by Robert Litan
Bold is certainly a word one associates with the late Julian Simon, who spent most of his academic career at the University of Maryland. If Simon were still alive, I’ll bet (and he was fond of bets) he wanted to be remembered for some of his other, controversial but far-sighted ideas, so I’ll give you a taste of those before returning to the story about his epiphany with applying auctions to the airline seat market.
Simon is likely best known for two mutually reinforcing propositions, and one related bet. The first proposition is that higher population growth is almost always good. More people means more brains devoted to solving the world’s problems, provided people are educated. The largest population growth in world history—the roughly six-fold increase between 1800 and today—coincided with the largest growth in per capita incomes and living standards, disproving any claims that population growth suppresses economic growth.
The second proposition is that the world has an inexhaustible supply of natural resources. One cannot simply extrapolate into the future the growth of current consumption of any commodity or seemingly scarce resource, such as oil, and then divide by the current estimate of the total worldwide reserves of that commodity or resource. As demand outruns current supplies, prices rise, which triggers additional investment to expand supplies, locate new reserves perhaps using new technologies (think of the recent shale oil and gas revolution enabled by the combination of hydraulic fracturing and horizontal drilling, discussed in Chapter 9), or to find alternatives. In addition, higher prices encourage consumers to reduce their demand, to become more efficient, and to use alternatives.
This basic economic principle, which economics students typically learn in any introductory course, was the basis for Simon’s classic rebuttal, The Ultimate Resource, to the now infamous Club of Rome study published by MIT scientists in the early 1970s that predicted the world would eventually run out of resources. The prediction was based on a simple extrapolation of past consumption trends indefinitely into the future.6
Simon didn’t merely write academic papers asserting these views; he put his money where his mouth (or pen) was. In 1980, he made a famous wager with Stanford environmental scientist Paul Ehrlich, who worried that the combination of rising population and resource scarcity would lead to soaring prices of basic commodities. Simon disagreed and made this bet: 10 years later, the prices of five metals would cost less than they were selling for in 1980; Ehrlich took the opposite position. Simon won.7
Simon’s urging of auctions as a way to clear a market with a fixed supply—in this case seats on a particular airplane—was related in a way to his skepticism about scarcity. Let prices do the work of allocating the seats even if the plane were oversold. Simon argued that by letting auctions clear the prices of full planes, airlines would work hard at booking more seats in advance, without fear of offending some ticketed passengers by not letting them on the plane, as was the custom before the auction practice became widespread.
Simon thought up his auction idea in the 1960s, when airline fares were strictly regulated, so at the same there was no way it could be implemented. Toward the end of the following decade, however, Simon took the notion to Alfred Kahn, who was then chairman of the CAB, which not only had long regulated airline fares but also approved which airlines could fly particular routes. You will learn a lot more about the remarkable Kahn in Chapter 9, which discusses the major business impact of his campaign to eliminate airline price and entry regulation and the CAB that oversaw it. Simon struck at just the right time, since Kahn and his colleagues on the board shared the predisposition of other economists who, except in rare cases of natural monopoly, argued that market forces should be allowed to determine prices rather than government regulators. Kahn and the rest of the CAB therefore were easily persuaded—though Kahn cleverly called the idea the volunteer bumping plan—and overbooked flights have never been the same.8 (The bumping plan did not differentiate seat prices by where they were located, which would have introduced an additional attribute and thus a source of service differentiation, into the auction. This may have been more theoretically pure, but in practice, the gate agents needed to fill planes quickly when they were overbooked so they could take off reasonably close to on time).
Unfortunately, some airlines did not let auctions really clear the airline seat market, but rather offered a take-it-or-leave-it deal, and thus some continued random bumping. Writing in 2010, the editors of the Wall Street Journal criticized the Department of Transportation, which had residual authority over the airline market after the CAB was abolished in 1978, for seeking to regulate what clearly should be an unregulated market. The Journal editors had the right answer for this problem: Get rid of the artificial rules, and go auction all the way.9
Airline seat auctions not only cleared seat markets where the airlines permitted them to do so, but they benefited the airlines and passengers in other not-so-obvious ways. Airlines gained from fuller flights, knowing that it was better to overbook and give a voucher to a few customers if they had to, in a way that made the customers happy. At the same time, fuller flights enabled the airlines to spread the fixed costs of flying their planes across more paying passengers, which allowed them to charge lower fares. One of Simon’s former colleagues writing in 2009 cited estimates that auctions of overbooked seats led to combined airline-passenger benefits of $100 billion over a 30-year period (a figure fully one-tenth of the trillion-dollar figure cited in the title of this book).10
Google and Online Ads
By far the most famous use of auctions in recent years is by Google for its online ads, which are a major source of that company’s revenues and profits. Two economists played an important role, one directly, the other as inspiration.
