The great trade collapse: What caused it and what does it mean? Richard Baldwin 27 November 2009 World trade experienced a sudden, severe, and synchronised collapse in late 2008 ââ¬â the sharpest in recorded history and deepest since WWII. This ebook ââ¬â written for the world's trade ministers gathering for the WTO's Trade Ministerial in Geneva ââ¬â presents the economics profession's received wisdom on the collapse. Two dozen chapters, written by leading economists from across the globe, summarise the latest research on the causes of the collapse as well as its consequences and the prospects for recovery.According to the emerging consensus, the collapse was caused by the sudden, severe and globally synchronised postponement of purchases, especially of durable consumer and investment goods (and their parts and components). The impact was amplified by ââ¬Å"compositionalâ⬠and ââ¬Å"synchronicityâ⬠effects in which international supply chains played a central role. The ââ¬Å"great trade collapseâ⬠occurred between the third quarter of 2008 and the second quarter of 2009. Signs are that it has ended and recovery has begun, but it was huge ââ¬â the steepest fall of world trade in recorded history and the deepest fall since the Great Depression.The drop was sudden, severe, and synchronised. A few facts justify the label: The Great Trade Collapse. It was severe and sudden Global trade has dropped before ââ¬â three times since WWII ââ¬â but this is by far the largest. As Figure 1 shows, global trade fell for at least three quarters during three of the worldwide recessions that have occurred since 1965 ââ¬â the oil-shock recession of 1974-75, the inflation-defeating recession of 1982-83, and the Tech-Wreck recession of 2001-02.Specifically: â⬠¢The 1982 and 2001 drops were comparatively mild, with growth from the previous yearââ¬â¢s quarter reaching -5% at the most. â⬠¢The 1970s event was twice that size, with g rowth stumbling to -11%. â⬠¢Today collapse is much worse; for two quarters in a row, world trade flows have been 15% below their previous year levels. The OECD has monthly data on its membersââ¬â¢ real trade for the past 533 months; the 7 biggest month-on-month drops among the 533 all occurred since November 2008 (see the chapter by Sonia Araujo and Joaquim Oliveira).Figure 1 The great trade collapses in historical perspective, 1965 ââ¬â 2009 Source: OECD Quarterly real trade data. The great trade collapse is not as large as that of the Great Depression, but it is much steeper. It took 24 months in the Great Depression for world trade to fall as far as it fell in the 9 months from November 2008 (Figure 2). The latest data in the figure (still somewhat preliminary) suggests a recovery is underway. Figure 2 The great trade collapses vs. the Great Depression Source: Eichengreen and Oââ¬â¢Rourke (2009), based on CPB online data for latest.It was synchronised â⬠¢All 104 nations on which the WTO reports data experienced a drop in both imports and exports during the second half of 2008 and the first half of 2009. â⬠¢Figure 3 shows how imports and exports collapsed for the EU27 and 10 other nations that together account for three-quarters of world trade; each of these trade flows dropped by more than 20% from 2008Q2 to 2009Q2; many fell 30% or more. Figure 3 The great trade collapse, 2008 Q2 to 2009 Q2 Sources: WTO online database.Figure 4 shows that world trade in almost all product categories were positive in 2008Q2, almost all were negative in 2008Q4, and all where negative in 2009Q1. The categories most marked by international supply chains (Mechanical and electrical machinery, Precision instruments, and Vehicles) saw some of the biggest drops, and detailed empirics in the chapter by Bems, Johnson and Yi finds that supply chains were hit harder controlling for other factors. The chart, however, shows that the falls were by no means extraordin ary large in these sectors.Figure 4 All types of goods trade collapsed simultaneously Source: Comtrade database. Manufactures and commodities Trade collapsed across the board, but it is important to distinguish between commodities and manufactures. The collapse in minerals and oil trade started from a boom time and fell faster than total trade (Figure 5). The reason was prices. Food, materials and especially oil experienced a steep run up in price in early 2008; the boom ended in mid 2008 ââ¬â well before the September 2008 Lehmanââ¬â¢s debacle. The price of manufactures, by contrast, was rather steady in this period (Figure 6).Figure 5 The great trade collapse and values: Food, oil, and manufactures Source: ITC online database. Since food, fuels, and raw materials make up about a quarter of global trade, these price movements had a big