Trade Liberalization and Economic Growth: An Examination of Cross-Country Indicators, Slides of Marketing

Download Trade Liberalization and Economic Growth: An Examination of Cross-Country Indicators and more Slides Marketing in PDF only on Docsity! JEFFREY D. SACHS Harvard University ANDREW WARNER Harvard University Economic Reform and the Process of Global Integration WHEN THE BROOKINGS Panel on Economic Activity began in 1970, the world economy roughly accorded with the idea of three distinct eco- nomic systems: a capitalist first world, a socialist second world, and a developing third world which aimed for a middle way between the first two. The third world was characterized not only by its low levels of per capita GDP, but also by a distinctive economic system that assigned the state sector the predominant role in industrialization, although not the monopoly on industrial ownership as in the socialist economies. The years between 1970 and 1995, and especially the last decade, have witnessed the most remarkable institutional harmonization and economic integration among nations in world history. While economic integration was increasing throughout the 1970s and 1980s, the extent of integration has come sharply into focus only since the collapse of com- munism in 1989. In 1995 one dominant global economic system is emerg- ing. The common set of institutions is exemplified by the new World Trade Organization (WTO), which was established by agreement of more than 120 economies, with almost all the rest eager to join as rapidly as possible. Part of the new trade agreement involves a codification of basic principles governing trade in goods and services. Similarly, the In- ternational Monetary Fund (IMF) now boasts nearly universal member- ship, with member countries pledged to basic principles of currency convertibility. Most programs of economic reform now underway in the developing world and in the post-communist world have as their strategic aim the 1 2 Brookings Paper-s on Economic Activity, 1:1995 integration of the national economy with the world economy. Integra- tion means not only increased market-based trade and financial flows, but also institutional harmonization with regard to trade policy, legal codes, tax systems, ownership patterns, and other regulatory arrange- ments. In each of these areas, international norms play a large and often decisive role in defining the terms of the reform policy. Most recently, China made commitments on international property rights and trade pol- icy with a view toward membership in the WTO, and membership in the world system more generally. Russian economic reforms are similarly guided by the overall aim of reestablishing the country's place within the world market system. In several sections of its April 1995 agreement with the IMF, the Russian government commits to abide by WTO princi- ples, even in advance of membership. The goal of this paper is to document the process of global integration and to assess its effects on economic growth in the reforming countries. Using cross-country indicators of trade openness as the measures of each country's orientation to the world economy, we examine the timing of trade liberalization, and the implications of trade liberalization for subsequent growth and for the onset or avoidance of economic crises. Of course, trade liberalization is usually just one part of a government's overall reform plan for integrating an economy with the world system. Other aspects of such a program almost always include price liberaliza- tion, budget restructuring, privatization, deregulation, and the installa- tion of a social safety net. Nonetheless, the international opening of the economy is the sine qua non of the overall reform process. Trade liberal- ization not only establishes powerful direct linkages between the econ- omy and the world system, but also effectively forces the government to take actions on the other parts of the reform program under the pres- sures of international competition. For these reasons, it is convenient and fairly accurate to gauge a country's overall reform program ac- cording to the progress of its trade liberalization. Our analysis helps to answer several debates concerning cross-coun- try growth patterns. Most important, we help to resolve the widely dis- cussed conundrum concerning economic convergence in the world economy. Long-held judgments about the development process, as well as the workhorse formal models of economic growth, suggest that poorer countries should tend to grow more rapidly than richer countries and therefore should close the proportionate income gap over time. The Jeffrey D. Sachs and Andrew Warner 5 Liberalization and Global Integration before 1970 One and one-half centuries ago, two close observers of the capitalist revolution in Western Europe made a pithy prediction about the course of global economic change. Marx and Engels correctly sensed the un- precedented efficiency of the industrial capitalism that had emerged. They predicted that as a result of superior economic efficiency, capital- ism would eventually sweep through the entire world, compelling other societies to restructure along the lines of Western Europe. In the pun- gent rhetoric of the Communist Manifesto they expostulated that: The bourgeoisie, by the rapid improvement of all instruments of production, by the immensely facilitated means of communication, draws all, even the most barbarian, nations into civilization. The cheap prices of its commodities are the heavy artillery with which it batters down all Chinese walls, with which it forces the barbarians' intensely obstinate hatred of foreigners to capitulate. It compels all nations, on pain of extinction, to adopt the bourgeois mode of production; it compels them to introduce what it calls civilization into their midst, i.e., to be- come bourgeois themselves. In one word, it creates a world after its own image.4 Marx and Engels got much disastrously wrong in their predictions, but they correctly sensed the decisive global implications of capitalism. As they foresaw, capitalism eventually spread to nearly the entire world, in a complex and sometimes violent process that dramatically raised worldwide living standards but also provoked social upheaval and war. It is often forgotten today, in the flush of the communist collapse after 1989, that global capitalism has emerged twice, at the end of the nineteenth century as well as the end of the twentieth century. The ear- lier global capitalist system peaked around 1910 but subsequently disin- tegrated in the first half of the twentieth century, between the outbreak of World War I and the end of World War II. The reemergence of a global, capitalist market economy since 1950, and especially since the mid-1980s, in an important sense reestablishes the global market econ- omy that had existed one hundred years earlier. The first episode of global capitalism, of course, came about as much through the instruments of violent conquest and colonial rule as through economic reform and the development of international institutions. Starting around 1840, Western European powers wielded their superior 4. Marx and Engels (1948, p. 225). 6 Brookings Papers on Economic Activity, 1:1995 industrial-and hence military-power to challenge traditional societies around the world. France began to colonize North Africa in the 1830s and 1840s; Britain forced its way into China in the Opium Wars, 1839-42; Britain and France defeated Russia in the Crimean War, 1854-56; and Britain completed the conquest of India in 1857. Among the populous societies of Asia and Near East, only Japan was able to mobilize social and political institutions to support market reforms, im- plementing history's first "shock therapy" reforms following the 1868 Meiji Restoration.s By the 1870s a global market had begun to take shape on the following economic lines. Western Europe and the United States constituted the main industrial powers. A major push toward industrialization, espe- cially in east-central Europe, followed the unification of Germany. Rus- sia began a period of rapid industrialization, partly through the building of foreign-financed railways across Russian Eurasia. Japan had begun its dramatic opening to the world economy through the adoption of capi- talist institutions and free trade. (Note that early Japanese industrializa- tion took place entirely under free trade, since the dominant Western powers imposed low Japanese tariff levels through "unequal treaties" that lasted until the end of the century.) Latin America, after a half cen- tury of postindependence upheaval, finally settled into market-based, export-led growth in the 1870s, based on raw materials exports and capi- tal imports (primarily for railroad construction). Africa, which lagged farthest behind, was gobbled up by the Western European powers in an orgy of imperial competition that reached its height between 1880 and 1910. Trade barriers remained low among these economies for several decades, from the 1860s to 1914.6 5. See Jansen and Rozman (1988) for a detailed analysis of the economic, political, and social reforms of the Meiji period. 6. The era of nineteenth-century free trade is usually dated from 1846, when Britain unilaterally liberalized by repealing the Corn Laws. (In fact, liberalization had begun ear- lier, with the abolition of export duties in 1842 and the reduction of import duties in 1842 and 1845.) The next decisive step was the Cobden-Chevalier Treaty of 1860, which liberal- ized British-French trade. The new German Reich was established by Bismarck on free trade principles and low tariffs in the early 1870s. It is often suggested that this free trade era ended in 1879 with a renewed wave of protectionism, starting with Bismarck's accep- tance of the famous tariff of bread and iron, which raised imports duties on agriculture and steel. Higher tariffs soon followed in France and Italy. In fact, even with these tariff increases, average tariff rates remained low until World War I, and nontariff barriers (for example, quotas and exchange controls) were virtually nonexistent. According to data assembled by Capie (1983, table 1.3, p. 8), average tariff revenues as a percentage of total Jeffrey D. Sachs and Andrew Warner 7 As in the late twentieth century, the emergence of the first global sys- tem was based on the interaction of technology and economic institu- tions. Long-distance transport and communications achieved break- throughs similar to those in the present.7 The Suez Canal, completed in 1869, and the Panama Canal, completed in 1914, dramatically cut inter- national shipping times, as did the progressive development of faster and larger steamships from the 1840s. New railways in India, Russia, the United States, and Latin America-often built with foreign finance- opened vast, fertile territories for settlement and economic develop- ment. The spread of telegraph lines and transoceanic cables from the 1850s linked the world at electronic speed. Military innovations, partic- ularly the breech-loading rifle in the 1840s, combined with mass-produc- tion made possible by industrialization, decisively shifted the military advantage to Europe. Medical advances, particularly the use of quinine as a preventative against malaria, played a pivotal role in the spread of European settlements, domination, and investment, especially in Af- rica. Without doubt, these technological breakthroughs were as revolu- tionary in underpinning the emerging global system as those of our own age. On the economic level, key institutions similarly spread on a global scale. International gold and silver standards became nearly universal after the 1870s, eventually embracing North and South America, Eu- rope, Russia, Japan, China, as well as other European colonies and inde- pendent countries. By 1908 roughly 89 percent of the world's population lived in countries with convertible currencies under the gold or silver standard.8 Basic legal institutions, such as business and commercial imports stayed below 10 percent in France, Germany, and the United Kingdom; between 10 and 20 percent in Italy; between 20 and 30 percent in the United States; and between 20 and 40 percent in Russia. 7. See Headrick (1981). 8. See Eichengreen and Flandreau (1994, p. 9). The countries on the gold or silver stan- dards in 1908 include, in Europe: United Kingdom, France, Belgium, Switzerland, Italy, Germany, Netherlands, Portugal, and Romania; in North America: United States and Canada; in Central America: Mexico, Nicaragua, Guatemala, Honduras, Salvador, and Costa Rica; in South America: Peru, Chile, Brazil, Venezuela, and Argentina; in Asia and the Pacific: India, China, Indonesia, Japan, Siam, Philippines, and Australia; and in the Middle East: the Ottoman Empire, Egypt, and Persia. The national currencies were con- vertible into gold in all cases except the following: Italy, Austria, Spain, Portugal, Nicara- gua, Guatemala, Peru, Chile, Brazil, and Venezuela. The Italian and Austrian currencies were stable though not convertible. 