Google’s extraordinary rise in so short a time to become one of the most valued, admired, and feared (by some) companies in the world is now well known—or easily found out from the multiple books written about the firm.11 One of those authors, Steven Levy, also has written an excellent account of how the company stumbled onto auctions as a way of selling advertising.12
According to Levy, Google’s cofounders, Sergei Brin and Larry Page, wanted advertising to be one way of monetizing the value of the keywords that users type into their Internet search engine algorithm, but no more important than the revenues they anticipated from licensing the search technology and selling servers. As it turned out, of course, advertising has become the overwhelmingly dominant source of Google’s revenues and profits.13
Early in the company’s history, the firm sold two kinds of ads. The ads at the top of the right hand side of a web page were displayed once a search term was entered and were sold in the traditional way, through human sales representatives who sold keywords, such as perfume or shoes, to specific companies or their advertising agencies. Advertisers would pay based on the number of views of their ads regardless of how many times users clicked through them. The second kind of ad, listed lower down the right side of the page, was sold directly online at a fixed price.
Levy reports that the two individuals Brin and Page had put in charge of the company’s advertising efforts, Salar Kamangar and Eric Veach, were barely out of college, and had majored in biology and computer science, respectively. Although neither had a background in economics or business, they thought it made more sense to auction the online ads at the bottom of the page rather than have Google guess their value by setting a price, while retaining a human sales force to sell the higher valued ads at the top of the page. Under their new AdWords Select system, small- and medium-sized businesses—the main online advertisers at the bottom of the search results page—would be asked to submit sealed bids for specific keywords in advance. Each time users typed in those keywords, Google’s algorithms would determine almost instantaneously the winning bids, and rank them in order on the page.
The order of placement, by the way, introduced an element of uniqueness into each ad, which fit the general nature of auctions, which tend to be used for differentiated
goods or services. In Google’s case, the exploding volume of searches—now into the billions a day—has created a vast market for slightly differentiated ads.
Kamanger and Veach came up with a significant innovation, however, in designing their auction, inspired by an auction-based ad system then being used by GoTO, an early search engine competitor of Google’s. As Levy reports, and various economists have pointed out, even sealed bid auctions can be gamed in order to avoid the winner’s curse—a bid that is substantially higher than the second-highest bid so that the winner feels like he or she has overpaid (which is often the case). This can lead each bidder to bid low; if all bidders behave this way, the seller will end up receiving too low a price. To avoid this outcome, Veach thought up on his own—reportedly without reading the economic literature—an alternative to the highest bid auction. The winning bidder would pay one penny more than the second-highest bid, with the rest of the ad order determined by the ranking of the other bids. This way bidders wouldn’t have to worry about overbidding, since the second place bid in effect would act as a safety net.
Without knowing it, Veach’s second-price auction essentially replicated an innovation for which a Canadian economist, William Vickery, won the Nobel Prize in Economics in 1996. Vickery’s Nobel win was bittersweet because, at the age of 83, he died three days after being named. In doing so, he became the only Nobel winner to have never actually received the prize, but it may have been some consolation to Vickery (while he was alive) and to his family, that the second-price auction widely was called the Vickery auction before his death and since.
Until Vickery came along, most of those who thought about auctions, including Walras, simply assumed that they worked best when whatever was being put up for sale went to the highest bidder. But Vickery had a very different and counter-intuitive notion: that the highest bidder still wins the auction, but pays that second highest price bid, a result that Veach at Google had reached largely for programming reasons. Vickery mathematically proved that this result holds more generally, both for bids conducted by sealed bids (where the bids are secret) and an open outcry or public auction, where by definition, the winner pays a bit more than the second-place bidder (but most likely less than the winner would have been willing to bid).14
It is not clear what motivated Vickery to study the auction process or to become interested in the subject, but we can take an educated guess by looking at other subjects that occupied his attention. Among the many topics he covered in his research, both macroeconomic and microeconomic in nature, he was concerned throughout his career about institutions that established proper incentives, especially under conditions of scarcity. For example, he is widely regarded as one of the fathers of charging for congestion, an idea that has been tried in a few places and, as I predict in Chapter 14, will be more heavily used in the future. Charging cars or planes during more congested times of the day is just another price-based way of allocating a fixed resource, much like using an auction to set a price for a unique item.
Shortly after agreeing to take the chief executive’s job at Google, Eric Schmidt persuaded Hal Varian, one of the profession’s leading experts on the Internet economy, to begin part-time consulting work for the company (the deal was proposed and consummated at a party in New York City in January 2002). Varian studied auctions and in his own work built on Vickery’s; see the following box for Varian’s background.
Hal Varian
Hal Varian’s migration to Google is an all-American success story. Having grown up on an orchard in Ohio (though one of his ancestors was the mayor of New York), he clearly had broader horizons for his future.
