Shock Diffusion: Does inter-sectoral network structure matter?

This paper introduces the concept of diffusion of shocks in a macroeconomic network consisting of inter-sectoral production linkages. Using sectoral and firm level data, the paper documents two empirical facts. First, sectoral output do not react contemporaneously to shocks in input sectors (it only reacts with a lag). Second, different sectors take different time horizon to respond to shocks to their input sectors. I incorporate these features in a model of production network to study the contribution of sectoral shocks to aggregate fluctuations. I show that if sectors have different reaction horizons it leads to diffusion of shocks through the network over time which prevents the inter-sectoral linkages to form the feedback loop structure essential to generate aggregate volatility. So, the impact of a given sectoral shock lingers over a longer time period but contributes less to aggregate volatility in any given period. After accounting for diffusion, the first order network interconnections still matter but the contribution of higher order interconnections to aggregate volatility gets diluted. Finally, I use a factor model to estimate the contribution of aggregate vs idiosyncratic sectoral shocks to aggregate fluctuations in US industrial production (IP) data. I find that in the case of a diffusion adjusted network model, the contribution of sectoral shocks to aggregate volatility is 27 percent and is of the same order of magnitude as in statistical factor analysis.