Airlines serve individual city-pair markets by constructing a network of flights where what happens in one market necessarily affects what happens in another. These effects become particularly evident when bad weather strikes and poor flying conditions in one place lead to canceled flights elsewhere. Yet network effects are generally ignored by antitrust enforcement authorities when evaluating proposed airline mergers. In this article, I outline both the factors behind and some costs resulting from this amnesia.
Economic markets are traditionally defined by price effects, whether in terms of cross elasticities of demand and supply or by the pricing ability of a hypothetical monopolist who controls all of the market’s supply. What matters in both constructs is how buyers respond to increased prices and how firms set prices accordingly. By following this approach, one finds wide agreement that airline markets should be evaluated by conditions within particular city-pairs. That is where prices are set and where consumer welfare in terms of relative prices is measured.
Modern policy-makers have adopted this approach and acted accordingly. What matters to them are the anticipated effects of a particular merger on prices; other considerations, such as those pertaining to quality levels, are largely ignored. Included in this latter category are factors derived from network effects. That approach is justified by the presumption that profit-making firms will organize their networks as efficiently as possible, which in turn leads to optimal levels of consumer welfare.
Unfortunately that presumption does not always follow. As Michael Spence taught us nearly forty years ago, profit calculations are made at the decision-making margin and refer to the marginal consumer, while consumer welfare calculations depend on all consumers and must reflect average valuations. Spence developed that principle by analyzing firm decisions specifically on setting the quality levels for their products, and there is an interesting application of his principles to airline services. What I have in mind is the most essential dimension of airline quality, which is arriving as close as possible to scheduled times, in bad weather as in good.
Airlines are not helpless in the face of bad weather conditions. They can improve this dimension of quality by building sufficient capacity into their networks. But doing so is expensive and may not be apparent to the marginal consumer on whom airline decisions ultimately depend. In such cases, the marginal costs may exceed the marginal revenues associated with improving this form of airline product quality, and the associated capacity levels are not then created.
In such circumstances, when bad weather strikes and planes at certain locations are grounded, there are fewer alternatives in their network to take up the slack. And passengers can suffer grievously from the low quality services that long delayed flights portend. Typically, complaints by consumers to airline representatives are met with the expected response of “don’t blame us; it’s the weather.” But of course, that answer is only partially correct. Even if airlines are right that many consumers, particularly those at the margin, would be unwilling to pay for this dimension of improved quality, their decisions are ultimately founded on a profit-making rather than on a consumer welfare optimizing calculus.
So if consumers cannot rely on the supply decisions of the carriers, what or who can they rely on? The answer to that query is competition. However, it is competition between networks and not between airlines in individual city-pair markets. In the “hub and spoke” route structures employed by the legacy carriers, an important dimension of network competition has long been the presence of competing hubs. And when mergers occur, the acquired carrier’s hubs are frequently closed. Let me offer some examples: United Airlines once had hubs in Kansas City (from Eastern Airlines), Greensboro (from Continental Lite) and Orlando (from Eastern and the original United), but these are all gone. American Airlines has closed at least six hubs in the last twenty years, including St. Louis (from TWA), and San Jose and Reno (from Reno Air). And Delta Airlines has closed prior hubs at Memphis, Dallas-Fort Worth and Orlando along with Denver, once a hub of Western Airlines which was acquired in the 1980s. There are other examples as well. These actions were all taken with the announced goal of reducing costs, which essentially means that the marginal costs of operating hub airline services from those hubs were higher than the additional revenues they would provide to their new owners.
The effect of more hubs, and perhaps also more planes to go with them, would mean more alternatives for replacing grounded flights when bad weather strikes. With fewer airlines and fewer hubs, there is necessarily greater geographic concentration. What this means, to quote an airline analyst, is that “the flexibility is gone in awful weather.” With more airlines, there would be more hubs and more planes at more locations; and as a result, more flexibility in the face of bad weather. This most essential dimension of airline product quality would improve.
Whatever the effects of hub closures for airline costs, the relevant question for antitrust policy is not that but rather their prospective implications for consumer welfare as reflected in average rather than marginal valuations. The appropriate conclusion to be drawn from these considerations is that policy judgments should focus on airline networks along with city-pair markets, and not exclusively on the latter as they are now.
 See for example the statement before the US Senate of Joel Klein, former head of the Antitrust Divison of the US Department of Justice: “relevant airline markets are likely to consist of scheduled airline service between a point of origin and a point of destination, generally referred to as city-pairs.” Statement of July 27, 2000, p. 21.
 A.M. Spence, “Monopoly, Quality, and Regulation,” Bell Journal of Economics, Vol. 6, 1975, p. 417,