not an opinion
Put a calendar date twelve months out to re-pull the same data: permits, vacancy, months' supply, and your own occupancy and rate. Decide in advance what result would change your hold decision. Writing the trigger down before you are emotionally attached to the property is the whole trick.
One more note on the data itself. Census construction figures are seasonally adjusted annual rates and get revised, sometimes meaningfully, in the following month's release. Treat a single month as noise. Track the trailing six-month average of permits and completions instead, and compare it to the same window a year earlier. The direction of production in your metro is a better input than any one print, and it is the part of the supply story you can actually check yourself in ten minutes.
Who this is and is not for
If you are buying your first rental and your plan depends on appreciation to make the numbers work, this article is not permission. It is the opposite. A contested national shortfall estimate is not a reason to buy a property that loses money monthly.
If you already own and you have been rattled by softer resale headlines, the supply picture is a reasonable argument for patience — provided your unit-level numbers hold. Entry-level construction has run far below historical norms for decades, and that does not reverse in a year.
If you are an operator with capacity to add a unit and the discipline to cap your price at underwritten value, constrained production in your target submarket is a fair thumb on the scale. Just keep it as a thumb, not the whole hand.
And if what you actually want is certainty about the number, I cannot give it to you and neither can anyone else. Pick decisions that survive being wrong about it.
Frequently Asked Questions
How many housing units is the United States actually short?
There is no single verified figure. Freddie Mac's research put the national deficit at 3.8 million units as of Q4 2020, up from 2.5 million in 2018, using a model that assumes a 13 percent target vacancy rate and adds back suppressed household formation. Other groups measure underproduction metro by metro and reach different totals. Anyone quoting one number as settled fact is overstating what the research supports.
Did the White House publish a report saying America is short 10 million homes?
I could not verify that. I found no report titled "Addressing America's Housing Shortage: A Path Forward" on whitehouse.gov and no 10 million unit federal figure I could source to a government domain. Until such a document is published and checkable, treat the 10 million number as unsourced.
If there is a shortage, why is inventory rising?
Because they measure different things. NAR reported months' supply of existing homes at 4.9 in August 2026, the highest in over ten years, which reflects current buyer demand against listings. A structural shortfall reflects accumulated underbuilding against household need over decades. High mortgage rates can slow buying while the long-run unit deficit is unchanged.
Should a supply shortage change how much I pay for a short-term rental?
No. Price from your underwritten revenue using trailing twelve-month comparable performance, then stress it. Treat supply constraint as upside you did not pay for. If a deal only works when you assume rent growth, you are buying a forecast, not a property. This is educational, not investment advice.
Which data should I actually track each quarter?
Four items: Census new residential construction permits and completions for your metro, Census rental vacancy, NAR months' supply, and your own occupancy and average daily rate. The first three tell you the weather. The last one tells you whether your business is working.
Most operators do not need a better macro opinion. They need someone to check their unit-level math before they sign. If you want that second set of eyes, book a complimentary one-on-one diagnostic call and bring your actual numbers — purchase price, comps, and your downside case.
Sources
Freddie Mac, Housing Supply: A Growing Deficit — 3.8 million unit deficit as of Q4 2020, the 2.5 million 2018 estimate, the 13 percent target vacancy assumption, and entry-level completion history.
U.S. Census Bureau and HUD, Monthly New Residential Construction — August 2026 permits (1,394,000), starts (1,275,000), and completions (1,128,000) at seasonally adjusted annual rates.
U.S. Census Bureau, Quarterly Residential Vacancies and Homeownership — Q2 2026 rental vacancy of 7.3 percent, homeowner vacancy of 1.2 percent, and the 65.0 percent homeownership rate.
National Association of Realtors, Existing-Home Sales — August 2026 sales down 2.0 percent and months' supply at 4.9, the highest in over ten years.
Up for Growth, 2025 Housing Underproduction in the U.S. — underproduction of 3.78 million homes in 2023, down from 3.85 million in 2022, and the metro-level method behind it.
Up for Growth, Housing Underproduction reports archive — the annual report series and datasets used for the metro-level underproduction approach.
