The formation of input–output architecture: Evidence from the European Union

Employing sector-level data on production input interlinkages within and across all countries of the European Union, I use a stochastic actor-oriented model to estimate the evolution of input–output architecture over the period 2001–2014. My behavioral model of network formation approaches these dynamics from the perspective of individual nodes and thus provides a framework to test alternative microfoundations of an endogenous production network. The proposed methodology resolves at least two fundamental problems of empirical investigations into the dynamics of production networks. First, it bypasses the endogeneity problem that arises when topological features of the network determine the formation or dissolution of network ties. Second, it accounts for the mutual dependence between the network and producers’ individual behaviors. Building on this model, I find that supplier heterogeneity in productivity, growth, and labor costs, the introduction of a common currency, and several structural network characteristics such as degree centrality (preferential attachment) as well as the tendencies to connect to indirect suppliers in the nearest network neighborhood (transitivity) and to form bilateral relationships (reciprocity) determine the evolution of the production network. At the same time, changes of the network structure feed back on productivity, growth, and labor costs as producers converge in behavior to their supply chain partners. My results, therefore, testify to additional dynamic network effects on aggregate outcomes that arise from the adjustment of the production network in response to idiosyncratic shocks.