How much will the airfare for your next vacation cost?
Once upon a time, this question was relatively simple to answer. Airlines set prices for given routes in a particular cabin that didn’t change, regardless of when you booked your ticket. Today, however, nearly all airlines use dynamic pricing—that is, they rely on complex algorithms to set fares that fluctuate. And airlines aren’t alone. Rideshare services, retail outlets, and event ticketing platforms have all adopted dynamic pricing to efficiently meet the ebbs and flows of consumer demand—and, of course, turn a profit.
Much of the existing research on dynamic pricing in airlines and beyond has shown it to have positive effects on what economists call welfare, a combination of what’s good for consumers and what’s good for companies. Welfare is a measure of both companies’ revenues and consumer surplus—the difference between what consumers were willing to pay and had to.
But much of that research doesn’t consider competition, which may have significant effects on how prices rise and fall in the leadup to a flight—and, ultimately, on welfare. A new paper by Yale SOM’s Kevin Williams, Aniko Öry, and Jose M. Betancourt, along with co-author Ali Hortaçsu of the University of Chicago, explores that thorny issue.
The researchers’ theoretical models reveal that dynamic pricing systems that factor in competitor behavior could result in airlines selling too many tickets too soon. However, dynamic pricing systems that ignore competitors and instead rely on internally determined pricing rules can avoid that trap, benefiting consumers and companies alike.
Many factors influence dynamic pricing for flights, such as the popularity of a particular route and whether it’s mainly used for business or leisure. Another significant element is time: ticket prices on popular routes might start high (because consumers typically plan vacations far in advance), dip for several weeks or months, then rise again close to the flight (because last-minute bookers are generally less price-sensitive).
“As soon as you add competition, dynamic airline pricing becomes infinitely more complicated, because at any point in time, everybody’s reacting to each other.”
While other research on dynamic pricing focused on routes only served by one carrier, Williams and Öry modeled a market in which two airlines serve the same route and compete for consumers. Their model drew on pricing and booking data from a major US airline, as well as third-party data on its competitors, from the first nine months of 2019.
The model showed that dynamic pricing in competitive markets can produce peculiar and unexpected effects. “As soon as you add competition, the problem becomes infinitely more complicated,” Öry explains, “because at any point in time, everybody’s reacting to each other.”
The researchers identified one especially noteworthy danger of dynamic pricing systems that incorporate competitor behavior. Imagine Airline A and Airline B are competing for customers on a particular route several months before departure. To attract customers, Airline A drops its price just a little. Airline B’s pricing algorithm detects that change and, to regain the upper hand, drops its price a little. Over time, the two airlines bid each other down to such a point that both sell too many seats too early, leaving too few seats available for late-booking travelers. This hurts the airlines (who rely on the high prices last-minute customers are willing to pay) and customers (who need that next-day flight).
“This is a really important point because it shows competition can actually not be good,” Williams explains. Normally, economists expect competition to improve market efficiency; here, “it can lead to inefficient outcomes.”
However, in analyzing their real-world data, the researchers discovered that the airline they studied wasn’t falling victim to this blunder. While the carrier may be monitoring its rivals in other ways, its pricing algorithm did not consider competitor prices at all. Instead, the airline’s system relied on a series of internally determined rules to set its prices (for example, adjusting prices at different times for leisure versus business routes). The researchers found that this approach results in a 4-5% revenue increase and 3% consumer surplus increase when compared to their theoretical system factoring in competitors’ prices.
Williams and Öry emphasize that more work needs to be done to understand the full effects of dynamic pricing under competition, which can be complex and unexpected.
“Dynamic pricing is both theoretically and empirically challenging,” Williams says. “This research emphasizes how much more complicated things can be and demonstrates the range of possibilities that we might see in all the markets that dynamic pricing is actually used in. One of the things that happens a lot in economics is the answer is, it depends. I think this work really emphasizes that.”