Abstract:
|
Experimentation and statistical inference in marketplaces is often complicated by general equilibrium effects. Through the channels of equilibrating factors such price and availability, a treatment provided to one subgroup of market participants can affect the decisions and outcomes of other participants. This talk presents a framework for the marketplace interference problem and then discusses its magnitude and strategies for mitigating it, including more sophisticated randomization schemes, model-based adjustments, and market-level experimentation.
|