Across 1,588 matched rate pairs at 151 US hotels in 12 cities, priced on 25 August 2026, the refundable rate cost a median +10.1% more than the identical non-refundable room (95% confidence interval 8.4% to 11.2%). In 15% of those pairs the refundable rate was cheaper, and in a further 2% the two were priced identically.
- Median premium
- +10.1%
- Flexibility was free
- 17%
- Matched pairs
- 1,588
- Rates checked
- 25 August 2026
95% CI 8.4% to 11.2%
of pairs — cheaper or same price
151 hotels · 10 cities
refreshed weekly
What was already known
Search for whether refundable hotel rates cost more and you get two very different classes of answer. Booking-industry blogs and revenue-management vendors offer ranges — 5–15% here, 10–20% there — with no sample size, no date and no method. Treat those as folklore.
Then there is one serious piece of work. Lacetera, Piga and Zirulia, writing in the Journal of Economics & Management Strategy, tracked around 25 million hotel room postings across cities in France, the UK, the US and Canada. They found that the gap between refundable and non-refundable prices does not shrink toward zero as the risk of cancellation falls, as you would expect if it were priced as insurance. It stays put, at roughly 10–15% of the price.
That is the prior work, and our headline is consistent with it — +10.1% sits just under the bottom of their range. What their data cannot tell a traveller is any of the things you would need at the moment of booking: it is not US city-level, it is not current, and it does not report how often the refundable rate turns out to be the cheaper of the two.
That last one is the finding that changes what you do. So that is what we measured.
The refundable rate is often the cheap one
In 17% of matched pairs — roughly one booking in six — booking with free cancellation cost the same or less than booking without it. Same hotel, same room, same dates, same board, same currency. The only difference was that one could be cancelled.
That is not evenly spread. In Honolulu flexibility was free on 42% of pairs. In Denver it was 8%. Same country, same week, same method.
It is frequent enough to be worth one extra click on every booking.
The average premium is real, and duller
Taken across every pair, flexibility does carry a price: a median +10.1%, with a 95% interval of 8.4% to 11.2%. That is close enough to what the industry has always claimed that the honest summary is: the most-repeated figure turns out to be roughly right.
In money, on the 2 adults, 2 nights stays in our sample: a median non-refundable total of $449 against $488 refundable — about $41 to keep the booking cancellable.
Roughly right is not the same as known, and an average hides the cases underneath it.
Where you go matters more than when you book
The spread between cities is wider than the spread between booking windows: 2.8 times between Nashville at +13.6% and New York at +4.9%.
How much and how often are not the same question, and the table separates them. A city can carry a low median premium and still rarely invert, or a high one and invert often. If you are choosing where flexibility is worth buying, the premium column tells you the price and the free column tells you the odds of not paying it at all.
Read the interval column before drawing conclusions from any single row. City medians rest on 11–19 hotels each, so several are wide, and 2 of the 10 span zero — meaning we cannot say from this scan whether flexibility costs anything at all there. The overall figure is on much firmer ground than any one city.
2 of the 12 cities we scanned are missing from the table. They cleared the pair count but not the 10-hotel floor, so their medians could not carry an interval and we do not publish a number we cannot bound. They are in the downloadable data.
| City | Median premium | 95% CI | Flexibility free | Hotels | Pairs |
|---|---|---|---|---|---|
| Nashville | +13.6% | 10.8% to 24.5% | 12% | 11 | 132 |
| Denver | +13.5% | 11.6% to 14.6% | 8% | 11 | 125 |
| Orlando | +12.9% | 11.5% to 15.8% | 12% | 15 | 178 |
| Miami | +10.2% | 6.5% to 13.5% | 15% | 13 | 188 |
| New Orleans | +10.2% | 6.1% to 11.8% | 16% | 11 | 108 |
| Phoenix | +8.9% | 3.3% to 14.1% | 10% | 13 | 153 |
| Chicago | +8.5% | 2.9% to 14.6% | 13% | 19 | 149 |
| Las Vegas | +6.7% | -1.2% to 11.2% | 29% | 12 | 151 |
| Honolulu | +5.8% | -17.9% to 10.6% | 42% | 11 | 116 |
| New York | +4.9% | 0.5% to 10.9% | 13% | 17 | 163 |
Booking early does not buy you cheaper flexibility
Standard advice says book early. For flexibility specifically the data leans the other way: the premium was at its lowest 60 days out (+9.0%) and at its highest 180 days ahead (+11.2%).
