Global trade started to slow down markedly in the course of 2011, after it bounced back from the Great Trade Collapse of 2008–2009.1 In 2012 and 2013 the growth rate of global trade volume reached only 3%, against nearly 7% in the pre-crisis period (2002–2007) and 6.8% in the period 1985–2007 (Figure 1). A remarkable observation about the slowdown is that world trade has actually grown at a slower pace than world GDP in these two years, whereas in the pre-crisis period global trade was growing more robustly than world GDP (Figure 2).2
This raises the question of the factors behind the global trade slowdown. In particular, it is important to properly disentangle the role of cyclical versus structural factors, given that they have very different implications for the outlook. If the slowdown merely reflects that of economic activity, which was very weak in recent years, then the projected pick-up in output growth should result in a rebound of trade flows. However, the slowdown in trade could also reflect deeper, structural factors, such as a rise in protectionism or a change in global production schemes throughout the world. If such factors are at play, the dynamics of global trade and GDP could change permanently, and rule-of-thumb elasticities commonly used for global forecasts may no longer be accurate.
Figure 1. World trade in goods
Figure 2. Global GDP and trade growth
Sources: CPB, WEO, and Banque de France staff calculations.
To address this question, this column proceeds in three steps. First, it presents some key stylised facts on the global trade slowdown. These stylised facts show that the trade slowdown was particularly pronounced in advanced economies, especially in the Eurozone. However, most regions of the world have been affected, including emerging market economies (EMEs). Second, it uses a trade model to gauge how much of the slowdown can be explained by cyclical developments. This model, outlined in Bussiere et al. (2013), relates real imports to relative import prices and to a novel measure of aggregate demand, which weighs the components of GDP according to their trade intensity. This model was estimated for a panel of 18 OECD economies and was updated until 2014Q2. Results suggest that most of the slowdown can be explained by cyclical factors. Third, the column turns to ‘structural’ factors often invoked in the context of the trade slowdown and concludes that the role of global production chains does not appear to be a major contributor to the slowdown, while protectionism may have played a limited role. Although the influence of such factors is hard to quantify, preliminary evidence suggests that they cannot be ruled out.3 This question has also been addressed in Constantinescu et al. (2014).
Key stylised facts
The trade slowdown seems to have a strong regional component. First, it affects advanced economies more strongly than EMEs (Figure 3). Second, there are noticeable differences even among the two groups of countries (Figure 4). Starting with advanced economies, Eurozone trade was particularly weak in 2012, before rebounding somewhat in 2013. Even though most of the slowdown of Eurozone trade comes from trade among member countries, extra-Eurozone trade also contributed negatively to global trade growth. Among EMEs, European EMEs have shown weaker trade than other regions, particularly in 2012, likely because of the spillover effects from sluggish Eurozone economic activity. Given that most of the slowdown can be attributed to advanced economies, the analysis presented in the following section focuses on a panel of 18 OECD countries.
Figure 3. World trade in goods (volume)
Figure 4. World trade in goods: Regional patterns (advanced economies and EMEs)
Source: National sources and Banque de France staff calculations.
The role of cyclical factors: Model results
We use the results of Bussiere et al. (2013), where import elasticities are estimated based on various demand measures: import intensity adjusted demand (IAD), domestic demand (DD), and GDP. Unlike the more traditional demand terms, IAD gives more weight to the import-intensive components of demand, as read from the OECD input-output tables. For this reason we use here the specification with IAD. The figure below compares the predictions of the model with actual trade volume growth. For the 18 advanced countries considered in the estimation, eyeballing the figure (Figure 5) suggests that the model does a reasonably good job in accounting for the observed slowdown in imports.
