Desktop version

Home arrow Business & Finance arrow Globalisation, comparative advantage and the changing dynamics of trade

Empirical methodology

The empirical model

The empirical methodology employed in this chapter is based on Chor (2010) who extends the aggregate Eaton-Kortum model of trade (Eaton and Kortum, 2002) to account for industry trade flows. In Chor (2010) the non-random component of productivity level of firms operating in a given industry is determined by the interaction between country and industry characteristics. He motivates this approach in the following way: “industries vary in the factors and institutional conditions that they need for production, and countries differ in their ability to provide for these industry-specific requirements.” The interaction approach draws on classical trade theories as well as on the recent body of empirical literature dealing with individual institutional sources of comparative advantage. For instance, Romalis (2004) interacted country-level measures of factor abundance with industry-level measures of factor intensities, as posited by the Hecksher- Ohlin-Samuelson theory. Braun (2003), Beck (2003) and Manova (2008) interacted country measures of credit availability with industry measures of dependence on external financing. Levchenko (2007) interacted a measure of input concentration with indicators measuring the quality of the rule of law. Nunn (2007) and Costinot (2009) conducted similar analyses of the rule of law using, respectively, measures of share of customised inputs and of job task complexity. Cunat and Melitz (2007) interacted a measure of labour market flexibility with a measure of industry sales volatility.

Modifying Chor’s notation to facilitate exposition the empirical model of bilateral exports at the industry level can be defined as follows:

where Хуг are exports of industry к from country i to country j in year t. dt and 3ft are, respectively, exporter fixed effects and importer-product-year fixed effects. The former type of fixed effects allow us to capture all unobserved exporter characteristics that are not interacted with any industry characteristics (such as the size of exporter’s GDP, its GDP per capita or exchange rate). The latter type of fixed effect terms account for all unobserved importer-product-year characteristics and in particular for any unobserved demand or, indeed, comparative advantage factors specific to a particular importer (e.g. the fact that a certain importer is an exceptionally significant demander of a specific commodity). With such a specification of fixed effects the variation in bilateral exports at industry level is left to be explained by relative differences in exporters’ abilities to produce certain goods which stem from interactions of exporter’s i characteristics with characteristics of industry к, as well bilateral factors such as distance In(disty'), common language( langу), common border (border^-), colonial relationship {colony^), which offer a natural benchmark for comparison of impacts for the policy and institutional variables.

The endowment, policy and institutional interaction terms are presented in the second

line of equation (1) with (-) x KINTk signifying the interaction of physical (or human)

capital-to-labour ratios in exporter i in year t with physical (or human) capital-intensity of sector k. The interactive terms PNit x PDNk signify interaction between the indictor of n-th institution or policy for exporter i in year t with an indicator of dependence of sector k on institution or policy n. One example of such an interaction from the existing literature would be an interaction of the World Bank index of labour market flexibility with an industry-level indicator of sales demand volatility as in Cunat and Melitz (2007).

Equation (1) embeds several earlier empirical specifications of determinants of exports proposed by the literature (e.g. the gravity model of trade) and allows including as many country and industry interactions as one is capable of measuring and handling econometrically. The approach decomposes determinants of trade flows and allows capturing how well the conditions in country i provide for the production needs of industry k. Consequently, estimation of parameters of equation (1) allows assessing the relative importance of various sources of comparative advantage in the sample. For instance, it allows determining whether differences in physical capital-to-labour ratios across the sample have been more important in determining the industry pattern of trade flows as compared to differences in financial development. In addition, the estimated parameters can be interpreted in the context of cross-country variation in country characteristics to shed light on trade implications of any potential future changes in these country characteristics on a ceteris paribus basis (e.g. trade effects of aligning a given country’s policy with an average or with the level of best performing peers).

< Prev   CONTENTS   Source   Next >

Related topics