Gravity models have been widely used to describe bilateral trade in goods. Portes and Rey [Portes, R., Rey, H., 2005. The Determinants of Cross-Border Equity Flows. Journal of International Economics, 65(2), 269–296.] applied this framework to cross-border equity flows and found that distance, which proxies information asymmetries, is a surprisingly very large barrier to cross-border asset trade. We adopt a different point of view and explore the complementarity between bilateral trade in goods and bilateral asset holdings in a simultaneous gravity equations framework. Providing different instruments for both endogenous variables, we show that a 10% increase in bilateral trade raises bilateral asset holdings by 6% to 7%. The reverse causality is also significant, albeit smaller. Controlling for trade, the impact of distance on asset holdings is drastically reduced.
The determinants of international asset holdings have recently received renewed attention. Existing theories are mostly based on portfolio choice models and put forward risk-sharing as the main motive for cross-border asset trade. However, this literature has been empirically extremely disappointing. Indeed, Capital Asset Pricing Models predictions do not fit data on international portfolios for two main reasons. First, those models were unable to replicate the size of the “home bias” in country portfolios. If twenty years ago the segmentation of financial markets could well explain the “home bias puzzle”, it is not likely to be the case today. Second, countries seem to invest much more in geographically close economies. Portes and Rey (2005) highlight the very large impact of geography on cross-border equity flows: when physical distance is doubled, capital flows are at least divided by two. To explain this surprising result, they argue that informational asymmetries lead to higher transaction costs between distant economies. Moreover, as they point out, since distant economies should be a better hedge for regional risk, this result is hard to justify in a world where investors want to diversify their risk. Those results suggest that barriers to international investment are still large, which is at odds with the popular view of intense and widespread financial globalization.
This puzzling effect of distance on capital flows leads to the following question: does distance directly affect international financial investment or does the negative impact of distance go through another feature of globalization? In this paper, we argue that distance affects bilateral asset holdings mainly through its impact on trade in goods. The argument is the following: assume that trade in goods is a powerful determinant of asset portfolios. In that case, since distance, understood as transport costs, reduces international trade in goods, it is likely to also reduce bilateral asset holdings. Indeed, we show that the “distance puzzle” documented by Portes and Rey is drastically reduced once we control for trade in goods. We find that the distance effect on asset holdings is at least divided by two. The remaining challenge is to explain why asset portfolios are induced by trade in goods.
Thus the second motivation of this paper is to analyze the complementarity between bilateral trade in goods and bilateral financial claims. Indeed, there are good reasons to think that trade in goods and trade in assets are closely related.2 First, due to information asymmetries, entrepreneurs may learn about each other by trading goods and this information facilitates trade in financial assets (and vice versa). Second, in the complete markets model developed by Obstfeld and Rogoff (2000), trade costs (transportation costs or other barriers to international trade) induce a bias in investors portfolios towards domestic securities and securities of their trading partners. As a consequence, country portfolios would reflect trade patterns. Lane and Milesi-Feretti (2004) test this model in an N-countries set-up and find the expected effects. However, the argument can easily be reversed: it may be that transaction costs in financial markets (pure transaction costs or informational costs) make agents exchange goods with countries with whom they can easily exchange securities. As a consequence, international investment patterns would impact trade flows.
Are those relations between trade and finance of first-order magnitude: in other words, can we still model international trade and international investment separately? We investigate this question empirically and the answer is an unambiguous no: we find a very robust and significant effect of trade on financial asset holdings. Moreover, the causality runs significantly in both ways although the impact of asset holdings on trade in goods is smaller.
In line with Portes and Rey (2005), we consider the “home bias” as given and focus on the determinants of geographical asset holdings using a “gravity equation” set-up3. We use a data set4 which breaks down international banking assets by countries5. We find that informational frictions decrease bilateral financial claims, institutional and cultural proximity affect positively international asset holdings and standard financial motives have marginal effects. Those results are consistent with the findings of Portes and Rey. However, we also show that bilateral trade patterns are a very strong determinant of bilateral asset holdings.
