Abstract: Firms often set prices using observational data in which historical prices are correlated with unobserved demand conditions. Instrumental-variable estimators, including generalized method of moments (GMM), can correct this endogeneity, but their estimates may remain unstable when samples are small or instruments generate limited exogenous price variation. Because price optimization can amplify estimation error, reducing demand-estimation bias does not necessarily increase revenue. We develop a robust GMM framework that incorporates residual estimation uncertainty directly into pricing decisions. A counterfactual-weight representation clarifies how the proposed approach limits the sensitivity of predicted revenue to noise in historical demand observations. We also provide tractable optimization algorithms and a bootstrap procedure for calibrating the uncertainty set. Across the numerical settings studied, robust GMM delivers higher average out-of-sample revenue than sequential GMM and the other robust benchmarks.

 

Bio：荣鹰博士现任上海交通大学安泰经济与管理学院教授、博士生导师。他于2010年8月回国执教于上海交通大学，此前在美国加州大学伯克利分校和里海大学从事博士后科研工作，并在上海交通大学和美国里海大学分别获学士、硕士和博士学位。荣鹰教授主要研究领域为服务运营管理、零售运营管理、新兴商业模型的运作以及数据驱动的优化模型。研究成果发表在Management Science、Operations Research、Manufacturing & Service Operations Management、Information Systems Research、Production and Operations Management、Naval Research Logistics、IIE Transactions等国际学术刊物上。