Financial crises and the global supply network: Evidence from multinational enterprises

Financial crises affect economic activity and may also distort global supply chains. This column estimates the network effects of crises by examining European multinational enterprises during the Great Recession in the 2000s. A larger network shock (captured using changes in risk premia across European countries) during the financial crisis led to lower growth in the number of affiliates, and parent firms also experienced lower revenue and employment growth. These effects were driven by parent companies with higher initial leverage. The results highlight the importance of a healthy financial system for international production organization.

Financial crises are conducive to sharp and persistent declines in economic activity (Schularick and Taylor 2012, Jordà et al. 2013). Most of the extant research efforts have focused on studying the economic effects of financial crises on the countries experiencing the crisis. However, the effects of financial crises also propagate to countries not directly exposed to the crisis. For example, Cravino and Levchenko (2017) emphasise the role that multinational enterprises play in propagating business cycle shocks. Yet, understanding how multinational enterprises help propagate financial crises across countries remains an open question. One may hypothesise that this is due to the shock spreading across multinationals’ affiliates in countries with differential exposure to the financial crisis. ¹ Also, propagation may occur due to the re-organisation of a multinational affiliates’ network following a financial shock. In Basco et al. (2024), we shed light on these questions by examining European multinational networks during the Great Recession.

The financial crisis in the euro area

The euro area experienced a salient financial crisis starting in the late 2000s. Figure 1 illustrates this financial distress in the euro area by reporting the monthly evolution of the ten-year government bond yields for Germany and Spain between 2001 and 2022. The formation of the euro area drove a convergence in interest rates, and both were almost identical in the early 2000s. This trend broke down in August 2007 when BNP Paribas froze funds related to US subprime mortgages. From this point onwards, the difference between Spanish and German bond yields (the risk premium) increased. The consensus is that this shock started the Great Recession (Cecchetti 2017). It was not until the ‘whatever-it-takes’ speech of Mario Draghi (July 2012), then-president of the ECB, that the risk premium started to decline. The figure is for Spain, but the same qualitative results emerge for all euro area countries, albeit with significant quantitative differences. We find large increases in risk premia in the periphery (e.g. Greece, Italy, Portugal) and lower in the core countries (e.g. Belgium or France). In our empirical exercise, we proxy for the severity of the financial shock by the difference in risk premium between these two events.

Figure 1 Financial disruption in euro area: Poster child

Notes: The two vertical lines are August 2007 and July 2012. The first corresponds to the announcement of BNP Paribas of freezing subprime-related funds. The second corresponds to the ‘whatever it takes’ announcement of Mario Draghi. Long-term interest rates are obtained from ECB Statistical Data.

The network shock

To compute the exposure of each parent to the financial crisis, we use the location of all affiliates in 2006 (before the BNP shock). In particular, the network shock is defined as the unweighted average of the change in risk premium in all countries where the network is present (including where the parent is located). ² This definition implies that the shock of parents in the same country will be different if they have affiliates in different countries. Since it is well-documented that multinational firms differ from domestic ones, we focus on multinationals. We also control for the industry of the parent. Therefore, we will be comparing, for example, within the car industry in Germany, a parent with affiliates in France and a parent with affiliates in Spain.

Financially hit parents reduce the number of affiliates growth and have lower sales growth

To perform our empirical analysis, we construct a panel of European multinational enterprises from 2003 to 2015. We use annual versions of the Amadeus database to assemble this panel. The details to construct this database and its representativeness can be found in Merlevede et al. (2015) and Merlevede and Theodorakopoulos (2023). Our dataset covers 29 European countries and 31 countries in total. One main advantage of this dataset is that we can track the location of affiliates over time. In addition, it is a long database, which may be needed to identify the effects since they could build slowly over time.

Our main result is that networks experiencing a larger shock had lower growth in the number of affiliates between 2006 and 2015. Quantitatively, a one standard deviation increase in the network shock translates into a 4.1% reduction in the growth rate in the number of affiliates. Figure 2 reports the effect of the network shock over time using the local projection method developed by Jordà (2005). Note that there is a clear negative trend, which is exacerbated after 2011.

We document that the effect is driven by vertical relationships (we find no effect for horizontal ones). We also find that the network of affiliates becomes more localised (both the distance between parent and affiliates and among affiliates decline), and parents choose to move away from the periphery. Lastly, we document that business complexity declines.

We also examine the effect of the network shock on the parents’ performance. We find that parents in more financially hit networks experience lower revenue and employment growth during this period. Quantitatively, one standard deviation increase in the shock reduces long-run revenues and employment growth by 7.6% and 10.7%, respectively.

Figure 2 Dynamic effect of the network shock on number affiliates growth

Notes: Network shock coefficients over time and 90% confidence intervals

Financial mechanism: The effects are driven by more initially leveraged parents

After documenting the effects of the network shock on the network, we investigate a potential mechanism behind these results. Given the financial nature of the shock, one main candidate is that credit constraints may be driving our findings. To test this hypothesis, we perform two main exercises.

First, we interact the network shock with the leverage of the parent at the onset of the financial crisis. Figure 3 reports the dynamic effect of the network shock on the growth in the number of affiliates for parents in the 10th and 90th percentiles of the leverage distribution in 2006. As can be seen, the effect is null among parents with very low initial leverage. In contrast, the trend and estimated effect are negative by the highly leveraged parents. We also show this effect is exacerbated in more financially dependent industries.

Second, we examine which affiliates are more likely to exit the network. We find that the leverage of the affiliate, by itself, does not increase the likelihood of being dropped. However, leveraged affiliates are more likely to be dropped from financially hit networks. In addition, this effect is exacerbated among more initially leveraged parents.  

Figure 3 Dynamic effect of the network shock for different levels of leverage

Notes: Network shock-leverage interaction coefficient evaluated at the 10th and 90th percentile of the leverage distribution with shaded 90% confidence intervals.

Source:cepr.org/voxeu