Optimizing Service using High Dimensional Panel Data

Amandeep Singh, Ken Moon, and Gad Allon, Operations, Information, and Decisions, The Wharton School

Abstract: We study how retail stores’ multi-dimensioned service levels affect consumers’ buying behavior in a spatial setting. To this end, we propose the Double Block-Lasso BLP estimator, which combines the double selection procedure introduced in Belloni, Chernozhukov, and Hansen (2014), with demand estimation methods set forth in Berry, Levinsohn and Pakes (1995). Under mild assumptions of structural stability and sparsity, we obtain consistent estimates of the influences that retails stores exert across various spatial markets. Using the estimated influence structure, we highlight the service quality interventions promising greatest joint impact on consumers’ buying patterns.