Competition Between Firms that Bundle Information Goods

Information goods, such as journal articles, require substantial initial investment, but low incremental reproduction and distribution costs. Advances in computing and digital communications have nearly eliminated these incremental costs. Thus, positive per-item prices are ineÆcient because they discourages consumption with value greater than marginal cost. Further, very low per item prices will not recover rst-copy costs and thus rms will not have an incentive to create new content. Bundling may be desirable in such a market: individuals pay no marginal fee for each item consumed and the producer is able to recover investment expenses through the bundled sales. Most previous research on bundling assumes a monopoly seller and e ectively homogeneous consumers. In this paper we analyze the pro tability and welfare properties of bundling information goods in a multirm setting. We also introduce a feasible but exible speci cation for heterogeneous consumer preferences. Modeling heterogeneity resolves unrealistic predictions about eÆciency and distribution from the prior literature. Introducing competition from a second rm selling imperfect substitutes results in much lower prices yet only a moderate pro t reduction. Bundling leads to particularly erce competition since consumers either buy all of a rm's collection or nothing. As a result, competition weakens, and sometimes reverses, the strong advantage known for monopoly bundling. We nd that although lower pro t under competition reduces the incentive to create new content, the distributional eÆciency gains in this di erentiated product market typically outweigh the welfare cost of reduced creation. Warrington College of Business, University of Florida, Gainesville, FL 32611, [email protected] .edu. Department of Economics and School of Information, University of Michigan, Ann Arbor, MI 48109, [email protected]. We appreciate helpful comments from Kai-Uwe Kuhn, audiences at the Federal Trade Commission, the University of California at Berkeley, the University of Florida, the University of Michigan, and the Telecommunications Policy Research Conference. We gratefully acknowledge support from NSF grant SBR-9230481 and two University Partnership grants from IBM.