Many important current policy debates involve the complex activities of multinational enterprises (MNEs) in time and space. For example, the recent escalation of the US–China tariff war will certainly have an effect on the reallocation decisions of MNEs, as the majority of Chinese exports come from foreign-owned MNEs located in China. Recent renegotiations of the US–Mexico–Canada agreement (formerly the North American Free Trade Agreement) focused on policies that primarily affect the MNE costs of offshoring (and onshoring), such as rules about minimum wages. And the behaviour of MNEs is a central source of the uncertainty surrounding Brexit. Policymakers in the UK are worried that MNEs could pull out from the island and reallocate towards other countries. In turn, EU authorities are worried that MNEs located there would also be negatively affected once barriers to accessing the UK market from continental Europe increase. Moreover, those effects may depend on different implementations of Brexit, which may have different short-run and long-run consequences on the behaviour of MNEs.
Providing sound answers to these and other similar questions requires a modelling approach to the decision-making process of MNEs that features not only a rich set of choices of production sites and destination markets (like the static models in Ramondo and Rodriguez-Clare 2013, Tintelnot 2017, Fan 2017, Arkolakis et al. 2018, Alviarez 2019 and Head and Mayer 2019), but also dynamic decisions. Only in this way can changes to different static and dynamic costs to the expansion of MNEs be evaluated and compared.
A new dynamic framework of multinational expansion
In recent research (Garetto et al. 2019), we present a dynamic quantitative model of MNE expansion that can be used to analyse the effects of policies that affect the cost of MNE operations, such as Brexit or the US–China tariff war. Our model captures in a tractable way the rich heterogeneity observed in the data, which is necessary for making predictions about the effects of proposed policy changes.
We assume that MNEs make decisions in a dynamic and uncertain environment, and that in doing so they face variable costs that change depending on how much is produced, or exported (e.g. labour costs, taxes, and tariffs), fixed per-period costs of production (and exporting), and sunk costs of starting production and export activities that are paid once and cannot be recovered. MNE affiliates can separate their locations of production and sales, and endogenously choose to enter or exit their host and export markets. These features are important for capturing the complex decision making of MNEs, yet including all of them can make it difficult to generate clear predictions about MNE responses to policy changes. By introducing a novel compound-option formulation of the problem of the MNE, we develop a model that incorporates the many features of MNE behaviour and generates clear-cut results.
Our approach builds on insights from the literature on investment under uncertainty (Dixit and Pindyck 1994). An MNE has the option of opening an affiliate in a foreign country, and once that option gets exercised, it can access a set of additional options, such as exporting from the affiliate to any other location. By breaking down the decision process in this way, we are able to solve a very complex problem in a tractable way.
All the assumptions we make in the model are supported by observations on the behaviour of US MNE affiliates located in the top ten receiving countries, over 25 years, from the US Bureau of Economic Analysis. These data allow us to estimate the model and to quantify the costs that drive the dynamic location choices of US MNEs. We find that, for example, the cost of opening a US affiliate in China is 110% higher than the cost of starting to export from that affiliate to other countries. In contrast, for Ireland, a very open business-friendly country, the affiliate opening cost is about half of the export cost.
Putting the theory to work: Predicting the effects of Brexit
We use our model to quantify the effects of raising trade barriers on MNE activities. Our data include affiliates of US MNEs located in the UK, Ireland, and Germany. Hence, we are well-positioned to perform our exercise using the context of Brexit. We model potential implementations of Brexit as increases in different types of export costs between the UK and other EU countries.
Figure 1 shows the evolution of the share of affiliates of US MNEs located in the UK, Germany, and Ireland, as well as the evolution of the share of those affiliates in the UK (Germany, Ireland) that export to the EU (UK), after permanently increasing the per-period export cost between the UK and EU countries by 20%. In our model, this increase in cost can be achieved by increasing either the variable export cost, the per-period fixed export cost, or the one-time sunk export cost.
