Network Origins of Aggregate Dynamics

There is growing evidence of producer-level idiosyncratic shocks affecting macroeconomic aggregates as they are propagated and amplified by input-output linkages, which are also known as the production network. Most of the recent models assume that this shock-propagation is instantaneous, however, in their seminal model, Long, and Plosser (1983) assumed it to be dynamic. In their model it can be characterized by the \textit{average propagation time}: the time it takes for an input supplier to pass on the effect of a shock to its direct customers.This paper is the first attempt to quantify the average propagation time of an economy by the structural estimation of a prediction by Long and Plosser (1983) that connects industry-level idiosyncratic TFP shocks to industry output by the production network. The estimation is carried out on annual data series of 66 industries of the US economy between 1997 and 2020. Preliminary findings suggest that the average propagation time for TFP shocks was around 6 months in past decades and became shorter (4 months) after the Great Recession. These results suggest that the propagation of TFP shocks stretches beyond the horizon of one year in just 2-3 propagation steps generating a correlation between subsequent annual GDP observations. Thus the network propagation of TFP shocks might explain aggregate dynamics. Based on simple calculations network propagation of TFP shocks can account for 85\% of the first-order auto-correlation of annual GDP in the US. These results also suggest the potential for optimally timed industry-level interventions during a recession