Introduction

Computable general equilibrium (CGE) models have been used extensively in trade policy analysis. Though they shed considerable light on the static welfare effects and structural adjustment of trade reform around the world, they fail to capture some important features of modern international trade.¹ The most striking is the extensive margin, i.e., the number of exporting firms and traded goods. In the standard CGE model with Armington’s (1969) national product differentiation, trade expands purely at the intensive margin: each exporter increases the size of its exports, but there is no change in the set of exporters. However, recent research has revealed the importance of the extensive margin for international trade. Empirical studies show that larger countries trade not only larger volumes, but also a wider variety of goods. Using data on shipments by 126 exporting countries to 59 importing countries in 5,000 product categories, Hummels and Klenow (2005) find that the extensive margin accounts for 60% of the increase of exports of larger economies, and about one third of that of the same countries. Eaton, Kortum and Kramarz (2004) examine firm-level export data of French firms and conclude that the number of exporting firms, rather than the amount exported by each firm, determines the variation across destinations. The extensive margin is also a crucial force behind trade expansion following liberalization. In a study of six different instances of liberalization, Kehoe and Ruhl (2003) find that trade in goods that were not traded earlier shows substantial growth following a decrease in trade barriers. A set of goods that accounted for only 10% percent of trade before the liberalization may come to account for 40% following it.

The absence of the extensive margin makes trade CGE models incapable of explaining the rapid world trade growth since the 1960s, leading to a quantitative puzzle of why modest decreases in tariffs generate a large expansion of trade (Bergoeing and Kehoe, 2001; Yi, 2003). For example, Kehoe (2005) uses data on actual changes in trade flows among Canada, the United States, and Mexico between 1988 and 1999 to evaluate the performance of three CGE models that were used in the early 1990s to estimate the impacts of NAFTA. He finds that these models dramatically underestimated the impact of NAFTA on the volume of regional trade, especially for Mexico. Mexico’s regional trade relative to GDP increased by over 1,000% in many sectors between 1988 and 1999, while the CGE models predicted changes in trade relative to GDP of less than 50% in most sectors.

The absence of the extensive margin in Armington-type CGE models also results in the wellknown “stuck on zero trade” problem (Kuiper and van Tongeren, 2006). The Armington specification has the effect of locking in pre-existing trade patterns and prevents the models from generating large changes in trade in sectors with little or no trade. Under this specification, if a country’s imports of a given product from another country are zero initially, they will always be zero, even after significant reductions of trade barriers. If imports are nonzero but small, they will remain small even if there are large changes in prices. This “stuck on zero trade” problem makes CGE models especially inappropriate for the least developed countries, which usually have limited trade with the rest of the world.

The Armington model typically specifies a constant elasticity of substitution (CES) utility function over the home and import goods, and explains the trade pattern by the relative prices of goods produced in different regions and the fixed Armington taste (share) parameters. These fixed taste parameters, usually obtained from the calibration using observed trade flows, are essentially a black box to the model. There is no economic theory underlying the choice of these parameters. As argued by Hillberry et al. (2005), just like error terms in econometric models, the Armington taste parameters serve the role of containing the unexplained variance in the bilateral trade flows in CGE models. There "error terms" tend to be large. Hillberry et al. (2005) examine the trade patterns of 50 commodities and find that the economic behavior modeled by the Armington model explains less than 20% of the variation in bilateral trade flows. Thus, in Armington-type CGE models, the trade patterns are largely determined by the fixed taste parameters, but these parameters are not explained by the model.

There have been some efforts toward incorporating the extensive trade margin into standard CGE models. Some keep the CES-based Armington structure, but seek to endogenize the Armington taste parameters. The MONASH and USAGE models developed by Monash University conduct "historical simulations" to estimate the historical trends in the movement of productivity and taste, and use them to update related parameters in the baseline forecasts (Dixon and Rimmer, 2002; 2003). Kuiper and van Tongeren (2006) propose an exogenous alteration of the Armingtion taste parameters based on separately estimated econometric gravity equations. Some CGE modelers opt for more general function forms in place of CES functions to model import demand. For example, Robinson et al (1993) and Weyerbrock (1998) use the AIDS function (Almost Ideal Demand System) while van der Mensbrugghe (2005) uses an extended CES function with minimum demand shifters a la LES (Linear Expenditure System) to capture the non-unitary income effects for import demand.

The recent trade theory incorporating firm-level heterogeneity offers an additional possibility for introducing the extensive margin into CGE models in a theoretically coherent way. In the models by Bernard, Eaton, Jensen and Kortum (2003), Melitz (2003), and Yeaple (2005), the patterns of trade are determined by variations in a number of factors, such as market size, number of firms, technology and trade barriers, rather than the fixed "taste" parameters. These models introduce the extensive margin as a result of the firms’ self-selection to export markets. They emphasize the interaction of trade costs and productivity differences across firms operating in imperfectly competitive industries. The existence of trade costs induces only the most productive firms to self-select into export markets. When trade costs decrease, new firms with lower productivity enter the export markets in response to the potential higher profits. On the other hand, the least productive non-exporting firms are forced to exit because of the increased import competition in domestic markets. Empirical evidence has largely supported the predictions by these new firm-heterogeneity trade models.²

One attractive feature of the new firm-heterogeneity trade models is that they provide an explicit microeconomic channel through which trade liberalization boosts aggregate productivity. Under these models, productivity gains originate from the reallocation of economic activity across firms within an industry, as low productivity firms exit and high productivity exporting firms expand their market shares following the trade liberalization. The productivity effects of trade liberalization are key factors for understanding the impact of trade liberalization, but are often missed in CGE models.³

The introduction of the extensive trade margin in CGE models has important implications for evaluating the welfare effects of trade liberalization. In CGE models with differentiation of national products, the simulated welfare changes of trade liberalization are dominated by the terms-of-trade effects associated with intensive export growth, i.e. expanding export quantity but lower export price of each variety (Brown, 1987). However, as mentioned by Hummels and Klenow (2005), if an export expansion is based more significantly on the extensive margin or high quality, such adverse terms-of-trade effects are no longer a necessary consequence.

In order to improve the ability of applied trade models to describe the facts of trade, this paper attempts to incorporate recent developments in heterogeneous-firm trade models into a global CGE model. Specifically, a firm heterogeneity global CGE model based on Melitz (2003) is used, and experimental simulations are carried out to illustrate its features. The remainder of the paper is organized as follows: the next section presents the Melitz (2003) model. Section 3 discusses the specifications of the heterogeneous firm CGE model and its calibration. Section 4 presents the simulation results of trade liberalization using the new CGE model, and compares them to those obtained from a standard Armington-type CGE model. Section 5 is a sensitivity analysis. The final section offers conclusions.

Download this Discussion Paper [ PDF 227.1KB| 29 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.