Catherine Tucker on Social Advertising

By TAP Guest Blogger

Posted on January 30, 2012

Is advertising more effective when it’s “social”? On the one hand, Facebook’s estimated market valuation is around $100 billion and 30% of US user’s online page-views are of social media, but, on the other hand, advertising on social media attracts only 5% of ad revenue. It is hard to know why or even how to begin to address this problem.
At the American Economic Association’s 2012 Annual Meeting in Chicago, Catherine Tucker, an Associate Professor of Marketing at MIT Sloan, presented new research that elegantly poses the question in a concrete way and takes an important step towards answering it. Her main finding is that adding a social component to advertising does make it more effective. Her design allows her to separately measure two mechanisms at play in social advertising that could, in principle, have very different effects from one another. The first mechanism is homophily, i.e. the tendency of people with similar interests to associate with one another, and the second is endorsement by one friend to another of a product.
To conduct this research, Tucker collaborated with a nonprofit organization to conduct a field test in the form of a Facebook advertising campaign. The charity in question helps provide secondary school education for young women in East Africa. To support this endeavor, they advertise their website on Facebook.
To determine the impact of “social” versus “nonsocial” advertising, Tucker compared the click-through rate (CTR) garnered by ads shown to a random sample of adult Facebook users to the CTR of ads that were shown to users who were friends (in the Facebook sense) of some user who had signed up as a “fan” of the nonprofit. This comparison reveals that a social ad is about twice as likely to be clicked as its nonsocial counterpart, and the significance of this finding is confirmed by various robustness checks.
What drives this apparent success of social ads? A useful source of variation helping Tucker get at this question is the fact that social ads may or may not be revealed to the user as such, depending on the privacy settings of the fan with whom she is friends. For example, suppose Alice is a fan of the charity and that she and Bill are friends. If Alice’s privacy settings are such that she does not share her interests with friends, then the ad shown to Bill will contain no mention of Alice.  If, however, Alice does share her interests with friends, then the ad Bill sees will indicate that Alice is a fan of the charity. While in both cases, the ad shown to Bill is considered to be social, in the latter case, it contains an explicit endorsement.
This form of endorsement turns out to be significant. While social ads containing no endorsement perform better than nonsocial ads, those including an endorsement perform even better.
One caveat to this result, which Tucker duly notes, is that since a user’s privacy settings may, in one way or another, reflect relevant characteristics of that user, the available data does not allow for “perfect” measurement of the effect of endorsement, holding all other factors constant. So, for instance, another interpretation could be that users who select less stringent privacy settings also tend to compensate for this openness about themselves by being more stringent in deciding whom to add as a friend. This could then lead them to have an overall group of friends whose interests are relatively more similar to theirs, yielding a stronger effectiveness of social ads, not caused by the endorsement per se, but rather by a particularly strong degree of homophily with their Facebook friends.
On the other hand, it also seems plausible to imagine that users who are generally more extroverted tend to be both more willing to share their interests and more willing to add slight acquaintances as friends, as compared to users who are generally more introverted. If this were true, then the impact of an endorsement would seem to be even greater than what the data suggests. As the above discussion hopes to convey, this work is both fascinating and clean, yet, due to the underlying complication of the phenomena at play, it has the potential to be a lightning rod for further study of many related questions.
Finally, it would be remiss not to mention that this presentation by Tucker was just one in a terrific lineup of speakers in the session on the Economics of Internet Markets. Jonathan Levin, of Stanford, who chaired the session, presented empirical work entitled, “Sales Mechanisms in Online Markets: What Happened to Online Auctions?”; Andre Veiga and Renato Gomes, each of Toulouse School of Economics, presented separate and impressive theoretical work, the former entitled “Multidimensional Heterogeneity and Platform Design” and the latter entitled “Price Discrimination in Many-to-Many Matching Markets”; Preston McAfee of Yahoo! and Jonathan Baker of American University gave enthusiastic discussions of each paper to conclude the session.
Alexander White is Assistant Professor at the Tsinghua University School of Economics and Management in Beijing, China. His research is mainly in Microeconomic Theory and Industrial Organization, with focus on search engines, two-sided markets and antitrust in network industries.