A fully committed sender seeks to sway a collective adoption decision through designing experiments. Voters have correlated payoff states and heterogeneous thresholds of doubt. We characterize the sender-optimal policy under unanimity rule for two persuasion modes. Under general persuasion, evidence presented to each voter depends on all voters’ states. The sender makes the most demanding voters indifferent between decisions, while the more lenient voters strictly benefit from persuasion. Under individual persuasion, evidence presented to each voter depends only on her state. The sender designates a subgroup of rubber-stampers, another of fully informed voters, and a third of partially informed voters. The most demanding voters are strategically accorded high-quality information.
NSF Award SES-1919329, ``Models of Complex Experimentation: Attribute Discovery, Contextual Experimentation, and Experimentation on Causal Graphs'' (2019-2022).
Most decisions– from a job seeker appraising a job offer to a policymaker assessing a novel social program– involve the consideration of numerous attributes of an object of interest. This paper studies the optimal evaluation of a complex project of uncertain quality by sampling a limited number of its attributes. The project is described by a unit mass of correlated attributes, of which only one is observed initially. Optimal sampling and adoption is characterized under both single-agent and principal-agent evaluation. In the former, sampling is guided by the initial attribute but it is unaffected by its realization. Sequential and simultaneous sampling are equivalent. The optimal sample balances variability of sampled attributes with the importance of neighboring unsampled ones. Under principal-agent evaluation, the realization of the initial attribute informs sampling so as to better influence adoption. Sampling hinges on (i) its informativeness for the principal, and (ii) the variation of the agent’s posterior belief explained by the principal’s posterior belief. Optimal sampling is not necessarily a compromise between the players’ ideal samples. I identify conditions under which mild or no ex-ante disagreement leads to excessively risky or conservative sampling. Yet, drastic disagreement always induces compromise.
"The Strategic Role of Public Information in the Rise and Growth of Ponzi Schemes: The Case of Albania, 1993-1997."
This paper examines how the strategic production of misguiding public information could have contributed to the unhindered flourishing of large-scale Ponzi schemes in Albania during the period 1993-1997. Using a global game framework, I study the pre-electoral incentives of an incumbent seeking reelection to under-investigate the quality of a project which benefits voter-investors in the short run. The incumbent balances the likelihood of a regulatory mistake with that of gaining electoral support among voter-investors. Such voters, who in the absence of the investment might disapprove of the incumbent, vote in her favor so as to ensure the continuation of the project. Under sufficiently strong electoral concerns, the incumbent benefits from highly uninformative public investigation followed by dampened regulatory action. In turn, lack of public information encourages excessive investment in the project. Historical evidence is brought to bear in arguing the relevance of this framework for the case of Albania and the implausibility of alternative historical explanations.
``Deliberate Sampling under Model Uncertainty"
A policymaker decides whether to scale up a social program based on the evidence collected from a limited number of small-scale tryouts in strategically selected communities. She cares about the effect of the program on all affected communities. Ex-ante comparable communities (e.g., communities with similar median household income) are expected to witness similar outcomes from the program. The policymaker is uncertain of both the particular outcome of the policy on untried communities and the structural parameter based on which she infers these outcomes from those in tried communities. That is, she is uncertain of the model of the world that should guide her inference process. The dual role of the collected evidence —namely, (i) to estimate the parameter and (ii) to extrapolate to untried communities— has interesting implications. Under model certainty, the same communities are sampled despite the particular value of the parameter. Yet, under model uncertainty, these communities are no longer optimal. The optimal sample consists of communities that are more similar to each other and of uniformly more extreme median income than under model certainty. Optimal sampling converges to that under model certainty as correlation among communities approaches zero.
``Gradual Exploration of Complex Alternatives: Diversify or Focus?"
An agent faces a set of ex-ante identical alternatives, of which she must acquire a single one. Each is characterized by a unit mass of correlated attributes. The agent has limited opportunities to sequentially sample individual attributes. At each sampling step, she faces a choice between diversified search (i.e., sampling attributes in previously unexplored alternatives) and focused search (i.e. sampling further an already well-explored alternative). By sampling an alternative further, the agent forgoes its current expected value. With only two alternatives, optimal sampling is deterministic and can be implemented simultaneously. Promising discoveries about an alternative imply unpromising prospects for the other. With larger pools, if an alternative shows particularly great promise, the agent allocates sampling towards other less promising alternatives, even if such sampling is not expected to reveal much new information.