We empirically and quantitatively study the impact of supply chain disruptions on U.S. businesses. Leveraging granular shipment-level data on the universe of U.S. seaborne imports with nearly 200 million observations, we construct a measure of disruptions at the individual firm level for the time period 2013-2023. We document a significant heterogeneity in disruption rates among U.S. public firms, with a notable increase observed in recent years. We introduce a notion of supplier capital and investigate the effect of supply disruptions on firms’ investment decisions. In the data, firms tend to increase investment in supplier capital following the shock, however, financially distressed firms exhibit a much weaker response. We develop a general equilibrium model with heterogeneous firms and with investment in supplier capital. We show that firms’ ability to accumulate supplier capital by making costly investment is an important margin of adjustment in the aftermath of such crises. Financial constraints help account for the heterogeneous treatment effect observed in the data. Two supply chain initiatives proposed by the U.S. government to mitigate disruptions are evaluated. Finally, we document a significant rise in supply disruptions in sectors critical to the U.S. economy and build an index of critical supply disruptions. We show quantitatively that firms relying heavily on imports of critical products experience a much larger decline in output following a disruption shock relative to firms which are not engaged in critical supply chains.
We develop a dynamic model of input-output networks that incorporates adjustment costs of changing inputs. Our closed-form solution for the dynamics of the economy shows that temporary shocks to upstream sectors, whose output travels through long supply chains, have disproportionately significant welfare impact compared to affected sectors’ Domar weights. We conduct a spectral analysis of the U.S. production network and reveal that the welfare impact of temporary sectoral shocks can be represented by a low-dimensional, 4-factor structure.