Taken one window at a time that gap is not much to lean on — it is about 2.2 percentage points and the intervals around individual windows overlap. What is harder to dismiss is the step: every window from 90 days out carries a higher premium than every window closer in, with no exceptions.
Either way it should break a tie rather than drive a decision — the city you are going to moves the number several times as much. And 180 days is the furthest ahead we priced, so none of this extends to booking a year out.
| Booked ahead | Median premium | 95% CI | Pairs |
|---|---|---|---|
| 7 days | +9.4% | 7.4% to 10.7% | 279 |
| 14 days | +10.1% | 8.9% to 11.9% | 264 |
| 30 days | +9.5% | 6.6% to 11.1% | 199 |
| 45 days | +9.2% | 6.0% to 11.2% | 161 |
| 60 days | +9.0% | 6.5% to 11.0% | 170 |
| 90 days | +11.0% | 9.2% to 12.9% | 183 |
| 120 days | +11.0% | 9.4% to 13.1% | 181 |
| 180 days | +11.2% | 6.7% to 13.4% | 151 |
Download the full dataset (JSON)
Why a refundable rate is ever cheaper
It looks like a pricing error. It is a supply chain.
A results page aggregates rates from different sources. A hotel sells inventory direct, to wholesalers, and onward again, and each link sets its own price and its own cancellation terms at its own moment. Price and refundability are decided by different parties at different times, so they come apart. The cheapest rate on a page is not systematically the most restrictive one.
There is a second reason the option is cheap to grant, though it is narrower than it first appears. Research in the International Journal of Hospitality Management, covering 2.2 million reservations across 628 Portuguese properties between 2022 and 2024, found that cancel-and-rebook — cancelling a booking to rebook the same stay at a lower rate — accounts for 0.94% of reservations. Ordinary cancellations are far more common than that, but they mostly free up a room the hotel can resell. It is the arbitrage that costs a hotel money, and the arbitrage is under one booking in a hundred.
How we measured it
The obvious approach is wrong. Comparing a hotel's cheapest refundable rate against its cheapest non-refundable rate usually compares a refundable standard room against a non-refundable suite, and calls the difference "flexibility".
Instead we only counted matched pairs: the same room, at the same hotel, for the same dates, with the same board type, offered both ways. The cancellation terms are the only thing that differs. Anything that failed that test was discarded.
Five rules did the rest of the work. Each one changed the answer:
- One observation per hotel, per stay
- Suppliers list one underlying rate pair under several room-name variants. Bellagio returned "Premier: King Bed", "Spa, Premier, King Bed" and two more, all at exactly the same two prices. Counted separately, that is four observations of one fact — and it over-weights whichever hotels list the most variants.
- Median, not average
- The distribution has a long tail on the cheap side. The average moves with a handful of outliers; the median describes what you would actually meet.
- The confidence interval resamples hotels, not observations
- The 1,588 pairs come from only 151 hotels — up to 24 apiece, across every booking window and stay shape. A hotel's gap between refundable and non-refundable is a pricing policy, near-identical across all of them, so those are not independent draws. Resampling pairs would report an interval roughly half as wide as the data supports. Ours resamples whole hotels — 151 independent units, which is the conservative choice. The premium survives it. A second decimal place would not.
- A floor before we report a breakdown
- Before we de-duplicated, Las Vegas came out at −5.9% on one sample and +5.8% on a larger one. Both passed the matched-pair test; neither had the room-name variants collapsed. De-duplication and sample size together decided the sign — which is why cities under 40 observations, and booking windows under 50, are collected but not published. A pair count was the only floor at first, which repeats the mistake the rule above fixes, so a city now also needs 10 distinct hotels: if its median cannot carry an interval, it does not ship.
