Gravity modeling in international economics has tended to focus on trade, particularly merchandise trade: the subsection on “Gravity models beyond trade in goods” in Head and Mayer’s chapter on gravity in the Handbook of International Economics occupies less than one page out of a total of 65. But globalization is generally defined as being broader than trade in just products or even services: dictionary definitions typically refer to cross-border movements of capital, people and information as well. How well does gravity work when we look at a broad range—12, to be precise--of cross-border interactions?
This paper provides some basic descriptive analysis that aims to address that question. It finds, first of all, that simple gravity models that focus on the sizes of economies and the geographic distance between them work fairly well across the range of interactions examined, i.e., gravity applies generally to globalization, not just to trade. Second, the additional types of distance customarily included in augmented gravity models of trade tend to boost explanatory power appreciably in the context of other types of interactions as well, i.e., to apply across the board when we take a broad view of globalization. Third, while some further increases in explanatory power can be achieved by customizing the additional explanatory variables to specific types of interactions, the gains are typically quite limited. Fourth, countries that are relatively distant from each with respect to a particular type of flow also tend to be distant as far as other flows are concerned (i.e., despite variations in the estimated coefficients on the distance variables, the ranking of countries that are close in terms of composite distance versus those that are far apart does not shift very much). Finally, geographic distance in particular is a fair proxy for augmented distance. Some implications of and extensions to the present analysis are also discussed.
Globalization, Gravity and Distance
Recent survey evidence confirms that popular opinions about globalization are still dominated by the notion of a flat world in which neither national borders nor distance matter very much. But empirical evidence suggests that international interactions are far more limited than such a worldview would imply and that the bulk of cross-border flows take place between countries that are close to each other culturally, administratively, geographically and economically: the “law of distance”.
The law of distance was originally based not on a theoretical model but on observed empirical regularities. The most interesting research on the factors underlying is contained in the thousands of studies that use “gravity” models to study international flows, particularly trade. The nomenclature highlights the analogy with Newton’s law of gravity: gravity models in international economics link interactions between countries to the product of their economic masses divided by some composite measure of distance. Such gravity models not only help us to understand why, for instance, the U.S.-Canadian trading relationship is the largest in the world; they also explain, in a statistical sense, two-thirds or more of all the variation in bilateral trade flows between all possible pairs of countries. As a result, Leamer and Levinsohn (1995) noted that the gravity model provides “some of the clearest and most robust empirical findings in economics”. And again, for trade—which has been studied the most and for which we have reasonably good historical coverage— the effects of geographic distance, in particular, have not decreased over time. In fact, Disdier and Head (2008) suggest that they may actually have gone up relative to a hundred years ago. Distance seems to be in robust good health rather than dead.
Most of the research on gravity modeling has focused on calibrating it in the context of merchandise trade and, far less frequently, for other types of cross-border flows (services, foreign direct investment, migration, tourism,…). But most of the time, individual analyses look at just one or two types of flows. As a consequence, they do not allow comparisons across different types of flows because they use different samples, estimation procedures and/or variables. In this work we rely on a detailed dyadic dataset (mostly) assembled for the DHL Global Connectedness Report (Ghemawat and Altman, 2012) to compare systematically sensitivity to distances of various sorts across 12 different types of cross- border flows.
Our estimation exercise covers (subsets of) the four broad dimensions of distance included in Ghemawat’s (2001) CAGE distance framework, which distinguishes among cultural, administrative, geographic and economic distances. Specifically, we consider the effects of different languages, absence of colonial ties, lack of trade agreements, geographical distance, noncontiguity or lack of a common border and differences in per capita income. While for some type of flows, one might expect various types of distance to diminish cross-border interactions, it is also possible to imagine the opposite happening for other types of flows (e.g., foreign direct investment versus trade) or for other distance categories (e.g., per capita income in the case of trade flows). So this analysis of what the effects actually are across multiple types of flows should add substantially to our understanding of globalization as a multidimensional phenomenon.