impact on aggregate trade figures. Countries dependent on commodity exports, in particular oil exporters, were among those that experienced t he greatest drop in exports (see the chapters Africa by Peter Draper and Gilberto Biacuana, and by Leonce Ndikumana and Tonia Kandiero, and on India by Rajiv Kumar and Dony Alex).The drop in manufactures trade was also massive, but it involved mostly quantity reductions. Exporters specialising in durable goods manufactures saw a particularly sharp decline in their exports (see chapters on Japan by Ruyhei Wakasugi and by Kiyoyasu Tanaka). Mexico, which is both an oil exporter and a participant in the USââ¬â¢s manufacturing supply chain, experienced one of the worldââ¬â¢s most severe trade slumps (see chapter by Ray Robertson). Figure 6 The great trade collapse and prices: Commodity vs. manufactures Source: CPB online database. CausesThe great trade collapse was triggered by ââ¬â and helped spread ââ¬â the global economic slump that has come to be called ââ¬Å"The Great Recession. 1 As the left panel of Figure 7 shows, the OECD nations slipped into recession in this per iod, with the largest importing markets ââ¬â the US, EU and Japan (the G3) ââ¬â seeing their GDP growth plummet more or less in synch. The US and Europe saw negative GDP growth rates of 3 to 4%; Japan was hit far worse. Figure 7 The current recession, OECD nations and G3, 2007Q1 ââ¬â 2009Q2 Note: G3 is US, EU and Japan. Source: OECD online data base.Why did trade fall so much more than GDP? Given the global recession, a drop in global trade is unsurprising. The question is: Why was it so big? The chapter by Caroline Freund shows that during the four large, postwar recessions (1975, 1982, 1991, and 2001) world trade dropped 4. 8 times more than GDP (also see Freund 2009). This time the drop was far, far larger. From a historical perspective (Figure 8), the drop is astonishing. The figure shows the trade-to-GDP ratio rising steeply in the late 1990s, before stagnating in the new century right up to the great trade collapse in 2008.The rise in the 1990s is explained by a nu mber of factors including trade liberalisation. A key driver, however, was the establishment of international supply chains (manufacturing was geographically unbundled with various slices of the value-added process being placed in nearby nations). This unbundling meant that the same value-added crossed borders several times. In a simple international supply chain, imported parts would be transformed into exported components which were in turn assembled into final goods and exported again, so the trade figures counted the final value added several times.As we shall see, the presences of these highly integrated and tightly synchronised production networks plays an important role in the nature of the great trade collapse (see chapters by Rudolfs Bems, Robert Johnson, and Kei-Mu Yi, and by Andrei Levchenko, Logan Lewis, and Linda Tesar). Figure 8 World trade to world GDP ratio, 1980Q1 to 2009Q2 Source: World imports from OECD online data base; World GDP based on IMF data. Emerging conse nsus on the causes Economists around the world have been working hard to understand the causes of this unusually large and abrupt shut down of international trade.The dozen chapters in Part II of this book summarise all the key research ââ¬â most of it done by the authors themselves. They do not all agree on all points, but a consensus is emerging. When sales drop sharply ââ¬â and the great trade collapse was a gigantic drop in international sales ââ¬â economists look for demand shocks and/or supply shocks. The emerging consensus is that the great trade collapse was mostly a demand shock ââ¬â although supply side factors played some role. The demand shock operated through two distinct but mutually reinforcing channels: â⬠¢Commodity prices ââ¬â which tumbled when the rice bubble burst in mid 2008 ââ¬â continued to follow world demand in its downward spiral. The price movements and diminished demand sent the value and volume of commodities trade diving. â⠬ ¢The production and exports of manufacturing collapsed as the Lehmanââ¬â¢s-induced shock-and-awe caused consumers and firms to wait and see; private demand for all manner of ââ¬Ëpostpone-ableââ¬â¢ consumption crashed. This second point was greatly amplified by the very particular nature of the demand shock that hit the worldââ¬â¢s economy in September 2008. Why so big? This consensus view, however, is incomplete.It raises the question: If the trade drop was demand driven, why was the trade drop so much larger than the GDP drop? The answer provided by the emerging consensus is that the nature of the demand shock interacted with ââ¬Å"compositionalâ⬠and ââ¬Å"synchronicityâ⬠effects