Figure 5. Import volumes: Data and model
To understand better the performance of these models and quantify the portion of the slowdown in trade that can be accounted for by cyclical factors, starting in 2012Q1, we compute the cumulative import volume growth in the data, that implied by the model, and finally that which would be implied by assuming the historical average growth rate of import volume. (See Figure 6 below.)4
Figure 6. Cumulative import volumes: Data, model, and linear trend
We find that applying the historical average growth rate yields a cumulative growth of 13.2%, and that the model predicts a growth of 8.6%, while the actual observed cumulative growth is 4.6%. Of the 8.6 percentage points gap between observed trade growth and the one based on historical averages, cyclical factors can account for 4.6 percentage points (54%).5,6
The role of non-cyclical factors
Non-cyclical factors are harder to quantify and provide evidence for compared to cyclical factors. Here we mainly discuss trends in global value chains (GVC) and protectionism.7 The rationale for such ‘structural’ factors is as follows. Since the 1980s global trade increased significantly faster than world GDP. One hypothesis for this development is that trade liberalisation gave trade a strong boost in the decades preceding the Global Crisis, which not only affected direct exports (from one country to another) but also contributed to an increase in the fragmentation of production across countries. More specifically, goods are not simply built in one country and exported to another, but are often produced through more complex value chains, with components built in one country, assembled in another, and exported to their final destination. Some fear that the Global Crisis put an end to this trend, possibly because of a rise in protectionism (or a deceleration of the trade liberalisation process), but also because producers may have realised that global production chains were too long and inefficient.
Global value chains (GVCs)
Data on GVCs rely on input-output tables, which come with a long lag.8 The latest numbers available are those for 2011. Hence, for the recent slowdown, one can only analyse proxy measures of GVC trade. A recent study by the World Bank (Ferrantino and Taglioni 2014) approximates GVC trade essentially by ‘imported intermediate goods’.9 This study shows that GVC trade may have been a driver of the Great Trade Collapse as it fell significantly more than total trade; however, during the recent slowdown, they seem to be moving together. Therefore, one cannot conclude that GVC trade is an important driver of the recent slowdown in trade.
One leading source of information on protectionism is the Global Trade Alert (GTA), an independent initiative coordinated by the CEPR. The GTA maintains a comprehensive database of trade measures since 2009 and colour codes each measure as red (almost certainly harms a foreign commercial interest), amber (likely to harm a foreign commercial interest), and green (trade liberalising or makes national policy more transparent).10
Below is a chart with the number of new measures that were already implemented at the time of writing (Figure 7). Overall, based on the figure below, one cannot argue that the number of protectionist measures declined in 2014Q2 (because of the potential upward revisions), but can comfortably say that during the recent slowdown, the number of protectionist measures remained around the levels observed during the Great Trade Collapse, or even slightly higher towards the end of 2012 and the beginning of 2013.
Figure 7. Number of new measures implemented
Source: Global Trade Alert.
A useful statistic on protectionism from the WTO is the imports covered by import-restrictive measures divided by total imports. An advantage of this approach over the number of measures from the GTA is that the WTO takes the announced trade measures (only traditional measures) and matches them to disaggregated imports data to get a sense of the fraction of imports that gets affected. Figure 8 plots the data starting in late 2008. Based on these findings, the WTO concludes that trade restrictiveness increased only modestly, but it accumulated over time as new measures more than compensated for the removal of old restrictions.
Figure 8. Trade covered by new import restrictive measures
One of the most useful indicators on protectionism that has longer time series and good cross-country coverage is the World Bank’s Temporary Trade Barriers Database (TTBD), which includes antidumping, global safeguards, China-specific transitional safeguard measures, and countervailing duties. Based on a trade-weighted measure, this indicator shows that the stock of such barriers has been increasing mildly during the recent slowdown.
Overall, on protectionism, there seems to be some small pick-up during the trade slowdown period based on the GTA numbers, the trade covered by new import-restrictive measures from the WTO, and the TTBD from the World Bank. Given the shortcomings of protectionist measures, it is hard to reach a definitive conclusion; however, one can comfortably argue that the last 2–3 years was not a period of extensive trade liberalisation, which could explain why trade is no longer rising so much faster than GDP.11
Conclusion: How do trade projections look… and should there be a policy response?
The latest IMF WEO projections assume 5.6% growth in world trade and 4.1% real GDP growth for 2019. These projections suggest that global trade is expected to recover from its low levels today and exceed GDP growth. Global trade should benefit in particular from the projected pick-up in investment in advanced economies.
Figure 9. Global trade volumes and real GDP growth: WEO projections (% change over 12 months)
Clearly, cyclical factors should disappear once the global recovery progresses further. Among the non-cyclical factors, we tend to think that only protectionism may be a concern if there are further signs that it is on the rise, going forward. By contrast, GVC trade can be seen as an optimal response to structural changes where policy action is harder to justify.
Authors’ note: The views expressed here are those of the authors and do not necessarily represent those of the institutions with which they are affiliated.
