Endogenous Spatial Production Networks: Quantitative Implications for Trade and Productivity

Working Paper
Published on 10 January 2021


In modern economies, production is organized in large-scale complex networks of firms trading intermediate inputs with each other. In this paper, Panigrahi develops a model of trade between multiple locations where input-output linkages between firms form endogenously. 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. Despite its rich structure, the researcher shows that the model can be estimated via maximum likelihood using a multinomial logit model of supplier choice for intermediate inputs. Panigrahi estimates the model using novel administrative data on transactions between Indian firms located across 141 districts. The estimated model reproduces new stylized facts about the margins of firms’ sales and trade that he documents using Indian micro-data on firm-to-firm linkages. Panigrahi proposes a procedure to conduct counterfactual analysis that accounts for the randomness in network formation between firms. Using the estimated model, he finds that a 10% decline in inter-state border frictions 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.


Piyush Panigrahi

University of California, Berkeley