Air fares are collected automatically from travel portals every ten minutes, across a fixed basket of the fifteen busiest domestic city pairs and five booking lead times. For each route and lead time the cheapest available fare is recorded — the price a traveller could actually have transacted at. Those prices are converted to a single index using the geometric (Jevons) method, weighted by each route’s share of scheduled seats. The resulting movement is airfare inflation, which is the input a consumer price index requires for air travel.
Index by city pair and booking lead time, against the reference period. Red is dearer than the reference period, blue is cheaper. T+7 means booked seven days before departure (the T−7 bucket in the problem statement). A cell at 100 means the cheapest fare bucket has not moved — hover any cell for the fares behind it. Blank cells had too few observations to price reliably and are excluded rather than estimated.
City pairs whose index has moved beyond ±15% from the reference period. The threshold is fixed in advance so that flagging is not a judgement made after seeing the data.
Index movement over the collection period, and average fare by day of week. Day-of-week effects are a known driver of airfare seasonality; the pattern firms up as more weeks are collected.
Each carrier's cheapest fare in a cell (route × lead time × day) against the cheapest anyone offered in that cell, over every observation. Controlling for route is what makes carriers comparable — a raw mean fare mostly reflects which routes a carrier flies.
The index is machine-readable. Central statistical systems can ingest it directly rather than transcribing from this page.
X-API-Key header. Every response
carries the reference period and the method used, so an ingested figure can never
be separated from how it was produced.