Globalisation has deepened economic inequalities between large cities and the rest. This column examines foreign trade integration in larger cities versus other regions across Brazil, China, France, and the US. Larger cities have higher ‘export intensity’ – they export more because they host more exporters, especially superstar firms, and are less sensitive to trade cost changes. Large cities ‘win’ from international trade, and growing urbanisation fuels globalisation.
Globalisation has been one of the key economic forces of the past half-century. It is credited with lifting millions of people worldwide out of poverty. However, the distributional consequences of globalisation have become a growing concern to policymakers in many advanced economies. A common narrative holds that it has widened economic disparities between prosperous metropolitan centres and the rest, generating well-paid jobs in big cities while displacing employment in smaller ones. These regional imbalances have been empirically tied to the recent backlash against trade openness (Davenport et al. 2020).
These political debates have sparked a renewed interest in the heterogeneous regional effects of international integration. Of course, the idea that trade affects regions differentially is far from new (Rodríguez-Pose 2012). Classic contributions emphasise the role of sectoral composition (Autor et al. 2013) or differential access to foreign markets (see Cosar and Fajgelbaum 2016, Storeygard 2016, Atkin and Donaldson 2015). However, these factors alone struggle to explain the growing metropolitan–non-metropolitan divide. In many contexts, large cities do not have dramatically different sector mixes or clearly superior access to international markets. And yet, their economic trajectories in the age of globalisation differ sharply.
In recent research (Bakker et al. 2024), we revisit patterns of foreign-trade integration within four countries: Brazil, China, France, and the USA. We show that larger cities are indeed more integrated in the global economy: they export a significantly larger proportion of their total output. In other words, larger cities have a higher ‘export intensity’. We show that this pattern is not driven by differences in industrial composition or access to foreign markets.
To rationalise these novel stylised facts, we build a model whose mechanism centres on firm heterogeneity across space (Baldwin and Okubo 2006, Gaubert 2018). Specifically, our model builds on the well-documented fact that large cities host a disproportionate share of the most productive firms and thus feature a thicker upper tail in their firm productivity distributions (Combes et al. 2012). Coupled with the fact that the most productive firms are more likely to engage in exporting (Melitz 2003), the explanation for the deeper integration of large cities in global markets becomes straightforward: large cities tend to export more because they host a disproportionate share of the firms most likely to engage in exporting.
We show that our mechanism has important implications along several dimensions: it affects how economies respond to trade shocks; it alters our estimates of the gains from trade; it establishes a link between spatial policies (e.g. planning) and international trade; and it affects our interpretation of the recent widening of spatial inequalities.
A new stylised fact: Exporting is disproportionately concentrated in large cities
A natural starting point is to ask whether cities differ in how integrated they are in the global economy. Using rich firm-level data from China and France, and supporting evidence from Brazil and the US, we document a robust pattern: exporting is disproportionately concentrated in large cities, even more so than overall economic activity. Figure 1 shows the relationship between city-level export intensity (i.e. total exports/total sales at the city level) and city size (measure by total city-level population). For all four countries, we find a positive and economically meaningful relationship: a doubling of city size is associated with a 17%–33% increase in city-level export intensity.
Figure 1 Export intensity and city size in Brazil, China, France, and the US
Notes: The figure shows a binned scatter plot between city-level log export intensity (exports relative to revenues) and log city size. Each dot (bin) represents 10 underlying cities for China, and five cities for the other three countries. Cities are defined in terms of Microregions for Brazil, Metropolitan Areas for China and the United States, and Employment Zones for France. The analysis considers cities with positive exports and at least 250 manufacturing firms for China and France. For Brazil and the United States, the analysis considers cities with a population above 100,000 inhabitants. All figures include the following controls: ln average distance to other domestic cities, ln distance to the border, ln distance to the coast, border dummies, and coastal dummies.
Crucially, the relationship is not driven by industrial composition. Controlling for detailed 4-digit industry shares barely attenuates the correlation. Geography also does not explain the pattern. Adding rich controls for coastal access, distance to borders, distance to other domestic cities, and transport infrastructure leaves the gradient virtually unchanged. Moreover, we can predict modern city size with the historical caloric suitability of crops in the surrounding area. Because this variable is uncorrelated with transportation infrastructure, this instrumental variable approach can filter out the potential role of market access. The estimated coefficients become, if anything, even larger.
