Examining Trade Response of Armington-Krugman-Melitz Encompassing Module in a CGE Model

Computable General Equilibrium (CGE) models have been widely used for quantifying economic impacts of free trade agreements and economic partnership agreements. For the recent examples, it is estimated that Trans-Pacific Partnership (TPP) will increase Japanese real GDP by 0.66%, according to Cabinet Secretariat (2013) in Japan. Pacific Economic Cooperation Council (2012) also estimated that the impact of TPP on Japanese real GDP would be 2.0% higher by 2020. Both of the estimates are based on simulation results obtained from global CGE model; the former uses the GTAP model (Hertel, (1997), and McDougall (2003)), and the latter develops their own global CGE model (Zhai (2008), and Petri et al. (2012)). The difference in the estimated economic effects seems to be large, however, it is not surprising since the components taken into their estimates are different. Petri et al. (2012) considers exhaustive components of liberalization; such as removing tariffs, reducing non-tariff barriers, liberalizing trade in services and foreign direct investment. On the other hand, Cabinet Secretariat (2013) estimates the impact of removing tariffs, thereby resulted in the lower estimate. Beside the difference in the components of liberalization, it is more interesting for us to ponder the difference in trade specification used in their global CGE model. Petri et al. (2012) define their trade module by following Melitz (2003) based on product differentiation at the firm level. The GTAP model has been using the conventional Armington (1969) specification based on product differentiation at the country level. Thus, we are interested in comparing different trade specifications in global CGE model and its implications for resulting estimates of economic impacts of trade liberalization.This paper introduces the AKME module following the modeling strategy in Dixon and Rimmer (2012) and Oyamada (2013). We modify the GTAP model (Hertel, 1997), which is a global CGE model widely used by researchers for quantifying policy impact. We redefine trade flow information stored in the benchmark GTAP Data Base, and implement a calibration procedure established in Oyamada (2013) and Oyamada (2014b). We run simulation of trade liberalization to draw a comparison between different trade specifications, decomposing the trade response in detail. Since there exits only a handful of attempts to compare the trade effects by examining the AKME module, we provide another results for further insights. Impacts of liberalization on regional trade are amplified as we switch trade specification from the standard GTAP model to Armington, Krugman, and Melitz in turn. By introducing “sourcing-by-agent”, we can decompose the simulation results on regional imports into agent specific demands, which is not available in the standard GTAP model. Also with the sourcing-by-agent, we can identify the intra-manufactured trade flows as the largest share. Further decomposition reveals that intensive margin trade effects are more pronounced in Krugman specification, whereas extensive margin trade effects are significant in Melitz specification. These decompositions clearly enrich our interpretation of trade liberalization.