The geography of gravity

ABSTRACT

Although geography has been considered an important factor in international trade, spatial heterogeneity has not been fully investigated in standard gravity models. This paper contributes to the literature by investigating how gravity works geographically in trade. The geographically weighted regression (GWR) reveals spatial variations in estimated parameters. GWR regression results suggest even though physical distance is the same, economic distance can be different based on the location. Our regression results on the impact of Kyoto protocol show that while a loss of competitiveness is observed among non-European developed countries, no loss of competitiveness is clearly seen in the European Union.

KEYWORDS:

• Gravity model
• geographically weighted regression
• spatial heterogeneity

Acknowledgments

The authors are grateful for the financial support of the Institute of Economic Research at Aoyama Gakuin University. Ko acknowledges financial support from the Grant-in-Aid for Young Scientists (B) (Grant Number 15K17095) from JSPS.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The data described in this article are openly available in the Open Science Framework at.

Notes

1. For example, Disdier and Head (2008) and Head and Mayer (2014) perform meta analysis by collecting a large set of estimates in the literature, to find that the elasticity of trade with respect to distance is -0.95 since 1990.

2. We also check the robustness by using the importer’s location. Regression results are robust.

3. In this case, there is an inconsistency with the OLS estimator because of the dependency between the error term of the transformed log-linear model and the regressors.

4. Nakaya et al. (2005) analyze the disease patterns resulting from spatially non-stationary processes.

5. AICc is a measure of the relative quality of statistical models for a given dataset.

6. We also perform estimation with literally zero trade data. However, we find that the estimated values are too broad and in several cases the signs are opposite for these pairs of countries. As already discussed in the literature, there are many problems with zero trade from the perspective of estimation and quality of data. For example, zeros can just be missing observations. Hence, in this paper, we show the results with positive trade flows.

7. For further details, see SST.

8. The PPML results are from SST.

9. Central America covers Costa Rica, El Salvador, Mexico, and Panama, and the northern part of South America includes Colombia, Ecuador, Suriname, and Venezuela.

10. The corresponding global estimate is -1.163.

11. We include Comoros, Madagascar, Malawi, Mauritius, South Africa, Zambia, and Zimbabwe.

12. We do not show the results with zero trade flow data. Empirically, we can overcome the zero flows problem with GWPR, but it turned out that the signs of the estimated results with zero flows are often opposite to the literature.

13. Decreased trade flows are excluded.

14. We do not control variables for endogeneity because our main interest is showing the existence of regional differences.

Additional information

Funding

This work was supported by the Institute of Economic Research at Aoyama Gakuin University [15K17095].