A0583
Title: On robust causal inference in models of firm productivity and efficiency in the presence of many environmental variables
Authors: Artem Prokhorov - University of Sydney (Australia) [presenting]
Valentin Zelenyuk - University of Queensland (Australia)
Christopher Parmeter - University of Miami (United States)
Abstract: A moment-based framework is provided for consistent estimation and normal inference for a firm's production function and inefficiency scores when relevant confounding factors are selected from a large set of variables using various machine learning tools such as lasso or deep neural networks. Connections are discussed between the estimator and the concept of moment and parameter redundancy and the specific case of a debiased lasso estimator is worked out.
