In the decade since deregulation, the route structure of the airline industry has changed dramatically. The economies of scope that can be captured by using the hub-and-spoke route network have led to the development of many new hubs and to the expansion of others. This transformation has had profound effects on competition and market structure in the industry. City-pair markets with a hub at one end point are less contestable than other markets [5] and, as a result, tend to have higher fares [3; 4]. The scope economies from operating a hub at a given city result in nonstop service being offered to that city from more points than would be possible without a hub [6; 11; 13; 16]. These scope economies also increase load factors and thus reduce average costs [1]. The location of hubs is critical to the performance of the industry. Given the powerful scope economies associated with hub operation, alternative hubs may be the only sources of competition on many low-density routes. And, as mergers increase the intra-airport concentration at existing hubs, the location of alternative hubs grows in importance. Even for routes that are currently served by only one carrier, the existence of a competitor's well-placed hub creates potential competition which reduces fares [4]. In this paper, we explore the determinants of airline hub location with the goals of finding a coherent explanation for the current pattern of hubs and identifying potential sites for future hubs. Though there is a vast literature on the hub-and-spoke route system, hub location has remained a virtually neglected topic. And very little of the literature that exists addresses the pattern of hub location. Schwieterman [18], for example, sought to explain the quantity of nonstop service (as opposed to connecting service) in the sixty largest population centers in the U.S. He found that the most important determinants of nonstop service were market size, length of haul and proximity to hubs. Jeng [12] studied the impact of market size, network size and number of cities on city-pair routings in a single-hub network, modeling which pairs would be served nonstop and which by connection. Bauer [2] modeled the decision to place a hub in a given city as a function of demographic and economic variables. Multicollinearity made interpretation of the individual variables difficult, but the model correctly categorized 87 percent of the 115 cities in his sample. Like Bauer, we seek the determinants of the pattern of hubs which has evolved in the deregulated airline industry. Our model, however, is richer, and our sample is considerably larger. We