Economics on the Fly

The first study that documents who pays what for airline travel.

Hubs and Spokes

When the U.S. government deregulated the airlines in 1978, it opened a tumultuous existence for the industry and gave birth to a controversy about what's good and what isn't for American travelers. Three deregulation legacies fuel the debate: a confounding pricing system, hub-and-spoke route structure, and domination of some hub airports by one or two airlines. Critics charge that this environment is the perfect breeding ground for monopoly, inefficient, costly operations and overpriced fares. Some suggest the government should reenter the aeronautical fray to increase competition by limiting the number of gates a carrier can lease at one hub or the number of departures it can schedule.

According to University of California, Berkeley economist Pablo T. Spiller, however, what may look like a quest to monopolize an airport is a natural consequence of operating efficiently. "This system is working," says Spiller, "and consumers are benefiting." Using data from the U.S. Department of Transportation and heavy-duty number crunching on Pittsburgh Supercomputing Center's CRAY C90, Spiller and his colleagues -- Steven Berry of Yale University and graduate student Michael Carnall of the University of Illinois -- completed the first study to sort through ticket purchase behavior by modeling supply and demand, and the results document who pays what for airline travel.

With the freedom brought by deregulation, the airlines quickly embraced hub-and-spoke network structures. As the industry learned on a limited basis prior to 1978, when a direct-route system was the norm, funneling passengers through a central location, or hub, where they can pick up connecting flights, offers the most logical means of moving large numbers of people to many cities many times a day.

Concentration of staff and aircraft at a hub often results in a carrier offering more departures to more destinations than carriers that base their operations elsewhere. Critics view this as behavior calculated to eliminate competition, and they charge that hub carriers have unreasonably high operating costs. Unlike other analyses, however, Spiller's study differentiates costs from markup, and the results show that at a given airport, a hub carrier enjoys 15-20 percent operational savings per passenger over a non-hub carrier at the same site. As its proponents argue, therefore, deregulation has fostered efficiency. "The critics," says Spiller, "are suggesting that the government tinker with the very structure that is allowing these savings to occur."

All's Fare: Modeling Airline Pricing

Part of the debate, says Spiller, represents an unwillingness to view airline prices like other consumer products. You cannot look at a ticket price and say, that's exorbitant, anymore than you can make the same claim of an automobile sticker price without considering the model, make, options offered, and demand for it at a given time. "You have to take into account what that fare represents -- for instance, coveted departure and return times during peak flying periods, whether or not you book at the last moment, whether or not you'll earn frequent flyer miles."

With these factors in mind, the airlines compete by offering distinct products, presented to consumers as fares. Each fare represents a different market -- a route connecting two cities -- and departure time. Unlike prior analyses of the industry, Spiller's model captures this labyrinthine system and the buying behavior it breeds. It sorts customers into two groups, business and tourist, comparing the purchasing behavior of both in choosing myriad products within the same market.

Marginal Cost vs. Distance and Density (4th Qtr, 1985)
The researchers modeled the marginal cost of a leg -- a single takeoff to landing -- as a function of flight distance and density (the number of passengers flown over the leg in a quarter). Density is a factor because larger aircraft used on dense legs lower per passenger costs. Shaded areas show cost decreasing with increasing density. On very short and very long distances marginal cost rises as density increases. Large aircraft are inefficient for short, low-density legs (takeoffs and landings gobble up fuel). Very long legs are usually covered by large aircraft whenever density justifies a direct flight, so there is little opportunity to reduce costs on these routes.
Focusing on data from 1985-93, the model shows travelers choosing from among 230,000 combinations of itinerary, fare and carrier in as many as 17,000 markets. Conceptually, says Carnall, the model "takes every product and compares it against the other products in that market, and tries to figure out, given the choices made, what people value. And it looks at these thousands of markets separately in order to make that determination."

The first modeling effort crunched data from the fourth quarter of 1985. During those three months, according to Spiller, tourists using a dominant hub carrier paid anywhere from 1-5 percent above passengers whose flights were booked with non-hub carriers. Business travelers flying hub carriers, however, paid nearly 20 percent more than their counterparts using non-hub carriers. "The problem with previous studies," says Spiller, "is they implied that all travelers who used a hub carrier were paying considerably higher prices. And we find that only the business traveler is paying premiums."

Good News for Consumers

"As you develop a hub," says Spiller, "your products become more attractive -- more direct flights, more frequent flights, more connections -- and with that, you gain ability to mark up prices, because those are product qualities that, according to our data, customers are willing to pay for." If airlines are exploiting anything, says Spiller, it is "this peculiar demand for large networks that business travelers have." Consumers are doing well, says Carnall: "For a tourist, the cost of a flight has actually come down a bit." Business travelers are paying more, but are offered more frequent departures and other perks.

Thus far the model has examined individually the fourth quarters for 1985 through 1993. The next objective is to examine consecutive years simultaneously, so that the researchers can get an idea of how the 1990 recession and other economic milestones, such as numerous carrier bankruptcies, affected cost and demand.

The modeling imposes very large memory requirements. Estimation of the smallest quarter (in terms of data) required 96 megabytes of memory and the largest year required 26 megawords. Because no existing software would accommodate the study objectives, initial work involved testing and tweaking, a process, says Carnall, that would have taken years without a supercomputer. The C90 could do three days of workstation processing in an hour. "On the C90, we get results in as quickly as a day."

For complex social policy analysis such as this, says Spiller, supercomputing is essential. "For many issues, the problems we're dealing with are more complex, so you have to start using more sophisticated methods. The supercomputer is there to provide that type of service."

Researchers: Pablo T. Spiller, University of California, Berkeley..
Hardware: CRAY C90
Software: User-developed code.
Keywords: airlines, fares, markup, pricing system, monopoly, deregulation, U.S. Department of Transportation, purchasing behavior, hubs, hub-and-spoke, route structure, network structure.
Related Material on the Web:
Airline Deregulation: Changes in Airfares, Service, and Safety at Small, Medium-Sized, and Large Communities -- GAO Report
Domestic Aviation: Changes in Airfares, Service, and Safety Since Airline Deregulation -- GAO Report
Airlines Research: Recent work by Pablo T. Spiller
Law & Economics Consulting Group: Pablo T. Spiller
Steve Berry Home Page
Projects in Scientific Computing, PSC's annual research report.

References, Acknowledgements & Credits