Transitivity in International Trade: Evidence from Colombia-U.S. Firm Relationships
Alejandra Martinez, Dennis Novy, Carlo Perroni · 2025
Read Transitivity in International Trade: Evidence from Colombia-U.S. Firm Relationships by Alejandra Martinez, Dennis Novy, Carlo Perroni (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-18893.txt.
Introduction
Alejandra Martinez, Dennis Novy, and Carlo Perroni test whether firms, like people, befriend their friends' friends. In social networks that tendency is called transitivity, and it clusters relationships. Their paper, posted to arXiv in December 2025, develops a way to detect and measure the same pattern in firm-to-firm trade, using systematic departures from conditional independence between trading relationships.
Applied to Colombia-U.S. exporter-importer data, the method finds transitivity is real and economically meaningful: sharing a trading partner makes two firms more likely to trade with each other. Counterfactual exercises show this clustering shapes how the whole network adjusts when costs shift. For trade policy, the lesson is that shocks travel along a web woven partly by introductions, so the structure of who knows whom belongs in the model.
Catalog record
| CREATOR | Alejandra Martinez, Dennis Novy, Carlo Perroni |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-18893
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Transitivity in International Trade: Evidence from Colombia-U.S. Firm Relationships Thanks: Novy gratefully acknowledges research support from the Economic and Social Research Council (ESRC grant ES/Z504701/1). We thank participants for helpful comments at the 2025 Midwest Trade Conference at Penn State, the Paris Trade Seminar/Sciences Po, Econometrics Society European Winter Meeting 2025. Martinez: School of Economics, University of Nottingham, Nottingham, UK and CITP; e-mail: [email protected]. Novy: Department of Economics, University of Warwick, Coventry, UK and CAGE, CEP/LSE, CEPR, CESifo, e-mail: [email protected]. Perroni: Department of Economics, University of Warwick, Coventry, UK and CAGE, CESifo, e-mail: [email protected]. Alejandra Martinez, Dennis Novy, Carlo Perroni Abstract A large literature has documented transitivity as a key feature of social networks: individuals are more likely connected with each other if they share common connections with other individuals. We take this idea to trading relationships between firms: firms are more likely to trade with each other if they share common trading partners. Transitivity leads to a clustered pattern of relationship formation and break-up. It is therefore important for understanding how firms meet and how shocks propagate through firm networks. We describe a method for detecting and quantifying transitivity in firm-to-firm transactions, based on systematic deviations from conditional independence across firm-to-firm relationships. We apply the method to Colombia-U.S. exporter-importer data and show in counterfactuals that transitivity is a significant and economically meaningful factor in how firm networks adjust to cost shocks. JEL Classification: D85, F10, F15, R12 Keywords: Agglomeration, Exporters, Firms, Importers, Networks, Search, Trade Costs 1 Introduction There is growing recognition that firm-to-firm relationships form complex networks rather than merely collections of independent buyer-seller pairs. In the context of international trade, firm-level networks play a central role in shaping cross-border flows of goods and services and in propagating shocks through global value chains ( Antràs and Chor 2013 ) . These relationships matter for performance: better supplier matches, for example, can reduce marginal costs and raise output ( Bernard et al. 2019 ) . Despite this emphasis on the network-like features of the relationships between firms, a large literature has relied on a dyadic paradigm of export link formation that omits explicit network effects ( Antràs and Costinot 2011 ; Bernard and Moxnes 2018 ; Eaton et al. 2022 ; Herkenhoff et al. 2024 ) , treating a firm’s decision to trade with one partner as being unrelated to its relationships with others—apart from general-equilibrium effects operating through the price system. This paradigm underpins both structural models and statistical approaches to firm matching, including the balls-and-bins model of Armenter and Koren 2014 . Even so, such frameworks have proven empirically successful in reproducing key qualitative features of production networks—such as their sparsity, skewed degree distribution, and variation across heterogeneous firms ( Bernard et al. 2022 ; Carvalho et al. 2021 ) —while retaining tractability. In contrast, research on social networks has shown that agent-to-agent links are typically not independent. For example, Banerjee et al. 2013 emphasize the role of shared social ties in supporting trust in loan arrangements; and Dustmann et al. 2016 document how referrals from prior employers mitigate information frictions in job matching. A common feature of these examples from the social networks literature is that agents are more likely to engage with one another when they share a connection with a third party, leading to the formation of triads . This transitivity mechanism is a hallmark of social networks ( Newman 2003 ; Jackson and Rogers 2007 ; Jackson 2008 ; Currarini et al. 2009 ) , corresponding to the familiar principle that “the friend of my friend is also my friend.” In this paper, we take this paradigm to firm-to-firm networks in international trade, using microdata on Colombian flower exports to the United States to test for the presence of transitivity and assess its quantitative importance. Investigating the role of transitivity in this context is not only of conceptual interest. How firms connect has direct implications for economic performance. It determines how bilateral trade costs are shaped by network structure and influences how trade responds to shocks. Transitivity provides a concrete mechanism for the emergence of agglomeration effects. In standard trade models, clustering is often introduced through reduced-form assumptions about external economies, whereas here it arises endogenously: if a seller’s connection to one buyer makes it more likely that another nearby seller also connects to that buyer—unlike in the firm matching models prevalent in the literature—then relational spillovers are embedded directly in link formation. This mechanism can help explain why trade networks exhibit localized clustering and why shocks to one set of links propagate unevenly through the system. In this sense, transitivity offers a microfoundation for the types of agglomeration forces long emphasized in economic geography. Transitive relationships also imply that link formation is path-dependent, potentially amplifying the clustering of gains or losses from policy interventions. They also influence network resilience and the pace of adjustment to external shocks. Yet, given that dyadic matching models can successfully explain key features of the data, any departure from the independence assumption requires empirical justification: it must be guided by patterns in the data which models featuring independent link formation are unable to explain. But detecting these patterns requires looking in the right place, with diagnostics that are specifically targeted to uncovering evidence of transitivity. One such diagnostic relates to the observed number of triads. If links are formed independently across dyads, then the expected number of triads in the network is pinned down by the number of observed dyads. An excess number of measured triads relative to this prediction indicates that a dyadic representation alone is insufficient to account for the observed network structure— no specification based solely on independent link formation is capable of explaining this discrepancy, irrespective of its particular formulation. Detecting symptoms of transitivity in real-world networks, however, is significantly less straightforward than the above simple triad count might suggest. Nodes differ markedly in size, and larger or more central firms naturally appear in more triangles, even in the absence of transitivity. The empirical challenge, then, is to construct a statistical test for transitivity that can account for the effects of observed heterogeneity in node characteristics. This requires comparing the actual number of closed triads in the network to a benchmark where each link is formed independently, but in a way that reproduces the empirical heterogeneity in dyadic linking probabilities. To this end, we build on the ideas in Graham 2017 , Dzemski 2019 , and Hughes 2022 , to develop a nonparametric simulation approach that generates a benchmark distribution of triads for heterogeneous nodes under conditional independence. Our test then examines whether the triadic closures observed in the data significantly exceed this benchmark. If so, we interpret the excess as evidence of transitivity, meaning that the link formation process violates conditional independence and is shaped, at least in part, by shared linkages with third parties. We apply this test to microdata on flower exports from Colombia to the United States. The dataset, provided by DIAN (Colombia’s National Directorate of Taxes and Customs), includes all transactions between exporters and U.S. importers from 2007 to 2019. We use this data to construct a bipartite network of Colombian exporters and U.S. importers and examine the