10 Brookings Papers on Economic Activity, 1:1995 the primary producers in Latin America and elsewhere was undermined by low and unstable commodities prices in the 1920s, and then was dev- astated by the Great Depression, which brought the utter collapse of the terms of trade, intense protectionism in Europe and the United States, and the end of capital inflows. Political upheaval accompanied economic and military upheaval. Most important was the Bolshevik Revolution in Russia in 1917, and the emergence of fascist states in Italy and Germany in the 1920s and 1930s respectively. In Latin America, the traditional political power of the landholders and mine owners was undermined by the collapse in the terms of trade. The free trade regimes of the late nineteenth century were replaced by a revolutionary regime in Mexico and authoritarian re- gimes in Argentina, Brazil, and Chile, that were heavily influenced by the state planning of the communist and fascist regimes in the Soviet Union and Europe. 16 Throughout the world, state planning, authoritari- anism, and militarism competed with limited government and market- based economies. Whether or not economic theory offered insights and predictions about these alternative strategies, political leaders felt com- pelled to push for new and radical experimentation. The changing zeitgeist is again decisively captured by Keynes, in his remarkable lecture "National Self-Sufficiency" delivered in Ireland in 1933, when the world was in the depths of the Great Depression. 17 In the lecture, Keynes rejects the commitment to free trade and the interna- tional harmonization of institutions, declaring the late-nineteenth- century experience a massive, and apparently inevitable, failure. In Keynes's view, the international system led to war, by stoking the com- petition among the leading powers: The protection of a country's existing foreign interests, the capture of new mar- kets, the progress of economic imperialism-these are a scarcely avoidable part of a scheme of things which aims at the maximum of international specialisation and at the maximum geographical diffusion of capital wherever its seat of own- ership. 18 For this reason, countries are best linked by ideas and culture, not economic and financial entanglements. Keynes writes: I sympathise, therefore, with those who would minimise, rather than with those who would maximise, economic entanglements between nations. Ideas, knowl- 16. See Thorp (1984) for very insightful essays on the country-by-country experience. 17. Keynes (1933). 18. Keynes (1933, p. 236). Jeffrey D. Sachs and Andrew Warner 11 edge, art, hospitality, travel-these are the things which should of their nature be international. But let goods be homespun whenever it is reasonably and con- veniently possible; and, above all, let finance be primarily national. 19 But perhaps more to the point, Keynes stressed, was the fact that countries simply demanded the right to experiment with new economic models, since the old ones no longer commanded respect and assent. He joined the chorus for experimentation, vividly exemplifying the end of intellectual faith in global capitalism by the 1930s: The point is that there is no prospect for the next generation of a uniformity of economic systems throughout he world, such as existed, broadly speaking, dur- ing the nineteenth century; that we all need to be as free as possible of interfer- ence from economic changes elsewhere, in order to make our own favourite ex- periments towards the ideal social republic of the future; and that a deliberate movement towards greater national self-sufficiency and economic isolation will make our task easier, in so far as it can be accomplished without excessive eco- nomic cost.20 Ironically, while Keynes would fundamentally reverse course himself, coming to see aggregate demand management and international institu- tions such as the IMF as the linchpins of a renewed global capitalist sys- tem, the genie of experimentation unleashed by the collapse of faith in market institutions lived on to dominate most of the world through much of the postwar era. The Tripartite World after 1945 At the end of World War II, the international economic system was in a shambles. International markets for trade in goods, services, and financial assets were essentially nonexistent. International trade was de- stroyed by currency inconvertibility and a web of protectionist mea- sures stemming from the Great Depression and World War II. When the IMF published its first comprehensive review of exchange rate arrange- ments in 1950, only five countries had established freely convertible cur- rencies under the standard of article VIII of the IMF Articles of Agreement: the United States and four Latin American countries pegged to the dollar, El Salvador, Guatemala, Mexico, and Panama.2' Switzerland, not then a member of the IMF, also had a convertible cur- rency. The IMF characterized another four countries as having effec- 19. Keynes (1933, p. 236). 20. Keynes (1933, p. 241). 21. IMF, Annual Report on Exchange Restrictions, 1950. 12 Brookings Papers on Economic Activity, 1:1995 Lively convertible currencies, even though they had not yet formally ac- cepted the obligations of article VIII: Cuba, Dominican Republic, Honduras, and Venezuela. As late as 1957, only two more countries had established convertibility subject to article VIII: Canada and Haiti. The members of the European Community established convertibility in 1958. Most other developing and socialist countries postponed the move for decades. While market-based economic linkages were methodically restored among the leading countries during the 1950s, most of the world's popu- lation lived in countries that chose fundamentally nonmarket economic strategies for development. Roughly one-third of the world's population lived in socialist countries (as measured by Kornai for the year 1986); another 50 percent or so lived in countries where governments pro- claimed a kind of "third way" between capitalism and socialism, state- led industrialization (SLI).2 In figure 1, we show the time profile of the opening of the world econ- omy in the postwar era, using the specific criteria for openness discussed below and in the appendix. The world economy was essentially closed after World War II, and only around 20 percent of the world's population lived in open economies by 1960. It was not until 1993 that more than 60 percent of the world's GDP, and more than 50 percent of the world's population, was located in open economies.23 The figure extends up to 1994, so that by our criteria, neither Russia nor China is part of the open system. If both of these countries cross the threshold to openness (and trade reforms in 1995 might well lead them to qualify), the proportion of openness by population would jump another 30 percent, to reach around 87 percent of the world's population; and the proportion of openness by GDP would jump by another 15 percent, to reach around 83 percent of the world's GDP (using 1975 weights in both cases). The governments of almost all the developing countries adopted either socialist or SLI policies after World War II. This was true of the 22. The population in socialist countries is measured by Kornai (1992, pp. 6-7) for 1986. The population under SLI is based on the authors' calculations using data from Sum- mers and Heston (1991). 23. Let PO, be the proportion of the world economies that are open in year t, as shown in figure 1. PO, is constructed as PO, = IVt'75Djt -1, where Dit -1 is a dummy variable set equal to one if the country is open as of year t - 1, and zero otherwise. W(75 is the weight of country i in the world in 1975. The weights are constructed using 1975 population data and 1975 real GDP data from version 5.6 of the data in Summers and Heston (1991). Jeffrey D. Sachs and Andrew Warner 15 Under these circumstances, the restoration of exchange rate convert- ibility required either a monetary reform (to remove the monetary over- hang); a temporary rise of prices and a currency devaluation (to absorb the monetary overhang through inflation); a long period of real economic growth to raise the demand for nominal money balances; or some combi- nation of the three. Milton Friedman cogently argued that a floating ex- change rate (combined with price decontrol) was the best policy: it would establish convertibility, and hence the monetary basis for free trade, automatically and immediately.26 Most countries, however, shied away from the temporary inflationary consequences that would have ac- companied such a move, even though they would have been one-shot rather than ongoing. Largely for this reason, the return to convertibility in Europe and many other parts of the world was delayed for more than a decade after World War II. In some countries, the consequences were even more long-lasting. In India, for example, various attempts to relax price controls and to rees- tablish free trade led to a spurt in prices and a subsequent reversal of the policy.27 For several years, half-hearted attempts at liberalization were reversed as the result of the accompanying price increases. During this period the ideology of state control gained in importance, as Jawaharlal Nehru consolidated his hold on the Indian leadership. Thus the initial macroeconomic pressures delayed the establishment of convertibility, while ideology and interest-group lobbying cemented the postwar poli- cies of inconvertibility, licensing, and protection. Even after the 1950s, macroeconomic instability continued to pro- duce closed trade practices in many countries. Throughout Latin America, populist fiscal policies repeatedly undermined the commit- ment to currency convertibility and allowed the emergence of significant black market premiums on the exchange rate in countries with a pegged rate regime. In this way trade practices were often closed inadvertently, as a result of overly expansionary macroeconomic policies, rather than 26. Friedman (1953) has typically been read as an argument for floating rather than fixed exchange rates. More fundamentally, it is an argument for convertibility (which auto- matically follows from floating rates) as compared to inconvertibility (which often accom- panies a fixed exchange rate regime). Friedman reasoned that a commitment to a fixed exchange rate would almost inevitably lead to balance-of- payments pressures and hence, multiple exchange rates or other forms of inconvertibility. This was certainly the pattern as of 1953. 27. See Tomlinson (1992). 16 Brookings Papers on Economic Activity, 1:1995 deliberately. Nonetheless, the economic effects were similar: a rise in the relative prices of import-competing goods, a reduction of imports, and a reduction of exports. INTELLECTUAL B ELI EFS. Several currents of intellectual belief coa- lesced to support the nearly worldwide adoption of SLI and socialist strategies. Following two world wars and the Great Depression, the lib- eral world view seemed to be indefensible. Capitalism had proved to be rapacious and violent, as J. A. Hobson and Lenin had predicted.28 Even Keynes had subscribed to this view in 1933. Moreover, the Marxist no- tion that profits were the result of the exploitation of labor was an ex- traordinarily enticing explanation for elites in the poorer countries, who could justifiably view the poverty of their own nations as the result of degradations committed by the richer nations. Moreover, as Keynes had seemingly demonstrated, capitalism was inherently unstable and needed the steadying rudder of the state, per- haps in the form of the nearly full nationalization of future investment. It should be remembered that banking, insurance, and much heavy in- dustry were nationalized in France (under Charles De Gaulle) and in Britain (under Clement Atlee) as well as in many other Western Euro- pean countries, and not just in the developing and socialist worlds. At the same time, there was a growing belief that coordinated, large- scale public investment was necessary to make a breakthrough to mod- ernizing industrialization. Paul Rosenstein-Rodan championed the strategy of the "big push," and Alexander Gershchenkron argued that the idea was supported by the history of nineteenth-century Europe, in which the countries lagging in industrialization increasingly relied on the state to catch up with the richer countries.29 The greater the gap at the start of industrialization, according to Gershchenkron, the greater was the state's role in mobilizing resources for the breakthrough. The appar- ent industrial successes of the Soviet Union, which had proved suffi- cient to defeat Nazi Germany, seemed to many observers to give ample confirmation of the technical possibilities of investment planning and state-led industrialization. These ideas, backed up by the new tech- niques of national income accounting, input-output analysis, and mathe- matical growth models, led to the widespread endorsement of develop- ment planning models in mainstream development economics. 