During the 2026 World Cup, host-city hotel rates rose 13% on average. Miami rose 30.3%. New York fell 4.5%. The average described nobody.
Before the tournament, the published forecast put host-city room rates up 7 to 25%. It was a reasonable band, and as an average it was roughly right: across the 16 host markets the rate premium came out at about 13%.
As a guide to any one city, it was wrong.
Living Lab measured each market on per-hotel fixed effects: each hotel against its own history, against a counterfactual period, rather than against a city average.
Miami booked 382,025 tournament room nights with a rate premium of 30.3%. Dallas booked 491,175 with a premium of 19.8%. New York booked the most room nights of any host, 537,655, and its rate fell 4.5%. The city absorbed the largest event on earth without moving on price. Las Vegas did not host a match; its flagship expo in the same period ran a rate 4.4% below the prior year.
Outcomes outside the forecast band, in both directions. The spread is the finding.
A national average would have told all 16 cities the same thing, and it would have been wrong for most of them.
A rate premium is a property of each city, not of the event. It depends on the city’s existing calendar, its hotel supply, its short-term rental stock, and its usual June. Miami had headroom and pricing power. New York had neither, because a June week in New York is already full.
A city that planned on the 13% average would have priced Miami too low and New York too high. The hotels that needed to hold rate and the hotels that needed to discount got the same advice.
The measure has to be per hotel, per event, per route, or it is not a measure. The cities that gained 30% and the city that lost 4.5% needed different decisions, and the average would have given them the same one.
The same discipline applies to every figure a government is offered. A national impact total says nothing about a region, a district, or a hotel. The decision is always local, so the evidence has to be.
Averages are how a headline is written. Decisions are made on distributions. Any supplier who gives you one number for 16 cities has saved themselves work and cost you the information you were paying for.
The measurement is in the test was the World Cup. The page for hotels and accommodation sets out how the same measurement serves a hotel group directly.
When you receive an event impact figure, ask for it at the level you will make decisions: per hotel, per district, per route. If it only exists as a city or national average, it has not been measured yet.
None of the three states what it includes, what period it covers, or whether the money stayed. A footprint figure is only useful when it carries its own definitions, and it is only defensible when it is kept separate from impact.
A hotel pipeline decision is usually made on an occupancy trend, extrapolated forward. That answers a different question from the one a planner, a developer, or an investor faces: is the market short of rooms overall, or short of rooms on the 18 nights a year when demand peaks?
A new route photographs well on launch day. Whether it is still flying in year three, without deep discounting, and still worth the airport incentive that started it, is a quieter question, and it is the one that comes up at renewal.
International visitors to the United States fell 4.2% in 2025, the first annual decline since the pandemic, while worldwide travel grew. The spending did not disappear. It moved. Five economies are absorbing most of it, and the shift now looks structural.
A major event is a rallying point for civic pride, and just as easily a rallying point for protest. The difference is usually whether residents can see what the event does for their street, in terms they recognise, before the road closures start.
During the 2026 World Cup, host-city hotel rates rose 13% on average. Miami rose 30.3%. New York fell 4.5%. The average was true and useless: it would have told all 16 cities the same thing, and been wrong for most of them.
Get the latest from the Fabrik newsroom