Baldwin, R (2009), The Great Trade Collapse: Causes, Consequences and Prospects, VoxEU.org eBook, 27 November 27.
Baldwin, R and J Lopez-Gonzalez (2013), “Supply-chain Trade: A Portrait of Global Patterns and Several Testable Hypotheses”, NBER Working Paper 18957.
Bussiere, M, G Callegari, F Ghironi, G Sestieri, and N Yamano, (2011), “Estimating Trade Elasticities: Demand Composition and the Trade Collapse of 2008–09”, American Economic Journal: Macroeconomics 5(3): 118–151.
Constantinescu, C, A Mattoo and M Ruta (2014), “Slow Trade,” Finance & Development 51(4).
Ferrantino, M J and D Taglioni (2014), “Global Value Chains in the Current Trade Slowdown”, World Bank Economic Premise 137.
Gawande, K, B Hoekman, and Y Cui (2014), “Global Supply Chains and Trade Policy Responses to the 2008 Crisis”, World Bank Economic Review.
Krugman, Paul (2013), “Should Slowing Trade Growth Worry Us?”, New York Times blog, 30 September. .
Sturgeon, T J and O Memedovic (2011), “Mapping Global Value Chains: Intermediate Goods Trade and Structural Change in the World Economy”, United Nations Industrial Development Organization, Development Policy and Strategic Research Branch Working Paper 05/2010.
 The Great Trade Collapse refers to the sudden and sharp drop in global trade in the wake of the 2008 financial crisis; see in particular Baldwin (2009) for an early analysis.
 The exact timing of the slowdown is hard to gauge. The year-on-year growth rate of world trade dropped from 7.4% in May 2011 to 3.7% the following month and remained low thereafter, so one may consider that trade decelerated in mid-2011.
 Other potential explanations for the trade slowdown, such as the role of trade finance and other factors affecting the supply side, are not tackled here to keep the column concise and tractable.
 The time period we consider in the calculation of the average growth is 1985–2014Q2. Note that it includes the Great Trade Collapse, during which import volume shrank by 10% in one quarter. Pushing the starting date of the average calculation to the mid-90s does not change our conclusion that about half of the trade slowdown can be explained by the model.
 If we calculate the same statistics starting in mid-2011 instead of 2012, the share of the slowdown explained by the model rises to 58%. The model tracks the data fairly well over the last two quarters of 2011. In fact, the model underestimates the slowdown mainly during two quarters – the last quarter of 2012 and the first quarter of 2013 – especially for a few observations, including for 2012Q4 the US, Germany, and Japan, and for 2013Q1 Australia, Netherlands, and the UK.
 We have also estimated the trade elasticity using the GDP model for several subsamples to identify any time variation in the value of the elasticity. Our panel regressions suggest that the elasticity estimated for the slowdown period was not significantly different than the one estimated using the entire sample, but this may be due to the fact that the slowdown period is very short and the coefficient is not tightly estimated.
 Of course one could think of additional structural factors, such as deep changes in consumption habits, e.g. associated with more home bias after the crisis, a protracted disruption of financial services hampering the distribution of trade finance, and other factors affecting the supply side. However, these other factors are probably harder to gauge and we leave it to future research to explore them.
 For a recent analysis of GVCs see for instance Baldwin and Lopez-Gonzalez (2013), Gawande et al. (2014), and Sturgeon and Memedovic (2011).
 This World Bank study takes disaggregated trade data for three sectors (apparel, electronics, autos), using Broad Economic Categories to label certain subsectors as ‘intermediate goods’ and using this as a proxy for GVC trade. They focus on these three sectors because they are known to be important in GVC trade. The main shortcoming of this approach is that if an imported intermediate good is used in the production of a final good that is consumed domestically, it should not be counted as GVC trade. In fact, GVC trade would be a subset of imported intermediate goods. To the extent that the domestically consumed final goods and those that are re-exported behave similarly, the approximation in terms of growth rates will be good.
 There are two disadvantages to GTA data. First, more measures are added to historical numbers as they are discovered, so the most recent observations will certainly be revised upwards significantly. Second, another potential concern is that the data include ‘behind-the-border’ measures in addition to those that are directly targeted to trade.
 Trade liberalisation cannot continue forever anyway. In particular, once tariffs have been brought to zero, they cannot be reduced any more – for tariffs also the zero lower bound is a constraint (even though non-tariff measures can always be phased out).