اثر خانه بازار و الگوهای تجارت دو جانبه: بازنگری در شواهد

This paper finds that the evidence for the home market effect (HME) found by Hanson and Xiang (AER, 2004) is sensitive to the way the dependent and the independent variables are constructed. Second, we also find that the HME evidence goes away when we estimate their difference-in-difference gravity model on a truncated sample of positive trade flows. With Eaton–Tamura–Tobit, Heckman, and Helpman–Melitz–Rubinstein estimation of the gravity equation using Hanson and Xiang's data, we are unable to find any evidence for the HME. Finally, the HME evidence is also absent for a sample of Canadian provinces' exports to U.S. states. All of our results, taken together, do not reject the existence of the HME in general but rather suggest that the HME results found by Hanson and Xiang may not be robust.

Hanson and Xiang (2004) develop a multi-sector, monopolistic competition model and use it to reveal a systematic relationship between the strength of the home-market effect and industry characteristics.1,2 The multisectoral nature of their model, by suggesting “treatment” and “control” sectors, allows them to devise a difference-in-difference gravity approach to empirically test the home-market effect. The home-market prediction is that industries with high transport costs and low substitution elasticities (more highly differentiated products) will tend to be more concentrated in large countries than industries with low transport costs and high substitution elasticities. Hanson and Xiang treat the former industries as “treatment” industries and the latter as “control” industries. Using this innovative approach, they are able to address major econometric concerns about earlier tests of the home-market effect, including possible correlation between industry demand and supply shocks and a failure to control for “remoteness” of exporting and importing countries, both of which can lead to biased coefficient estimates. Because Hanson and Xiang's approach provides a novel and potentially quite useful methodological breakthrough, we examine the robustness of their findings to changes in data handling, changes in sample, and changes in estimation procedure. Overall, the weight of evidence from reasonable amendments to the Hanson–Xiang methodology and from a large number of robustness check runs against the presence of a significant home-market effect in trade flows.

In this paper, we reexamine closely the empirical evidence in support of the home-market effect in Hanson and Xiang (2004). We first show that their difference-in-difference gravity specification produces estimates that are sensitive to the way the independent variable is created when a constant term or a squared term of the independent variable of interest is included in the regression. Our empirical analysis using Hanson and Xiang's sample shows that the difference-in-difference gravity specification in which the dependent variable or/and the independent variable is constructed using different configurations of exporter pairs yields completely conflicting results on the HME.