In recent decades, multinational production has grown even faster than international goods trade to become, by some measures, the most important means through which firms serve foreign markets. Foreign multinational affiliates now account for about one-third of gross output in many developed countries (e.g. Alviarez 2013).
By definition, multinationals encompass production facilities in multiple countries, and thus it is a natural conjecture that they have an impact on the international transmission of economic shocks. Indeed, it appears that the most closely integrated countries experienced an increase in business cycle synchronisation over the past three decades (Kose et al. 2008, Hirata et al. 2013), the time period coinciding with the dramatic growth in multinational production. However, the relationship between multinational firms and the transmission of shocks is not yet well understood. At the micro level, there is limited empirical evidence on how the activities of the different parts of a multinational are interrelated at business cycle frequencies. At the macro level, it is yet to be established whether multinational production matters quantitatively for aggregate co-movement.
In a recent paper, we use novel firm-level data and a quantitative multi-country model to examine the role of multinational firms in aggregate business cycle transmission (Cravino and Levchenko 2016). Our data cover several million domestic and multinational firms operating in 34 countries over the period 2004-2012. The key feature of the dataset is that it contains information on domestic and foreign ownership. Hence, for the first time in this context, the operations of parents and affiliates are observed in the same dataset as well as through time and in a broad cross-section of countries. This information allows us to study micro-level, cross-country co-movement between the different parties of the multinational corporations. At the same time, the data cover the bulk of economic activity in our sample of countries, making it possible to aggregate the firm-level results and derive their implications for business cycle co-movement.
Our analysis goes from micro patterns to macro implications in three stages. First, at the firm level, we document strong co-movement between multinational affiliates and their parents: a 10% growth in the sales of the parent is associated with a 2% growth in the sales of the affiliate. This correlation is computed after controlling for sectoral and aggregate trends using source-sector-destination-sector-year fixed effects, so that it captures the role of linkages within the multinational firm.
The firm-level estimates show that units of the same firm co-move together at the business cycle frequency. However, precisely because they are obtained controlling for very detailed aggregate trends, they may not capture transmission of shocks that are common across parent firms in the source country. With this in mind, in our second step we aggregate multinational sales to the source-destination level (e.g. combined sales of all US multinational affiliates operating in the UK), and estimate whether the variation in source-destination growth rates is driven by source-specific or destination-specific factors. Source-specific factors account for about 10% of the variation in bilateral growth rates, compared to 18% accounted for by destination-specific factors.
The third step of our analysis assesses the quantitative importance of this phenomenon for aggregate business cycle transmission using a multi-country model that can be taken to the data. In the model, co-movement between multinational firms and their foreign affiliates occurs because the productivity of the affiliate is affected by the productivity of the parent. In particular, the productivity of foreign affiliates is a combination of a source-specific and destination-specific component. The relative importance of the source versus the destination component is governed by a crucial parameter that we discipline with the data. In the model, the extent to which multinationals contribute to the transmission of shocks across countries is driven by:
- What share of the firm's technology shock originates in the source versus the destination country;
- The distribution of bilateral multinational shares in the economy; and
- General equilibrium effects.
We estimate that between 20% and 40% of the foreign affiliates' shocks originate in the source country. The multinational production shares are taken directly from the data. Finally, the magnitude of the general equilibrium effects depends on a composite parameter that combines the elasticity of substitution across intermediates and the Frisch labour supply elasticity. We benchmark these parameters using micro estimates of these elasticities, and check the sensitivity of the results to alternative values.
We use the calibrated model to conduct quantitative exercises to measure the importance of multinational firms for the transmission of shocks across countries. First, we track the propagation of productivity shocks across countries. The first row of Table 1 summarises the impact of a 1% shock to productivity in the rest of the world (excluding the country in question). A 1% productivity shock in the rest of the world as a whole raises productivity by 0.12% in the average country, and by as much as 0.2-0.35% in the most integrated countries such as Ireland, the Netherlands, and Slovakia.
Table 1. Simulated percentage change in destination GDP due to a 1% shock to the source
Not surprisingly, the external impact of individual source country shocks is considerably smaller. This is because while all foreign multinational affiliates account for a large fraction of aggregate output in the typical country (nearly one-third), multinational affiliates from an individual country tend to be much less important in aggregate. As evidenced by the remainder of Table 1, a shock that increases GDP by 1% in one of the four most important source countries – the US, Germany, the UK, and France – raises output in the rest of the sample by between 0.01% and 0.02%. Shocks to other source countries have a negligible impact, since multinational affiliates from other source countries tend to have small output shares in a typical destination.
We also use the model to compute the business cycle correlation between each pair of countries assuming that the primitive productivity shocks are uncorrelated. This is an assessment of how much correlation can be generated purely by propagation of shocks through multinationals under the observed levels of multinational activity. The variation in model-implied correlations is driven entirely by the pattern of multinational output shares. On the one hand, in most country pairs bilateral multinational shares are small, and thus the model generates little business cycle co-movement – the mean model-implied correlation is 0.01 in the full sample of country pairs. On the other hand, for country pairs involving either a major source or a major recipient of multinational firms, the model generates between 10% and 25% of the correlation observed in the data. In addition, in the cross-section of country pairs the model-implied correlations have a positive and highly significant relationship to the GDP correlations in the data.
Our main takeaway from these exercises is that the combined impact of all foreign multinationals is small but significant, accounting for about 10% of the productivity shocks in a typical country and leading to a somewhat more synchronised international business cycle. The impact is highly heterogeneous across countries. The transmission of shocks and positive business cycle correlations induced by multinational presence are clearly detectable for the country pairs involving the most important source and destination countries. On the other hand, aggregate interdependence between most individual country pairs is minimal, since most bilateral shares are small.
Alviarez, V (2013) “Multinational production and comparative advantage”, December, mimeo, University of Michigan.
Cravino, J and A A Levchenko (2016) “Multinational firms and international business cycle transmission,” Quarterly Journal of Economics, forthcoming.
Hirata, H, A Kose and C Otrok (2013) "Regionalization vs. globalization", IMF Working Papers 13/19, International Monetary Fund.
Kose, M A, C Otrok, and E Prasad (2012) "Global business cycles: Convergence or decoupling?", International Economic Review, 53(2): 511-538.