Unpacking sources of comparative advantage: A quantitative approach

https://doi.org/10.1016/j.jinteco.2010.07.004Get rights and content

Abstract

This paper develops an approach for quantifying the importance of different sources of comparative advantage, by extending the Eaton and Kortum (2002) model to predict industry trade flows. In this framework, comparative advantage is determined by the interaction of country and industry characteristics, with countries specializing in industries whose production needs they can best meet with their factor endowments and institutional strengths. I estimate the model parameters using: (i) OLS; and (ii) a simulated method of moments procedure that accounts for the prevalence of zeros in the bilateral trade data. I apply the model to explore various quantitative questions, such as how much distance, Ricardian productivity, factor endowments, and institutions each matter for country welfare in the global trade equilibrium.

Introduction

The past few years have seen a resurgence in empirical work on sources of comparative advantage, namely those forces, such as country differences in productivity or factor endowments, that determine patterns of specialization and trade. On the Ricardian model, it is only recently that Eaton and Kortum (2002) showed how an appropriate parametrization for the underlying distribution of productivity levels can deliver a tractable expression for trade flows in a general multi-country setup. The good fit of their model to the manufacturing trade data delivered an important piece of evidence on the role of productivity differences in determining comparative advantage (Eaton & Kortum, 2002, Costinot and Komunjer, 2007).1 Separately, several studies have reaffirmed the role of factor endowments and the Heckscher–Ohlin framework, showing in particular that countries' relative endowments are informative of their pattern of trade (Debaere, 2003, Romalis, 2004).2 Moving beyond this neoclassical focus, a recent cluster of work has identified the influence of country institutions on international trade. Such institutional sources of comparative advantage include: financial development (Beck, 2003, Manova, 2008), the security of contract enforcement (Levchenko, 2007, Nunn, 2007, Costinot, 2009), and labor market flexibility (Cuñat and Melitz, 2009).3

This paper seeks to quantify the importance of these different sources of comparative advantage for country welfare within a common framework. I present an extension of the Eaton–Kortum (EK) model that goes beyond aggregate trade volumes to explain the cross-country pattern of specialization and industry trade flows. In the model, the productivity level of firms is composed of a systematic and a stochastic component, where the systematic component is driven by the interaction between country and industry characteristics. The motivation for this is intuitive: industries vary in the factors and institutional conditions needed for production, and countries differ in their ability to provide for these industry-specific requirements. Comparative advantage therefore stems from such country–industry matches.

At heart, this specification draws on a recent body of work that identifies comparative advantage from such interactions between country and industry characteristics. Romalis (2004) applied this logic to test for Heckscher–Ohlin forces: by interacting countries' relative factor abundance with an industry measure of factor intensities in production, he showed that countries capture a larger US market share in industries that use their abundant factors more intensively.4 The literature on institutional determinants of trade has also adopted this empirical strategy, by applying or constructing novel measures of an industry's dependence on specific institutional conditions. Beck, 2003, Manova, 2008 interacted country measures of private credit availability with an industry measure of external capital dependence, to show that countries with better financial development export more in industries that rely heavily on external financing.5 Similarly, several studies have shown that countries with better rule of law export relatively more in industries that are more exposed to holdup problems or other institutional frictions, as measured by input concentration (Levchenko, 2007), the share of customized inputs (Nunn, 2007), or job task complexity (Costinot, 2009).6 Cuñat and Melitz (2009) have further demonstrated that countries with flexible labor markets facilitate exports in more volatile industries, these being the industries that benefit most from being able to adjust employment margins regularly.

The model that I present in Section 2 provides a common interpretation for the estimation being performed in this recent literature, by embedding these specifications within the multi-country setting of the EK model. Conveniently, the model delivers an analytic expression for trade flows at the industry level that resembles a gravity equation, which incorporates a role for distance barriers, Ricardian forces, Heckscher–Ohlin forces, and institutional determinants in explaining trade volumes. The model can thus be readily taken to the data.