to greatly exaggerate the movement of the trade-to-GDP ratio. Compositional effect The compositional effect turns on the peculiar nature of the demand shock. The demand shock was very large, but also focused on a narrow range of domestic value-added activities ââ¬â the produ ction of ââ¬Å"postponeableâ⬠goods, consumer durables and investment goods.This demand drop immediately, reducing demand for all related intermediate inputs (parts and components, chemicals, steel, etc). The compositional-effect argument is founded on the fact that postponeables make up a narrow slice of world GDP, but a very large slice of the world trade (Figure 9). In a nutshell, the common cause of the GDP and trade collapse ââ¬â a sudden drop in the demand for postponeables ââ¬â operated with full force on trade but diminished force on GDP due to the compositional difference.The large demand shock applied to the near-totality of trade while only applying to a thin portion of GDP. Here is a simple example. 2 Suppose exports consisted of 90% ââ¬Å"postponeableâ⬠(consumer and investment electronics, transport equipment, machinery and their parts and components). GDP, however, consists most of non-tradeables (services, etc). Taking postponeablesââ¬â¢ share in US GDP to be 20%, the pre-crisis situation is: When the sales of postponeables slumps by, say, half, the numerator falls much more than the denominator.Assuming that â⬠otherâ⬠continues growth in trade and GDP by 2%, the post-crisis trade to GDP ratio is Exports have fallen 44. 8% in this example, while GDP has fallen only 8. 4%. In short, the different composition of trade and GDP, taken together with the specific nature of the demand shock, has resulted in trade falling more than 5 times as fast as GDP. See the chapter by Andrei Levchenko, Logan Lewis, and Linda Tesar for a careful investigation of this logic using detailed US production and trade data; they find that the compositional effect accounts for most of the US trade drop.The chapter by Joseph Francois and Julia Woerz uses US and Chinese data to argue that the compositional effect is key to understanding the trade collapse. 3 Figure 9 Composition of world goods trade Source: WTO online database for 2007. Sync hronicity effect The synchronicity effect helps explain why the great trade collapse was so great in an even more direct manner; almost every nationââ¬â¢s imports and exports fell at the same time. There was none of the averaging out that occurred in the three other postwar trade drops. But why was it so synchronised?There are two leading explanations for the remarkable synchronicity. The first concerns international supply chains, the second concerns the ultimate cause of the Great Recession. The profound internationalisation of the supply chain that has occurred since the 1980s ââ¬â specifically, the just-in-time nature of these vertically integrated production networks ââ¬â served to coordinate, i. e. rapidly transmit, demand shocks. Even a decade ago, a drop in consumer sales in the US or Europe took months to be transmitted back to the factories and even longer to reach the suppliers of those factories.Today, Factory Asia is online. Hesitation by US and European cons umers is transmitted almost instantly to the entire supply chain, which reacts almost instantly by producing and buying less; trade drops in synch, both imports and exports. For example, during the 2001 trade collapse, monthly data for 52 nations shows that 39% of the month-nation pairs had negative growth for both imports and exports. In the 2008 crisis the figure is 83%. For details on this point, see Di Giovanni, Julian and Andrei Levchenko (2009), Yi (2009), and the chapters by Rudolfs Bems, Robert Johnson, and Kei-Mu Yi, and by Kiyoyasu Tanaka.The second explanation requires a bit of background and a bit of conjecture (macroeconomists have not arrived at a consensus on the causes of the Great Recession). To understand the global shock to the demand for traded goods, we need a thumbnail sketch of the global crisis. How the subprime crisis became the global crisis The ââ¬Å"Subprime Crisisâ⬠broke out in August 2007. For 13 months, the world viewed this as a financial crisi s that was mainly restricted to the G7 nations who had mismanaged their monetary and regulatory policy ââ¬â especially the US and the UK.Figure 3 shows that world trade continued growing apace in 2007 and early 2008. The crisis metastasised from the ââ¬Å"Subprime Crisisâ⬠to the global crisis in September 2008. The defining moment came when the US Treasury allowed the investment bank Lehman Brothers to go bankrupt. This shocked the global financial community since they had assumed no major financial institution would be allowed to