Research

"Attention Costs and Strategic Communication"

Abstract:  This paper studies how the value of sender commitment depends on the nature of a receiver’s costly attention. I consider a common-interest sender–receiver game in which the receiver imperfectly perceives messages and chooses attention optimally. I compare listening costs, which depend on the mechanical accuracy of perception, with processing costs, which depend on mutual information, and allow attention to be either selective across messages or non-selective. The sender’s message strategy affects both the benefits and, under processing costs, the costs of attention. When attention is non-selective, commitment provides only an equilibrium-selection benefit: binary action recommendation is already implementable without commitment. With selective attention and listening costs, commitment can create a strict surplus by allowing the sender to credibly distinguish states with the same optimal action but different stakes, thereby directing attention more effectively. The analysis also suggests that this benefit is weaker under processing costs, which directly penalize finer message distinctions.

"Speech, Silence, and Costly Attention"

Abstract:  This paper studies communication when a receiver must pay a fixed cost to learn the content of a sender’s message. I consider a sender–receiver game with common, state-dependent preferences, non-verifiable cheap talk, and silence versus talk observed for free. When attention costs are sufficiently low, full communication is feasible; otherwise, attentive equilibria impose strong restrictions on speech. Although appealing to the sender as a signaling device, silence forces any possible communication to be necessarily coarse and interval-based, while commitment power would allow fine communication and sustain attention at higher costs. I characterize equilibria with and without attention, including informative no-attention equilibria that exploit the distinction between silence and talk. I compare preferences across equilibria and show that the sender can use when to speak, and not just what to say, to raise attention incentives.

"Contextual Inference with Research"

Abstract:  This paper studies how consumers’ ability to conduct costly research affects a firm’s incentives to reveal information through its product menu. I introduce consumer research into the framework of Kamenica (2008), where a firm observes aggregate preferences over product variants, while some consumers are uninformed about both their own preferences and the aggregate taste distribution. The firm cannot commit to an arbitrary information structure; instead, its menu acts as an endogenous signal. Focusing on the two-variant case, I characterize the unique pure-strategy perfect Bayesian equilibrium, subject to a refinement of off-path beliefs. Introducing research weakly increases equilibrium menu size, but does not eliminate signaling: the firm may still offer an informative singleton menu to discourage further research. The firm and informed consumers are weakly better off with research, while uninformed consumers are weakly better off ex ante, though not necessarily for every realization of aggregate tastes.

"Sequential Learning with Predecessor Choice"

Abstract:  This work analyses a model of sequential learning with predecessor choice. I consider two states of the world and two types of agents, who differ only in their initial (subjective) beliefs about the state of the world. Agents can choose to observe one of their immediate predecessors, either of their own type or the opposite. After choosing a predecessor, agents receive a private signal and learn the action of their chosen predecessor, as well as the action of their predecessor’s predecessor, and so on. Then they take their own action. I ask: under which conditions would agents prefer to observe their opposite type across generations, leading to assimilation between the two groups? I find that with two signals, agents are indifferent between predecessor types; introducing an uninformative signal creates the possibility of strict preferences for either type. Asymptotically, I find that agents will always prefer to observe their own type; however, early generations may choose to observe their opposite types. I derive conditions under which this occurs, and characterize decision rules specifying the action taken by these agents. Using simulations, I discuss comparative statics for the point at which preferences switch.