Section 3 estimates the trade flow expressions using a fixed effects ordinary least squares (OLS) specification. For the empirical implementation, I assembled a dataset of bilateral industry trade flows, distance measures, as well as country and industry characteristics for a sample of 83 countries and 20 manufacturing industries, including a comprehensive set of all the country–industry interaction terms from the papers cited above. I find strong evidence for the importance of factor endowments, financial development, legal institutions, and labor market regimes as sources of comparative advantage, even when all interaction terms are run in one regression. This represents a first exercise (to the best of my knowledge) at jointly verifying the significance of this extensive a list of trade determinants, while also facilitating a first-pass comparison of their quantitative importance.

While OLS provides a useful baseline, it nevertheless suffers from the drawback that zero trade observations are dropped when the log of trade flows is the dependent variable. These zeros constitute about two-thirds of the data, and discarding this information could systematically bias the OLS coefficients (see Santos-Silva & Tenreyro, 2006, Helpman et al., 2008, among others).7 To address this, Section 4 pursues a minimal modification of the model to generate zero trade predictions. Specifically, I impose a bounded support on the distribution of the stochastic productivity component, so that a country with a low systematic productivity may thus never receive a large enough productivity shock to be able to export a good to a given market. As we however lose closed-form expressions, I instead implement a simulated method of moments (SMM) procedure to obtain a separate set of parameter estimates, by matching key statistical moments in the actual data with those simulated from the model (Pakes and Pollard, 1989).8

Using the SMM estimates, I briefly explore two sets of quantitative exercises in Section 5. A first set of counterfactuals relates to distance and geography. The model implies a sizeable average increase in country welfare (15.7%) from a hypothetical reduction of distance measures to their minimum value, comparable to what EK (2002) find for their OECD sample (16.1%–24.1%). Second, the framework allows us to assess the relative importance of the various sources of comparative advantage for country welfare. I explore this by shutting down the relevant terms in the empirical model that capture each comparative advantage force, namely the ability of a country to benefit from relative producer price differences across industries vis-à-vis other countries stemming from the source of comparative advantage in question. The simulations indicate that each of the individual Ricardian, Heckscher–Ohlin and institutional determinants are comparably important, with the impact of neutralizing each being similar to doubling the distance markup faced by the country.

This paper falls within a broader body of research seeking to quantify the importance of different determinants of trade flows. Past work has examined for example the welfare effects of moving to a zero-gravity world (EK, 2002), border effects (Anderson and van Wincoop, 2003), and tariff liberalization (Lai and Trefler, 2002, Lai & Zhu, 2004, Alvarez & Lucas, 2007). The approach I develop here goes a step further in enabling the researcher to explore the role of country and industry characteristics in influencing the pattern of comparative advantage. In doing so, I build on recent studies which have sought a more holistic view on the determinants of trade by incorporating both Ricardian and Heckscher–Ohlin forces within a common setting (Harrigan, 1997, Morrow, 2008, Burstein and Vogel, 2009). In this regard, a closely related paper is Shikher (2010), who also extends the EK model to the industry level. Empirically, Shikher calibrates the technology parameters to fit the output and trade data, whereas the approach that I take will instead link these productivity parameters to observable country and industry characteristics.

The roadmap of the paper is as follows. Section 2 extends the canonical EK model to the industry level. Section 3 presents the OLS results. I modify the model in Section 4 to incorporate the zero trade flows, and re-estimate it with the SMM procedure. Section 5 briefly explores some counterfactuals. Section 6 concludes. The Data appendix at the end of this paper provides an overview of the dataset. Further details on the data variables and the SMM estimation are documented in an online Supplementary Appendix for interested readers.

Section snippets

The basic setup

Consider a world with n = 1, …, N countries. There are K + 1 industries, indexed by k = 0, 1, …, K. Industry 0 denotes non-tradables, which is a homogeneous good sector. The tradable sectors (k  1) feature differentiated products, where the continuum of varieties in each industry is indexed by jk  [0, 1]. (The measure of varieties in each industry is normalized to 1.) I proceed to build the model in stages.

Estimation by OLS

I turn now to the task of estimating the model. It turns out that the regression model for trade flows implied by Eq. (11) resembles closely that in existing empirical work based on standard OLS methods. The OLS exercise which follows thus provides a basis for comparison and corroboration with the current literature on sources of comparative advantage. We will address the potential bias from the omission of zero trade flows later in Section 4 using the SMM procedure.

