Notes on the global condition: Have we ever been global? Tyler Cowen and the “distance puzzle”

Tyler Cowen argued in a recent op ed that the impact of the backlash against deglobalization on global trade might not be as serious as many fear. His reasoning was original. We should not fear deglobalization Cowen argued, because we have exaggerated the extent of long-distance trade integration in the first place.

“Globalization probably won’t contract much because globalization, at least in the form of international trade, didn’t proceed as far as many people had thought. In fact, the ability of trade to surmount barriers of distance probably never went up in the first place.  … Just as the internet did not herald the “death of distance” but instead led to a concentration of talent in the San Francisco Bay Area, New York City, London and other megacities, so have contemporary technologies kept trade geographically clustered. The pessimistic reading of trade clustering is that human beings simply have not spread their wings very far. But these days, I find the gravity equation to be a comfort. Given that our ability to trade across great distances has not outraced our ability to trade nearby, I am not expecting any kind of a major trade snapback or correction. The evolution of trade, rather than throwing out fragile, delicate spokes, has instead made some fairly hardy connections, sturdy enough it seems to survive Trump’s rhetoric. … When trying to understand globalization looking forward, we must resist the temptation to believe the most puffed up reports about what we achieved in the past. The bright side is this: The continuing relevance of distance may be part of what is keeping our current world stable.”

So geography trumps Trump. This at least is the conclusion that Cowen draws from a wave of economic studies using so-called gravity equation models. They purport to show that when examined econometrically, the elasticity of trade with regard to distance is unchanged since the 1960s, or even, perhaps,  increasing. For every 1000 km of distance between trading partners, trade declines by as much, or more than it did decades ago.

The result is so counterintuitive that it has provoked more pushback than Cowen credits. But working through the argument is worthwhile because it reveals some interesting dimensions of globalization.

It is crucial to take the basic Cowen point on board. To a far greater extent than we generally credit, countries trade with their neighbors. This is reinforced by regional trade agreements. A huge portion of global trade is accounted for by trade between the US-Canada and Mexico in NAFTA and within the EU. But the effect operates even where there is no general regional trade agreement as between China and its big neighbors i.e. South Korea, Japan, Taiwan.

But then it gets more complicated: The elasticity of trade with regard to distance, is a function of the elasticity of trade with regard to cost and the elasticity of cost with regard to distance. These factors might potentially pull in different directions. If costs fall particularly fast for short-distance or regional trade, then the effect of a general cost reduction may be to increase shorter distance trade.

Recently, poor countries have tended to trade less over very long distances. In some cases this may be due to disproportionate reductions in local trade costs. More worryingly it might indicate the selective inclusion of middle income rather than truly low income countries in complex global value chains. Czechoslovakia and Mexico are key to transnational automotive production, not Mali.


This atrophying of very long distance rich-to-poor country connections is compounded by the dramatic decline in the significance of postcolonial and neo-imperial trading patterns. Those connections still had a major impact on global trade in the 1960s. In industries like cocoa/chocolate which I blogged about for Valentine’s they still do.

Some trade or no trade? Handling cases where there is zero trade requires some fiddly math, so in some of the gravity model tests, the instances of zero connection are excluded from the sample. That biases the result, because instances of zero trade tend to be very long distance especially lower down the income scale. Once you include the declining number of zero trade pairings, some studies show that the trade over very long distances has indeed been increasing since the 1960s.


But, finally, for my money the cleanest and most definitive resolution of the “distance puzzle” comes from a study by Yoto V Yotov, which points out that what the majority of studies have been looking at is not the effect of trade costs but relative international trade costs. If we allow for a general tendency to trade with closer trade partners, the question being asked is actually whether new cost-reducing technologies have been biased towards long distance or short distance international trade. The conclusion that the elasticity in distance has not changed is really a statement about the relatively even impact of globalization across all international trade. If we compare not between international trade flows but between international and domestic trade flows then the distance puzzle evaporates. Globalization has happened.

In the author’s words:

“In this note, I solve the distance puzzle and I demonstrate that globalization is in fact present in the gravity model.1 The idea is simple and intuitive. To develop my argument, I capitalize on the properties of the theoretical gravity model from Anderson and van Wincoop (2003) who show that the structural gravity system can only identify relative trade costs. This means that existing studies, which use international trade data only, measure international trade costs relative to other international trade costs. At each point of time, the effect of distance on international trade between a given pair of countries is estimated relative to the effect of distance on international trade for another pair of countries. Thus, assuming that the effects of globalization are spread evenly among the different pairs of trading partners, it should not be surprising that the estimates of the effects of distance and trade costs are stable over time. In short, my explanation for the persistence in the distance estimates from empirical gravity models is that these indexes are obtained exclusively from international trade data and, therefore, they cannot capture globalization.

To solve the puzzle, I argue that the appropriate measure of globalization is the increase in international economic integration relative to the integration of internal markets. Consequently, we need to evaluate the effects of bilateral distance and international trade costs relative to the effects of internal distance and internal trade costs in order to capture the effects of globalization. This intuitive measure should not be subject to the distance puzzle because external trade costs must have fallen relatively more than internal trade costs. The reason is that, though technological and communication breakthroughs affect both internal and external trade costs, institutional developments have been more significant when it comes to changes in external trade costs. Mechanically, the simple adjustment that solves the distance puzzle is based on a stricter adherence to the structural gravity theory, which requires that internal trade and internal distance are accounted for in standard gravity estimations. Once this is done and the effects of international distance are measured relative to the effects of internal distance, the distance puzzle disappears.”

Once we compare international with domestic trade rather than one international pair v. another, then the distance elasticity of trade does appear to have fallen consistently, independently of specification, by a substantial amount, in the region of 37 % (to be overly precise).

Together with Ingo Borchert, Yotov has recently expanded on the methodology of the earlier paper to produce a fascinating study of globalization that allows a comparison of declining distance elasticities across countries. It shows a remarkable pattern of differentiation with middle income countries benefiting most from the distance-reducing-and-trade-enhancing effects of globalization. For rich countries specializing in high value production that is less sensitive to trade costs, the effect dwindles. But the really alarming finding is about the poorest countries. In the US, perhaps unsurprisingly, the distance effect is particularly modest.


As the authors conclude “In combination, the four panels in Fig. 2 confirm that the estimated changes in gravity distance elasticities are indeed reflecting globalization effects and the associated changes in how and what countries are trading. In particular, one driving force appears to be a shift towards more sophisticated goods whose exports are less sensitive to distance frictions. Yet with the notable exception of China, low income countries as a group were markedly less transformative in that regard than middle income countries, raising the prospect of further decoupling the poorest nations from the dynamics (and gains) of globalization.”

Upshot: Yes, geography matters. Yes, trade with those closest to us is most important. But our eyes have not been deceiving us, the distance-annihilating effect of globalization is not an optical illusion. Those countries that have experienced the most rapid economic growth in the middle of the global income distribution have benefited from it. The poorest countries remain largely excluded from this integration effect. Globalization is not reducible to the natural logic of proximity. For better or worse globalization is a econo-techno-political artifact. As such it is vulnerable to the caprices of the government of the richest and most powerful country in the system. Those countries that will be most exposed to any fallout are not the very rich, but the rapidly rising middle income economies, on which the optimistic global narrative of the last few decades (such as it is) is based.

related posts