Serena Ng (Columbia)

Event Date

Blue room, Social Sciences & Humanities,1113

"Skewed Fluctuations and Propagation Through Production Networks: An Empirical Investigation"


Recessions tend to be deeper than expansions and negative skewness is a prevalent feature of macroeconomic time series. Skewness may arise exogenously because the common shocks are not symmetrically distributed, or endogenously as symmetric sectoral shocks propagate through locally through production networks. Previous theoretical work often studies these two possibilities in isolation and precludes the interaction of local and global type co-movements. We nests both possibilities in a reduced-form, factor-augmented network model that allows for heterogeneous spillovers. The framework is used to analyze data from real gross output of 43 sectors over the sample 1957-2019. When the data is the sum of a network, a common, and an idiosyncratic component, the skewness coefficient is not simply the sum of the skewness of the three components. Unless the components are independent, higher order covariations of the three components (also known as co-skewness) can contribute to skewness. We  find that while the common factors are skewed, the production network component tends to be more skewed. The dominant source of skewness is traced to the co-skewness terms, in particular, the comovements between higher order variations of the common factors and those of the network component. We construct and estimate a multi-sector general equilibrium model with a production network to help us interpret our results.

Event Category