28. See Hobson (1902) and Lenin (1926). 29. See Rosenstein-Rodan (1943) and Gershchenkron and Nimitz (1952). Jeffrey D. Sachs and Andrew Warner 17 Export pessimism combined with the idea of the big push to produce the highly influential view that open trade would condemn developing countries to long-term subservience in the international system as raw materials exporters and manufactured goods importers. Comparative advantage, it was argued by the Economic Commission of Latin America (ECLA) and others, was driven by short-run considerations that would prevent raw materials exporting nations from ever building up an industrial base. The protection of infant industries was therefore vital if the developing countries were to escape from their overdepen- dence on raw materials production. These views spread within the United Nations system (to regional offices of the United Nations Eco- nomic Commission), and were adopted largely by the United Nations Conference on Trade and Development (UNCTAD). In 1964 they found international legal sanction in a new part IV of the General Agreement on Tariffs and Trade (GATT), which established that developing coun- tries should enjoy the right to asymmetric trade policies. While the de- veloped countries should open their markets, the developing countries could continue to protect their own markets. Of course, this "right" was the proverbial rope on which to hang one's own economy! More radical anti-capitalist views fueled Marxist-inspired revolu- tions in nearly two dozen countries during the postwar period. Forrest Colburn offers a masterful evocation of the underlying ideas and sym- bols common to these revolutions.30 He puts great stress on the role of ideas, rather than the political economy in motivating the revolutionary leaders: The trajectory of contemporary revolutionary regimes illuminates why, at least in poor countries, the choices of political elites are so consequential. In many such countries, political elites are not significantly constrained by either the in- stitutions and norms of government or by civil society. Thus, the time for experi- mentation and implementation of ideas can be dangerously compressed.31 STATE BUILDING. In his classic analysis of European mercantilism, Eli Heckscher argued that mercantilist trade and industrial policies were a crucial mechanism by which new nation states consolidated their polit- 30. The list, as provided by Colburn (1994, p. 8) is as follows: Afghanistan (1978), Alge- ria (1962), Angola (1975), Benin (1972), Bolivia (1952), Burkina Faso (1983), Burma (1962), Cambodia (1975), China (1949), Cuba (1959), Egypt (1952), Ethiopia (1974), Grenada (1979), Guinea-Bissau (1974), Laos (1975), North Korea (1948), South Yemen (1967), Viet- nam (1945). The Iranian revolution (1979) was inspired by Islamic, not Marxist, principles. 31. Colburn (1994, p. 103). 20 Brookings Papers on Economic Activity, 1:1995 nopoly purchases of agricultural output at below-market prices; and the tilt of the internal terms of trade in favor of urban (largely government) workers, and away from peasant cultivators.38 The basic Heckscher-Ohlin-Samuelson (HOS) and Ricardo-Viner (RV) models of trade give some notion of which economic groups in a society should favor trade protection (either import taxation or export taxation), and which should lean toward open trade. According to the HOS model, the move from autarky to trade favors the abundant factors of production and reduces the real income of the scarce factors of pro- duction. Thus the relatively scarce factors of production in an economy should tend to be in favor of autarkic policies. The RV theory highlights the implications of factor immobility between sectors. When capital or labor cannot move between sectors, the immobile factors should tend to favor protection for their own sector, irrespective of the overall scarcity or abundance of specific factors of production. Firms with sunk capital in the import-competing sector, and workers with skills specific to that sector, should tend to favor protection of the sector. Ronald Rogowski and others have examined relative factor intensi- ties to assess the pressures for and against free trade in the postwar era.39 Most of the Asian economies tend to have high labor-to-land ratios (land is the scarce factor), suggesting that workers would tend to favor free trade (in order to benefit from the export of labor-intensive goods and the import of inexpensive food), while landowners would tend to favor protection (to raise the price of foodstuffs in the local economy). In Latin America and Africa, where labor is scarce and land is abundant, we would expect the reverse: landowners should be on the side of free trade (to raise the export price of foodstuffs), and urban workers should be interested in protection (against the import of labor-intensive goods and the export of foodstuffs). Of course, the relative power of the various interests to influence trade policy will depend on a myriad factors, including the capacity of competing groups to organize politically and the institutions for political competition (for example, elections or military rule). In Latin America from the 1950s to the 1980s, for example, protectionism tended to be fa- vored during democratic periods, since workers (who, as the scarce fac- tor, favored protection) could outvote landowners; free trade, on the 38. See Bauer (1954) and Bates (1988). 39. Rogowski (1989). Jeffrey D. Sachs and Andrew Warner 21 other hand, was typically promoted by authoritarian governments, sid- ing with large landowners and mineowners.40 In many parts of the devel- oping world, especially Latin America and Africa, political power has been disproportionately concentrated in urban areas, thereby adding to the political weight of labor relative to landowners and turning the trade regime more protectionist.41 It might seem that a labor-intensive economy would tend to lean more readily toward free trade than would a land-intensive (or resource-inten- sive) economy. Postwar governments have tended to respond more to labor interests than landowner interests, whether as the result of the search for votes, or the fear of labor unrest, or the urban bias promoted by government-sector workers. If labor interests are indeed the deter- mining factor, then trade liberalization would come more readily in Asia than in Latin America or Africa. But as already noted, interest group politics has hardly been decisive. Some labor-intensive economies, such as the South Asian countries (India, Bangladesh, Pakistan, and Sri Lanka) were long protectionist, while labor-scarce Chile became the first sustained free trader in Latin America (although, notably, under a military regime). The Classification and Timing of Trade Policies The outcome of these various forces produced an overwhelming turn toward socialism or SLI in the developing world during the 1940s and 1950s, which was only gradually reversed over the next forty years. According to our classifications, shown in tables 1-5, seventy-eight de- veloping countries outside of the Soviet bloc chose some form of in- ward-looking development strategy in the postwar period. Of these, forty-three had opened their economies by 1994 (see table 2) and thirty- five were still closed as of 1994 (see table 3). Although developed coun- tries typically started open and remained open throughout the period, 40. One example is Peru, which maintained open trade during the Odria dictatorship, between 1948 and 1956. Trade remained relatively free during the democratic presidency of Manuel Prado (1956-63), but then turned gradually more protectionist under Fernando Belaunde Terry (1963-68). Peru finally embarked on autarkic, socialist policies under a left-wing military dictatorship led by Juan Velasco Alvarado (1968-75). Conversely, it was the Pinochet dictatorship in Chile after 1973 which ended decades of protectionism. For details, see Skidmore and Smith (1984). 41. See Lipton (1976) and Bates (1981). 22 Brookings Papers on Economic Activity, 1:1995 Table 1. Developing Economies That Have Always Been Opena Country Year of independence Barbados 1966 Cyprus 1960 Hong Kong not applicable Malaysia 1963 Mauritius 1968 Singapore 1965 Thailand never colonizt-d Yemen Arab Republic 1918 Source: See appendix. a. Since independence, where applicable. we have found eight other developing countries that followed this pat- tern (see table 1) and thirteen that had episodes of temporary liberaliza- tion (these periods are identified in parentheses in tables 1-5). A parallel process of liberalization was underway in the developed economies, although integration was typically achieved in the 1950s and 1960s, rather than the 1980s and 1990s (see table 4). Note that for the purposes of this paper, we define developed economies as all countries with a real GDP of $5,000 or more in 1970, according to the purchasing- power adjusted data in Summers and Heston.42 This criterion results in a few classifications that are not standard, namely that Ireland, Greece, and Portugal are classified as developing countries, while Trinidad and Tobago and Venezuela are classified as developed. But these unusual classifications have little impact on our main conclusions. Our categorization and timing of trade liberalization are fundamental to tables 1-5 and the subsequent empirical work. We judge a country to have a closed trade policy if it has at least one of the following character- istics: 1. Nontariff barriers (NTBs) covering 40 percent or more of trade. 2. Average tariff rates of 40 percent or more. 3. A black market exchange rate that is depreciated by 20 percent or more relative to the official exchange rate, on average, during the 1970s or 1980s. 4. A socialist economic system (as defined by Kornai).43 5. A state monopoly on major exports. 42. Summers and Heston (1991). 43. Kornai (1992). Jeffrey D. Sachs and Andrew Warner 25 Table 4. Developed Economies with Year of Opening Year of Countiy openinga Australia 1964 Austria 1960 Belgium 1960 Canada 1952 Denmark 1960 Finland 1960 France 1959 Germany 1959 Israel 1985 Italy 1959 Japan 1962 Luxembourg 1959 Netherlands 1959 New Zealand 1986 Norway 1960 Spain 1960 Sweden 1960 Switzerland 1950 or earlier Trinidad and Tobago closed United Kingdom 1960 United States 1950 or earlier Venezuela (1950-59) (1989-92) Source: See appendix. a. Dates in parentheses identify temporary liberalizations. the periods of temporary openness were often characterized by sus- tained economic growth at a higher level than during the subsequent pe- riod of closure. Finally, in table 5 we show the liberalization record of the post-socialist European economies, but in this case we rely on EBRD standards of openness. The five criteria are chosen in order to cover all of the major types of trade restriction. Tariffs and nontariff barriers (mainly quotas) are most obvious. We rely mainly on UNCTAD data for these classifications. The black market premium (BMP) is a measure of exchange control: a large BMP is evidence of the rationing of foreign exchange, which tends to be a form of import control. The socialist classification is used as an indica- tor to cover countries like Poland and Hungary, which relied on central planning rather than overt trade policies (for example, tariffs) to main- tain a closed economy. Export controls are symmetrical with import controls in their effects on closing an economy, as A. P. Lerner first es- 26 Brookings Papers on Economic Activity, 1:1995 Table 5. Post-communist Countries with Year of Opening Yeas of Countuy opening Hungary 1990 Poland 1990 Bulgaria 1991 Czech Republic 1991 Slovak Republic 1991 Slovenia 1991 Albania 1992 Estonia 1992 Romania 1992 Croatia 1993 Latvia 1993 Lithuania 1993 Belarus 1994 Kyrgyzstan 1994 FYR Macedonia 1994 Moldova 1994 Armenia closed Azerbaijan closed Georgia closed Kazakhstan closed Russia closed Tajikistan closed Turkmenistan closed Ukraine closed Uzbekistan closed Yugoslavia closed Source: European Bank for Reconstruction and Development (1994). tablished.45 The sub-Saharan African countries relied extensively on ex- port monopolies on foodstuffs, in part to maintain low domestic prices of food for urban residents. Returning to tables 1-5, we show that only a few developing countries have been continuously open from the start of the postwar era, or from the start of their independence: Barbados, Cyprus, Hong King, Malay- sia, Mauritius, Singapore, Thailand, and the Yemen Arab Republic. Many others embarked on a path of inward-oriented growth in the 1950s or 1960s that was subsequently reversed in the 1970s or later. (Bolivia, Ecuador, and Jamaica closed quite late in the period: 1978, 1983, and 1973, respectively.) Some of the first closed economies to open trade were three East Asian countries: Taiwan (1963), South Korea (1968), 45. See Lerner (1936). (- rq m o o o= 6 6 o= rq m - o = rn ) CD CD CD CD CD " rn rn ? o o cn rq - " \C rn ON t- 00 CD ') r n "t o b tr tr q C- - ,C "t : o o CD :t - rn C' " r4 C) t oNo L~ N0'o1-00u000r1 N'~-r 1 cnOO. N*r < n1.0.o ou<tN1- to4 A > rq C, tf .