|
Article Excerpt
Abstract This study examines the features and determinants of intra-industry trade (IIT), horizontal IIT (HIIT) and vertical IIT (VIIT) between Portugal and the European Union in the period 1996-2002, using a static and a dynamic panel data analysis. The findings indicate that Portuguese VIIT increased significantly during the period in accordance with the values expected for a developed country. The regression results show that there is evidence supporting the explanation of VIIT by Heckscher-Ohlin's (HO) theory and that Portugal has comparative advantages in low-quality differentiated products. The findings support the theory that, in general, there is no positive statistical association between HIIT and HO variables. The central theme of this paper is to show that it may be preferable to use the GMM approach in empirical studies of IIT rather than pooled OLS, fixed effects or random effects estimators. The results also suggest that the GMM system estimator obtains more reasonable parameter estimates than the first-differenced GMM estimator.
Keywords Intra-industry trade * Horizontal intra-industry trade * Vertical intra-industry trade * Comparative advantage * Dynamic panel data * GMM-SYS Estimator * GMM-DIF Estimator
JEL Classification C20 * C30 * F12 L10
Introduction
In a multi-country, multi-product and multi-factor world, one may expect to generate intra-industry trade (IIT) on a multilateral basis. Indeed, the majority of the empirical studies are on a multilateral basis, although a number of such studies are based on bilateral trade (Loertscher and Wolter 1980; Bergstrand 1983; Greenaway et al. 1995). The empirical work is used to "test" industry-specific and/or country-specific determinants of IIT. These studies have generally found more empirical support for country characteristics (factor endowments, income levels, level of development, trade imbalance, distance) than for industry characteristics (i.e., market structure, minimum efficient scale, product differentiation and foreign direct investment). Estimated coefficients on proxies for product differentiation and scale economies have often been insignificant or have had the wrong sign. Greenaway et al. (1994, 1995) argue that this may be the result of not separating horizontal differentiated from vertical differentiated trade. Theoretical literature suggests that horizontal and vertical IIT have different determinants, particularly that vertical IIT can be explained by the traditional Heckscher-Ohlin (HO) factor proportions theory.
This study examines IIT, horizontal IIT (HIIT) and vertical IIT (VIIT) between Portugal and the European Union(EU15), using a balanced panel with 21 industries for the period 1996-2000 (for other features of empirical work, namely the evolution of IIT, HIIT and VIIT indexes, the period is 1996-2002). By making the distinction between HIIT and VIIT, this study is expected to perform a more targeted testing of the industry hypotheses.
In static panel data models, Pooled OLS, fixed-effects (FE) and random-effects (RE) estimators are used (Hummels and Levinshon 1995; Zhang et al. 2005). The problems with this type of applied work arise because in these models, serial correlation, heteroskedasticity and endogeneity of some explanatory variables occur and the estimators used do not take this into account. The solution for these econometric problems was found by Arellano and Bond (1991), Arellano and Bover (1995) and Blundell and Bond (1998, 2000), who developed the first-differenced GMM (GMM-DIF) estimator and the GMM system (GMM-SYS) estimator. (1) The GMM-SYS estimator is a system containing both first-differenced and levels equations. In addition to using instruments in levels for equations in first differences, it uses instruments in first differences for equations in levels (Arellano and Bover 1995). The GMM-SYS estimator is an alternative to the standard first-differenced GMM estimator.
In dynamic panel data models, the GMM-SYS estimator eliminates the unobserved industry-specific effects through the equations in first differences. The GMM-SYS estimator also controls for the endogeneity of the explanatory variables. A standard assumption on the initial conditions allows the use of the endogenous lagged variables for two or more periods as valid instruments if there is no serial correlation (Blundel and Bond 1998, 2000). If we assume that the first differences of the variables are orthogonal to the industry-specific effects, this additionally allows the use of lagged first differences of variables for one or two periods as instruments for equations in levels (Arellano and Bover 1995; Blundell and Bond 1998, 2000). The validity of instruments is tested using a Sargan test of the over-identifying restrictions and serial correlation. First-order and second-order serial correlation in the first-differenced residuals is tested using ml and m2 statistics (Arellano and Bond 1991). The GMM system estimator is consistent if there is no second-order serial correlation in the residuals (m2 statistic). The dynamic panel data model is valid if the estimator is consistent and the instruments are valid.
As far as we know, dynamic panel data analysis as not been used in empirical studies of IIT. However, in recent intra-industry studies, production functions, firms' growth, income growth and exports, economic growth, productivity spillovers from foreign direct investment or from multinational corporations, most of the researchers use a dynamic panel data model (Arellano and Bond 1991; Blundell and Bond 2000; Goddard et al. 2002; Agiomirgianakis et al. 2002; Badinger and Breuss 2004; Cuaresma and Worz 2005).
To estimate the dynamic models, we apply the methodology of Blundell and Bond (1998, 2000). The results presented in this paper are generally consistent with the predictions of the theory of intra-industry trade. The regression results demonstrate that there is strong statistical evidence supporting the explanation of VIIT by HO theory. Furthermore, the evidence suggests that Portugal has comparative advantages in low-quality products, as was expected. The results also suggest that HIIT is not explained by comparative advantage determinants (non-qualified labor and physical capital intensity). This was also expected, in accordance with the theory.
The central theme of this paper is to apply the new methodology to IIT studies and to show that better results can be achieved using the GMM approach, rather than OLS, fixed-effects or random-effects estimators. The GMM estimators have the comparative advantage, based on their potential for obtaining consistent parameter estimates even in the presence of measurement errors, omitted variables and endogenous right-hand-side variables. Moreover, our results confirm that compared to the first-differenced GMM estimator, the GMM-SYS estimator obtains more reasonable parameter estimates. The GMM-SYS estimator is preferable to the standard GMM-DIF estimator because it reduces finite-sample biases associated with the latter.
The remainder of the paper is organized as follows. The second section reviews the theoretical literature of IIT models. The third section presents the indexes, the explanatory variables and the sources. The fourth section reports the evolution of the ITT, HIIT and VIIT between Portugal and the European Union over the period 1995-2002. The fifth section presents the static and dynamic panel data models of IIT, HIIT and VIIT and analyzes the estimation results. The final section concludes.
Theoretical Literature
The literature on IIT began to appear in the 1960s with Verdoorn (1960) and Balassa (1965, 1966). These authors became aware that certain developed countries exported and imported products in the same product categories. This phenomenon occurred in the years following the formation of the European Economic Community (EEC). However, it only started to receive increasing attention after Grubel and Lloyd (1975) had introduced an index to measure IIT. After these studies, there was a wide acceptance of the idea that IIT was a more intense phenomenon between countries with similar income levels, a similarity reinforced by the economic integration process. Thus, the traditional HO model could not explain this trade between similarly endowed countries.
The pioneering work in intra-industry models is due to Krugman (1979, 1980), Lancaster (1980), Helpman (1981) and Eaton and Kierzkowski (1984). All these models consider that products are horizontally differentiated (different varieties of a product are of a similar quality). Neo-Chamberlinian models, such...
|