go under. Many of the remaining financial institutions were essentially bankrupt in an accounting sense, so no one knew who might be next. Bankers stopped lending to each other and credit markets froze.The Lehman bankruptcy, however, was just one of a half dozen ââ¬Å"impossible eventsâ⬠that occurred at this time. Here is a short list of others:4 â⬠¢All big investment banks disappeared. â⬠¢The US Fed lent $85 billion to an insuran ce company (AIG), borrowing money from the US Treasury to cover the loan. â⬠¢A US money market fund lost so much that it could not repay its depositors capital. â⬠¢US Treasury Secretary Paulson asked the US Congress for three-quarters of a trillion dollars based on a 3-page proposal; he had difficulties in answering direct questions about how the money would fix the problem. The hereto laissez-faire US Securities and Exchange Commission banned short selling of bank stocks to slow the drop in financial institutions stock prices. It didnââ¬â¢t work. â⬠¢Daniel Gros and Stephano Micossi (2009) pointed out that European banks were too big to fail and too big to save (their assets were often multiples of the their home nationsââ¬â¢ GDPs); â⬠¢Congress said ââ¬Å"noâ⬠to Paulsonââ¬â¢s ill-explained plan, promising its own version. As people around the world watched this unsteady and ill-explained behaviour of the US government, a massive feeling of insecurity formed.Extensive research in behavioural economics shows that people tend to act in extremely risk averse ways when gripped by fears of the unknown (as opposed to when they are faced with risk, as in a game of cards, where all outcomes can be enumerated and assigned a probability). Fall 2008 was a time when people really had no idea what might happen. This is Ricardo Caballeroââ¬â¢s hypothesis of ââ¬Å"Knightian Uncertaintyâ⬠(i. e. the fear of the unknown) which has been endorsed by the IMFââ¬â¢s chief economist Olivier Blanchard. Consumers, firms, and investors around the world decided to ââ¬Å"wait and seeâ⬠ââ¬â to hold off on postponeable purchases and investments until they could determine how bad things would get. The delaying of purchases and investments, the redressing of balance sheets and the switching of wealth to the safest assets caused what Caballero has called ââ¬Å"sudden financial arrestâ⬠(a conscious reference to the usually fatal medi cal condition ââ¬Å"sudden cardiac arrestâ⬠). The ââ¬Å"fear factorâ⬠spread across the globe at internet speed. Consumers, firms and investors all feared that theyââ¬â¢d find out what capitalism without the capital would be like.They independently, but simultaneously decided to shelf plans for buying durable consumer and investment goods and indeed anything that could be postponed, including expensive holidays and leisure travel. In previous episodes of declining world trade, there was no Lehman-like event to synchronise the wait-and-see stance on a global scale. The key points as concerns the trade and GDP collapse: â⬠¢As the fear factor was propagating via the electronic press; the transmission was global and instantaneous. â⬠¢The demand shock to GDP and the demand shock to trade occurred simultaneously. ââ¬Å"Postponeableâ⬠sector production and trade were hit first and hardest. There are a number of indications that this is the right story. First, g lobal trade in services did not, in general, collapse (see the chapter by Aditya Mattoo and Ingo Borchert). Interestingly, one of the few categories of services trade that did collapse was tourism ââ¬â the ultimate postponeable. Second, macroeconomistsââ¬â¢ investigations into the transmission mechanisms operating in this crisis show that none of the usual transmission vectors ââ¬â trade in goods, international capital flows, and financial crisis contagion ââ¬â were esponsible for the synchronisation of the global income drop (Rose and Spiegel 2009). Supply-side effects The Lehman-link ââ¬Å"sudden financial arrestâ⬠froze global credit markets and spilled over on the specialized financial instruments that help grease the gears of international trade ââ¬â letters of credit and the like. From the earliest days of the great trade collapse, analysts suspected that a lack of trade-credit financing was a contributing factor (Auboin 2009). As the chapter by Jesse Mora and William Powers argues, such supply-side shocks have been important in the past.Careful research on the 1997 Asian crisis (Amiti and Weinstein 2009) and historical bank crises (see the chapter by Leonardo Iacovone and Veronika Zavacka) provide convincing evidence that credit conditions can affect trade flows. The Mora and Powers chapter, however, finds that declines in global trade finance have not had a major impact on trade flows. While global