Network of Firms and the Aggregate Economy: Evidence from Costa Rica

You are here

Student Research Report, Summer 2016

A presentation to the Central Bank of Costa Rica. (Photo courtesy of Jose Vasquez.)
A presentation to the Central Bank of Costa Rica. (Photo courtesy of Jose Vasquez.)

Network of Firms and the Aggregate Economy: Evidence from Costa Rica

General Motivation of the Project

There has been a recent effort in better understanding the role of firm linkages in production and aggregate stability. This comes to challenge the original idea of Lucas (1977), who argued that small shocks to particular firms would not matter in the aggregate because the Law of Large Numbers would average out those shocks. In contrast, the more recent view presents the idea that a shock to a single firm (or sector) could have a larger impact if it also affects other firms connected to this firm through a network of input-output linkages.

These new ideas have immediate applications in studying aggregate fluctuations, systemic risk and contagion, international trade shocks, industrial organization, economic geography, among others. In general, it is widely applicable in all areas of economics dealing with supply-chains of production. However, most of the progress in the area has been theoretical and the testable implications of these models have been mainly explored only at the sector level due to data unavailability. So far, almost all papers have focused on input-output matrices to proxy for networks among sectors in the economy. The frequency and detail of these data are also limited. For example, for the US there is yearly information for only 15 industries or decennial for 389 industries.

Research Idea

I am working in this project jointly with my UC Berkeley colleague Isabela Manelici and Alonso Alfaro, who works at the Research Department of the Costa Rican Central Bank. We want to study whether Costa Rican firms experience productivity increases when they become part of the supply-chain of a multinational firm or when they are connected to a firm receiving Foreign Direct Investment (FDI). It is often argued that FDI can encourage productivity spillovers to domestic firms in developing countries. The superior technology of these foreign firms could lead to new technology adoption in the host country or could increase domestic productivity through knowledge flows. These arguments are pillars for policy making in many developing countries – such countries not only reduce barriers to FDI and foreign firms’ entry but also offer special incentives (like tax holidays) to attract them.

Costa Rican policy makers have adhered to this trend and have put in place an aggressive strategy to attract foreign firms to the country. However, empirical evidence thus far on whether FDI leads to productivity growth is still open to criticism. Data unavailability or the lack of a credible research design are frequent concerns. We want to overcome these difficulties using administrative records of transactions among all Costa Rican firms. With this data I will be able to construct and understand the complete network of production of the Costa Rican economy and in particular, the supply-chain connecting domestic and multi-national firms. To study the causal effect on domestic firms we make use of a Matching Program used by the Ministry of Commerce to establish contacts between foreign firms and domestic suppliers. This program creates short-lists of potential supplier candidates for each major multi-national entrant. We plan to use “losers” from the short-list as a control group for those firms which ultimately established business relations with the entrant foreign firms. Next steps involve the study of mechanisms through which spillovers occurred.

A New Dataset: Data from Costa Rican Tax Records

As noted above, further progress on this area of research is limited by data availability. However, the way Costa Rican corporate taxes are collected brings a new design that pushes the frontier on these topics.

We use a novel firm-level dataset coming from confidential administrative records covering all Costa Rican firms. The Ministerio de Hacienda (the Costa Rican Ministry of Finance) collects yearly information on all business-to-business transactions that have a value larger or equal than 2.5 million colones (approximately 5,000 $US). Providing this information is compulsory to all firms. Ministerio de Hacienda uses this data to prevent corporate tax evasion. Costa Rica imposes increasing average tax rates on profits as a function of firms’ revenue (for more details on the Costa Rican tax structure see Bachas and Soto (2016)). To prevent firms  from inflating their costs or deflating their revenues, the Ministerio de Hacienda requests all firms to report the exact amount of their transactions and the identity of their seller or buyer. Then, the Ministerio de Hacienda cross-checks the veracity of their reports. For example, if Firm A reports selling (buying) $X to Firm B, then the Ministerio de Hacienda checks if Firm B reports buying (selling) $X to Firm A.

Accomplished Objectives of the Exploratory Trip

The nature of the dataset is highly confidential and the usage is very restrictive even within the Ministerio de Hacienda and the Costa Rican Central Bank. The purpose of the trip was to negotiate access to the data with the authorities of the Central Bank. For this reason, it was important to explain the benefits of the project for policy making and to propose a protocol which would allow us to conduct the research using these data while protecting the confidentiality of the firms involved. 

The negotiation process took several steps from which I gained several insights that will enormously benefit my career in the future. The first step was to initially approach the institution at the end of 2015. I decided to discuss first with the research department of the Costa Rican Central Bank because they understand very well the importance of academic research and challenges to be faced when conducting it. It was also important to convince them that the topic of research was not only of interest in academic circles but also very relevant for the Central Bank and the Costa Rican Government in general.

A second step was to ask the director of the research department to consult on the proposed collaboration with the highest hierarchy of the Bank. For this I had to write a formal request and a research proposal in Spanish (available upon request). It was very important for the Central Bank to have the guarantee of our commitment to foremost protect the confidentiality of the data. 

Their acceptance of our proposal allowed me to have access to the data from their secured computer inside their main building in San Jose, Costa Rica and to interact with researchers and directors of the Bank, which I found as a mutually enriching experience.

The third step was to present my progress to the authorities of the Central Bank towards the end of my stay. This allowed me to show them the preliminary data work, to discuss the conceptual progress in defining the research idea and to inform them about the next challenges of the project. Only then, I felt confident enough to ask for the formalization of the collaboration starting this summer until the completion of the project. The usual length of these projects from the start to the moment of publication could take several years. For this reason it is  important that all parts involved are fully committed to it.

The fourth and final step was to discuss a protocol in which the collaboration could take place remotely from now onwards. My teaching responsibilities during the academic year substantially restrict my geographic mobility. For this reason, it was important for me to establish a communication channel through which we could make constant progress in the data work without being in person in the Central Bank. The confidential nature of the data precluded me from taking even a sample of it. Given these circumstances I would produce codes and programs in different statistical/econometric software that I would write from Berkeley and our institutional coauthor (Alonso Alfaro) will adapt them to be compiled in the secured computer. Fortunately, the Central Bank agreed on sharing all aggregated results with us as long as they do not compromise the identity of any of the individual tax payers.

Next Steps

At this point we are finishing to merge different sources in order to create a final dataset on which the main analysis will be conducted. The data contains approximately 40 thousand Costa Rican firms per year between 2008 and 2015, records almost 4 million transactions. The dimensionality of the data could be one of its main strengths but at the same time imposes a huge burden on the initial construction of the ready-to-use data. This is work in progress and could take several months.

The next step would be to actually analyze the data. We already count with an empirical strategy that we will implement and we are conducting data analysis to understand the mechanisms through which productivity increases can occur. This will be the most exciting part but at the same time the most challenging one. We are excited to get to that point and hopefully have results that can inform policy making in Costa Rica.


Bachas, P. and Soto, M. (2016). Not(ch) your average tax system: Corporate taxation under weak enforcement. Working Paper.

Lucas, R. E. (1977). Understanding Business Cycles. In Carnegie-Rochester Conference Series on Public Policy, North-Holland, vol. 5, pp. 7–29.

You are here