The Setback

Method

How these numbers are built.

Every report runs on a pro forma model with a stated approach, a full list of inputs, and a sensitivity test against its own conclusion. None of that belongs in the middle of an argument, so it lives here instead.

R-01 · Washington, D.C.

The technical backing for the Washington, D.C. report: does the conclusion hold up under different assumptions, where every input comes from, and exactly how the model works.

Does this hold up?

A model that only produces one answer is not a model. Each material input moved to the ends of its defensible range, one at a time:

AssumptionLowBaseHigh
Hard cost, $/SF4.38% (at $220)3.72%2.95% (at $370)
Achievable rent2.98% (at $2,113)3.72%4.70% (at $2,900)
Land4.00% (at $1.5M)3.72%3.42% (at $3.5M)
Operating expenses per unit4.14% (at $7,500)3.72%3.40% (at $11,000)
Construction loan rate3.76% (at 6.5%)3.72%3.67% (at 9.5%)
Exit cap rate3.66% (at 4.75%)3.72%3.78% (at 6.00%)
Soft cost %3.88% (at 15%)3.72%3.57% (at 25%)
Vacancy3.84% (at 5.0%)3.72%3.59% (at 9.0%)

No single input, moved to its most favorable defensible end, clears 6.50 percent. The two that move the answer most, hard cost per square foot and achievable rent, are both cost-structural. Neither is a zoning variable.

To confirm the model is not rigged to fail: set every input to its most favorable end simultaneously and apply the full reform package. Return on cost is 8.45 percent, comfortably clearing. The model can produce a feasible building. Washington's current conditions cannot.

What would actually close it

At the regulatory ceiling, holding everything else at base: rent would need to reach $3,240 a month, a 32 percent increase, in a market where rents fell 3.8 percent last year.

Residual land value is negative $1.5 million. A merchant builder cannot pay for the dirt. They would need to be paid to take it.

That second number is the whole report in one line. At the current cost of construction and the current level of rents, a well-located D.C. site zoned for forty apartments is a liability rather than an asset to anyone whose business is building apartments.

Input manifest

Status flags: VERIFIED: published figure from a named source. VERIFIED_RANGE: base case sits inside a published range. DERIVED: computed from verified inputs. ESTIMATE: professional judgment, the priority validation queue.

#InputValueStatusSource
1Units40ASSUMPTIONModel design
2Unit mix and sizes484 / 697 / 996 SFVERIFIEDRentCafe / Yardi Matrix, Aug 2026
3Efficiency ratio80%ESTIMATEReflects two-stair mandate
4Land cost$2,400,000ESTIMATE~$205/land SF at FAR 3.0; 26% below D.C. asking average of $278/SF
5Hard cost, shell and core$275/GSFVERIFIED_RANGERLB QCR Q1 2026, D.C. multifamily $220–370/SF
6Structured parking cost$35,000/spaceVERIFIEDD.C. Policy Center, below-grade
7Parking ratio0.50/unitESTIMATEMinimums bind on 96% of D.C. residential lots
8Soft costs20% of land + hardESTIMATEPriority validation
9Entitlement timeline14 monthsESTIMATEPriority validation
10Construction timeline18 monthsESTIMATEPriority validation
11Construction loan rate8.00%VERIFIED_RANGESOFR ~4.8% + 275–400 bps; multifamily construction average 7.6%
12Achievable rent$2,450/moVERIFIEDRentCafe / Yardi Matrix, citywide average $2,446
13Inclusionary zoning set-aside10% of unitsVERIFIEDD.C. Policy Center, program requires 8–12.5% of residential floor area
14IZ rent, 60% MFI$1,550/moESTIMATERefine against DHCD 2026 IZ schedule
15Vacancy and credit loss7.0%DERIVEDYardi Matrix occupancy 94.2%; Northmarq vacancy 5.2%
16Other income3% of GPRESTIMATEIndustry convention
17Operating expenses$9,500/unit/yrESTIMATEPriority validation
18Property tax rate0.85%VERIFIEDD.C. OTR, Class 1A, $0.85 per $100
19Exit cap rate5.25%VERIFIED_RANGECBRE H1 2026, D.C. Class A stabilized infill 4.75–5.5%
20Required developer spread125 bpsESTIMATEStandard merchant-build screen

Priority validation queue: soft cost percentage (#8), entitlement timeline (#9), construction timeline (#10), operating expenses (#17). Land (#4) is now triangulated against published asking prices and one transaction, and the conclusion is independent of it.

Full source list

Permits and production

Policy and regulation

Costs, rents, capital markets, land

Limitations

The model is a merchant-build screen. A long-hold owner, a mission-driven developer with patient capital, or a project stacking low-income housing tax credits faces different arithmetic. The 6.50 percent hurdle is the right test for the marginal private dollar, not for every dollar.

Site-specific costs (remediation, historic façade retention, protected trees, public-space exactions) are excluded. Each is real in Washington and each pushes the answer further negative. The model is optimistic, not pessimistic.

Four inputs remain professional estimates rather than published figures. They are flagged in the manifest and will be updated against primary records.

The 2026 permit uptick is unresolved. If the second half of 2026 sustains the May–June pace, part of this analysis will need revisiting. I will publish an update when the full year is available rather than quietly leave the earlier framing in place.

Approach

The model asks one question. Build the building, rent it, run it for a year, and what does it earn as a share of what it cost?

That is the return on cost. It gets compared to what an investor needs: the rate a finished building of this type trades at, plus 1.25 percentage points to compensate for the risk of building it. In Washington that comes to 6.50 percent.

This is a single-year snapshot, not a full cash flow projection. It is the screen a developer runs before a deal gets a second meeting, and it is the right test for whether something gets built.

Three details are easy to get wrong, so they are stated plainly:

  • Construction interest is charged on half the loan. A construction loan pays out in stages as work progresses, so the average balance over the build is roughly half the total. Charging interest on the full amount for the full term overstates the cost by about double.
  • Entitlement interest is charged on the full land price. Land is bought in one go, so the whole balance accrues interest from day one.
  • Vacancy is deducted. No building is 100 percent leased and paying. Models that skip this step overstate income and therefore overstate value.

Property tax is solved by iteration. The tax depends on the assessed value, the assessed value depends on the income, and the income depends on the tax. The model loops until the three agree.

Calculator sources, all cities

The pro forma calculator runs the same R-01 engine across Washington, D.C., Los Angeles, Detroit, San Francisco, New York City and Miami, plus a frictionless benchmark. Land, hard cost, rents, timelines, taxes and exit caps are set per city in src/data/city-proforma.json; unit mix, efficiency, and construction financing are held constant across cities so only local conditions vary. Sources not already listed above, for the other five cities:

Land is an estimate in every city but D.C., where it was triangulated against asking prices and one transaction. Detroit hard cost is a proxy; RLB does not cover the market. Property tax rates for the four non-D.C. cities are effective-rate approximations, not published statutory figures, because those cities' effective rates depend on assessment practice. Entitlement timelines for Detroit, Miami and New York are professional estimates. Each of these is flagged ESTIMATE in city-proforma.json and on the calculator page.