VoxEU Column Global economy International trade

International business cycle co-movement through the lens of individual firms

The business cycles of countries with greater bilateral trade and multinational production linkages are more closely correlated. But the meaning of this empirical relationship is not well understood. Some contend that these linkages allow for the transmission of shocks across countries, while others argue that countries that trade more with each other are similar in other ways and are thus subject to common shocks. Using data from France, this column examines the properties of international co-movement at the firm level. Even after controlling for common shocks, there is still substantial evidence of transmission of shocks through trade and multinational linkages. Furthermore, trade linkages matter more than multinational ones, especially when it comes to the aggregate impact. 

Policy and journalistic discourse on the state of the world economy is increasingly focused on the consequences of macroeconomic interdependence between countries. Most recently, for instance, there has been a great deal of discussion about what the turmoil in China means for the rest of the world, and the Federal Reserve’s decision to hold rates on 27 January 2016 has been widely attributed in part to the weak state of the world economy. However, others are sceptical that China is big enough to have a significant impact on the world cycle (Blanchard 2016).

Can shocks in one country have a substantial impact on GDP growth in another? It has been known for a long time that countries that exhibit greater bilateral trade and multinational production linkages have more correlated business cycles (Frankel and Rose 1998, Kleinert et al. 2015). And yet, the meaning of this empirical relationship is not well understood.

Taken at face value, the positive association between bilateral trade and multinational linkages and co-movement is often interpreted as evidence of transmission of shocks across countries through those linkages. However, as argued by Imbs (2004), an alternative explanation is that countries that trade more with each other are similar in other ways, and thus subject to common shocks. Under an extreme version of this view, the trade linkage variable does not reflect the intensity of transmission of shocks, but rather is simply a stand-in for the prevalence of common shocks. Furthermore, even if one accepts the transmission of shocks interpretation of the relationship between trade and co-movement, the coarse nature of the cross-country setting makes it difficult to learn about its micro underpinnings.1

International co-movement at the firm level: New evidence

In a recent paper (di Giovanni et al. 2016), we examine the properties of international co-movement at the firm level and its aggregate implications using data covering the universe of French firm-level value-added, destination-specific imports and exports, and cross-border ownership over the period 1993-2007. Examining cross-border co-movement at the firm level has two advantages relative to the traditional approach of looking directly at GDP correlations.

  • First, at the micro level, the data allow for precise measurement of trade and multinational linkages – by firm×country – and to control for common shocks using appropriate fixed effects.

This lets us establish much more firmly that the positive relationship between trade and co-movement is due at least in part to transmission of shocks at the firm level.

  • Second, at the macro level, our approach captures the implications of firm heterogeneity in both size and the extent of international linkages for aggregate co-movement.

Larger firms are disproportionately more likely to trade internationally and own affiliates in foreign countries. Indeed, in most countries international trade flows are dominated by only a handful of large firms. An emerging research agenda in closed-economy macro has argued convincingly that modelling and measuring shocks at the micro level (to firms and sectors), and linkages between them, is essential for understanding aggregate fluctuations.2 If large firms and firm-to-firm linkages matter for aggregate fluctuations, a natural conjecture is that they will matter as much if not more for cross-border co-movement.

We begin by estimating a specification inspired by Frankel and Rose (1998) that relates the correlation of firm total value-added growth with foreign GDP growth to firm-level direct linkages with that country. The data contain, for each firm and potential partner country, four types of direct linkages: importing from it, exporting to it, being a France-based affiliate of a multinational firm headquartered in that country, and being a French firm with a foreign affiliate in that country. Because the sample includes many firms and countries, estimation controls for both firm and country effects. Country effects in particular absorb the common aggregate shocks affecting France and each foreign country.  

Main findings

  • The main micro finding is that trade linkages at the firm level are significantly associated with increased co-movement between an individual firm and the country with which it trades.

An import link increases the correlation by 0.012, and an export link by 0.005. This is large relative to the average correlation between an individual firm and foreign GDP, which is 0.024 for directly connected firms, and essentially zero for non-directly connected ones. By a similar token, affiliates of foreign multinationals operating in France have a 0.01 higher correlation with their source countries. 

We then examine the macro implications of the micro-level findings. We start with the observation that the aggregate business cycle correlation between France and another country is simply an appropriately weighted sum of the correlations of firm-level total value added with that country.

