Periodic Reporting for period 4 - TRADENET (Firm-to-Firm Trade Networks)
Berichtszeitraum: 2021-09-01 bis 2023-07-31
Over the last 20 years, the production of manufacturing goods has been increasingly fragmented along “Global Value Chains” (GVCs). It is estimated that trade within these GVCs now represents about 50% of the aggregate volume of products exchanged in international markets. The TRADENET project digs into the microeconomic structure of these GVCs: Who are the firms involved in GVCs? How do they match in international markets? How do firms involved in GVCs bargain over the price of their products? I argue that the microeconomic structure surrounding these productive structures has consequences for their resilience to shocks. The reason is that these international production relationships are a channel for the propagation of shocks across countries. In such a production network, any perturbation affecting one node of the network will then travel along the chain and affect all the firms involved in the same value chain. An unfortunate example of this transmission of shocks is the massive disruption in the functioning of GVCs in the aftermath of the Covid pandemic.
Three characteristics of the functioning of these production structures explain that the transmission of shocks affecting specific parts of the network has massive aggregate consequences. First, firms involved in GVCs are typically very large, “granular” firms. Second, GVCs are poorly diversified, explaining that a small number of firms can become very central in the functioning of GVCs. Third, firms involved in international production have largely adopted just-in-time production processes, which implies that these production structures lack inventory buffers in case a bad shock occurs. The TRADENET project studies these microeconomic features and their aggregate consequences using a combination of theoretical modeling, reduced-form evidence and econometric estimates. The project mostly exploits granular firm-to-firm trade data, in the cross-section and over time. The cross-sectional structure of these data is used to characterize the (lack of) diversification of production networks. The time dimension can be exploited to trace the consequences of a specific shock, such as firms’ exposure to China during the early lockdown of the country in January 2020. Finally, the use of administrative data for a specific country, France, allows me to aggregate the micro-level evidence over the whole population of firms and study the consequences for aggregate outcomes.
Three characteristics of the functioning of these production structures explain that the transmission of shocks affecting specific parts of the network has massive aggregate consequences. First, firms involved in GVCs are typically very large, “granular” firms. Second, GVCs are poorly diversified, explaining that a small number of firms can become very central in the functioning of GVCs. Third, firms involved in international production have largely adopted just-in-time production processes, which implies that these production structures lack inventory buffers in case a bad shock occurs. The TRADENET project studies these microeconomic features and their aggregate consequences using a combination of theoretical modeling, reduced-form evidence and econometric estimates. The project mostly exploits granular firm-to-firm trade data, in the cross-section and over time. The cross-sectional structure of these data is used to characterize the (lack of) diversification of production networks. The time dimension can be exploited to trace the consequences of a specific shock, such as firms’ exposure to China during the early lockdown of the country in January 2020. Finally, the use of administrative data for a specific country, France, allows me to aggregate the micro-level evidence over the whole population of firms and study the consequences for aggregate outcomes.
In the first part of the project, I have developed theoretical models that I can bring to the data to study the matching of firms with their suppliers in international markets and the dynamics of their relationship, conditional on a match. The underlying assumption is that finding international suppliers for specific inputs is not easy, which we represent using the concept of “frictional product markets”. I then exploit statistics on the likelihood that a firm finds a new supplier for one of its inputs to measure the magnitude of these “frictions”. The existence of these frictions also means that firms are somewhat “locked” in firm-to-firm relationships, which has consequences for price bargaining. This part of the TRADENET project has already led to two publications in peer-reviewed journals and one working paper that is in the process of being published.
In the second part of the project, I instead take the structure of the firm-to-firm network as given and study the consequences of this structure for the propagation of shocks across countries. I rely on administrative data covering the universe of French firms, their balance-sheet information and a detailed view of their international activities. Based on these data, I first show how the international activity of a few large firms can be a source of exposure to foreign shocks. While a minority of firms participate to international trade, they are larger than the average. Their international activity exposes them to foreign shocks, which I show can explain around one third of the comovement of the French economy with foreign countries. In theory, international trade is a vector of risk diversification. In practice, the structure of international trade is much less diversified than would be needed for the diversification to be effective. The majority of exporting firms, including the largest ones, have export portfolios that are concentrated on one or two main partners. I show that this lack of diversification is a major source of the volatility of trade in the aggregate. Finally, another source of risk that internationally active firms are exposed to is a risk to their supply chain. Leveraging upon the Covid experience, I show that firms that were exposed to China in January 2020, when the country entered into an early lockdown, performed systematically worse than other firms involved in GVCs that were not exposed to China. I show that the transmission is weakened among firms with inventory buffers. The part of the TRADENET project has led to five publications in peer-reviewed journals, several policy notes and blog columns, and a number of media and policy interventions.
In the second part of the project, I instead take the structure of the firm-to-firm network as given and study the consequences of this structure for the propagation of shocks across countries. I rely on administrative data covering the universe of French firms, their balance-sheet information and a detailed view of their international activities. Based on these data, I first show how the international activity of a few large firms can be a source of exposure to foreign shocks. While a minority of firms participate to international trade, they are larger than the average. Their international activity exposes them to foreign shocks, which I show can explain around one third of the comovement of the French economy with foreign countries. In theory, international trade is a vector of risk diversification. In practice, the structure of international trade is much less diversified than would be needed for the diversification to be effective. The majority of exporting firms, including the largest ones, have export portfolios that are concentrated on one or two main partners. I show that this lack of diversification is a major source of the volatility of trade in the aggregate. Finally, another source of risk that internationally active firms are exposed to is a risk to their supply chain. Leveraging upon the Covid experience, I show that firms that were exposed to China in January 2020, when the country entered into an early lockdown, performed systematically worse than other firms involved in GVCs that were not exposed to China. I show that the transmission is weakened among firms with inventory buffers. The part of the TRADENET project has led to five publications in peer-reviewed journals, several policy notes and blog columns, and a number of media and policy interventions.
Within the project, I developed new models and techniques for the analysis of production networks in an international context. This line of the project takes its inspiration from the analysis of the labor literature, which extensively relies on models assuming the matching of workers and firms is frictional. The parameters of these models are estimated using panel data on workers and their mobility from firm to firm. My project uses similar models and techniques. I argue that the matching of firms in international markets is also frictional, and I use observed switches of firms across potential suppliers of the same product to measure the magnitude of these frictions. Although the duration of trade relationships had already been studied in the literature before, few researchers have exploited panel data on firm-to-firm relationships and use those to estimate the parameters of their model.
The novelty of my approach in the second part of the project lies in the use of exhaustive administrative data to link firm-level exposure to foreign markets with aggregate outcomes. Working with the whole population of firms helps identify which dimensions of firms’ international activity matter for their exposure to foreign shocks. Aggregating across firms allows quantifying the consequences of these micro-level choices for the economy as a whole.
The novelty of my approach in the second part of the project lies in the use of exhaustive administrative data to link firm-level exposure to foreign markets with aggregate outcomes. Working with the whole population of firms helps identify which dimensions of firms’ international activity matter for their exposure to foreign shocks. Aggregating across firms allows quantifying the consequences of these micro-level choices for the economy as a whole.