credit markets in general did freeze up, trade finance declined only moderately in most cases. If anything, US cross-border bank financing bounced back earlier than bank financing from other sources.In short, trade financing had at most a moderate role in reducing global trade. Internationalised supply chains are a second potential source of supply shocks. One could imagine that a big drop in demand combined with deteriorating credit conditions might produce widespread bankruptcies among trading firms. Since the supply chain is a cha in, bankruptcy of even a few links could suppress trade along the whole chain. The chapters by Peter Schott (on US data), by Lionel Fontagne and Guillaume Gaulier (on French data), and by Ruyhei Wakasugi (on Japanese data) present evidence that such disruptions did not occur this time.They do this by looking at very disaggregated data (firm-level data in the Fontagne-Gaulier chapter) and distinguishing between the so-called ââ¬Å"intensiveâ⬠and ââ¬Å"extensiveâ⬠margins of trade. These margins decompose changes in trade flows into changes in sales across existing trade relations (intensive) and changes in the number of such relations (extensive). If the supply-chain-disruption story were an important part of the great trade collapse, these authors should have found that the extensive margin was important.The authors, however, find that the great trade collapse has been primarily driven by the intensive margin ââ¬â by changes in pre-existing trade relationships. Trad e fell because firms sold less of products that they were already selling; there was very little destruction of trade relationships as would be the case if the extensive margin had been found to be important. This findings may be due to the notion of â⬠hysteresis in tradeâ⬠(Baldwin 1988), namely, that large and sunk market-entry costs imply that firms are reluctant to exit markets in the face of temporary shocks.Instead of exiting, they merely scale back their operations, waiting for better times. Protectionism is the final supply shock commonly broached as a cause of the great trade collapse. The chapter by Simon Evenett documents the rise in crisis-linked protectionist measures. While many measures have been put in place ââ¬â on average, one G20 government has broken its no-protection pledge every other day since November 2008 ââ¬â they do not yet cover a substantial fraction of world trade. Protection, in short, has not been a major cause of the trade collapse so far.Prospects The suddenness of the 2008 trade drop holds out the hope of an equally sudden recovery. If the fear-factor-demand-drop was the driver of the great trade collapse, a confidence-factor-demand-revival could equally drive a rapid restoration of trade to robust growth. If it was all a demand problem, after all, little long-lasting damage will have been done. See the chapter by Ruyhei Wakasugi on this. There are clear signs that trade is recovering, and it is absolutely clear that the drop has halted. Will the trade revival continue?No one can know the future path of global economic recovery ââ¬â and this is the key to the trade recovery. It is useful nonetheless to think of the global economic crisis as consisting of two very different crises: a banking-and-balance-sheet crisis in the over-indebted advanced nations (especially the US and UK), on one hand, and an expectations-crisis in most of the rest of the world on the other hand. In the US, UK and some other G7 na tions, the damage done by the bursting subprime bubble is still being felt.Their financial systems are still under severe strain. Bank lending is sluggish and corporate-debt issuances are problematic. Extraordinary direct interventions by central banks in the capital markets are underpinning the economic recovery. For these nations, the crisis ââ¬â specifically the Subprime Crisis ââ¬â has caused lasting damage. Banks, firms and individuals who over-leveraged during what they thought was the â⬠great moderationâ⬠are now holding back on consumption and investment in an attempt to redress their balance sheets (Bean 2009).This could play itself out like the lost decade Japan experienced in the 1990s (Leijonhufvud 2009, Kobayashi 2008); also see the chapter by Michael Ferrantino and Aimee Larsen. For most nations in the world, however, this is not a financial crisis ââ¬â it is a trade crisis. Many have reacted by instituting fiscal stimuli of historic proportions, but their banks and consumers are in relatively good shape, having avoided the overleveraging in the post tech-wreck period (2001-2007) that afflicted many of the G7 economies.The critical question is whether the damage to the G7ââ¬â¢s financial systems will prevent a rapid recovery of demand