  • For the ten large trading partners of France in our sample, we show that the large directly connected firms are important in accounting for aggregate co-movement. While on average only about 8% of firms are directly connected to a particular market, these directly connected firms account for the 56% of total value added.

We use the conditional relationship between direct linkages and firm-level correlations to compute the change in the aggregate correlation between France and each foreign country that would occur if direct linkages at the firm level disappeared. The aggregation exercise combines information on the estimation-based predicted change in the correlation at the firm level with firm-level weights.

  • If direct linkages at the firm level were severed, the aggregate correlation would fall by 0.091 on average in our sample of 10 partner countries.

This is a non-negligible change relative to the observed correlations between France and its main trading partners of 0.29 on average over this period. Figure 1 displays the change in aggregate correlation predicted by our estimates against the observed level of aggregate correlation. Countries with the highest observed aggregate correlations also exhibit greater direct linkages, and thus severing those linkages leads to larger reductions in aggregate correlations. Since our data allow us to estimate the coefficients on trade and multinational links separately, we can also check which ones matter more for generating aggregate co-movement. It turns out that the trade linkages are about nine times more important in generating aggregate co-movement than multinational linkages, accounting for 0.083 of the overall 0.091 effect.

Figure 1. Predicted change in aggregate correlation

Concluding remarks

To summarise, even after controlling for common shocks, there is still substantial evidence of transmission of shocks through trade and multinational linkages. Among those linkages, trade linkages appear to matter more than multinational ones, especially when it comes to the aggregate impact. These results provide a firm empirical basis to the frequent presumption that cross-border linkages do propagate shocks across countries, and that this matters for the aggregate economy.


Acemoglu, D, U Akcigit, and W Kerr (2015), “Networks and the Macroeconomy: An Empirical Exploration,” April, Forthcoming, NBER Macroeconomics Annual.

Acemoglu, D, V M Carvalho, A Ozdaglar, and A Tahbaz-Salehi (2012), “The Network Ori- gins of Aggregate Fluctuations,” Econometrica, September, 80 (5), 1977–2016.

Blanchard, O J (2016), “The price of oil, China, and stock market herding,” VoxEU.org, 18 January.

Carvalho, V M and X Gabaix (2013), “The Great Diversification and its Undoing,” American Economic Review, August, 103 (5), 1697–1727.

Carvalho, V M and B Grassi (2015), “Large Firm Dynamics and the Business Cycle,” CEPR Discussion Paper 10587.

di Giovanni, J and A A Levchenko (2012), “Country Size, International Trade, and Aggregate Fluctuations in Granular Economies,” Journal of Political Economy, December, 120 (6), 1083–1132.

di Giovanni, J, A A Levchenko, and I Mejean (2016), “The Micro Origins of International Business Cycle Comovement,” CEPR Discussion Paper 11036.

Frankel, J A and A K Rose (1998), “The Endogeneity of the Optimum Currency Area Criteria,” Economic Journal, 108 (449), 1009–25.

Gabaix, X (2011), “The Granular Origins of Aggregate Fluctuations,” Econometrica, 79 (3), 733–772.

Imbs, J (2004), “Trade, Finance, Specialization, and Synchronization,” Review of Economics and Statistics, 86 (3), 723–34.

Johnson, R C (2014), “Trade in Intermediate Inputs and Business Cycle Comovement,” American Economic Journal: Macroeconomics, 6 (4), 39–83.

Kleinert, J, J Martin, and F Toubal (2015), “The Few Leading the Many: Foreign Affil- iates and Business Cycle Comovement,” American Economic Journal: Macroeconomics, 7 (4), 134–159.

Kose, M A and K-M Yi (2006), “Can the Standard International Business Cycle Model Explain the Relation Between Trade and Comovement,” Journal of International Economics, 68 (2), 267–295.


1 This lack of understanding is reinforced by the quantitative literature, which has struggled to capture the relationship between trade and co-movement. Kose and Yi (2006) and Johnson (2014) show that even quite sophisticated international real business cycle models fail to generate the observed positive association, dubbing it the “trade-co-movement puzzle”.

2 e.g. Gabaix (2011), Acemoglu et al. (2012), di Giovanni and Levchenko (2012), Carvalho and Gabaix (2013), and Carvalho and Grassi (2015).

1,785 Reads