Daniel Chen
Daniel Chen
Research Interests
- Theory and Finance
Job Market Paper
This paper develops a dynamic general equilibrium model of the market for attention. Digital platforms compete for the attention of consumers by investing in the quality of their services, which they provide for free. Platforms then sell the attention, in the form of advertisements, to firms in the product market via auctions that use consumer data for targeting. We characterize outcomes in the product market, ad revenue, and platform investment in the unique stationary equilibrium. When data is more informative for all platforms, typically product consumption improves but ad revenues and investment decline. When platforms are more interoperable, investment rises but product consumption worsens. Compared with the first best, investment can be either too high or too low. The model predicts variation in ad prices, bid pacing, and delay in the matching of a firm to a consumer and relates these to platform market power. It also predicts that platforms that are data-rich relative to their rivals typically have higher market shares, ad prices, and investment.
Publications
We model a simple market setting in which fragmentation of trade of the same asset across multiple exchanges improves allocative efficiency. Fragmentation reduces the inhibiting effect of price-impact avoidance on order submission. Although fragmentation reduces market depth on each exchange, it also isolates cross-exchange price impacts, leading to more aggressive overall order submission and better rebalancing of unwanted positions across traders. Fragmentation also has implications for the extent to which prices reveal traders’ private information. While a given exchange price is less informative in more fragmented markets, all exchange prices taken together are more informative.
Working Papers
We consider the design of a market for a single asset where a finite number of risk averse traders may trade to share risk from asset endowments. We derive the direct mechanisms that maximize a linear combination of expected revenue and allocative efficiency. We find that the first best allocation is Bayesian-Nash implementable with ex-ante budget balance if and only if the expectations of traders’ endowments are proportional to their risk capacities. We show that an optimal direct mechanism has an indirect implementation by a double auction with side payments. Thus there may be cause for regulation of side payments and potential to use them as effective policy tools.
An agent acquires information dynamically until her posterior belief about an unknown binary state reaches either an upper or a lower threshold. The agent can choose any signal process subject to a constraint on the rate of "uncertainty reduction." We show that learning strategies can be ordered by time risk---the dispersiveness of the distribution of time that a threshold is reached. We construct a strategy that maximizes time risk and one that minimizes time risk. Under the time-risk maximizing strategy, posterior beliefs evolve according to a Poisson process. The posterior belief either jumps to the threshold that is closer in Bregman divergence or drifts towards the other threshold. Under the time-risk minimizing strategy, the posterior belief reaches a threshold at a deterministic time.