\CN I ?t C ,\ ON \C CD .n O:t C) 00 .00 rq rS CN,< CN, It CD ON C n ', t C ) 00 rq \C It C l ) eclm C', ) Csl X r5 o o o C o C u u u uu uu u C ._~~~2 N~~~~~~~~~ L~~~~~~~~~ ~~ OOni *ON0~~~~0N0 x - dcx d3__4< , rU r *<rnOn .N< O N .0<~r *' QQoo;== i-niUU UUU ~~~~~~ ~ ~ ~ ~ ~ ~ ~ ~~~2 Cq no n ( OO N C) o It 1 C 0r no 0 t N -m X N ~~~~~~- .r0 .o oo4F-rn -- .rrioC C ~~~~~ 0C,C = ~ C ~ 0 C4 DC toL (N4 *0 0 oo (N ( O *N (O C) C0 C) o X .t~~~~~~~~~~~~~~~~~. CDc )C C~ oCv4 I g OD 30 o o o o o o o e' 'r O N 03 03 E OO~~~~~~~4O t-0 o v >. >> 03 0 0 t 4)> > _ 00 * - E 00 E E o 00 Z -= 000 ? =? ? a> E E 0 0Z- U ? *L==CC0 4) CZ CZ t9 cL Q Q Q Q ~~> > ct E E ? > 3 .. > > > > X 3 31 32 Brookings Papers on Economic Activity, 1:1995 Table 7. Trade Policy Indicators for Asian Economies Black market Black market Quota Average Country premium (1970s)a premium (1980S)b coveragec tariffJU Hong Kong 0.00 0.00 0.00 0.00 Korea 0.09 0.09 0.10 0.14 Singapore 0.00 0.00 0.01 0.02 Taiwan 0.05 0.06 0.38 0.07 Indonesia 0.03 0.04 0.10 0.14 Malaysia 0.00 0.00 0.05 0.09 Thailand 0.00 0.00 0.06 0.29 All developing countries Median 0.17 0.18 0.14 0.19 Mean 0.44 0.70 0.26 0.22 Source: See appendix for complete definitions of variables and sources. a. The variable BMP70, measuring the black market exchange premium, averaged over the 1970s. b. The variable BMP80, measuring the black market exchange premium, averaged over the 1980s. c. The variable OWQI, indicating coverage of quotas on imports of intermediates and capital goods. d. The variable OWTI, indicating average tariffs on imports of intermediates and capital goods. and Indonesia (1970). It has become fashionable to argue that East Asian countries are not really open or market-oriented, and that, in fact, they systematically "got the prices wrong" to spur industrial growth.46 It is surely true that Korea, Taiwan and Indonesia are not laissez faire, but they and their neighbors in Southeast Asia, Thailand and Malaysia, have been more open to trade than other developing countries, based on ob- jective indicators of trade policy, shown in table 7. All of the East Asian economies have low or zero BMPs; all but Thailand have low tariff rates; and all but Taiwan have low NTB coverage. Moreover, the Thai tariffs and the Taiwanese NTBs are moderate, not extreme. In a later paper we intend to specify a detailed model of the timing of liberalization during the postwar period. Here we simply test a few of the simplest propositions that arise from political economy considerations: that timing should be related to the relative endowments of labor and land, the size of the economy, the per capita income, and perhaps the previous political history (for example, number of years since indepen- dence). As described above, we would expect the transition to openness to be faster in land-scarce and labor-abundant economies, since it is plausible that governments will tend to be more responsive to the inter- ests of labor over landowners. We would also expect the transition to 46. See Wade (1990) with regard to Hong Kong, South Korea, Singapore, and Taiwan, and Amsden (1989) with regard to South Korea. Jeffrey D. Sachs and Andrew Warner 35 tial episode of temporary liberalization. In almost all cases, these are countries that had a tradition of open trade, which was resurrected im- mediately following the Second World War. Table 9 lists these countries and the dates of the temporary liberalization as well as the average growth rates, both during the open period and the subsequent closed pe- riod. We find that in twelve of the fifteen cases, average growth in the open period exceeded that in the subsequent closed period. This is im- portant because it suggests that the eventual decision to close the econ- omy was generally not caused by slow growth during the open period, but rather by political and ideological shifts within each country. This is corroborated by economic histories of these countries, which rarely give slow growth as a reason for the policy switch. In the notes to the table, we also report high average growth rates in two later temporary liberalizations, in Sri Lanka and Venezuela. Overall, we find little direct evidence that slow growth played an important role in ending these epi- sodes of liberalization. The Impact of Postwar Global Integration on Economic Performance, 1970-89 In this section we show that during the period 1970-89 open econo- mies outperformed closed economies on three main dimensions of eco- nomic performance: economic growth, avoidance of extreme macroec- onomic crises, and structural change. In the process we demonstrate the close relationship between economic integration and economic conver- gence, that is, poor countries tend to grow faster than richer countries, as long as the poor and rich countries are linked together by international trade. Poor, closed economies have often performed significantly less well than the richer countries. For the purposes of this section, we define a country as open if it satis- fies the five policy criteria for the duration of the 1970s and 1980s. Coun- tries that were closed during part of this period but subsequently liberal- ized are treated as closed economies. In the following section, we pick up the trail of those economies by examining the effects of relatively late trade liberalization on economic performance. Openness and Growth During 1970-89, we find a strong association between openness and growth, both within the group of developing and the group of developed 36 Brookings Papers on Economic Activity, 1:1995 Table 10. Developing Country Growth and Openness, 197089a Always Not always Growth rate' open open Average growth >3.0 1 1 4 Average growth <3.0 4 70 Source: See appendix. a. In a test of independence the chi square is 41 (significance level = 0.000). b. The growth variable is G7089, the real annual per capita growth in GDP over 1970-89, described in the appendix. countries. Within the group of developing countries, the open econo- mies grew at 4.49 percent per year, and the closed economies grew at 0.69 percent per year. Within the group of developed economies, the open economies grew at 2.29 percent per year, and the closed economies grew at 0.74 percent per year.48 We may also classify the data in a different way to focus on growth within the open and closed groups. Within the closed group, average growth is about the same for the poorer developing countries (0.69 per- cent) as the richer developed countries (0.74 percent). However, within the group of open economies, the developing countries grew faster (4.49 percent) than the developed countries (2.29 percent). This suggests that within the group of open economies, both developing and developed, we should tend to observe economic convergence. Another way to look at this is provided by table 10 which shows the frequency of growth rates above and below 3 percent per year for closed and open developing countries. Eleven of the fifteen open economies grew at more than 3 per- cent per year, while only four of seventy-four closed economies achieved such growth. For this cross-tabulation, x2 = 41 (p < 0.000), so that we may reject the null hypothesis of no difference in growth rates between the closed and open economies. Figure 2 shows the average annual growth rates for a group of forty always-closed developing economies and a group of eight always-open developing economies during the period 1965-90. As can be seen, the 48. Four developed economies were closed for part of the period: Israel, New Zealand, Trinidad and Tobago, and Venezuela. While three of these countries are not nor- mally classified as developed, on a purchasing power parity basis, their per capita GDP exceeded $5,000 in 1970, and thus they qualify as developed by the standard used in this paper. Jeffrey D. Sachs and Andrew Warner 37 Figure 2. Average Growth of Eight Always Open and Forty Always Closed Economies, 1966-90 Growth ratea 0.15 0.14 0.13 - 0.12 - 0.11 _ 0.10 _ 0.09 /Open 0.08 - 0.07 - 0.06 - 0.05 - 0.04 - 0.03 - loe 0.02 -----Closed- 0.01 -%00 0.00 % -0.01 -0.02' -0.03- -0.04- -0.05 I l l I 1965 1970 1975 1980 1985 Source: Authors' calculations using version 5.6 of the data in Summers and Heston (1991). a. Figure shows three-year moving averages. always-open economies outperformed the always-closed economies in every year. The open economies were clearly more susceptible to the external shocks of the first half of the 1970s (the breakdown of Bretton Woods, worldwide inflation, and the OPEC oil price increases), but then bounced back. Note that the average per capita growth rates of the open economies at the end of the 1980s was about the same as during the sec- ond half of the 1960s, around 5 or 6 percent per year. The closed econo- mies, by contrast, evidence a long-term slowdown in growth (2 or 3 per- cent per year in the late 1960s, around 0 percent per year in the late 1980s). The data suggest that the absence of overall convergence in the world economy during the past few decades might well result from the closed trading regimes of most of the poorer countries. We now investigate the issue of convergence in greater detail. Our starting point is figure 3. We graph on the x-axis the 1970 level of per capita GDP of our sample of countries and on the y-axis, the growth of per capita GDP for 1970-89.49 49. The exact definitions of the data and the selection of countries are described in the appendix. 40 Brookings Papers on Economic Activity, 1:1995 ogy to borrow from abroad, the poorest countries are unable to bridge the gap in technology and knowledge.53 Barro and Sala-i-Martin have introduced the related notion of condi- tional convergence, in which countries differ in their long-run per capita income levels, with each country tending to grow more rapidly the greater is the gap between its initial per capita income level and its own long-run per capita income level.54 Formally, country i is assumed to have the long-run per capita income level yi*, and initial per capita in- come level yi. The rate of growth, ji, is assumed to be an increasing func- tion of the gap between yi* and yi: (2) Ys = - (Yi*-Yi) A positive value of B is said to signify conditional convergence. In turn, yi* is proxied by certain "structural" variables, Zjfl such as the initial level'of human capital, according to an equation y,* = EY]i Z,i Barro and others then estimate a regression equation of the form: (3) yi = B (EY1 zj> - Y1) They tend to find a negative and significant coefficient for initial income, yi, and significant coefficients on several of the structural variables, Zi.55 Like Baumol, Barro concludes that "a poor country tends to grow faster than a rich country, but only for a given quantity of human capital; that is, only if the poor country's human capital exceeds the amount that typically accompanies the low level of per capita income. "156 More re- cently, Robert Barro, Gregory Mankiw, and Xavier Sala-i-Martin state that the "substantially different steady states . . . can reflect the effects of disparities in preferences and government policies on the saving rate, fertility, and the available production technology."57 In summary, there have been three dominant explanations offered in 53. Baumol, Nelson, and Wolff (1994, p. 65). 54. See Barro (1991) and Barro and Sala-i-Martin (1991, 1992a). 55. The finding of conditional convergence is now fairly well established in the empiri- cal literature. A number of studies have found this result using post-World War II data with different conditioning variables. The list includes Barro (1991), De Long and Sum- mers (1991), King and Levine (1993), Levine and Renelt (1992), Mankiw, Romer, and Weil (1992), and Sachs and Warner (1995). 56. Barro (1991, p. 409). 