Endogenous Spatial Production Networks: Quantitative Implications for Trade and Productivity

Working Paper
Published on 29 June 2021

An earlier version of this working paper was published in January 2021.

Abstract

In modern economies, production is organized in large-scale complex networks of firms trading intermediate inputs with each other. Larger Indian firms selling inputs to other firms tend to have more customers, tend to be used more intensively by their customers, and tend to have larger customers. Motivated by these regularities, Panigrahi proposes a novel empirical model of trade featuring endogenous formation of input-output linkages between spatially distant firms. The empirical model consists of (a) a theoretical framework that accommodates first order features of firm-to-firm network data, (b) a maximum likelihood framework for structural estimation that is uninhibited by the scale of data, and (c) a procedure for counterfactual analysis that speaks to the effects of micro- and macro- shocks to the spatial network economy. In the model, differences in production costs across firms arise not just from differences in productivity but also from finding the most cost-effective suppliers of intermediate inputs. Firms with low production costs end up larger because they find more customers, are used more intensively by their customers and in turn their customers lower production costs and end up larger themselves. The model is estimated using novel micro-data on firm-to-firm sales between Indian firms. The model’s fit is good. The estimated model implies that a 10% decline in inter-state border frictions in India leads to welfare gains ranging between 1% and 8% across districts. Moreover, over half of the variation in changes in firms’ sales to other firms can be explained by endogenous changes in the network structure.

Authors

Piyush Panigrahi

University of California, Berkeley