Information Frictions, Investment Promotion and Multinational Production: Firm-Level Evidence
Countries make use of a wide range of policies to attract multinational firms but identifying the effect of such policies is difficult. Combining firm-level data on both the location of these firms’ foreign affiliates and detailed service-specific information from Costa Rica’s investment promotion agency (IPA) over time, we find that IPA support significantly increases the probability that a multinational firm establishes its first affiliate in the country but has generally no impact on the expansion of its presence thereafter. We then show that this effect is primarily driven by the resolution of information asymmetries. It is stronger for IPA information services and on multinational firms from countries and in sectors facing more severe information frictions.
 Automatic Product Classification in International Trade: Machine Learning & Large Language Models
Accurately classifying products is essential in international trade. In this paper, we apply and assess several different algorithms to automatically classify products based on text descriptions. To do so we use agricultural product descriptions from several public agencies, including customs authorities and the USDA. We find that while traditional machine learning models tend to perform well within their training dataset, their precision drops dramatically when implemented outside. In contrast, large language models, such as GPT 3.5 and GPT 4 show a consistently good performance across all datasets, with accuracy rates ranging between 60-90% in all datasets. Our analysis highlights the valuable role of LLMs in product classification at scale and at enhancing the categorization of unstructured data.
 Linkages with Multinationals: The Effects on Domestic Firms’ Exports
In this project we examine whether and how linking up with multinational firms results in improved export performance for domestic firms, using a unique dataset from Uruguay that includes data on firm-to-firm purchases and sales both within and across countries. Our estimation results indicate that selling to a multinational firm is associated with a significant increase in the probability that a domestic firm starts to export, especially to a country where the respective multinational firm is headquartered or has an affiliate. This estimated effect is larger when the multinational firms themselves sell abroad and when the linkage intensity is higher.
 Financial Constraints to Exporting: Evidence from Rwanda’s Export Growth Fund
Export-led growth has long been seen as a key to unlocking structural transformation. In pursuance of this goal, governments often enact industrial policies to encourage firms to overcome barriers to exporting. Despite their prevalence, we have little rigorous empirical evidence of the impact of these policies. In this project, we propose one of the first randomized control trials of industrial policy. We evaluate Rwanda’s Export Growth Fund (EGF), which provides large subsidized loans to exporters and potential exporters at preferential interest rates. We aim to generate exogenous variation in loan take-up through a randomized door-to-door marketing campaign.
 The Value of Organic Certifications
Consumer taste for organic products has grown, increasing demand and costs along agricultural value chains. We estimate the impact of the world’s largest agricultural certification system – USDA organic – on exporters from one of the world’s most important organic production regions – Latin America. For that purpose we use a novel dataset which combines the universe of agricultural export transactions from 10 countries in Latin America with the universe of firm-level organic certifications granted by the US Department of Agriculture since 2012. Our results indicate that certified firms increase exports to the USA relative to non-certified firms. Spillover effects are positive within firms, as certified firms increase exports of certified products to other destinations and non-certified products to the USA. Spillover effects are negative across firms: the more firms are already certified in the USA, the less that exports increase following certification.
IDB Blog Post on early version.
 Information Constraints in Supplier Linkages in Rwanda
In low-income countries multinational firms often do not source from local suppliers, importing a very high share of their inputs. Informational frictions likely deepen existing capacity constraints: large corporations are often unaware of local supplier availability and cannot easily verify their reliability. In this project, we aim to examine the key constraints that prevent large firms from sourcing domestically in Rwanda. We will focus on the role of information constraints related to: (i) identifying reliable suppliers, (ii) assessing the quality standards of potential suppliers and/or (iii) the lack of local knowledge about the exact quality requirements of multinational firms. Our final aim is to conduct a randomized controlled trial (RCT) of a firm-matching platform to connect multinational buyers established in Rwanda with local suppliers.