Factor models; trade volatility; diversification.

C38, L25, F14

This paper analyzes the sources of export volatility estimating a dynamic factor model on transaction-level data. Using an exhaustive dataset covering all French export transactions over the period 1993-2017, we reconstruct the latent factor space associated to global and destination-specific macroeconomic cycles by means of a modified expectation maximization algorithm to accommodate both the sparsity and the high dimensionality of the micro time series. Thus while paving the way for a novel application of dynamic factor models to microeconomic analysis, we provide a decomposition of the volatility of aggregate export and firms growth rates, highlighting structural spatial patterns and drawing attention to the role of geographical diversification for the mitigation of risks related to firms’ export activities.