Friend-of-a-Friend in Production Networks: Micro Estimates and Macro Implications

The friend-of-a-friend effect is the idea that when two firms share a common trading partner, a new link between them is likely to form. We quantify this effect in production networks, where the shared partner acts as relational capital facilitating new connections. First, we develop a general equilibrium (GE) model that endogenizes firms’ link formation and incorporates the friend-of-a-friend mechanism. We show that the GE model simplifies to a dyad-level logit specification, enabling us to estimate the friend-of-a-friend effect using a quadruple-based conditional logit that controls for buyer and supplier fixed effects. Analyzing a dynamic panel of Japanese firm-to-firm transactions provides strong evidence of the friend-of-a-friend effect, with a magnitude comparable to other important factors like physical distance and sectoral proximity. Finally, we evaluate the macroeconomic impact through a counterfactual analysis within a calibrated GE model. Results indicate that removing the friend-of-a-friend effect decreases welfare by 0.6% and changes the propagation of firm-level shocks by altering the network structure.