Econometrics for the Firm (FIRMMETRIX)

Lead Research Organisation: University College London
Department Name: Economics

Abstract

The research this proposal describes will tackle existing challenges with the empirical assessment of firms' production processes and interactions. Firms' decisions and interactions lie in the background of the current debate on important themes such as inequality, market power, productivity, and the origins of macroeconomic fluctuations. The quantification of the main drivers of firms' behaviour is a key ingredient to many metrics colouring those debates. The identification of the causal links between firms' inputs and outputs and their market behaviour is thus fundamental as imprecisions here will steer conclusions on those big societal themes. Yet, (a) modern procedures to estimate production possibilities from conventional data rest on critical assumptions on how and when firms choose inputs that may unreliably drive estimates; (b) important multi-firm negotiation models lack clear guidelines on the data requirements for their empirical validity; and (c) systematic market definition protocols - key for research and competition policy - remain a challenge. This proposal offers ground-breaking research addressing these issues.

The proposal incorporates new and increasingly available data to address the issues above; develops the necessary tools to take on the econometric challenges that these new data bring; and shows through substantive applications how this approach helps. The proposal covers methodological considerations and empirical applications on the analysis of firm interaction models in four workstreams that complement and build on one another. The workstreams will: (i) incorporate data on firm expectations and business networks to address the issues in (a) in two separate workstreams; (ii) formally explore the data demands for the accurate assessment of firm-to-firm relationship models noted in (b); (iii) and bring together machine learning methods and economics for market segmentation to address (c).

Publications

10 25 50