57. Barro, Mankiw, and Sala-i-Martin (1995, p. 103). Jeffrey D. Sachs and Andrew Warner 41 the literature for the absence of convergence shown in figure 3. The first holds that productive technology is intrinsically kind to the technologi- cal leader: the rich tend to grow richer as a result of increasing returns to scale in one form or another.58 The second holds that convergence is a fact of life, but only among countries with a sound human capital base for using modern technology. The third holds that currently poor coun- tries have a low long-term potential income level (yi*), though countries do tend to grow faster the greater is the gap between their current in- come and their own long-run potential. The first two interpretations, and possibly the third, would be pro- foundly pessimistic for the poorer countries today, since they suggest that the poorer countries will be unable to close the gap with the richer countries. The conditional convergence hypothesis is ambiguous on this fundamental point. If the low long-term potential income of the poor countries that it posits is due to preferences and initial skill levels, then it too is profoundly pessimistic. In this case the hypothesis is akin to Baumol's convergence club. On the other hand, if the low long-term po- tential income is due to bad policies, then convergence could still be achieved by policy changes. We suggest that the most parsimonious reading of the evidence is that convergence can be achieved by all countries, even those with low ini- tial levels of skills, as long as they are open and integrated in the world economy. In this interpretation, the convergence club is the club of economies linked together by international trade: thus the OECD, the European Community, the late-nineteenth-century economies, the U.S. states, and the Japanese prefectures all tend to show convergence. In terms of the conditional convergence hypothesis, we argue that the ap- parent differences in long-term income levels are not differences due to fundamental tastes and technologies, but rather to policies regarding economic integration. The role of policy choices in convergence is dramatically evident in figures 4 and 5, where we divide the sample in figure 3 into groups of open and closed economies. Figure 4 shows that the open countries dis- play a strong tendency toward economic convergence, and that the countries with initially low per capita income levels grow more rapidly 58. Increasing returns to scale is shorthand for a wide variety of technological possibil- ities, such as learning by doing, spillovers in knowledge accumulation, agglomeration economies among suppliers of specialized inputs to production, etc. 42 Brookings Papers on Economic Activity, 1:1995 Figure 4. Growth and Initial Income, Open Economies, 1970-89 Annual growth per capita (percent) 8 * Korea,R. of * Taiwan 6 Singapore* * Hong Kong Yemen . Indonesia a * Mauritius 4 - Thailand * * Malaysia * Cyprus * Portugal * J~a4n Ireland Norwai. Finland e Canada 2 } * Jordan (*a eBarbados t *Austri * Luxembourg 2 - Gec *1 Belgium United States.* Denm?ark Ge Netherlands * Wzerland . 0 -2 A I I -2 0 2 4 6 8 10 12 14 1970 GDP per capita (thousands of 1985 dollars) Source: Authors' calculations using version 5.5 of the data in Summers and Heston (1991). than the richer countries.59 The closed economies in figure 5 do not dis- play any tendency toward convergence. In fact, they are clearly the source of the failure of convergence noted in figure 3. Even more strik- ing, there is not a single country in our sample (which covers 111 coun- tries and approximately 98 percent of the non-communist world in 1970) which pursued open trade policies during the entire period 1970-89 and yet had per capita growth of less than 1.2 percent per year (Switzerland had the lowest growth, at 1.24 percent). And not a single open devel- oping country grew at less than 2 percent per year (Greece, at 2.38 per- cent, and Jordan, at 2.58 percent, are the lowest)! 59. The open economies also exhibit convergence in the sense of having a declining dispersion of GDP over time (sigma-convergence in Barro and Sala-i-Martin's terminol- ogy). For the open economies the standard deviation of the log of GDP was 0.83 in 1970 and 0.75 in 1989. Jeffrey D. Sachs and Andrew Warner 45 emphasized above, most developing countries started out as closed economies. And the few that had temporary episodes of liberalization had high growth during the open period. It is therefore hard to argue that slow growth caused the turn to closed policies. Rather, it seems that for reasons unrelated to growth performance, the developing world in 1970 was sorted into a large group of closed economies and a much smaller group of open economies. Twenty-five years later, sufficient time has passed for us to see the effects of this fundamental policy choice on growth. Our evidence so far suggests that being open to international trade has been sufficient to achieve growth in excess of 2 percent for devel- oping countries.6' What of necessity? Are there many countries that closed and yet achieved high economic growth? There are four devel- oping countries that are classified as closed during the period and yet had per capita growth of more than 3 percent per year during 1970-89: Botswana, China, Hungary, and Tunisia. Botswana failed to qualify on the basis of its black market premium for the 1970s, but did qualify for the 1980s. It passed all other criteria. Overall, therefore, the policies have been relatively open, especially in the 1980s. Moreover, since around 80 percent of Botswana's exports are diamonds, and a remarkably small proportion (less than 5 percent) of the labor force is in agriculture, Botswana avoided the anti-agricultural bi- ases that affected most of sub-Saharan Africa. It is relatively straightforward to account for Hungary and Tunisia. Their successful growth is more apparent than real. Both countries pur- sued statist development strategies that produced growth in the 1970s and financial crises in the 1980s and early 1990s. In both cases there was a serious downturn in growth at the end of the 1980s, as these financial crises hit the government. Considered over a slightly longer time period, these countries would not look like successes, and therefore would not be anomalous. In our view, China is the only puzzle, although it is essentially consis- tent with the importance of open trade. It is indeed true that China has violated most of the rules: high black market premiums on the yuan, ex- 61. Of course, our indicators of openness are associated with other market-based re- form policies, which makes it difficult to identify the precise contribution of trade as com- pared to other policies. A more precise statement is that open policies together with other correlated policies were sufficient for growth in excess of 2 percent during 1970-89. 46 Brookings Papers on Economic Activity, 1:1995 tensive reliance on trade quotas, and a socialist ownership structure. Nonetheless, the country has experienced a boom. We believe that China's success is strongly related to its particular economic structure at the onset of its market reforms at the end of the 1970s. In particular, China was a very poor economy in 1978, with three- quarters of the labor force in peasant farming. The essence of Deng Xiaoping's reforms at the end of the 1970s was to free the peasant econ- omy from state controls, even while maintaining the state's grip on the nonpeasant, state-owned sector (which covered just 18 percent of the labor force). With respect to international trade, the economy was es- sentially liberalized for nonstate firms, especially those operating in the Special Economic Zones (SEZ) in the coastal areas. Even though the currency remained inconvertible, and many state enterprises remained subject to rationing of imports, the nonstate enterprises (including joint ventures and foreign firms) were generally able to import their inputs nearly duty free, and to export processed goods to world markets. The result was a remarkable export boom, based heavily on labor-intensive operations. Shang-Jin Wei presents clear evidence that trade liberaliza- tion played an important role in China's growth.62 Thus, China's "two-track approach" (decontrol of the peasant sector and continued control of the state sector) was sufficient to unleash eco- nomic development and a labor-intensive export boom, even though it did not solve the many problems of poor performance in the state-owned sector. Some analysts have also argued that its boom is fragile and could still be stopped by the macroeconomic instability characteristic of many economies part way between planning and a market economy.63 We now turn briefly to a regression analysis, to confirm and deepen these basic findings. Various regression estimates are reported in table 11, showing the relationship between initial income in 1970 and subse- quent growth between 1970 and 1989. We see in regression 1 the absence 62. To quote his conclusions: "I have found some clear evidence that during 1980-90 more exports are positively associated with higher growth rates across Chinese cities. In the late 1980s, the contribution to growth comes mainly from foreign investment. Further- more, the contribution of foreign investment comes in the form of technological or mana- gerial spillovers across firms as opposed to an infusion of new capital. Finally, the superb growth rates of coastal areas relative to the national average can be entirely explained by their effective use of exports and foreign investment." (Wei, 1995, p. 74.) Also, see Lardy (1994) for further discussion of China's recent experience with international trade. 63. See Sachs and Woo (1994) for further details of the two-track approach in China and the current macroeconomic problems. Jeffrey D. Sachs and Andrew Warner 47 of convergence for the entire sample of countries (the coefficient on ini- tial income is positive rather than negative, and is statistically insignifi- cant). In regression 2 we see the strong evidence of convergence within the set of open countries. The coefficient on initial income suggests that each percentage point rise in per capita income in 1970 reduces subse- quent annual growth by 0.014 percentage points. Each doubling of 1970 income reduces annual growth by 0.95 percentage points (= ln(2) x 1.368). In regression 3, we confirm the absence of convergence among the nonqualifying countries. We have also found that this result on the importance of openness for growth is robust to the presence of several other possible explanatory variables. In regressions 4 and 5 we illustrate this with Barro's growth regression, since it is particularly well known.M4 In regression 4, we rep- licate the Barro regression on cross-country growth for our sample and time period. We see it performs as expected, showing conditional con- vergence (a negative, significant coefficient on initial income), positive (although not significant) effects of educational attainment, positive ef- fects of the investment-to-GDP ratio, and negative effects of measures of political instability. Regression 5 estimates the same equation but includes a dummy vari- able for openness (OPEN = 0 for a closed economy, 1 for an open econ- omy). When we add OPEN, we find that the open economies grow, on average, by 2.45 percentage points more than the closed economies, with a highly statistically significant effect. Comparing the rest of the re- gressors with the estimates of regression 4, the effect of investment de- clines and the initial education levels are even less significant. This is consistent with our view that the growth rate over this period was deter- mined less by initial human capital levels than by policy choices. Our finding that openness plays an important role in a Barro-style cross- country growth equation is consistent with much recent research, in- cluding work by Surgit Bhalla, J. Bradford De Long and Lawrence Sum- mers, David Dollar, and Ross Levine and David Renelt.65 Indeed, some 64. See Barro (1991). 65. See Bhalla (1994), De Long and Summers (1991), Dollar (1992), and Levine and Renelt (1992). More specifically, De Long and Summers use several measures of outward orientation and price distortions. Levine and Renelt use the black market premium, the number of revolutions and coups, a socialist dummy, a civil liberties index, and measures of openness based on Leamer (1988). These studies examine the marginal contribution of the variables on the right side of a regression equation; none use these variables to sort countries into groups and examine the groups separately. 