«Information, Privacy, and the Internet: An Economic Perspective Susan Athey Stanford Graduate School of Business1 Contents Introduction 5 1. ...»
There is a continuum of outcomes as to whether advertising is a value-enhancing part of the user experience, essentially a complementary form of content (search ads, classifieds, sponsored links to news, sponsored posts in Twitter) or whether advertising is purely annoying, distracting, misleading, or even dangerous (e.g.
distributing malware; see Edelman (2011) for a number of realworld examples). In some cases, such as product listing services or shopping sites, all content is a form of advertising. Though some business models of advertisers and publishers are based on trickery and misleading consumers (see Edelman (2011), many websites have the incentive to provide ads relevant to the consumer’s intent, as these monetize better and create a better user experience.
The general industry trend has been towards making advertising more relevant to the user’s context as well to the user’s interests (independent of context). Industry reports suggest that such changes often increase the value created for both users and advertisers by an order of magnitude, and an ever-increasing proportion of advertising on the web utilizes some sort of user data or segmentation. There are several important categories for how
user and contextual data can be used:
Getting users information that they will value: both tailoring the ad and targeting the user Tailoring the ad to the user’s context, where users prefer different types of ads when they are doing research, consuming entertainment, or browsing social media Publishers tailoring their own content as well as advertising to the user. For example, news websites may rank articles differently based on user history. They may
also use browsing information to target ads out of context:
for example, a general news page might serve different ads after you visit the finance or technology page Observe the differences in outcomes that you might expect if “cookies” can be used only within a website (first-party cookies), or if they can be used across websites (third-party cookies). If the same publisher owns a wide range of content, that publisher can use information from a consumer’s web browsing on commercial parts of the site (e.g. the autos or finance or technology pages) to place more efficient advertising on less commercial parts of the site (hard news). On the other hand, if a website focuses only on hard news and does not have access to other information about the consumer’s web browsing, it will be at a distinct disadvantage with regard to monetization as well as user experience.
An article in Business Insider about the shift by advertisers to
behavioral targeting campaigns (Maher, 2010) echoes this point:
If agencies continue to spend more of their budgets on behavioral targeting campaigns smaller niche publishers will have a hard time competing for these dollars with the portals and networks with much larger audiences. This will cause smaller publishers to join networks that offer behavioral targeting and split revenue with them, cutting into their margins.
Another very important use of data in online advertising is in measuring “attribution.” An advertiser needs to understand the return on the advertising investment in order to make choices about the most cost-effective way to advertise. This can be a challenging problem if a user might see many different ads from the same advertiser on different websites and in different contexts.
The more difficult it is for an advertiser to track user views of their ads across websites, the more difficult it is for the advertiser to allocate advertising dollars efficiently. A typical example cited in an
industry report is given here:
One Adobe client, a hospitality and entertainment group, realized that their apps were driving sales through other online and offline channels. They only realized this once they stopped obsessing on the last click before a sale, and tracked customers across channels.
Potential inefficiencies deriving from the inability of advertisers to track users’ views of their ads are likely to continue to grow in importance. A Google Research report stated that 90% of people move between devices to accomplish a goal. Firms and services that find a way to keep track of users across devices will be at an advantage in terms of helping advertisers place ads efficiently.
Measuring the Benefits of Data How can we quantify the value of data for the efficiency of advertising? There are few universally applicable measures, since the value of additional data depends on how much data was used initially. However, a number of studies shed some light on this issue. Beales (2010) reported that the price of behaviorally targeted advertising was almost 3 times the price of untargeted advertising, reflecting the value attributed to reaching the right consumers.
Goldfarb and Tucker (2011) showed that the implementation of the 2002 E-privacy directive in the European Union, which restricted the use of targeting techniques, reduced the effectiveness of ads in the European Union by 64% relative to the rest of the world. The impact was larger for general websites (such as Yahoo.com) relative to more targeted websites (such as cars.com), illustrating in practice the fact that access to data (and thus privacy regulation) can have implications for the nature of businesses and content providers that can be successful. In this case, broad, general interest websites are disadvantaged.
Data is also very important for customizing products and personalizing product offerings. Indeed, the pioneering work on data-driven marketing by credit card firm Capital One in the United States was based around using data and experimentation to get the right credit card offer to the right consumer (Clemons and Thatcher, 1998). This approach helped Capital One grow from a new entrant to a major player in credit cards. Today, from voice recognition to personalized recommendations through sites like Amazon.com, the more data available, the more accurate predictions can be. Overall, it appears that the value created by “big data” to improve recommendations has just scratched the surface, as the quality of the algorithms matures, but also because recommendation engines are bringing in more and more data from a variety of sources, including Twitter and other social media (Booker, 2013).