- Anything past ±60% is thrown out
- Beyond that the two rates are almost certainly different products sharing a name. It removed 8 observations on this run — small enough not to matter, large enough to distort a mean if left in.
One wholesale supplier's live inventory (the same supply Travorro books), US properties rated 3 stars and above with at least 200 reviews, up to 30 hotels per city. That is a real and specific slice of the market — mid-range to upscale, branded, well-reviewed — and it is not an independent audit of the whole industry. A hotel offering only refundable rates, or only non-refundable, contributes nothing to it.
We also have an obvious interest in the conclusion: we sell hotel bookings, and "check both rates" is an argument for using a booking site. That is exactly why the method is on this page and the raw snapshot is downloadable above, rather than being a one-off you have to take on trust.
Coverage: 151 hotels across 12 cities, 8 booking windows from 7 days to 180 days, and 3 stay shapes. Scanned 25 August 2026.
The scan reruns weekly, and this page refuses to display a snapshot more than three weeks old rather than show you a number that has stopped being true. If you can break the method, we want to hear it.
What this does not measure
This is a measurement of price, not of promise. We compared what the two rate types cost. We did not test whether a rate labelled "free cancellation" actually returns your money.
That distinction matters, because the non-profit Consumers' Checkbook collected close to 2,000 rates across 25 travel sites in 2023 and found the label is often unreliable — many hotels refunding only part of a prepayment, and at least one "Flexible Cancellation" rate whose deadline had already passed at the moment it was offered.
So a cheap refundable rate is not automatically a good one. Read the deadline and the refund amount, which are two separate things: the deadline says when you lose the right to cancel, the refund amount says how much you get back once you do.
Scope, plainly: 151 hotels across 12 US cities, 1,588 matched pairs, USD, scanned 25 August 2026. One market, one week. It does not describe Europe, it does not describe hostels, apartments or independents outside that supply, and it is not a forecast.
The label is barely policed — with one exception
Most enforcement in this industry has been aimed at how prices are displayed rather than at whether a cancellation promise is kept. The Texas Attorney General reached a $9.5m settlement with Booking Holdings in August 2025 over fees hidden until checkout — the fifth in a series of state settlements, with a sixth against Hyatt following that December. The UK's competition regulator extracted undertakings from six booking sites in 2019, covering pressure selling and discount claims as well as hidden charges. Australia's competition regulator won a A$44.7m penalty against trivago in 2022, after a 2020 finding that its room-rate rankings misled consumers.
There is one regulator that has gone at the cancellation label directly, and it went at exactly the thing this article measures. In April 2020 the Hungarian Competition Authority fined Booking.com HUF 2.5bn, finding among other things that advertising accommodation with "free cancellation" was misleading — because the option was time-limited at many properties and because the rates carrying it were priced above the same accommodation booked without it. In July 2024 it fined the company a further HUF 382.5m for continuing to use the claim.
That is one authority, in one member state. Everywhere else the hook is thinner than it looks. The FTC's fee rule, in force since May 2025, makes it unlawful to misrepresent "the nature, purpose, amount, or refundability of any fee or charge" — but the Commission dropped the proposed requirement to affirmatively disclose refundability, and nothing in the rule requires anyone to disclose a cancellation policy at all.
What to do with this
- Check both rates every time
- Around one booking in six, flexibility costs nothing — 42% of the time in Honolulu, against 8% in Denver. Where you are going shifts the odds, but nothing tells you about the specific booking in front of you, so the only way to catch it is to look at both.
- The premium pays for itself if you are more than about 9.2% likely to cancel
- At +10.1% on top, a refundable rate is the better expected-value bet once your chance of cancelling exceeds roughly one trip in eleven. A leisure trip with fixed dates and paid flights sits well under that. A work trip waiting on someone else's calendar does not, and a stay booked six months out is a longer window for life to happen in. That, not the percentage, is the number to argue with yourself about.
- Buy flexibility in the middle distance, if it is close
- The premium bottomed out 60 days before check-in, and every horizon from 90 days out was dearer than every one closer in. It is a tiebreaker, not a strategy — booking further ahead simply stops helping.
- Compare the premium to the whole amount at risk
- A tenth on top to make a stay cancellable is a different proposition on a two-night city break than on a two-week trip. The premium scales with the booking; so does what you lose.
