A new NFL stadium is sold on a page of projections — how long it’ll last, how many jobs it’ll make, what the bond will cost. Voters approve the page. Then time grades it. The projections were never binding; the bond always was. This lab is the audit nobody runs at the vote: what the paper promised, against what the years revealed.
This is the fifth lab in How Constraints Actually Work. A rule with a gap. A part nobody tested. A ruler nobody re-checked. A baseline nobody observed. This one is a feasibility study — the page of numbers a stadium is sold on at the vote. Projected useful life, promised jobs, economic impact, the bond that pays for it. The public approves the page. But only one line on it is a promise anyone is legally bound to: the bond schedule. The rest is forecast.
Between 2016 and 2027, six NFL stadiums were built or approved on public money — U.S. Bank, Mercedes-Benz, SoFi, Allegiant, the new Bills stadium, the new Titans stadium — each with a feasibility study behind the vote. A decade of receipts is now in. The buildings mostly stand; the projections did not survive contact with time. A revenue stream forecast at $34M a year came in at $1.7M. A “10,000 jobs” headline was a ~6,842-job study. A “first-class through 2039” obligation got a building replaced at 28.
None of that requires anyone to have lied. The bond got signed and honored. The projection was made in good faith and simply wasn’t binding — and nothing on the ballot marked which numbers were which. That’s the fifth thing to check before you trust a plan: which line here is a promise someone has to keep, and which line is a forecast that costs nothing to miss?
● Sourced — a figure quoted to an official statement, a government release, or reporting, cited on the last tab. Dates, public dollars, bond maturities, the numbers actually printed at the vote.
◐ Derived / inferred — a ratio or read the lab computes from sourced numbers (realized ÷ stated life; public $ per promised job), or an ex-post economic reading drawn from the general subsidy literature. Useful, but it’s an argument, not a receipt.
◯ Unverified — a figure the sources didn’t pin down (some bond terms in years). Marked, never guessed.
“I’ve signed enough paper to know the trick. On the ballot, the jobs number and the bond number sit in the same font, and that’s the con — not a lie, a font. One of them is a contract and the other is a wish, and they’re dressed the same. Find the bond schedule. That’s the number somebody has to answer for in thirty years. Everything sitting above it is what got you to say yes.”
“Now this — this right here — is a generational investment. World-class. Transformational. Ten thousand good jobs, hundreds of millions in impact, put our city on the map for a generation, and folks, here’s the beautiful part: at no cost to the hardworking taxpayer. It’s all right there in the study. You have my word.” [He can go six hours on this without stopping for water. Count the numbers he just named — jobs, impact, ‘no cost’ — every one is a forecast. Count the one he didn’t name: the bond schedule, the only figure anybody signs. That’s the whole fork — hot air on top, a signature underneath. He is the archetype, not a person; the target is the pitch, not a party.]
Every row is a stadium; every row opens to its two ledgers — ◐ what the paper promised at the vote, and ● what time revealed. Click a column header to sort (public dollars, how far the debt runs, realized ÷ stated life, public $ per promised job). Click a row to open it. The derived columns are the whole point — they’re the audit the feasibility study never showed you.
| Stadium ▼ | Opened ▼ | Public $ (M) ▼ | Debt runs to ▼ | Realized ÷ stated ▼ | Public $ / promised job ▼ |
|---|
Sort by “public $ / promised job.” Then find the same figure in the feasibility study that sold each deal — you won’t, because nobody prints the subsidy divided by the promise. Now take a decision in your own town on the ballot this year: a stadium, a convention center, a road. Separate its page into two columns — the lines someone is contractually bound to, and the lines that are forecasts — and decide which column you’re actually voting on. There’s no trick to it. There’s only reading the whole page instead of the headline. Same shape at household scale: rent-vs-own and a compounding scholarship over in The Business Sampler, and productive-vs-predatory credit in The Bank on Wheels — money drawn out over time, where the honest cost hides in the schedule.
The other four walls each hide a failure — a gap, an untested part, a stale ruler, an unobserved baseline. This one hides in plain sight, on the ballot, in matching type. A feasibility study is a real document doing real work; the bond inside it is a genuine promise. The trouble is only that the forecast lines carry no penalty for being wrong, and nothing marks them apart from the lines that do. A published useful life and a published FEMA discharge and a published Brexit cost are the same animal as a published stadium projection: a confident number whose confidence lives in the paperwork, not the outcome — and the bond is the one line the paperwork can’t fudge, because somebody signs it.
Why it matters. This lab is not anti-stadium and not pro-stadium. Whether a city should subsidize a team is a real argument with real people on both sides. The lab is about one narrower, checkable thing: at the moment of the vote, the page mixes contractual promises with free forecasts in identical type, and the reader is rarely told which is which. Read the whole page, separate the two, and you can have the argument honestly.
Opening years, total and original costs, public dollars and shares, bond maturities/pledge end-years, the projected figures printed at each vote (pull-tab revenue, job counts, economic-output and tax projections), and the historical debt-vs-demolition timelines are quoted to official statements, government releases, and reporting, listed below.
The workbook’s two computed columns — realized ÷ stated life and public $ per promised job — are the lab’s arithmetic on sourced numbers, shown as an audit the feasibility studies never present. The old-Nissan “~70%” (28 of 40 contracted years) is a derived ratio, not a published figure. The reading that promised economic benefits generally did not materialize rests on the broad subsidy literature (Bradbury, Coates & Humphreys; Noll; Matheson), not a venue-by-venue audit — there rarely is one. “Debt outlived the building” is stated cleanly only for Veterans Stadium and the Kingdome; for the modern venues the accurate phrasing is “the public revenue pledge runs decades past construction” (Atlanta to 2050, Minneapolis to 2046), and U.S. Bank is the counter-case where state debt was retired ~20 years early.
The Metro Nashville revenue bonds are a sourced 30-year term (approved 2023 → ~2053). Still not pinned to an official statement: the final maturity of Erie County’s GO serial “Bills Bonds” (Series 2024B) and the Tennessee state $500M GO term — both live in the offering documents on EMMA (login-gated), and are left as blanks rather than guessed. Several packages are appropriation- or settlement-funded (Minnesota; New York’s $600M Seneca casino money), where “bond term” doesn’t cleanly apply — which is itself the tell on the Bills deal: appropriation funding has no maturity schedule to hold anyone to.
The scoreboard is not the subject; the paperwork is. The dollars and the maturities and the numbers printed at the vote were recorded — those are receipts. The ratios I built from them are arguments: honest arithmetic, but arithmetic, and I’ve marked them so. If a figure here is stale or a bond term surfaces, email User Zero — corrections get acknowledged right here.