Didier Laussel & Ngo V. Long & Joana Resende, 2020, RAND Journal of Economics, vol. 51(3), p. 650-675.
THE RESEARCH PROGRAM
Advances in digital technology mean firms are increasingly able to collect and process huge amounts of consumer-specific data, allowing them to classify consumers on the basis of their purchase histories and browsing histories. Is this good or bad for firms and/or consumers? The “behavior-based price discrimination” literature has concentrated on the oligopoly case and distinguished two effects on profits: a direct, positive one when the rival’s strategy is fixed and a strategic, negative one from increased competition. Their conclusion is that the overall effect is negative. We investigate the same issue using a dynamic model of a non-durable good monopoly, where competition is between the different successive selves of the monopolist. The answer, we find, depends on consumers’ profiling accuracy.
The literature on third-degree price discrimination (when firms offer different prices to different groups of identifiable consumers with no arbitrage opportunity) has overlooked the following important issue. When customers are heterogeneous in their willingness to pay (WTP), their purchase histories are endogenously determined by the firm’s dynamic pricing policy. The number of distinct market segments based on the monopolist’s grouping of repeat customer types depend on his current, past, and future pricing policies. Anticipating the firm’s future prices and grouping strategy, lower-type customers have an incentive to defer their first purchases until later periods in order to receive a better deal. The firm counters this incentive by offering higher informational rents to the new customers it wishes to serve in each period. Under these circumstances, a firm’s ability to acquire customer information is detrimental to its profit.
Our article shows that a “Curse of Knowledge” arises within a coarse information setting, shedding light on optimal dynamic pricing in several real-world set-ups (subscription-based business models, telecommunications industry, streaming music industry, online video industry, online betting sector, and the pricing of football/baseball tickets). In our model, a monopolist is able to recognize former customers only from the timing of their first purchase, and uses this coarse information to engage in third-degree price discrimination. His use of customer information for intertemporal price discrimination is shown to reduce his aggregate profit below the level he would otherwise get.
We compare the equilibrium dynamics arising in this coarse customer-preference information model to the equilibrium dynamics in the polar case of full information acquisition (FIA), in which the monopolist can use consumers’ purchase history to uncover their exact WTP and thus personalize pricing. These two cases lead to diametrically opposed conclusions concerning the equilibrium. Under FIA, the monopolist gains from his ability to acquire full information, and his profit is even greater than that obtained by a full commitment monopolist under the coarse information scenario.
We study the Markov-perfect equilibria (MPE) of a game in which both firm’s strategy and consumers’ expectations depend on the number of consumers who have already bought the good for the first time.
Our characterization of the MPE under coarse information reveals that (i) eventually, the whole market is covered (in sharp contrast with the static equilibrium outcome), and (ii) a shortening of the commitment interval results in a fall in the firm’s aggregate profit. In extreme cases, as the commitment period tends to zero, the profit vanishes, reminiscent of the Coase Conjecture for the durable good monopolist. The monopolist’s profit under the MPE is strictly lower than that obtained when he is able to commit initially to a specified sequence of prices. Paradoxically, it is therefore bad for the firm to know its consumers’ histories and use this information to implement price discrimination.
The analysis raises the question of the possible non-monotonicity of the relationship in a dynamic context between firms’ profits and consumers’ profiling accuracy. We tackle this question generally in a new paper (Laussel, D., Long, N. V., & Resende, J. (2020), Consumers’ profiling accuracy and monopoly profits: a non-monotonic relationship) in which we show that knowledge is good only if sufficiently refined, and that a little knowledge can be more harmful than no knowledge at all. Another issue under investigation is privacy: how our conclusions may be affected and how the monopolist’s profits change when consumers are able to manage their identity by hiding it from the firm, either ex ante or ex post.
© Photo by Sunward on Adobe Stock
→ This article was issued in AMSE Newletter, Fall 2020.