2 Competition and Welfare in Search and Online Advertising Internet search and online advertising are two examples of “multisided markets.” This means that analyzing competition in these markets is quite complex. Since multi-sided markets are so central to competition in internet businesses, it is worthwhile to understand some of the principles of multi-sided markets in greater detail.
A multi-sided market is a business that brings together distinct
groups of customers to interact. Here are some common examples:
Perhaps surprisingly, the economic theory of multi-sided markets is very recent, as almost all of it was developed after the year 2000.
Much of the theory was originally motivated by legal cases, particularly credit cards.
Multi-sided markets differ from multi-product firms because the sides of the markets are typically distinct entities: consumers and advertisers, for example. Each side cares about the behavior of the other side of the market (whether they use the platform and how much, for example), but does not directly care how much the other side is charged. This leads to different incentives for the platform.
Economists found that the more realistic and rich theory of platform markets was a much better lens for understanding the behavior of firms in platform markets than trying to simply apply existing theories of multi-product pricing. The new theory is useful even though the line between a traditional market and a multisided market is sometimes blurry. For example, Amazon may be considered a traditional firm in its book business; or it can be considered a multi-sided firm, matching publishers to consumers.
In the end, whether it is necessary to use the theory of multi-sided markets depends on the question and the context.
Another key feature of many multi-sided markets (and one that helps determine whether this is a useful lens) is the presence of indirect network effects: one side of the market cares about the activity of the other side of the market. For example, advertisers care about how many users are on a platform when they decide whether to advertise there. These differ from “standard” network effects where users want to directly interact with others on the platform (e.g. telephones and fax machines exhibit “standard” network effects).
Rysman (2009) provides an accessible introduction to multi-sided markets. Evans and Schmalensee (2013) provide a thorough overview of the economics literature on two-sided markets, with an emphasis on antitrust. The latter article provides more detail about theories of competition as well as strategic behavior.
One initial result highlighted in both studies is that pricing is often very asymmetric in multi-sided markets. Prices on one side depend on how “elastic” demand is (how price-sensitive that side of the market) as well as the externality that side has on the other side. If one side is relatively price-sensitive, and it is very important to the other side, then prices are likely to be quite low, perhaps even below cost.
Another important point highlighted in the literature is that behavior on one side of the market impacts welfare and competition on all sides of the market. A firm might engage in exclusive behavior on one side of the market, gaining market power there; but the payoff could come from extracting surplus on the other side of the market. For example, a search engine might write a long-term exclusive contract with a publisher to send all of its search traffic to that search engine. This contract might give the search engine access to a large volume of users. This in turn might give the search engine market power in the search advertising market, allowing the search engine to raise prices on advertisers.
The exclusive behavior occurs in the publisher side of the market, but the harm occurs on the advertiser side of the market.
One very interesting set of results about competition in multi-sided markets concerns a particular set of stylized assumptions about the “homing” behavior of the different sides of the market, which refers to whether the individuals use multiple platforms or a single platform. For example, in mobile phones, many consumers purchase only one phone (single homing) while many application developers port their apps to multiple platforms (multi-homing).
The literature usually starts from extreme assumptions, for example that all users single home (by assumption) in media markets, and that all advertisers multi-home (by assumption). Of course, the real world is more complex.
The theoretical result about what happens when two platforms compete, when one side (“users”) single-homes and the other fully multi-homes is quite stark. Working backwards, once a platform has attracted a set of users, the fact that the other side uses all platforms (by assumption) means that the platform can charge a monopoly price to the multi-homing side. Anticipating that, the platform is willing to pay up to the per-user monopoly profit to attract the single-homing side. If the good has zero marginal cost (like broadcast media), then the good will be given away for free to the single-homing side, and the firm will make investments to attract the single-homing side. This kind of stark result predicts that competition for single-homers will be intense, where the motivation is to extract revenue from the multi-homing side of the market. The Android mobile phone operating system is free to users, but searches conducted through the Google search engine raise money from advertisers. (For further reading on multi-sided markets in various industries, see Eisenmann (2008); Eisenmann, Parker, and Van Alstyne (2006); Haigu and Yoffie (2009); or Lee (2011).) Of course, few real-world markets fit this framework exactly, but it is a useful starting point for understanding the kinds of pricing patterns we see in practice.
Competition in Online Advertising Markets: A Multi-sided Markets Perspective Online advertising markets are complex. There are not many comprehensive articles that cover all of the relevant economic and technical background. For some initial reading, see Evans (2009).
Search advertising is perhaps easier to begin with. Levy (2011) provides a detailed history of the development of Google and Google’s online advertising. A description of search advertising auctions is given by Varian (2006), though the market has continued to evolve and become more complex over time. The U.S.
Department of Justice (2008) press release on the Google-Yahoo!
proposed agreement outlines some facts and assessments of the impact of competition in this market. The U.S. Department of Justice (2010) press release on the Microsoft-Yahoo! search alliance provides further background on the importance of economies of scale for competition.