50 Brookings Papers on Economic Activity, 1:1995 of our data, such as the tariff and black market premium data, have been used in previous studies, including those by Jong-Wha Lee and Levine and Renelt.66 Our treatment of these data differs from the earlier studies in two important ways: first, by our construction of a single indicator measure of openness (built on several underlying variables); and sec- ond, by our examination of growth performance within the subset of open economies, as well as between closed and open economies. To our knowledge, no earlier studies have pointed out that corivergence applies to the worldwide subset of open economies.67 In regression 6, we add a dummy variable, POL, to account for ex- treme political conditions detrimental to long-term investment. The variable POL takes a value of 1 when any of the following conditions applies: -A socialist economic structure, according to the list of countries compiled by Kornai.68 -Extreme domestic unrest caused by revolutions, coups, chronic civil unrest, or a prolonged war with a foreign country that is fought on domestic territory. -Extreme deprivation of civil and political rights, according to the Freedom House index reported by McMillan, Rausser, and Johnson.69 We see that the POL variable is statistically significant at the 10 per- cent level (t = 1.986), suggesting that property rights, freedom, and safety from violence are additional determinants of growth.70 This find- ing is in accord with other recent studies, including work by A. S. Ales- ina and others, Barro, Bhalla, and Jakob Svensson.71 In other regres- sions, not reported here, we have experimented with the three individual items in the POL index, and have found that each one plays a role in the growth process. 66. See Lee (1993) and Levine and Renelt (1992). 67. Note also that we find no evidence for significant interactions between the open- ness variable and the other regressors that would diminish the explanatory power of openness. 68. Kornai (1992). 69. McMillan, Rausser, and Johnson (1994). 70. Note, however, that the set of countries with POL = 1 is a subset of the set of closed economies. Therefore use of the POL variable as an additional criterion to classify countries would give the same set of countries as using the OPEN variable alone. The same is true for the set of countries that had annual inflation rates above 100 percent for any year between 1970 and 1989. 71. See Alesina and others (1992), Barro (1991), Bhalla (1994), and Svensson (1994). Jeffrey D. Sachs and Andrew Warner 51 Table 12. Effects of Growth on Human and Physical Capital Accumulationa Independent Dependent variable variable INV7089b DS YRc DPYRd Constant -0.144 0.118 0.151 (-2.232) (3.144) (4.566) OPENe 0.054 -0.004 -0.003 (3.328) (-0.502) (-0.360) LGDP70f 0.047 -0.010 -0.016 (6.412) (-2.137) (-4.283) Summarv statistics R~2 0.519 0.069 0.250 Mean dependent variable 0.182 0.049 0.025 SER 0.063 0.033 0.027 Sample size 113 97 90 Source: Authors' regressions based on data described in the appendix. a. The numbers in parentheses are t statistics. b. IN7089 is the ratio of public and private investment spending to GDP, averaged over the period 1970-89. c. DSYR is average accumulation of secondary schooling over the period 1970-85. Specifically, DSYR [ln(SYR85) - ln(SYR70)/15], where SYRxx is person years of secondary schooling divided by the total population over age fifteen. d. DPYR is accumulation of primary schooling, calculated in the same manner as DSYR. e. OPEN is a dummy variable set equal to one for open economies. f. LGDP70 is the natural log of GDP per capita in 1970. We have shown above that the labor-to-land ratio has been a determi- nant of the timing of liberalization among developing countries. In re- gression 7 we include this ratio as a possible independent determinant of growth, to check whether openness is acting simply as a proxy for rela- tive factor endowments. The variable is insignificant, while the open- ness variable maintains its magnitude and statistical significance. We have found strong evidence that protectionist trade policies re- duce overall growth when controlling for the other variables. Since poor trade policies might also affect the rates of investment relative to GDP and the rates of human capital accumulation, we would expect poor poli- cies to have indirect adverse growth effects as a result of slower accumu- lation of capital, both physical and human. In regressions 8-10 in table 12, we therefore check whether open and closed economies differed sys- tematically in the rates of capital accumulation, once we control for ini- tial income. In regression 8 we find that the open economies had signifi- cantly higher investment-to-GDP ratios, and that OPEN raised the investment ratio by an average of 5.4 percentage points.72 Interestingly, 72. Levine and Renelt (1992), using trade shares as a measure of openness, also find that investment shares are higher in more open economies. This is one of the few findings that they classify as robust, using extreme bounds analysis. 52 Brookings Papers on Economic Activity, 1:1995 there is also some evidence that richer countries have higher investment rates than poorer countries. In regressions 9 and 10, we ask whether the increase in educational attainment between 1970 and 1985 was different for the two subsets of countries. We find no evidence that the closed economies had less im- provement in the coverage of primary and secondary education than did the open economies. It is clear, though, that the more developed econo- mies had less improvement in educational coverage than did the poorer countries (as evidenced by the significant negative sign on initial income in both regressions). Based on the regression analysis, we may make four conclusions: -There is strong evidence of unconditional convergence for open countries, and no evidence of unconditional convergence for closed countries .7 -Closed countries systematically grow more slowly than do open countries, showing that "good" policies matter. -The role of trade policy continues after controlling for other growth factors, as in a standard Barro cross-country growth equation. -Poor trade policies seem to affect growth directly, controlling for other factors, and to affect the rate of accumulation of physical capital. Trade Policy and Changes in the Export Structure One of the original arguments for SLI was the promotion of manufac- turing exports. Raul Prebisch and other economists worried that raw materials exporters that maintained free trade would be unable to indus- trialize, and would therefore be vulnerable to long-term adverse move- ments in the terms of trade between primary and manufactured goods. Import substitution, it was argued, would give time for domestic indus- try to develop and to improve productivity, perhaps sufficient to gener- ate manufactured exports in the distant future. Paul Krugman gave an influential exposition of this infant-industry argument in a formal model of increasing-returns-to-scale production resulting from learning by doing.74 73. An alternative, and formally equivalent, way to state our conclusion is that conver- gence is conditional on policies, not on structural variables (for example, initial income or level of education). We therefore argue against the notion of a low-income "development trap," since open trade policies (and correlated market policies) are available to even the poorest countries. 74. Krugman (1987). Jeffrey D. Sachs and Andrew Warner 55 closed economies display almost no change at all in export structure dur- ing the nearly-twenty-year interval examined in the regressions, since the estimate of ,3 is always insignificant. In regression 4 we show that these conclusions also hold when the regression is estimated with the developed countries added to the sample. Trade Policy and Macroeconomic Crises Jeffrey Sachs argued in 1985 that the outward orientation of the East Asian economies had saved them from the developing country debt cri- sis that ravaged Latin America.75 Now, ten years later, it is possible to reassess his hypothesis with a greater time span and a larger number of country observations. Is there evidence that openness to trade helped to avoid macroeconomic crises in the 1980s? To address this issue we clas- sify countries according to their trade orientation in the 1970s and then examine whether the countries that were open in the 1970s were less likely to experience a severe macroeconomic crisis in the 1980s and 1990s. For these purposes, we define a severe macroeconomic crisis by any one of the following three occurrences: -A rescheduling of foreign debt in the Paris Club (official creditors) or the London Club (commercial bank creditors). -Arrears on external payments (including debt servicing), as re- ported by the IMF. -An inflation rate in excess of 100 percent per year. We expect that closed economies will be more likely than open econ- omies to fall into one or more of these crises, for several related reasons. First, and most important, closed economies often borrowed heavily from foreign sources in order to overcome economic stagnation caused by the deeper problem of poor economic policies.76 The reliance on debt was a temporary expedient, and resulted in a debt crisis when creditors withdrew support from further lending. Second, closed economies ori- ented investment toward nontraded goods, and thus lacked the foreign exchange earnings to service the debts. Third, closed economies tended to have a higher level of state involvement in the economy, including the ownership of state enterprises. Loss-making state enterprises added sig- nificantly to the overall fiscal burden of many governments in the 1980s, contributing to the onset of high inflation and foreign debt crises. 75. Sachs (1985). 76. See Sachs (1994). 56 Brookings Papers on Economic Activity, 1:1995 Table 14. Developing Country Openness and Macroeconomic Crisisa Macroeconomic No macroeconomic Openness crisis in 1980s crisis in 1980s Open in 1970s 1 16 Not open in 1970s (i) 59 14 Source: See appendix. a. In a test of independence the chi square is 34.8 (significance level <0.000). There are seventeen developing countries that had an open trade pol- icy in the 1970s. Of these, only Jordan succumbed to a macroeconomic crisis after opening: debt reschedulings in 1987 and 1992, and external payments arrears in 1993. The first episode of macroeconomic difficul- ties followed a sharp cutback in foreign aid from the oil-rich states of the region as a result of the collapse of world oil prices in 1986. Following the 1990 Gulf War, Jordan experienced a more serious macroeconomic shock, which cost it heavily in remittance and export earnings. There were seventy-three closed developing countries in the 1970s. Of these, as many as fifty-nine experienced a severe macroeconomic cri- sis: forty-nine had a debt crisis; fifty had external payments arrears; and nineteen had inflation above 100 percent (most manifested more than one of these crises). Table 14 summarizes the relative frequencies of openness and macroeconomic crisis. A x2 test on the null hypothesis of independence between trade policy in the 1970s and macroeconomic cri- sis in the 1980s is rejected at the 0.000 level. Rather than focus on the large majority of countries that succumbed to crisis, it is easier to assess the fourteen that did not: Bangladesh, Bots- wana, Burundi, China, Colombia, Hungary, India, Iran, Nepal, Papua New Guinea, Rwanda, Sri Lanka, Tunisia, and Zimbabwe. Of these, Botswana had opened its economy by 1979; Colombia maintained very cautious and moderate policies both in trade and in finance; both Hun- gary and India, in fact, flirted with a debt crisis which was narrowly averted; China began the 1980s with very little debt because it had bor- rowed little during the Cultural Revolution of 1966-76; Bangladesh, Bu- rundi, Nepal, and Rwanda are among the world's poorest countries and have little, if any, access to loans on commercial terms, which has prob- ably saved them from generating a debt crisis. Moreover, Burundi and Rwanda have been subject to extreme internal unrest. Jeffrey D. Sachs and Andrew Warner 57 Recent Reforms and Economic Performance The previous section compared countries with long-standing policies of open trade during the nineteen-year period 1970-89, with countries that were closed during some or all of the period. In this section, we ex- amine the growth effects of trade liberalization in developing countries that have opened their economies since 1975. By our assessment, there are thirty-eight non-communist reformers that have opened their econo- mies since 1975 and sustained the opening until 1993. Venezuela at- tempted trade liberalization between 1989 and 1992, but the policy was then reversed. Another thirty-six countries in our sample did not even achieve a temporary liberalization during 1980-93. In addition to this group of countries, we also examine the recent growth performance of the twenty-five post-communist economies of eastern Europe and the former Soviet Union, to see how growth performance relates to trade reform and overall economic reform. Here, we stress again that trade reform is almost always accompanied by a much broader range of reforms, including macroeconomic stabili- zation, internal liberalization (for example, the end of price controls), legal