Is air traffic a demand signal or misleading noise? It's both. The skill lies in telling apart how much of each.
One of the most tempting ideas for a hotelier is this: "If I count the planes arriving in my city, I can know in advance how full my rooms will be tomorrow." It sounds logical — after all, someone has to reach the city before they can stay in a hotel. But this simple intuition is both correct and dangerously incomplete.
This article treats three questions separately: (1) Are air traffic and hotel occupancy genuinely related? (2) Is it a direct causal link or an indirect shadow? (3) How much of this signal is misleading?
1. The link is real — and measurable
The existence of the relationship is not in dispute. Both academic literature and industry data point the same way.
The academic side: In a study by Harvard and MIT researchers Farronato and Fradkin, published in the American Economic Review, the number of passengers arriving in a city is used directly as an instrumental variable for accommodation demand. In other words, when econometricians look for a signal strong and exogenous enough to stand in for demand, one of the first things they reach for is the number of arriving (not returning, but arriving) air passengers. That tells us the relationship is not merely intuitive but statistically robust.
The industry side: Aviation data provider OAG positions the flight schedules and capacity of 900-plus carriers as a "forward-looking demand indicator," noting that new route launches and capacity increases directly grow hospitality demand. Lighthouse (formerly OTA Insight) reported measuring a positive correlation between flight search volume and hotel occupancy in its own internal research. They even quantified the timing: in an August 2022 Singapore example, the demand signal could be seen ~145 days earlier in flight searches and ~80 days earlier in hotel searches.
So air traffic is a leading indicator. It can flag that demand is forming before hotel reservations have even begun to arrive.
2. Not a direct effect — an "access" signal
Here comes the most critical conceptual distinction. Flight capacity does not create hotel occupancy; it merely makes it possible.
A seat is not a demand — it is an opportunity for access. The flight schedule says: "X people can come to this city this week." Hotel occupancy measures: "How many of those people came, stayed, and stayed in a hotel." Every "and" in that chain is a point of leakage.
That's why air data, at best, draws the upper bound of the supply/access side. The actual demand — the real number of people who will fill your rooms — is a subset that sits below this ceiling, often far below. Hospitality.today's 2026 review emphasizes exactly this: flight search is the earliest signal and captures intent — but it does not guarantee that hotel bookings will follow.
3. The misleading share: six factors that break the link
Here are the main leaks that wedge themselves between the "plane" and the "room," explaining why the signal is unreliable on its own:
a) Transit / connecting passengers. The biggest distorter. Not everyone landing at an airport stays in that city; most simply change planes. Istanbul Airport's hard data is striking here: of the 84.4 million passengers it handled in 2025, roughly 48% were transfer (connecting) passengers, and only 52% were true origin–destination travelers. So nearly half of the airport's traffic boards another plane without spending a single night in the city. Anyone mistaking raw passenger counts for demand sees almost double the reality.
b) Load factor. Capacity ≠ passengers. A plane having 200 seats does not mean 200 people arrived. In Heathrow data the average load factor is ~0.79 — meaning about one in five seats flies empty. Computing from "scheduled seats" overstates actual arrivals.
c) Average length of stay. The same number of passengers can produce vastly different "room-nights." In Istanbul the average stay is ~2.2 days; at a resort destination it might be 7. Even with passenger numbers held constant, a change in length of stay completely changes the number of nights a hotel is filled. Air data cannot see this multiplier.
d) Purpose and type of stay. Not everyone arriving is a hotel customer: business travelers who finish their day and fly back, people visiting friends and relatives who stay in homes (the VFR segment), second-home owners, residents returning to their own city... all of these are counted as "arriving passengers" yet contribute nothing to hotel occupancy.
e) Non-hotel supply — Airbnb and short-term rentals. Even someone who arrives and genuinely stays overnight may not fill your hotel; part of the demand shifts to short-term rentals. Research shows Airbnb affects hotel price (ADR) more than occupancy — that is, the demand "exists," but it can erode the hotel's share of it.
f) Origin/nationality mix. The same total passenger count produces very different hotel behavior depending on the nationality breakdown. A thousand people from a high-spend, long-stay market are not the same, for a hotel, as a thousand from a day-trip- or transit-heavy market.
Practical takeaway: how to use the signal
Throwing away air traffic data would be wrong — but treating it as the only truth is even more wrong. The right use looks like this:
- Use it for direction, not for level. "Capacity next month is up 15% year over year" is a valuable directional signal. But reading it directly as "I'll be 15% fuller" is a mistake.
- Use it as a leading indicator, alongside OTB. Air data warns you before demand turns into bookings; actual on-the-books (OTB) data measures reality. Using both together strikes the "early warning + real measurement" balance.
- Adjust for transit share and origin market. Instead of raw airport passenger counts, use origin–destination (O&D) passengers where possible, and origin-market breakdowns where possible. At major connecting hubs like Istanbul, this adjustment changes everything.
- Don't conflate what you measure with what you infer. Air data is a measurement (how many seats, how many passengers). "Therefore my hotel will fill up" is an inference. The common industry error is presenting the measurement as if it were the inference.
Industry data also reminds us: a 10% improvement in forecast accuracy can translate to roughly a 3% revenue increase. So there is a real financial payoff to calibrating the signal correctly. But that gain goes to those who use the signal not raw, but cleaned of its misleading share.
In short: There is a real, measurable, and early relationship between flights landing in a city and hotel occupancy — but it is not direct causation; it is an access signal full of leaks. The plane tells you how wide the door has opened; not how many people walked in and stayed. A good revenue manager knows this difference and uses the signal not in its raw form, but cleaned of transit, load factor, length of stay, and non-hotel supply.
Sources
- Farronato, C. & Fradkin, A. (2022). The Welfare Effects of Peer Entry: The Case of Airbnb and the Accommodation Industry. American Economic Review, 112(6). (Arriving air passengers as an instrumental variable for demand)
- OAG — Sky to Suite: Leveraging Aviation Data to Maximize Hospitality Growth (2024)
- Lighthouse / OTA Insight — Forecasting Hotel Demand: Is Flight and Hotel Search Data a Reliable Indicator of Market Demand? (2022)
- Hospitality.today — Forward-looking Data for Hotel Forecasting: What's Available in 2026 (2026)
- Harvard Business School — Forecasting Airport Transfer Passenger Flow (load factor and Heathrow data)
- ScienceDirect — Revisiting Airbnb's disruption of hotels (2025) and Investigating the whole picture: Airbnb supply vs hotel performance (Airbnb's ADR vs OCC effect)
- DHMİ / İGA / Anadolu Agency — Istanbul Airport 2025 figures (84.4M passengers, 48% transfer / 52% O&D)
- Republic of Türkiye Ministry of Culture and Tourism — Istanbul Tourism Statistics Report (average length of stay ~2.2 days)