Essays on Shock Propagation in Economic Production Networks: Applications to U.S. Oil Price Episodes and Green Jobs

Abstract

This dissertation is composed of three essays that explore how dynamics in economic production networks can lead shocks to have differential impacts for producing entities and for workers. In Chapter 1, I propose a set of indicators that aim to capture, and predict, such differential impacts for producing entities based on both the configuration and the evolution of network connections. In terms of the latter, I focus specifically on how producing entities use their inputs as either substitutes or complements at the pairwise level. I confirm the indicators' predictive power by leveraging synthetic data produced by a computational model. To bridge this conceptual work and empirical applications, I also propose and test an approach that categorizes production processes' input pairs as substitutes or complements based on percentage changes in the usage of those inputs. In Chapter 2, I apply these ideas to the context of historical oil price episodes to construct a set of indicators for U.S. manufacturing industries over the period 1968-2018. Leveraging a regression framework, I find substantial heterogeneity in industry outcomes during oil price episodes that were tied both to industries' places in the network and to the changing input usage of the industries around them. The results also suggest that many oil price increases were at least partially demand-driven and that the supply-side pass-through of prices was due, at times, to industries' inability to substitute away from higher-priced petroleum products. In Chapter 3, I explore the relationship between green jobs and occupational transitions. Using data for the United States for 2001-2012, I train machine learning models that relate economic changes at the state-industry level to historical occupation-to-occupation flows. I leverage these models to explore how several scenarios of economic change could impact a set of 84 focus occupations, which were identified in the previous literature as either green or brown based on their titles, the tasks they perform, and/or the industries in which they are employed. I find that the models' predictions have both similarities and differences with these earlier categorizations, suggesting that the distinction between green and brown jobs may be blurred when viewed from the perspective of the broader networked economy.