reform, and often extensive privatization. This has been especially clear in the post-communist countries. In almost all cases, trade reform has been part of the overall institutional harmonization with the ad- vanced market economies. Our results cannot, therefore, distinguish between the effects of trade policy per se, and the effects of other parts of the policy package that accompany the trade measures. While we view the direct effects of trade liberalization (increased competition, specialization, and reduced rent-seeking) as important contributory fac- tors for growth, we put off attempting to tease out the specific contribu- tions to growth of the various parts of a standard reform program until a future study. We note also that the very-short-term growth consequences of a trade reform will depend importantly on the inherited structure of the econ- omy. In the post-communist countries of eastern Europe and the former Soviet Union, the long period of central planning left the economy with an enormously overgrown heavy industrial sector, evident, for exam- ple, in the high levels of coal and steel production relative to the rest of 60 Brookings Papers on Economic Activity, 1:1995 by an average of 1.09 percentage points per year, and in the more distant future [T + 3, N], by an average of 1.33 percentage points per year.78 The near-term gain is significant at (p = 0.10), while the long- term gain is statistically significant at (p = 0.05). The increase in average growth is larger when compared with the years immediately preceding the trade liberalization, since average growth rates are lower in those years than in the years [T - 10] through [T - 3] (by 0.88 percent per year on average).79 Trade Liberalization and Growth in Transition Economies The countries of central and eastern Europe have been undertaking market reforms, including trade liberalization, since 1990, while the countries of the former Soviet Union have been undertaking market re- forms since 1992. It is obviously extremely premature to draw strong conclusions regarding the effects of these reforms on the restoration of economic growth. Nonetheless, at least some evidence can be adduced from the five or more years of reform experienced by some parts of the region. 78. Other statistics for the regression are the number of observations, N = 548, ad- justed R2 = 0.149; and the F statistic for the overall regression, F = 3.45 (p < 0.001). The only country dummy that is statistically significant is Botswana, with a dummy variable of 8.14, t = 4.585. 79. It is worthwhile responding to two possible criticisms of these results. First, it could be objected that if growth outcomes were purely random, and countries reformed only when growth fell below a critical threshold, then although we would tend to observe higher growth after reform, it would be incorrect to attribute the higher growth to the re- form. However, we stress that we are comparing growth after the reforms with growth in the distant rather than the immediate past, and further, that our period for the distant past spans seven years. Second, it is possible that countries may have sorted themselves randomly as reform- ers and nonreformers. If some grew and others did not, and those that did not closed up again and thus were eliminated from our group of reformers, we would be left with a biased sample of reformers with high growth. But we have found few examples of countries that experienced slow growth after true reform. For example, economies that were temporarily open in the 1950s and 1960s and subsequently closed again, tended to have high growth rates during the liberal episode. We have also found that certain countries that are some- times cited as recent reformers, such as the Dominican Republic in the early 1980s and Nigeria between 1986 and 1992, actually did not reform sufficiently (by our criteria), while others that did reform temporarily, such as Venezuela, experienced rapid growth during the episode of liberalization. Hence we find few examples to suggest that sample selection bias is an important issue when examining the growth performance of recent reformers. Jeffrey D. Sachs and Andrew Warner 61 We are aided in this process by a recent review of the reform experi- ence conducted by the European Bank for Reconstruction and Develop- ment (EBRD).80 The EBRD ranked each country of the region according to the extent of trade liberalization, as well as several other key cate- gories of reform, such as privatization, regulatory reform, and fiscal re- form. Using these indicators, it is possible to group the countries by the intensity of reform as well as the timing of its onset, as we do in table 16. The intensity of reform is measured on a scale from 1 to 4, with a higher number greater intensity. On the basis of this categorization, we ask two questions: first, whether intensive reformers exhibit more or less decline in cumulative GDP between 1989 and 1994; and second, whether inten- sive or early reformers enjoy a faster turnaround in economic growth, and thereby achieve positive GDP growth by 1994. Table 16 shows that all of the strong trade reformers had achieved positive economic growth by 1994, while none of the other countries had done so. On average, the strong reformers also experienced a smaller cumulative loss of GDP between 1989 and 1994, though there is consid- erable variance in the data. We must stress, however, that since all the countries of central and eastern Europe and the Baltic states are classi- fied as strong trade reformers, while none of the states of the former So- viet Union is, we cannot distinguish adequately between the specific role of trade policy and the many other differences (geography, politics, resource endowments) between the two regions that might help to ex- plain the differences in growth performance. At the least we can high- light that the data are consistent with the notion that strong trade re- forms have produced a faster turnaround in growth and a smaller cumulative decline. More powerful tests of this hypothesis will have to wait until more time has elapsed. Conclusions The world economy at the end of the twentieth century looks much like the world economy at the end of the nineteenth century. A global capitalist system is taking shape, drawing almost all regions of the world into arrangements of open trade and harmonized economic institutions. As in the nineteenth century, this new round of globalization promises 80. European Bank for Reconstruction and Development (1994). 62 Brookings Papers on Economic Activity, 1:1995 Table 16. Growth Rates of the Transition Economies Percent Strength of Year of Cumulative trade trade growth Growth Countiy reform reform 1989-94 1994 Strong reforms Hungary 4 1990 - 17.94 2.00 Poland 4 1990 - 9.23 5.00 Bulgaria 4 1991 - 26.41 1.40 Czech Republic 4 1991 - 15.49 3.00 Slovak Republic 4 1991 - 19.53 5.00 Slovenia 4 1991 - 13.26 5.00 Albania 4 1992 - 22.89 7.00 Estonia 4 1992 - 29.15 5.00 Romania 4 1992 - 30.79 3.00 Croatia 4 1993 - 31.04 1.00 Latvia 4 1993 - 39.52 3.00 Lithuania 4 1993 - 55.44 2.00 Average - 25.89 3.53 Moderate reforms Kyrgyzstan 3 1994 -42.30 - 10.00 Russia 3 closed -47.29 - 15.00 Average -42.61 - 12.50 Weak reforms FYR Macedonia 2 1994 -51.30 -7.00 Moldova 2 1994 - 54.30 - 25.00 Armenia 2 closed - 61.60 0.00 Kazakhstan 2 closed - 51.01 - 25.00 Uzbekistan 2 closed - 11.75 - 3.00 Average -45.99 - 12.00 Weakest reforms Belarus 1 1994 - 35.93 - 22.00 Azerbaijan 1 closed - 54.32 - 22.00 Georgia 1 closed - 85.35 - 35.00 Tajikistan 1 closed - 70.37 - 25.00 Turkmenistan 1 closed - 38.29 - 20.00 Ukraine 1 closed - 51.36 - 23.00 Average - 55.94 - 24.50 Overall average - 38.63 - 7.58 Source: European Bank for Reconstruction and Development (1994, 1995) with national sources for Bulgaria for 1994. Jeffrey D. Sachs and Andrew Warner 65 Variables Used to Classify Countries as Open or Closed The following are the variables used to classify countries as open or not open for the period 1970-89, and their sources: BMP70 and BMP80 Black market exchange rate premium, averaged over the 1970s and 1980s respectively. Source: Cowitt (1986) with updates from World Bank data (supplied by Ross Levine). BMP Dummy variable equal to 1 if either BMP70 > 0.2, orBMP80 > 0.2. EXM Dummy variable equal to 1 if a country had a score of 4 on the export marketing index in the World Bank study, Adjustment in Africa (Husain and Faruqee, 1994, p. 238). A score of 4 means that the country had extreme distortions resulting from its export marketing board. These boards are state-run monopsonies that typically purchase agricultural products at prices much below world prices, and then resell them at world prices. The study covered African countries only. Soc Dummy variable equal to 1 if the country was classified as socialist in Kornai (1992, table 1.1). O WQI Variable indicating coverage of quotas on imports of intermediates and capital goods. It is the own- import weighted nontariff frequency on capital goods and intermediates. Includes licensing, pro- hibitions, and quotas. It is taken from Barro and Lee (1994) who, in turn, rely on UNCTAD data. The period covered is 1985-88. OWQID Dummy variable equal to 1 if OWQI > 0.4. OWTI Variable indicating average tariffs on imports of intermediates and capital goods. It is the own-im- port weighted average tariff rate on capital goods 66 Brookings Papers on Economic Activity, 1:1995 and intermediates. It is taken from Barro and Lee (1994) who, in turn, rely on UNCTAD data. Pe- riod covered is 1985-88. OPEN Dummy variable equal to 1 for open economies. It is equal to 0 if a country scored a 1 on either the BMP variable, the SOC variable, the EXM vari- able, or the OWQID variable. If a country had some missing values and was not otherwise ex- cluded, some effort was made to classify it as either a 0 or a 1, as is discussed below. The tariff variable is not used in forming the OPEN variable because it is redundant: all countries with OWTI > 0.4 are already classified as closed on other grounds. If there were insufficient data to make a judgment, the country was assigned a ''missing" value. The reasoning behind the adjustments we made to the OPEN variable is as follows. First, we know that we lack cross-country tariff and quota coverage data for the 1970s, so we were especially concerned to find countries that had restrictive trade practices in the 1970s, but had re- formed by the mid-1980s, and thus appeared open by the OWQI and OWTI variable. We found one case, Morocco, and changed the OPEN variable to 0. Other adjustments were made for different reasons. South Africa followed an inward-looking development strategy throughout the 1970s and 1980s, and this was reinforced from outside as the rest of the world gradually tightened trade sanctions on South Africa in 1985. (Source: Lachman and Bercuson, 1992.) Hence, South Africa is rated as a closed economy. Lesotho and Swaziland were members of the South- ern African Customs Union and thus were open in relation to the south- ern African region but closed in relation to the rest of the world. Since these economies are small relative to the South African market, we con- sider them inherently ambiguous cases and assign them "missing" val- ues (Lesotho is a high-growth economy and Swaziland is a low-growth economy). Botswana is also a member of the customs union but is rated closed, based on a high black market exchange rate premium. Haiti was rated as closed, based on the extensive evidence for restrictive trade practices in Lundahl (1992). Luxembourg is a member of the EEC and Jeffrey D. Sachs and Andrew Warner 67 is rated as open. New Zealand is rated as closed, based on evidence that quantitative trade restrictions covered more than 40 percent of imports in 1981 and 1983. (Source: Laird and Yeats, 1990, table 4.2.) Australia is rated as open, based on Caves and Krause (1984). After these adjustments, there remained nine small countries for which we had insufficient data to make an informed assessment on the OPEN variable: Cape Verde, Comoros, Liberia, Iceland, Fiji, Malta, Panama, Seychelles, and Suriname. Therefore, whenever the analysis in the paper requires the OPEN variable, the sample size is reduced to 111. The following variables are used to define the POL dummy variable. RIGHT Dummy variable equal to 1 if a country scored a 6 or above (higher means more repressive) on either the political rights index or the civil liberties index in MacMillan, Rausser, and Johnson (1994, ta- ble 1, pp. 8-10). EDU Dummy variable equal to 1 in cases of extremely disruptive unrest. Intended to capture disruptive internal or external wars, coups, and revolutions. Constructed by the authors using several indica- tors. First, countries were assigned a I if they scored 0.6 or higher on the REVC70 or REVC80 index in Barro and