Spatial effects in real networks: measures, null models, and applications.
This paper provides striking confirmation of the restrictions of the structural gravity model of trade. Structural forces predicted by theory explain 95% of the variation of the fixed effects used to control for them in the recent gravity literature, fixed effects that in principle could reflect other forces. This validation opens avenues to inferring unobserved sectoral activity and multilateral resistance variables by equating fixed effects with structural gravity counterparts. Our findings also provide important validation of a host of general equilibrium comparative static exercises based on the structural gravity model. JEL Classification: F10, F15, R10, R40. ∗A much earlier version of a portion of this work was presented at the NBER ITI meetings, Spring 2010 and the Venice Trade Costs conference, June 2010. We thank participants for helpful comments, especially Dave Donaldson, Keith Head and Michael Waugh. We also thank Arthur Lewbel for helpful comments. †Contact information: James E. Anderson, Department of Economics, Boston College, Chestnut Hill, MA 02467, USA; Yoto V. Yotov, Department of Economics, Drexel University, Philadelphia, PA 19104, USA. The extensive gravity model literature moved from pulp fiction to high brow shelves with the development of the structural gravity model by Anderson and van Wincoop (2003) and its success in explaining the border puzzle posed by McCallum (1995). This paper provides the first empirical test of structural gravity. The results are a gold standard benchmark. Structural gravity forces account for 95% of variation in product/importer/time and product/exporter/time fixed effects estimated from empirical gravity equations for 18 manufacturing sectors and 76 countries from 1990-2002. Similar results are found in a robustness check on different and perhaps special data for 28 goods and services sectors in Canada’s provinces from 1997-2007: 96% of variation is explained. These results provide an empirical justification for comparative static applications of structural gravity. Perhaps more important, they justify inference of unobservable multilateral resistances and unobservable or distrusted sales and expenditure variables from estimated fixed effects and structural gravity restrictions. Gravity estimation following Feenstra (2004) usually features importer and exporter country fixed effects as controls in trade flow equations. Econometric problems of exogeneity and omitted variables are demolished when fixed effects replace the theoretically indicated size and multilateral resistance variables. Another potential advantage is that this specification is agnostic as to whether the fixed effects are explained solely or at all by the structural gravity forces. Bilateral trade cost proxies such as distance have consistently estimated coefficients; but if the structural model is valid, much structural information is lost — the bank building is blown up to get at the safe inside. The methodological novelty of this paper is to compare the patterns of fixed effects in the rubble to a separate reconstruction of structural patterns predicted by theory. Estimated “The gold standard” is a pervasive metaphor in medical and health research for the most certain medical knowledge or best test, the meaning we intend. The metaphor suggests the highly probable fixity of exchange rates in the gold standard era. The connotation of crisis when the test is failed is not intended. Comparative static applications start with Anderson and van Wincoop’s own comparative static experiment with the effect of removing the Canada-US border barrier, and have proliferated in the subsequent literature, especially with variations on the Eaton-Kortum (2002) model that nests structural gravity within a Ricardian production model.