Periodic Reporting for period 3 - TRADENET (Firm-to-Firm Trade Networks)
Période du rapport: 2020-04-01 au 2021-08-31
The increasing fragmentation of production processes at the international level has created large networks of firms located in various countries and linked together through contractual input-output relationships. These contractual linkages in turn give raise to many firm-to-firm international flows of goods and services, capital and knowledge which are often studied at the aggregate level but are rarely analyzed from a disaggregated point of view. The purpose of this project is to analyze the structure of these firm-to-firm networks, in the data and in theoretical models. In particular, we want to understand better how the constitution of these “global value chains” over the last twenty years has affected the functioning of modern economies. There are at least two reasons why these global value chains are an interesting object of analysis. First, they are huge. Global Value Chains are estimated to be responsible for at least half of the overall volume of trade and more than twenty percent of world production. As a consequence, understanding the determinants of aggregate trade and output requires to dig deeper into these value chains. Second, the structure of these value chains is extremely specific. They are usually concentrated on a few very large firms. And decisions on how to structure these value chains seem rather persistent over time. One consequence of these structures is that there are likely to give raise to strong propagation mechanisms among firms belonging to the same value chain. A good example of these propagation mechanisms is the 2011 tsunami in Japan, that has affected a relatively small number of firms directly exposed to the natural disaster but has had consequences on many firms, beyond Japan, that were indirectly exposed through their value chain.
In this project, we want to start from the most disaggregated level, using models and data of firm-to-firm networks, explore the statistical properties of these networks, and study what they imply for various aggregate outcomes.
In this project, we want to start from the most disaggregated level, using models and data of firm-to-firm networks, explore the statistical properties of these networks, and study what they imply for various aggregate outcomes.
A first part in the project consists in studying the structure of firm-to-firm networks, theoretically and in the data. The objective is to build theoretical models that can replicate the observed structure in the data. These models rely on the assumption that the underlying market in which transactions between these firms take place is frictional. These frictions come from the lack of information about the firms involved in these markets. They can also arise from information asymmetries, when a firm outsources a production step to an outside partner, thus transferring knowledge about its own production process without being able to fully control what the partner does. Finally, frictions can also be the consequences of relationship-specific investments, for instance when an input supplier needs to invest to customize its product to the particular partner it is serving. All these market imperfections are well-known to complicate the matching between firms and can deliver network structures that are relatively sparse, concentrated, and persistent, exactly as we observe in the data (Lenoir, Martin and Mejean, CEPR Discussion Paper 13442, 2019).
Beyond being useful to replicate the observed structure of firm-to-firm networks, such models can also help understand why shocks affecting these networks may have an impact that is quite different from what is predicted by models of trade in frictionless markets. For instance, the impact of uncertainty shocks is magnified in firms’ networks. The reason is that investment decisions by one firm in the network have consequences for all firms directly or indirectly connected to the firm. Episodes of high uncertainty are thus especially costly in these networks, a question that we study in Martin, Mejean and Parenti (2019, mimeo), using the Brexit vote as a natural experiment of a raise in uncertainty. Another characteristic of frictional markets is that they give raise to dispersed prices, a phenomenon that we study in Fontaine, Martin and Mejean (CEPR Discussion Paper 13960, 2019). We document the large dispersion of prices set by French firms in EMU markets and show that they reveal a strong degree of price discrimination. French firms operating in frictional trade markets maximize their profit by charging different firms with different prices.
Another consequence of the network structure of international trade is that it helps propagate shocks across countries, thus generating business cycle comovements. We study the micro origin of these business cycle comovements in di Giovanni, Levchenko and Mejean (AER, 2018). The analysis exploits highly disaggregated data that allow identifying firms’ linkages with foreign countries and their role as a propagation channel for various shocks. From a normative point of view, these propagation mechanisms are potentially welfare-enhancing as they help diversify risk across countries. What the microeconomic structure of trade networks reveals is that diversification opportunities are de facto limited because firms concentrate their sales on a small number of partners (Kramarz, Martin and Mejean, Forthcoming in the Journal of International Economics). As a consequence, they are left quite exposed to idiosyncratic demand risks, which in turn generate a substantial amount of volatility.
Beyond being useful to replicate the observed structure of firm-to-firm networks, such models can also help understand why shocks affecting these networks may have an impact that is quite different from what is predicted by models of trade in frictionless markets. For instance, the impact of uncertainty shocks is magnified in firms’ networks. The reason is that investment decisions by one firm in the network have consequences for all firms directly or indirectly connected to the firm. Episodes of high uncertainty are thus especially costly in these networks, a question that we study in Martin, Mejean and Parenti (2019, mimeo), using the Brexit vote as a natural experiment of a raise in uncertainty. Another characteristic of frictional markets is that they give raise to dispersed prices, a phenomenon that we study in Fontaine, Martin and Mejean (CEPR Discussion Paper 13960, 2019). We document the large dispersion of prices set by French firms in EMU markets and show that they reveal a strong degree of price discrimination. French firms operating in frictional trade markets maximize their profit by charging different firms with different prices.
Another consequence of the network structure of international trade is that it helps propagate shocks across countries, thus generating business cycle comovements. We study the micro origin of these business cycle comovements in di Giovanni, Levchenko and Mejean (AER, 2018). The analysis exploits highly disaggregated data that allow identifying firms’ linkages with foreign countries and their role as a propagation channel for various shocks. From a normative point of view, these propagation mechanisms are potentially welfare-enhancing as they help diversify risk across countries. What the microeconomic structure of trade networks reveals is that diversification opportunities are de facto limited because firms concentrate their sales on a small number of partners (Kramarz, Martin and Mejean, Forthcoming in the Journal of International Economics). As a consequence, they are left quite exposed to idiosyncratic demand risks, which in turn generate a substantial amount of volatility.
Up to now, research in this project has made contributions along two main lines. The first one relies on structural models of frictional good markets to represent interactions between firms in international markets. These frictions are well-known to be an important barrier to international trade but the literature so far has hardly taken them into account in models of international trade. In Lenoir, Martin and Mejean (CEPR DP 13442, 2019), we have shown that it is possible to build and estimate tractable models of frictional trade markets and that these models deliver new insights regarding the determinants of international trade. By borrowing tools from the labor literature, we aim to further develop this line of research and provide further insights on the matching of firms in international markets.
The second contribution is on the identification of transmission mechanisms for shocks through firm-to-firm trade relationships. In di Giovanni, Levchenko and Mejean (AER 2018) as in Kramarz, Martin and Mejean (JIE forthcoming), we have exploited highly granular data to identify such propagation mechanisms. In both cases, the structure of firms’ networks is taken as given. Further work needs to be produced to understand if and how firms in turn react to the shocks by reshuffling their trade portfolio.
The second contribution is on the identification of transmission mechanisms for shocks through firm-to-firm trade relationships. In di Giovanni, Levchenko and Mejean (AER 2018) as in Kramarz, Martin and Mejean (JIE forthcoming), we have exploited highly granular data to identify such propagation mechanisms. In both cases, the structure of firms’ networks is taken as given. Further work needs to be produced to understand if and how firms in turn react to the shocks by reshuffling their trade portfolio.