Five US airline mergers, 7,936 routes where the merging carriers overlapped, and 350 million ticket records. Every prediction below is computed from information available before each deal closed, and scored against what fares actually did afterwards.
Effect on the average fare of an overlap route relative to matched control routes, by quarter from the merger closing. The quarter before closing is the omitted baseline. Grey where the 95% interval covers zero.
Nothing happens for four quarters. The single operating certificate — the point at which two airlines become one for regulatory and fare-setting purposes — lands four to six quarters after closing. A retrospective that stopped at one year would have found nothing.
Overlap routes sorted into deciles by each prediction, with the average realised fare effect in each decile. A tool that works should climb from left to right.
The increase in HHI is the screen the 2023 Merger Guidelines are built on. Across its whole range the realised effect is flat.
GUPPI — the value of sales diverted to the merger partner — needs demand estimation and is rarely the first thing computed. It sorts the routes almost perfectly.
Change in the realised fare effect for a one-standard-deviation increase in each prediction, with merger fixed effects and standard errors clustered on the route.
The same stacked design, one deal at a time, each against its own matched controls.
The Guidelines flag a merger when it raises HHI by more than 100 points in a concentrated market. This is what other thresholds would have done on this data.
Overlap route. A city pair where both merging carriers each held at least 1% of passengers in the four quarters before the merger was announced. Measured before the announcement, because carriers move capacity once a deal is public.
Control route. A route where exactly one of the two carriers was selling. Both groups get whatever the merger did to the combined airline's costs and network; only overlap routes lose a competitor. Controls are then matched to overlap routes on distance, traffic, concentration, fare, nonstop share and pre-merger fare trend.
HHI and its increase. The sum of squared market shares on a 0–10,000 scale. A merger of two firms raises it by twice the product of their shares. It needs nothing but quantities, which is its appeal and its limit.
GUPPI. The value of the sales a carrier diverts to its merger partner when it raises its fare, as a fraction of that fare. It requires a demand system, and the one used here is calibrated rather than freely estimated — see the write-up on demand estimation. Its ranking of routes is nearly invariant to that calibration; its level is not.
Realised effect. The change in a route's average real fare from the eight quarters before closing to quarters +5 through +12, minus the same change on its matched controls.