Cross-city differences in the upper tail of firm productivity distributions are key for explaining the stylised fact
To rationalise our novel stylised fact, we develop a model that combines elements from urban economics with those from international trade. In the model, cities exogenously differ in their firm productivity distributions, while market access to both foreign and domestic locations is symmetric across cities. Workers choose locations based on wages and the cost of housing.
A key insight from our model is that simple mean shifts in city-level productivity distributions cannot account for the higher export intensity in larger cities. This is because such shifts push up local wages, which generates a cleansing effect that is symmetric for exporters and non-exporters. As a result, the productivity threshold to serve the domestic market rises in the same proportion as the productivity threshold for exporting. Thus, exports relative to overall city output do not change, leaving city-level export intensity unaffected.
Instead, differences in the upper tails of the city-level firm productivity distributions are crucial to deliver our stylised fact. Intuitively, the firms from the upper tail of the productivity distribution in each city are exporters (as they are competitive in international markets). Thus, if large cities have thicker upper tails, they feature a higher fraction of those high-productivity exporting firms. The fraction of firms above the exporting threshold is higher in larger cities, which then accounts for their greater export intensity. Crucially, this theoretical requirement implied by our model is supported by the data: we show that larger cities in China and France indeed feature firm productivity distributions with thicker upper tails, confirming the pattern first documented by Combes et al. (2012).
The bilateral relationship between international trade and economic geography
Our model provides several novel insights for the joint determination of international trade flows and the spatial distribution of economic activity within countries. Perhaps most pertinent to the debate on the impact of globalisation, our model yields a systematic large-city-bias of trade integration. In a world in which large cities are more likely to host superstar firms, which are in turn more likely to be exporters, any reduction of international trade costs benefits these large cities. As exporting firms expand due to reductions in trade costs, labour demand in the locations that host these firms expands, attracting workers, but also raising wages and housing prices. This can help explain the disproportionate success of large and productive cities in the recent era of globalisation.
Our simple model also predicts an inverse link from internal geography to international trade. Any reallocation of workers towards large cities (e.g. due to relaxing zoning restrictions) tends to raise national export activity. This mechanism reinforces the case for easing planning constraints in large and successful cities.
All in all, our modelling work yields one overarching conclusion: globalisation and large cities are complements. Trade liberalisation tends to make large cities larger and more expensive, while geographical policies that allow large cities to grow tend to deepen international trade integration.
City size and trade elasticities
Our model yields a unique, distinguishing prediction that we can test in the data: exports from large cities react less elastically to changes in trade costs. Intuitively, with a larger share of firms already exporting, the impact of the novel exporters following a reduction in trade costs is smaller.
We test this prediction of our model using Chinese firm-level data. We study the response of city-level exports in China to the granting of Permanent Normalized Trade Relations by the US around the time of China’s WTO accession. In line with our model predictions, we find that exporting in smaller cities responded sharply to the change in US tariff uncertainty produced by the policy, whereas the exports of the largest cities responded only moderately.
Coupled with our earlier results, these findings suggest two additional insights. First, the geographic concentration of economic activity in large cities not only deepens globalisation, but also makes it more resilient: large cities export more, and their exports are less sensitive to changes in trade costs. This implication also has important consequences for a potential era of trade disintegration.
Secondly, and more subtly, our model suggests that a key parameter of many existing trade models – the (country-level) trade elasticity – is endogenous to trade costs. Specifically, our model predicts that the country-level trade elasticity is an export-weighted average of city-level trade elasticities, which are lower in larger cities. Thus, as trade costs increase and population gradually shifts from large cities to small cities, the country-level trade elasticity increases. This dynamic has knock-on implications for the gains from trade, which we study quantitatively in the last section of our paper.
Rethinking the geography of globalisation
Our research suggests globalisation and the geographic concentration of economic activity have deep complementarities rooted in firm heterogeneity. This complementarity implies that trade liberalisation does not only reallocate activity across sectors and countries, but also across space with countries – shifting population and economic activity toward already large, high-wage cities. In this sense, our findings complement Baldwin’s (2006, 2016) account of globalisation’s ‘great unbundlings’, whereby falling transport and coordination costs enabled production and consumption to separate geographically, fostering industrial concentration. We add that these forces operate disproportionately in larger cities.
Understanding globalisation and its implications increasingly requires understanding the geography of production within countries. Internal geography plays a significant role in shaping a nation’s participation in the global economy and the response of its economy to trade-related shocks.
Source : VOXeu
