NBER Working Papers · 2026 · Ingvil Gaarder、Lancelot Henry de Frahan、Magne Mogstad、Alexander Torgovitsky、Oscar Volpe
Abstract
We revisit the classic identification problem of separating supply and demand for a homogeneous good using data from multiple markets. We allow markets to be heterogeneous according to unobservables, a feature that arises if there are unobservable differences in consumer preferences or firm technology. We develop a new identification analysis based on hypothetical market types. We use this analysis to show how nonparametric, economically motivated assumptions carry empirical restrictions for a wide range of target parameters, including elasticities, but also welfare parameters, such as consumer surplus. Then, we develop computationally tractable methods for implementing partially identified linear random coefficients models in which the slopes of supply and demand are heterogeneous. We apply these methods to estimate the welfare impact and incidence of sales taxes in the United States.