Why is your rent so high?

A stylized, directionally-honest model of one housing market, built to teach the mechanism, not an appraisal, a forecast, or investment advice. Metro-level averages hide huge neighborhood variation. Where we estimate, we say so; where research gives a range, we pick the middle and cite it.

HUD's Fair Market Rent: the 40th percentile of local gross rent (rent + basic utilities) for a 2-bedroom, the standard used for housing vouchers. It is deliberately a typical rent, below the average of new listings you see on rental sites. The ACS median gross rent shown beside it covers all unit sizes and all tenants, including long-settled ones.

Cost floor = operating costs + property taxes + annualized construction cost + an unconstrained land price. That is what a competitive market would charge if homes were easy to add. Construction: national $/sqft × a metro cost index × a capital annualization rate (cap rate + reserves). Taxes: effective rates on the replacement basis (Lincoln Institute). This mirrors the "replacement cost" logic of Glaeser & Gyourko's zoning-tax work.

Premium = observed rent − cost floor, floored at zero. We split it between geography and regulation by weighing Saiz (2010) undevelopable-land shares against WRLURI-2018 restrictiveness plus typical permitting time. The split is illustrative: geography and rules interact (scarce land makes every rule bind harder). Cities where rent sits below the floor get a market-discount line instead: that's why little gets built there.

“How many homes is my city short?” has no single agreed answer. Published national estimates range from roughly 1.5 million to 5.5 million homes (Freddie Mac near 3.8M, Up for Growth near 3.9M, Zillow about 4.5M, the National Association of Realtors about 5.5M) because each one defines “enough” differently. So the definition matters more than the arithmetic.

We measure the deficit by price, not by headcount: how much more housing would it take to bring rent down to what building and running a home actually costs? That inverts the same supply response the lab uses: homes needed = stock × [(rent ÷ cost floor)^(1/elasticity) − 1]. It needs no assumption about which past year was “balanced.”

We only publish a deficit for cities carrying a scarcity premium. Where rent already sits at or below building costs, extra homes have nothing to squeeze out of the price, so quoting a shortage there would be meaningless. Those cities get the honest answer instead: no measurable deficit, or a surplus.

The big advantage is that it doesn't care who moved in. Counting people understates shortage exactly where shortage is worst: in an expensive city, the people priced out never appear in the population figures. Prices capture them; headcounts don't.

The cost is sensitivity to the elasticity. At −1.5 the deficit is larger, at −2.5 smaller, which is why we publish a range rather than a single figure. A vacancy-based target would give far smaller numbers, and we avoid it deliberately: vacancy barely moves in markets where rents have doubled, so it badly understates the gap.

Download the full methodology note (PDF): the derivation worked through step by step, a sensitivity table, why our national figure sits above the familiar 3.8–5.5 million estimates, and where the method stops working. It travels with this page; no download from anywhere else.

Building permits measure approvals, not homes. Some are never built; others are offset by homes lost to demolition, fire, or conversion to another use. Nationally, net additions run about 83% of permits, so counting permits alone quietly overstates how much housing a city actually gained.

The chase chart therefore counts net homes. We apply a national loss rate that falls in fast-growing metros and rises in shrinking ones, where old stock comes down faster than new stock goes up. It's an estimate (the Census measures this through its Components of Inventory Change program, which runs infrequently and not at the metro level), so treat the exact percentage as approximate and the direction as solid.

All supply-side levers sum in housing-stock space, then one price map applies: +1% stock ⇒ ≈ −2% rent (see elasticity below), phased over 10 years. Cost-side levers lower the cost floor of new units and blend into the market as new supply arrives. Three regimes, honestly modeled: in scarcity markets supply levers bite and cost levers barely move market rent; at-cost markets track the (falling) cost floor but shrug off extra permits, because supply already answers demand there; below-replacement markets respond to neither, because demand, not supply, is their binding constraint. Downzoning (NIMBY mode) raises rents in every regime. Real-world anchors: Auckland's 2016 upzoning left rents ≈ 26–33% below trend after 8 years; Austin's 2021–23 permit surge cut asking rents ~7–12%; Minneapolis 2018–22 grew rents far slower than its state.

Be precise about what this number is. It is our working assumption, a central value consistent with a body of evidence that new supply lowers rents, not a figure any single paper reports. Been, Ellen & O'Regan survey that evidence in Supply Skepticism (2019, updated 2023); Mast (2021) traces migration chains out of new market-rate buildings; Asquith, Mast & Reed (2023) measure rents on the blocks around new construction; Li (2022) studies NYC towers. Those establish the direction and give a sense of magnitude. None of them estimates a metro-wide elasticity of −2, and we don't claim they do, which is exactly why the deficit ships as a range rather than a number.

Rents: HUD FMR (FY2015 & current). Median rent, income, population, vacancy, transit share: Census ACS. Permits: Census Building Permits Survey (metro, 2010–2024). Zoning restrictiveness: Wharton Residential Land Use Regulatory Index (Gyourko, Hartley & Krimmel 2018 metro means). Geographic constraint: Saiz (2010 QJE) undevelopable-land shares. Property taxes: Lincoln Institute 50-State Property Tax Comparison effective rates. Construction cost indices and permitting times: published city cost indices and industry surveys, flagged as estimates. Fields we estimated for a metro are marked est in the data file.

These constants, the formulas above, and your metro's data row fully determine every number on this page. Change any of them and the receipt reprints.

FMR lags asking rents by ~1–2 years and lowballs hot submarkets. New York's property-tax system (class shares, abatements) resists a single effective rate. Gross rent includes utilities; our operating line does too, so the receipt stays consistent. Archetype cities are illustrative composites. The geo/reg split is a modeling choice, not a measurement. Rent control, vouchers, and submarket dynamics are out of scope for v1.