Shoulder-season price cliff

Why does a room's price halve in the week either side of the peak when the town is still nearly full?

Move a holiday one week earlier and the same room, in the same hotel, with the same view, costs half as much. What is strange is what you find on arrival: the town is busy. Restaurants have queues, the car park fills, the hotel is nearly full. Demand has not halved — nothing like it.

So the obvious story, that price follows demand, does not fit. Demand fell a little and price a lot. Why should a small change in the week produce a large change in the rate?

Begin with a correction everything else depends on. Ask why the town is nearly full in the shoulder week. It is nearly full at half price. Would it have been at the peak price? Almost certainly not. Occupancy is not an independent fact sitting alongside the price; it is the result of it. So the question is not "why does price fall when demand hasn't" but "what sets the rate at which the town fills".

To answer that, stop thinking about the average room and think about the last one. A hotel deciding whether to release a cheap rate is not asking whether the rate covers its costs — almost any rate does, since putting a guest into an already-open hotel costs little. It asks: if I sell this room now at the low rate, what am I giving up? That quantity has a name in revenue management, the displacement cost or bid price, and it is the entire mechanism.

Peak week. The hotel forecasts more people wanting rooms than it has rooms, so every room will sell. Selling one now at a discount does not add a sale; it replaces one that would have happened at the full rate. The displacement cost is nearly the whole full rate, so the discount is worth nothing and cheap buckets stay shut. Price is then not set by cost at all but by rationing, rising until the number of people still willing equals the number of rooms.

Shoulder week. Now the forecast says some rooms will go unsold. What does the marginal room displace? Nothing: unsold, it is worth zero forever. Its displacement cost collapses to what leaving it empty would avoid — housekeeping, laundry, breakfast, utilities, channel commission — and any rate above that is worth taking. Every hotel in town runs the same arithmetic on the same night, so competition pushes the rate towards that floor.

Do you see where the cliff comes from? These are not two points on a smooth curve but two pricing regimes, and what decides which you are in is not "how much demand" but "is expected demand above or below capacity". That is a comparison with a threshold in it. Cross it and the displacement cost drops from nearly the full rate to nearly nothing in one step, taking the advertised price with it.

Two structural facts sharpen the step, both boring and both real. Hotels do not price continuously; they open and close discrete rate buckets, so when the forecast crosses the line a whole bucket opens at once and the cheapest rate falls in a jump even if willingness to pay moved smoothly. And demand itself has a step in it: school terms begin on a date, not on a gradient.

Two neighbours ask different questions. tourism-seasonality.md asks why the year has a peak and why the peak price must carry twelve months of fixed cost — the amplitude. hotel-price-discrimination.md asks why two guests on one night pay differently, which is fences sorting buyers — the spread. Neither explains why the change between adjacent weeks is a step rather than a ramp. The fixed point of difference is the threshold: the operative variable is not the level of demand but whether it sits above or below capacity.

A misconception worth naming is that the shoulder discount is a marketing sweetener to tempt people into quieter weeks. That reads the causation backwards. Nobody decided the shoulder week should be half price; the forecast crossed capacity, the displacement cost collapsed, and the rate fell out of the arithmetic.

Two honest limits. Forecasts are uncertain, so the switch smears over a few days rather than falling on one, and the cliff is sharpest where demand is most predictable. And this is the mechanism, not a measurement: how far the price falls depends on local avoidable cost and how many hotels compete, and I would not put a number on it.

The falsification test is sharp, though. This account says price tracks the forecast relative to capacity, in advance. So a week forecast full should stay dear even if it ends up half empty, and a week forecast quiet should be cheap from the start even if it later fills. If instead rates only fall after rooms have gone unsold — reactive discounting rather than forward displacement pricing — the mechanism is wrong. Likewise, a hotel that genuinely expects to sell out in the shoulder week and still halves its rate would refute it outright.

• Why a hotel with empty rooms will still refuse a short booking, and what displacement cost has to do with it.
• What a hotel does when the regime it priced for turns out to be the wrong one.

Every term the collection defines is gathered in the glossary.