Three Papers on Strategic Decision Making in Digitally Mediated Markets.

Abstract

Technological developments have been reshaping existing markets and giving rise to new ones. In my dissertation, I address several questions that emerged as a result of these changes. In the first two chapters, I study strategies in two different markets: a simultaneous ascending auction (SAA) and an information-good market. The challenge in both cases is analytical intractability of the problem of finding equilibrium strategies. I therefore apply heuristic methods and computer simulation to incrementally search for improvements relative to known leading strategies. For SAAs with complementary goods, in which bidders often face an exposure risk, the result of this search procedure is a price-predicting strategy with self-confirming predictions. If the goods are substitutes, the results suggest that a simple demand-reduction strategy yields the highest expected surplus. In chapter two, I apply similar methods to study bundling strategies in an information-good market. I restrict my attention to three bundling schemes: pure unbundling, pure bundling, and mixed bundling, where the latter is a choice between the two pure schemes. I find that when consumers' tastes are highly diverse, the relative profitability of the schemes is defined by the consumer preference distribution, even when the marginal costs are zero. This holds for a duopoly as well as a monopoly. Under duopoly, restricting firms to pure schemes may result in higher equilibrium profits relative to mixed bundling. I also provide comparison of the market efficiency, profits and consumer welfare under a monopoly and a duopoly. In chapter three, I take the perspective of a market designer and conduct a human-subject experiment to study learning of Bayesian Nash equilibrium. The study is motivated by the increasing need to test game-theoretic predictions in the relatively new online environment. I focus on three treatment variables: available information, existence of a potential, and supermodularity of the game. The findings are largely consistent with the theory of supermodular and potential games. In addition, the evidence suggests that there is a strong interaction effect between supermodularity and information, which I attribute to differences in the nature of irrationality that prevails under the different information conditions.