In order to address the issue of reverse causality (between bilateral trade and bilateral asset holdings), the use of good instruments is crucial. We use some geographical variables (excluding distance6) and data on bilateral transport costs to instrument bilateral trade in goods. Another set of instruments for bilateral financial asset holdings is required: using data on bilateral tax treaties (fiscal taxation of foreign capital and bilateral agreements on double-taxation) and some institutional proximity variables, we provide reasonable instruments which allow us to properly address the reverse causality issue. We estimate that a 10% increase in bilateral trade induces a 6% to 7% increase in bilateral financial assets holdings so that the effect of trade in goods on asset portfolios is quantitatively important. Conversely, a 10% increase in bilateral financial asset holdings induces a 2.5% increase in bilateral trade. This empirical methodology also allows us to identify the channel through which some variables affect bilateral trade (resp. bilateral holdings of financial assets): as mentioned before, we find that distance affects country portfolios mainly through its impact on trade. This suggests that globalization has gone much further on the financial side than on the real side. Finally, as a by-product, we find some interesting results on the “distance puzzle”7 in the gravity equation of international trade: depending on the methodology, we reduce the impact of distance on trade in goods by around 20%. In order to test the robustness of our findings, we reestimate our gravity models using the same empirical methodology but with a different data set (the “Coordinated Portfolio Investment Survey”) which breaks down securities holdings by countries8. All our results are confirmed qualitatively and quantitatively.
In Section 2, we give some insights on the standard gravity models in international trade in goods and international asset portfolios: although it is not the main point of the paper, we reestimate the standard “gravity equations” for comparison purposes. In Section 3, we properly address the question of the complementarity between international financial asset holdings and trade flows and give the estimates for the system of two simultaneous gravity equations. We analyze our main results and comment on the “correlation puzzle” that emerges from the empirics: we find that, even after controlling for trade and distance, investors still hold more financial assets from countries whose returns are positively correlated with their domestic stock market9. We also discuss the links between our empirical results and the existing theory on the complementarity between trade in goods and trade in assets. In Section 4, we conclude.
We bridge two strands of literature: international trade in goods on the one hand and international
asset portfolios on the other. Numerous papers have shown that international trade in goods can be
very well described by gravity models and some recent papers have pointed out that international
asset portfolios could also be described by this kind of models: if the distance between two countries
doubles, bilateral asset holdings are almost divided by two. This far from negligible impact seems
somewhat puzzling, since geography should not shape asset trade in a globalized world.
Portes and Rey (2005)
justifies the impact of distance on asset flows by information costs,
distance acting as a proxy for the informational asymmetries. We chose here to investigate another
idea, namely that trade in goods and asset holdings are mutually reinforcing. The strong impact of
distance on asset holdings is the consequence of the complementarity between trade in goods and
trade in assets.
Using bilateral data on international trade flows and international banking claims, we have
examined what remained of the effect of distance once we take into account the fact that trade in
goods and bilateral financial claims are mutually determined. The set of instruments we use to
identify the system is a crucial aspect of this study. We have used geographical variables and new
data on transport costs to instrument trade in goods. To instrument asset holdings, we followed
La
Porta et al. (1997, 1998)
and used data on legal environments; to the standard legal environment
data, we added a set of variables we built, which describe some aspects of the bilateral fiscal
relationships between countries (bilateral withholding taxes on dividends and interests, and fiscal
agreements). This methodology allows us to estimate precisely the effect of bilateral trade in goods
on bilateral asset holdings: this effect is found to be quantitatively important since a 10% increase
in bilateral imports lead to a 6% to 7% increase in bilateral asset holdings. Bilateral asset holdings
also enhance trade in goods but the latter effect is found to be much smaller.
Our results show that only trade in goods has an undisputable gravity structure, i.e. a structure
in which distance (understood as a proxy for transportation and transaction costs) is a majordeterminant. The system we have estimated shows that distance affects asset holdings mainly
through its impact on trade in goods: in the asset part of the system, the magnitude of the distance
puzzle is at least reduced by 60%.
The existing scenarios (Obstfeld and Rogoff's consumption hedging, Rose's sovereign risk)
cannot be formally eliminated so far, even if we have shown that some of our results cast doubt on
each one of them. Another story based on common transaction costs on financial markets and
goods markets could be a more natural match to our result. For example, in line with Portes and
Rey's paper, it might be that both trade in assets and trade in goods are subject to some common
information costs making trade in goods and trade in assets complementary. Those information
spillovers (from goods markets to financial markets) would need to be very large to have the
observed effects but we do think that this explanation is a large part of the story. However, we do
not pretend to provide a full-fletched theory of what we point out in the data. The robustness and
the strength of our empirical results shed light on the necessity to model trade and financial
linkages together. It is a new challenge for the economic theory.
Furthermore, our framework leads to another puzzle: the higher the correlation between two
countries stock returns, the larger the volume of asset trading between the two. This result still
holds true once we control for trade in goods. This reinforces the need for a theoretical study on
the interactions between trade in goods, trade in assets and diversification.
Finally, these results raise some interesting questions about the coherence of liberalization
policies. We show in this paper that trade in goods and in assets reinforce each other. Trade
policies and capital account liberalization cannot be considered independently. Therefore, these
policies should be thought of by policymakers in a common and single perspective.