The remainder of the paper is organized as follows. In Section 2, we describe the flows analyzed in this paper and prior work on them. In Section 3 we discuss appropriate specifications of the gravity model for different types of flows and present the main results: first, based on a general specification for all types of flows and then for specifications with variables customized to to each type of flow. Section 4 shows that despite the differences in the estimates described in Section 3, bilateral flows from each country are often concentrated in the same partners and that composite measures of distance tend to correlate well with geographic distance. Section 5 concludes.
In their chapter on gravity equations prepared for the forthcoming Handbook of International Economics, Head and Mayer (2013) show how two of the variables central to gravity models, GDP and geographic distance, play a key role in explaining part of the variation of the magnitude of bilateral trade flows. The authors present two scatter charts of bilateral trade flows (exports and imports) between Japan and EU countries. Since geographic distance within the EU is small relative to its geographic distance from Japan, differences in geographic distance between EU countries and Japan should not play a major role in variation across these bilateral flows. Similarly, since other potential sources of differences such as tariffs, language, religion, currencies or colonial linkages do not vary across the EU countries when considered in relation to Japan, the magnitude of the flows is expected to be linked mainly to the economic size of each EU country, which turns out to be the case. Head and Mayer (2013) also present some evidence about the relationship between distance and France’s exports and imports (normalized by the partner’s GDP) to Francophone countries, former colonies, members of the EU or the Eurozone and other countries. They characterize this simple analysis as capturing what they call the ‘spirit’ of the gravity equation.
In this subsection, we supply a similarly simple demonstration that these empirical regularities apply not just to trade flows but to all 12 types of flows considered here. These flows can be classified into four types or pillars: trade, capital, information and people. Within the trade pillar, we are analyzing merchandise and services exports. For capital flows we are analyzing FDI outward stocks and outflows as well as portfolio equity assets and portfolio long term debt. Information flows are covered by studying printed publication exports, outgoing phone calls and the number of patents in the patent office with origin in a different country. Finally, for people interactions we are including the emigration intensities, international tertiary students and international tourists’ arrivals. Farther information about the data sources are reported in the appendix and next section provides a more detailed description of the flows analyzed. We are considering positive flows for the period 2005-2011 and we have restricted the sample to 97 countries1 which currently account for 92% of the world’s GDP and 88% of its population.
Figure 1 shows USA’s flows with each partner on the vertical axis and the partner’s gdp on the horizontal axis –relative to the Japan’s values-, in order to display the relation between the magnitude of the flow and the partner’s size. The variables represented are the average values for the period 2005 – 2011 and are in logarithmic form, so the lines in the graph can be predicted values from a simple regression of the log of each flow on the log of the partner’s gdp and the slope of the line can be interpreted as the elasticity.
Despite some differences across different types of flows, for the case of the USA flows, the partner’s output have a clear positive impact in the magnitude of all kind of flows. However, these diagrams are slightly different from the ones in Head and Mayer (2013), since the group of countries considered have different characteristics in respect to the country of interest -the USA, in our case- that can affect the magnitude of the bilateral flows independently of the gdp. For example, Canada and Spain are two countries with similar economic sizes, but USA’s flows to Canada are significantly larger than the ones between the USA and Spain. Capital flows are an exception in this respect, since they are quite close. In fact, without controlling for the impact of other factors as the fact that Canada is contiguous to the USA, that they have the same legal origin and a trade agreement in force, it seems that there are other factors that push the outgoing FDI flows and stocks from the USA to Canada and the Spain so close in the graph. Then, although these diagrams are used to illustrate some simple correlations, to get a better understanding of this kind of relations we will need to control for other variables.
Figure 1. USA flows and partner’s GDP. Average 2005-2011.
Similarly, Figure 2 shows Spain’s bilateral flows on the vertical axis and the bilateral distance with each partner on the horizontal axis. In this case, the effect of distance is clearly negative in some cases, but the difference between the observed and the predicted values are higher than in Figure 1 for the gdp. In this respect, and even when we focus on countries of similar size as Argentina and Thailand, and that are similarly far from Spain, we can find very different performance in terms of the magnitude of their flows with Spain: flows between Spain and Argentina are larger than flows between Spain and Thailand. It seems that other factors as the colonial linkages, the similar cultures and the fact that Spain and Argentina share the same official language have an impact in the magnitude of the flows between Argentina and Spain in comparison with other countries that do not have these similarities with Spain.