Optimal Targeting of Television Advertisements∗

In this paper, we study the targeting of advertising for the largest media expenditure for most firms—television. While television advertising was once used to “broadcast” to the widest possible audience, today with the explosion of channels and show content, it has the potential to be targeted to a narrower audience. In this paper, we investigate this potential in the context of congressional elections and present a method that estimates the effect of advertising. We find that advertising significantly influences candidate choice, but has a relatively small effect on turnout. We also develop and apply an approach for identifying the optimal allocation of media buys across television programs. Our results suggest that advertising can be effectively targeted, that candidates generally target in a way consistent with the most desirable generic strategy, and that the total effect of advertising is to increase relative vote shares roughly 2% over not advertising. By applying the technique we have developed campaigns could on average double the effectiveness of their television advertising campaigns. Consistent with prior expectations, close races and incumbents come closest to the optimal benefit of advertising. ∗We’d like to thank Eric Bradlow, Jean Pierre Dubé, Paul Ellickson, Ron Goettler, Brett Gordon, Wes Hartmann, Günter Hitsch, Sanjog Misra, Ryan Moore, and Ron Shachar; participants of seminars at Carnegie Mellon University, the Booth and Harris Schools at University of Chicago, the Simon School at University of Rochester, and UC San Diego; and audience members at the 8th Invitational Choice Symposium (2010), Marketing Science (2010), the Econometrics Society Summer Conference (2011), SICS (2011), Polmeth (2011), and QME (2011). †Simon School of Business, University of Rochester. [email protected] ‡Department of Political Science, University of Rochester. [email protected] 1