and a restoration of confidence that will re-start the investment engine. In absence of a crystal ball, the chapter by Baldwin and Taglioni undertakes simple simulations that assume trade this time recovers at the pace it did in the past three global trade contractions (1974, 1982 and 2001). In those episodes, trade recovered to its pre-crisis path 2 to 4 quarters after the nadir.Assuming that 2009Q2 was the bottom of the great trade collapse ââ¬â again an assumption that would require a crystal ball to confirm ââ¬â this means trade would be back on track by mid 2010. Forecasts are never better than the assumptions on which they are built, so such calculations must be viewed as what- if scenarios rather than serious forecasts. Implications What does the great trade collapse mean for the world economy? The authors of this Ebook present a remarkable consensus on this.Three points are repeatedly stressed: â⬠¢Global trade imbalances are a problem that needs to be tackled. One group of authors (see the chapters by Fred Bergsten, by Anne Krueger, and by Jeff Frieden) sees them as one the root causes of the Subprime Crisis. They worry that allowing them to continue is setting up the world for another global economic crisis. Fred Bergsten in particular argues that the US must get its federal budget deficit in order to avoid laying the carpet for the next crisis.Another group points to the combination of Asian trade surpluses and persistent high unemployment in the US and Europe as a source of protectionist pressures (see the chapters by Caroline Freund, by Simon Evenett, and by Richard Baldwin and Daria Taglioni). The chapter by Oââ¬â¢Rourke notes that avoiding a protectionist backlash will require that the slump ends soon, and that severe exchange rate misalignments at a time of rising unemployment are avoided. â⬠¢Governments should guard against compliancy in their vigil against protectionism.Most authors mention the point that while new protectionism to date has had a modest trade effect, things need not stay that way. The chapter by Simon Evenett is particularly clear on this point. There is much work to be done before economists fully understand the great trade collapse, but the chapters in this Ebook constitute a first draft of the consensus that will undoubtedly emerge from the pages of scientific journals in two or three yearsââ¬â¢ time. Footnotes 1 See Di Giovanni and Levchenko (2009) for evidence on how the shock was transmitted via international production networks. This is drawn from Baldwin and Taglioni (2009). 3 Jon Eaton, Sam Kortum, Brent Neiman and John Romalis make similar arguments with data from many nations in an unpublished manuscript dated October 2009. 4 See the excellent timeline of the crisis by the New York Fed. 5 Caballero (2009a, b) and Blanchard (2009). References Auboin, Marc (2009). ââ¬Å"The challenges of trade financingâ⬠, VoxEU. org, 28 January 2009. Baldwin, Richard (1988). ââ¬Å"Hysteresis in Import Prices: The Beachhead Effectâ⬠, American Economic Review, 78, 4, pp 773-785, 1988.Baldwin, Richard and Daria Taglioni (2009). ââ¬Å"The illusion of improving global imbalancesâ⬠, VoxEU. org, 14 November 2009. Bean, Charles (2009). ââ¬Å"The Great Moderation, the Great Panic and the Great Contractionâ⬠, Schumpeter Lecture, European Economic Association, Barcelona, 25 August 2009. Blanchard, Olivier (2009). ââ¬Å"(Nearly) nothing to fear but fear itselfâ⬠, Economics Focus column, The Economist print edition, 29 January 2009. Caballero, Ricardo (2009a). ââ¬Å"A global perspective on the great financial insurance run: Causes, consequences, and solutions (Part 2)â⬠, VoxEU. rg, 23 January 2009. Caballero, Ricardo (2009b). ââ¬Å"Sudden financial arrestâ⬠, VoxEU. org, 17 November 2009. Di Giovanni, Julian and Andrei Levchenko (2009). â⬠International trade, vertical production linkages, and the transmission of shocksâ⬠, VoxEU. org, 11 November 2009.Freund, Caroline (2009a). ââ¬Å"The Trade Response to Global Crises: Historical Evidenceâ⬠, World Bank working paper. Gros, Daniel and Stefano Micossi (2009). ââ¬Å"The beginning of the end gameâ⬠¦Ã¢â¬ , VoxEU. org, 20 September 2008. Kobayashi, Keiichiro (2008). Financial crisis management: Lessons from Japanââ¬â¢s failureâ⬠, VoxEU. org, 27 October 2008. Leijonhufvud, Axel (2009). ââ¬Å"No ordinary recessionâ⬠, VoxEU. org, 13 February 2009. Rose, Andrew and Mark Spiegel (2009). ââ¬Å"Searching for international contagion in the 2008 financial crisisâ⬠, VoxEU. org, 3 October 2009. Yi, Kei-Mu (2009), ââ¬Å"The collapse of global trade: Th e role of vertical specialisationâ⬠, in Baldwin and Evenett (eds), The collapse of global trade, murky protectionism, and the crisis: Recommendations for the G20, a VoxEU publication.