Estimation by simulated method of moments (SMM)

In keeping with the Ricardian spirit of the EK model, the approach I pursue is instead to view the zero trade flows as arising from large productivity gaps between countries, which prevent low productivity countries from exporting to particular markets. This requires a slight modification of the model to generate zero trade predictions. The underlying parameters can then be re-estimated by matching moments of trade flows simulated from the model with the corresponding moments from the actual

Counterfactuals

With the SMM estimates, we can now fully parameterize the model and explore several counterfactual exercises. It should be emphasized that the scenarios considered here are inherently hypothetical: While it is easy to mechanically perturb the model along a dimension of interest, one cannot in reality reduce the physical distance between countries or neutralize a country's ability to leverage on its endowments or institutions. The effects computed below should instead be seen as a rough gauge of

Conclusion

This paper presents an approach for quantifying the importance of different sources of comparative advantage within a common modeling framework. To understand patterns of specialization, I present an extension of the multi-country model of Eaton and Kortum (2002) to explain trade flows at the industry level. The model expresses comparative advantage as a function of country–industry matches, so that countries specialize in those industries whose production needs they can best meet with their

Acknowledgements

I am grateful to the editor, Daniel Trefler, as well as to Pol Antràs, Gary Chamberlain, Elhanan Helpman, and Marc J. Melitz for their patient advice. Thanks also to Leon Berkelmans, Filipe Campante, Arnaud Costinot, Jonathan Eaton, Doireann Fitzgerald, Hiroyuki Kasahara, Samuel Kortum, Kalina Manova, Brent Neiman, Nathan Nunn, Adrian Ow, Parag Pathak (who started as a co-author), Natalia Ramondo, Roberto Rigobon, Kartini Shastry, and Jeffrey Williamson, as well as seminar audiences at Harvard,

References (60)

  • J. Anderson et al.

    Gravity with gravitas: a solution to the border problem

    The American Economic Review

    (2003)
  • J. Anderson et al.

    Trade costs

    Journal of Economic Literature

    (2004)
  • R. Baldwin

    Determinants of the commodity structure of US trade

    The American Economic Review

    (1971)
  • R. Baldwin

    Determinants of trade and foreign investment: further evidence

    The Review of Economics and Statistics

    (1979)
  • E.J. Bartelsman et al.

    NBER-CES Manufacturing Industry Database

    (2000)
  • T. Beck et al.

    A new database on financial development and structure

    World Bank Economic Review

    (2000)
  • T. Beck

    Financial dependence and international trade

    Review of International Economics

    (2003)
  • B. Becker et al.

    Financial Development, Fixed Costs and International Trade

    (2007)
  • D. Berkowitz et al.

    Trade, law, and product complexity

    The Review of Economics and Statistics

    (2006)
  • A. Bernard et al.

    Plants and productivity in international trade

    The American Economic Review

    (2003)
  • J. Botero et al.

    The regulation of labor

    Quarterly Journal of Economics

    (2004)
  • A. Burstein et al.

    Globalization, Technology, and the Skill Premium

    (2009)
  • A. Costinot et al.

    What goods do countries trade? New Ricardian predictions

  • Cuñat, A., Melitz, M.J., 2009. Volatility, labor market flexibility, and the pattern of comparative advantage. Journal...
  • D. Davis et al.

    An account of global factor trade

    The American Economic Review

    (2001)
  • A. Deardorff

    Testing trade theories and predicting trade flows

  • P. Debaere

    Relative factor abundance and trade

    Journal of Political Economy

    (2003)
  • A.K. Dixit et al.

    Theory of International Trade

    (1980)
  • R. Dornbusch et al.

    Comparative advantage, trade, and payments in a Ricardian model with a continuum of goods

    The American Economic Review

    (1977)
  • J. Eaton et al.

    Technology, geography, and trade

    Econometrica

    (2002)
  • Cited by (219)

    • Robots, tasks, and trade

      2023, Journal of International Economics
    • Immigration and occupational comparative advantage

      2023, Journal of International Economics
    View all citing articles on Scopus
    View full text