They developed through his schooling, and in particular, like another famous economist, Paul Krugman, the noted New York Times columnist and Nobel Prize winning economist, Varian was entranced with Isaac Asimov’s Foundation Trilogy (I was, too, but Varian and Krugman took their enchantment to much greater heights). Varian was especially struck by Asimov’s theme that society could be structured through mathematical relationships. When Varian went to MIT as an undergraduate, he took courses in psychology and political science before ending up with economics, which he realized was the subject most closely aligned with Asimov’s vision: It purported to model people’s behavior mathematically in a way that none of the other disciplines did.
Varian went from MIT to Berkeley, where he earned his MA in mathematics and then his PhD in economics. He has had an unconventional, and highly productive, life ever since, working his way through the academic ranks at MIT, Stanford, Oxford, the University of Michigan, and then back to Berkeley, where he became the first dean of Berkeley’s School of Information Sciences. In the process, he turned what had been a modestly boring field of library science into the study of managing large bodies of data, what the world now knows as Big Data.
Since the mid-2000s, he is most widely known as a highly successful corporate economist who made the transition from academic life to being an entrepreneur with ease, first as a consultant to and later as chief economist for Google. Had Varian gone to college today, he might have been a neuroscientist or perhaps a computational geneticist, fields that today also employ highly mathematical techniques to model human behavior. Google and his fellow economists are glad Varian grew up in an earlier era and became an economist instead.
When Varian asked Schmidt what he wanted advice about, Schmidt suggested that Varian begin by taking a look at the auction process that Kamanger and Veach had designed. Levy reports, and Varian confirmed to me, that after studying Veach’s auction process, he told Schmidt that Google had designed its auction perfectly. Later, Varian also wrote that Google’s adoption of what is also known as a second-price auction had “nothing to do with Vickery auctions: It was primarily an engineering choice design.”15 Varian’s response nonetheless made it easier for Google’s managers to take the next logical step: converting all of Google’s ads to an auction-based system and, in the process, dispensing with any human salespeople. This seemed very risky at the time because no other business sold its ads this way, nor were advertisers accustomed to such a system. Google had to persuade—actually teach—them how to use an auction. Many resisted at first, but eventually, the ad auctions became highly successful and the main source of the company’s revenue.
Google adopted several important adjustments to its auction process to ensure that ads are targeted to the right audiences. One amendment is to add a quality adjustment score, so when a user types in the words “Oklahoma City restaurants” the quality adjustment process, which is part of Google’s secret sauce, assigns the highest scores to all restaurant ads that include the words “Oklahoma City.” This modification weeds out restaurants from all locations outside Oklahoma City. Knowing that Google will do this encourages all restaurants to bid more than they might otherwise, which maximizes ad revenue for Google not only through the higher prices advertisers are willing to pay, but through the greater number of clicks on the ads, which trigger payments by advertisers to Google.16
Google’s auction also differs from the standard Vickery auction, which is designed only for one winner. Instead, Google auctions off positions, with a first place winner, second place right below, third place below that, and so on. Accordingly, while the company goes with the second-highest price for the winning auction, the other lesser prices are charged, in order, to those bidders who win those lesser positions. Google constantly is fine-tuning its auction procedures, just as they fine-tune its search algorithm, using simulators to quickly get to the core of a problem rather than a series of A/B tests (discussed in Chapter 6) that can be quite expensive.
An entire industry has grown around Google’s auction process, with advisers charging advertisers on how best to respond to Google’s auctions. Varian has done his best to help advertisers participate in Google’s auctions by explaining the company’s auction procedures and how to participate in the auctions in a highly viewed YouTube video (Google bought YouTube in 2006), which I highly recommen
d for readers interested in the subject. One important takeaway from the video lecture cannot be overemphasized: the importance of concentrating on the incremental cost of the ads and the value an advertiser expects to receive from them rather than the absolute value.
Remember from Chapter 2 that prices are set by the last incremental purchaser, not by all those who have gone before. This simple insight is critical to advertisers bidding for positions on Google and for the company’s own marketing efforts. For advertisers, the amount they should bid for any position should reflect the incremental revenue they expect to gain from any additional clicks for which they have to pay Google, not the total revenue for that position. To bid any more puts advertisers in a position where what they have paid is exceeded by any additional revenue they may gain from being a slot or two higher in the advertising pecking order.
By the same reasoning, Varian explains, Google’s marketing efforts to interest more advertisers concentrate not on the already thick markets for most popular words, where there are many advertisers and the spread between their bids is narrow, but rather in thin markets for less-popular words, where the additional bids, if calculated according to the incremental value principle, are likely to yield the company more revenue. Admittedly, this type of thinking may run counter to some business instincts, which are to double down on customer segments that are already producing much revenue. The more sensible course, as a general rule, is to focus marketing in segments that are less popular but where the upside gain, compared to the incremental costs of marketing, is greater. Because it is not always easy to make these calculations, which require data and the analytical power to interpret them, Varian has dubbed “marketing as the new finance.”
