Granularity, Network Asymmetry, and Aggregate Volatility

I evaluate two competing theories for microfoundations of aggregate fluctuations. The network hypothesis suggests industry-level shocks propagate across the inputoutput (IO) network of the economy, resulting in aggregate fluctuations. The granular hypothesis suggests idiosyncratic shocks to very large firms result in aggregate fluctuations. My main contribution is to connect the two aggregate fluctuation hypotheses for the first time and theoretically and empirically quantify the contributions of each to volatility. The network hypothesis depends crucially on certain plants being essential suppliers to the economy. However, they may be essential suppliers due to their productivity and not any underlying input-output requirements, which means productivity may be the source of both the granularity and network hypotheses. To disentangle these relationships, I document a plant-plant input-output network, then develop a model in which productivity and the exogenous IO network can vary independently and both combine to determine the observed IO network. Finally, I calibrate the model to uncover the underlying IO network and then investigate the empirical relationship between the uncovered IO network and aggregate volatility. I find (i) the observed plant-plant IO network is very asymmetric, (ii) productivity doesn’t vary enough to explain the observed IO network, (iii) and therefore the true underlying IO network explains the majority of the plant size distribution and 34% of aggregate volatility.