- Read the deadline, not just the label
- "Free cancellation" describes a window, not a guarantee. Check the date you stop being able to cancel and what you get back after it — and if you booked through a third party, check who actually holds your money.
Travorro shows the refundable and non-refundable price for the same room side by side, with the total including taxes and fees — so the comparison in this article is one you can make on every search.
Compare hotel ratesFrequently asked questions
- Are hotels with free cancellation more expensive than non-refundable?
- On average yes, but far less reliably than people assume. Across 1,588 matched rate pairs at 151 US hotels in 25 August 2026, the refundable rate carried a median premium of +10.1% for the identical room on identical dates. In 17% of those pairs, though, the refundable rate cost the same or less.
- How much more does free cancellation cost?
- A median of +10.1% (95% confidence interval 8.4% to 11.2%) in our 25 August 2026 scan of 151 US hotels, or about $41 on a typical 2 adults, 2 nights stay. It varies far more by city than by booking date — 2.8 times between Nashville (+13.6%) and New York (+4.9%).
- Is free cancellation worth paying extra for?
- It depends on how likely you are to cancel, and the break-even is calculable. At a median premium of +10.1%, paying for flexibility is the better expected-value choice once your chance of cancelling exceeds about 9.2% — because the premium buys back the entire stay, not just the difference. Below that, the non-refundable rate wins on average.
- Which US cities have the biggest gap between refundable and non-refundable rates?
- In our 25 August 2026 scan the widest gaps were Nashville (+13.6%), Denver (+13.5%), Orlando (+12.9%). The narrowest were New York (+4.9%), Honolulu (+5.8%), Las Vegas (+6.7%). Reversals cluster differently again: in Honolulu the refundable rate cost the same or less on 42% of pairs.
- Does booking early make refundable rates cheaper?
- No — if anything the reverse. The premium was lowest 60 days before check-in (+9.0%) and highest 180 days out (+11.2%), and every horizon from 90 days out was dearer than every one closer in. The effect is small enough to be a tiebreaker rather than a strategy, and we priced up to 180 days ahead, so it does not extend to booking a year out.
- Can I book a free-cancellation rate and rebook cheaper later?
- Some people do, but far fewer than you would guess. Research covering 2.2 million reservations across 628 properties found that cancel-and-rebook — cancelling in order to rebook the same stay at a lower rate — accounts for 0.94% of reservations. It also assumes the cheaper rate is still there when you look, and that you cancel before the deadline. Treat the refundable rate as insurance you are unlikely to claim on, not as a strategy.
- Does free cancellation always mean a full refund?
- No. The label describes a window, not a guarantee. A 2023 study by the non-profit Consumers' Checkbook found many rates advertised as flexible refunded only part of a prepayment. The cancellation deadline and the refund amount are separate terms and both need reading.
- Why is a refundable rate sometimes cheaper than a non-refundable one?
- Because a booking page aggregates rates from several supply chains, and each link sets its price and its cancellation terms independently — so the two come apart. The cheapest rate on a page is not systematically the most restrictive one.
Sources
- Lacetera, Piga & Zirulia, “Sticky price for declining risk? Business strategies with ‘behavioral’ customers in the hotel industry”, Journal of Economics & Management Strategy 34(1), 2025
- Kmiecik & Antonio, “Cancel, rebook, save: Revenue leakage from price cuts in hotels”, International Journal of Hospitality Management 135, 2026
- Consumers’ Checkbook, “Hotels’ ‘Refundable’ and ‘Flexible’ Rates Often Not What They Seem”, May 2023
- Hungarian Competition Authority (GVH), decision fining Booking.com HUF 2.5bn, April 2020
- Hungarian Competition Authority (GVH), further fine of HUF 382.5m for continued use of the claim, July 2024
- US Federal Trade Commission, Rule on Unfair or Deceptive Fees (16 CFR Part 464), effective 12 May 2025
- Texas Attorney General, $9.5m settlement with Booking Holdings over undisclosed fees, August 2025
- UK Competition and Markets Authority, undertakings from six hotel booking sites, February 2019