Figure 1 The effects of Brexit
Note: ‘High X’ refers to an increase in the barrier X from/to the UK to/from country j. ‘All’ refers to increasing all three export barriers from/to the UK to/from country j at once. Country j refers to Ireland, Germany, and France. Results are shown as deviations from the baseline scenario with no Brexit. Brexit happens in the model at t=15.
Our model predicts a static effect, a dynamic effect, and an aggregate price effect of Brexit.
- First, export activities between the UK and the EU become more costly, so that sales from UK-based affiliates to the EU, as well as sales from EU-based affiliates to the UK, decline, lowering the incentive to open affiliates in the UK and in other EU countries, due to the higher cost of exporting among those countries.
- Second, increases in trade costs affect the affiliate decision to enter and exit export markets.
- Finally, increases in trade costs have the effect of raising prices not only in the UK, but also in the EU, encouraging more firms to export from the UK into those markets.
Our quantitative results combine the effects of these three forces, with the strength of each effect depending on the type of costs that experiences the change.
The upper panels of Figure 1 show that higher trade costs between the UK and other EU countries always reduce the share of affiliates of US MNEs based in the UK. Higher trade costs between Germany and the UK also reduce the incentives to locate in Germany, and the share of affiliates of US MNEs located in Germany declines. In contrast, the share of affiliates of US MNEs located in Ireland slightly increases following an increase in variable trade costs, due to the price effect. The lower panels of Figure 1 show the effect of Brexit on export participation rates of British (German, Irish) affiliates of US MNEs to the EU (UK). The effects are opposite depending on the type of cost that increases.
The observed decline in export participation when either of the recurring costs increases comes from affiliates that stop exporting. Conversely, the non-monotonic response (first increase, then decrease) in export participation driven by an increase in the one-time sunk export cost comes from a pronounced decline in affiliate export exit rates. Quantitatively, the effects can also be very different depending on the type of Brexit shock. For instance, while increasing the one-time export cost has a minimal effect on the share of German affiliates of US MNEs that stop exporting to the UK, an increase in the per-period fixed export cost can reduce that share by 15%.
If Brexit leads to an increase in all types of trade costs (‘deep Brexit’), as expected, the effects would be much more pronounced (solid line in Figure 1). This case also highlights that effects could be different in the short versus the long run. Figure 1d shows that the share of British affiliates of US MNEs that export to the EU would first increase and then decrease, reaching a decline of around seven percentage points relative to the pre-shock levels after 15 years.
Finally, we are able to evaluate the potential changes in real income coming from changes in aggregate prices driven by Brexit. The largest loss is obviously experienced under deep Brexit, with a decrease in real income that goes from 0.7% in the short run to more than 0.8% after 15 years. Losses are highest for Ireland (around 1% of real income), whose economy is deeply connected to that of the UK.
One has to keep in mind that the effects described above are driven exclusively by the behaviour of US MNEs. Brexit would presumably also affect local firms, local exporters, and other non-US MNEs, magnifying the results.
The Brexit example illustrates the importance of considering the global structure of the MNE in time and space for accurately assessing the consequences of trade shocks for MNE expansion and contraction. Distinguishing different types of trade frictions is also key to understanding the dynamic responses of firms and real income changes that are the result of much discussed protectionist measures.
Dixit, A K and R S Pindyck (1994), Investment under uncertainty, Princeton University Press.
Alviarez, V (2019), “Multinational production and comparative advantage”, Journal of International Economics 119: 1-54.
Arkolakis, C, N Ramondo, A Rodriguez-Clare and S Yeaple (2018), “Innovation and production in the global economy”, American Economic Review 108(8): 2128-2173.
Fan, J (2017), “Talent, geography, and offshore R&D”, working paper, Penn State University.
Garetto, S, L Oldenski and N Ramondo (2019), “Multinational expansion in time and space”, NBER working paper 25804.
Head, K and T Mayer (2019), “Brands in motion: How frictions shape multinational production”, American Economic Review 109(9): 3073-3124.
Tintelnot, F (2017), “Global production with export platforms”, Quarterly Journal of Economics 132(1): 157-209.