Lee (1994). In addition, the fol- lowing countries were assigned a 1 because of conflicts: Angola (sixteen-year civil war); Bu- rundi (Hutu rebellion 1973-74, resulting in an esti- mated 160,000 deaths); Chad (repeated battles with Moslem rebels in the north); El Salvador (twelve-year civil war); Ethiopia (extended war with Somalia over control of the Ogaden region of Ethiopia); Guatemala (repeated conflicts between the military rulers and the guerrilla army of the poor); Iran and Iraq (war in the early 1980s); Is- rael; Mozambique (protracted civil war in the 1980s); Nicaragua (civil war); Panama (U.S. inva- sion in the late 1980s); Somalia (see Ethiopia); Sri Lanka (repeated violence with the Tamil and Sin- 70 Brookings Papers on Economic Activity, 1:1995 TL The dependent variable is a dummy taking the value 1.0 if the country had opened by 1970 and stayed open. Table 11 GDP70 Real GDP per capita in 1970 (1985 international prices) from Summers and Heston version 5.5. GDP89 Real GDP per capita in 1989 (1985 international prices) from Summers and Heston version 5.5. G7089 Real per capita growth rate of GDP per year: G7089 = [ln(GDP89) - ln(GDP70)]/19. Note that this variable is calculated differently for a few countries, as listed at the beginning of this ap- pendix. SEC70 Secondary school enrollment rate. Source: Barro and Lee (1994). PRI70 Primary school enrollment rate. Source: Barro and Lee (1994). GVXDXE Ratio of real government "consumption" spend- ing net of spending on the military and education, to real GDP. Source: Barro and Lee (1994) who, in turn, used Summers and Heston version 5.5. REVCOUP Number of revolutions and coups per year, aver- aged over the period 1970-85. Source Barro and Lee, 1994. ASSASSP Number of assassinations per million population per year, 1970-85. Source: Barro and Lee (1994). PPI70DEV The deviation of the log of the price level of in- vestment (PPP investment divided by exchange rate relative to the United States) from the cross- country sample mean in 1970. Source: Authors' calculation based on the PISH5 price data in Barro and Lee (1994). Jeffrey D. Sachs and Andrew Warner 71 INV7089 Ratio of real gross domestic investment (public and private) to real GDP, averaged over the pe- riod 1970-89. Source: Barro and Lee (1994) who, in turn, used Summers and Heston version 5.5. Table 12 DS YR Average accumulation of secondary schooling over the period 1970-85. Specifically, DSYR = [ln(S YR85) - ln(S YR70)]115, where S YRxx is per- son years of secondary schooling divided by the total population over age fifteen. DP YR Accumulation of primary schooling, calculated in the same manner as DS YR. Table 13 X71 Primary export intensity in 1971. Ratio of primary exports to total exports in 1971, with both numer- ator and denominator expressed in nominal dol- lars. Primary exports are defined as agriculture, minerals, fuels, and metals. These correspond to SITC (revision 1) categories 0, 1, 2, 3, 4, and 68. Source: For all countries except Taiwan and South Africa, World Bank, World Tables, 1994. Data for Taiwan were obtained from the Taiwan Statistical Data Book (Republic of China, 1993). Data for South Africa include exports of raw dia- monds and gold and were obtained from South Af- rica's Bulletin of Statistics, December 1972 and June 1992. Data for Singapore were estimated as 0.01, based on GDP and labor force data indi- cating that Singapore produces no mining, no pri- mary energy, and only a very small amount of agriculture, forestry, and fishing products. The data for Bangladesh are for 1972 rather than 1971. The data for Cameroon were set to 1.0; they ex- ceeded 1.0 using the published data. 72 Brookings Papers on Economic Activity, 1:1995 X89 Primary export intensity in 1989, calculated in the same manner as X71. Table 14 The growth data for the recent reformers are real per capita growth from the World Tables of the World Bank. We did not use the data pro- vided by Summers and Heston because we needed recent growth data. Background on Country Classifications Algeria Never open. The black market premium averaged 350 percent during 1985-90. Some trade liberal- ization began in the second stage of its reform pro- gram in 1991, but implementation was interrupted by political turmoil. Source: World Bank, Trends in Developing Economies, 1994 (TIDE), p. 6. Angola Never open. A protracted civil war has plagued the country since independence. Source: TIDE, p. 12. Argentina Open since 1991. The average nominal tariff level for manufacturing was 141 percent in 1958 (Little, Scitovsky, and Scott, 1970, p. 163). The liberal- ization in 1976-80 did not sufficiently reduce ef- fective rates of protection (estimated at 88 percent for all of manufacturing in 1980, from Cavallo and Cottani, 1991, table 3.19). The dating of the liber- alization in 1991 is based on TIDE, p. 17. Australia Open since 1964. Australian tariffs were high by OECD standards, but the mean tariff did not ex- ceed 40 percent. Source: Unpublished data from the World Bank. The date of liberalization is based on the evidence of the gradual relaxation of quantitative restrictions and import licensing in IMF, Annual Report on Exchange Restrictions, various issues. Jeffrey D. Sachs and Andrew Warner 75 Chad Never open. Chad received a score of 4 on its ex- port marketing board and thus is considered closed before 1990. In 1990 this rating was re- duced to a 3, so that Chad would potentially be open from this date. However, because there is considerable discussion of the lack of progress on trade reforms in TIDE, p. 93, we do not classify Chad as a reformer. Chile Open since 1976. Chile is classified as closed in the 1950s, based on the accounts of import prohi- bitions, licensing, and multiple exchange rates in various issues of the IMF's Annual Report on Ex- change Restrictions covering the years 1950-61. For the 1960s, Chile is not rated as open because the mean black market premium was 54 percent. The 1976 dating for the liberalization is based on Dornbusch and Edwards in Bosworth, Dorn- busch, and Laban (1994, pp. 84-85); as well as Pa- pageorgiou, Michaely, and Choksi (1991, vol. 7, figure 2.3). China Never open. Rated as socialist in Kornai (1992). Trade policies have been progressively liberal- ized since 1978, but the trading system was still rife with quantitative restrictions at least through 1994. (See text for further discussion.) Colombia Open since 1986. Colombia has had a complicated mixture of tariffs and quantitative restrictions since 1931. Its classification as closed is based on the fact that the index of trade liberalization in Garcia Garcia (1991) is fairly constant between 1950 and 1983, as well as evidence that average tariffs exceeded 40 percent in 1962, 1971, and 1973 (Diaz-Alejandro, 1976, p. 108). The liberalization episodes in 1954, 1966, and 1979 were too short to qualify as sustained liberalizations. The dating for the opening is based on evidence in Garay (1991) that average tariffs rates fell below 40 percent in 1986 and have stayed low up to the present. 76 Brookings Papers on Economic Activity, 1:1995 Congo Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). There was an attempt at liberalization in 1987, but it did not go far enough (TIDE, p. 117). The export mar- keting board was still rated as a 4 in 1990. Costa Rica Open 1952-61, closed 1962-85, open since 1986. In the 1950s Costa Rica had no exchange restric- tions on foreign payments and no import licensing (IMF, Annual Report on Exchange Restrictions, various issues). Imports could be obtained freely at an exchange rate that was 17 percent more de- preciated than the rate at which exports had to be surrendered to the central bank. In 1960 Costa Rica joined the Central American Common Mar- ket (CACM), so 1961 is chosen as the date of closure. The mean common external tariff in the CACM was 40 percent in 1966 (Carnoy, 1972, p. 14). Costa Rica had a mean black market premium in excess of 20 percent in the period 1960-64. The mean external tariff was 53 percent before 1986 (World Bank, 1992a, p. 86). The dat- ing for the reform in the 1980s is based on the de- cline in the black market premium to 1 percent (1985-89) and the 1986 tariff liberalization, which reduced the mean tariff to 26 percent (World Bank, 1992a, p. 86). Cote d'Ivoire Never open. Received a score of 4 on its export marketing board, and thus is considered closed before 1990 (Husain and Faruqee, 1994, p. 238). In 1990 this rating was reduced to a 3. However, there are still extensive nontariff barriers, sched- uled to be phased out by 1995. Source: TIDE, p. 125. Cyprus Open since independence. There are some infant industry tariffs but the mean has never exceeded 20 percent. Trade liberalization has been helped Jeffrey D. Sachs and Andrew Warner 77 by the European Community's Mediterranean policy. Cyprus can export most products at low tariffs to the EEC and reciprocates with low tar- iffs on EEC products. There are no quantitative restrictions. Source: Wilson (1992). Denmark Open since 1959. Although Denmark was not an original member of the EEC or the EFTA, its trade policy was harmonized with the rest of Eu- rope. Dating is based on timing of convertibility throughout Europe. Dominican Never open. The liberalization episode of Republic 1981-86 did not go far enough. Another liberaliza- tion started in 1991, but has not progressed sig- nificantly. Source: TIDE, p. 135. Ecuador Open 1950-82, closed 1983-90, open since 1991. The dating of the initial liberal phase is based on the IMF's Annual Report on Exchange Restric- tions from the early 1950s, which states that im- port licenses were, "in most cases," issued freely, and De Janvry, Sadoulet, and Fargeix (1991, p. 58), who report implied trade taxes for the period 1970-85. Extensive trade reform was started in 1990. By 1991 virtually all the nontariff restric- tions had been eliminated. The maximum tariff was 35 percent in 1990. Source: Economist Intelli- gence Unit, Country Report 3, 1991; and TIDE, p. 140. Egypt Never open. The state-led development planning and import-substituting industrialization policies were established in the mid- to late 1950s, under Nasser. We lack hard data to gauge whether Egypt was open in the years immediately after World War II. Egypt has certainly not been open since 1960, the start of the first five-year period for which we have black market premium data. Be- tween 1960 and 1980 the mean premium was 83 80 Brookings Papers on Economic Activity, 1:1995 port on Exchange Restrictions, various issues). In 1960, Guatemala joined the Central American Common Market (CACM), so 1961 is chosen as the date of closure. The mean common external tariff in the CACM was 40 percent in 1966 (Car- noy, 1972, p. 14). The mean external tariff was 53 percent 1966-86 (World Bank, 1992a, p. 86). The election of a civilian government in 1985 started a period of reform. The 1988 dating is based on TIDE, p. 196. Guinea Open since 1986. Not open before 1986 due to a rating of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). In 1990 the World Bank reports a black market premium of 8 percent and gives Guinea a rating of 1 (most liberal) on its export marketing system. The 1986 dating is based on TIDE, p. 200, which reports a compre- hensive dismantling of state-led development in- stitutions, including external trade protection. Guinea-Bissau Open since 1987. Not open before 1987 due to a rating of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). In 1990 the World Bank reports a black market premium of - 2 per- cent and gives Guinea a rating of 1 (most liberal) on its export marketing system. The 1987 dating is based on TIDE, p. 205. Guyana Open since 1988. A high mean black market pre- mium (298 percent) disqualifies Guyana between about 1975 and the late 1980s. Prior to 1988 there was an extensive list of import prohibitions and restrictions, which have since been greatly re- duced (World Bank, 1993b, p. 32). In 1991 Guy- ana adopted the Caribbean Community (CARI- COM) common external tariff schedule, with rates that average well below 40 percent. The 1988 dating is based on the assessment in TIDE, p. 210, that this was the decisive year of reform. Jeffrey D. Sachs and Andrew Warner 81 Haiti Never open. Extensive tariffs and quantitative re- strictions protected domestic manufacturing from 1949 through 1986 (Lundhal, 1992, p. 407). There are special export zones in Haiti where firms are allowed to import intermediate products, assem- ble them, and then export, but these represent a small fraction of the economy. Since 1986 liberal- ization has been extremely slow. In 1990 the black market premium was still 40 percent (Lundahl, 1992, p. 418). Honduras Open 1950-61, closed 1962-90, open since 1991. Honduras assumed the obligations of article VIII in 1950. In the 1950s and early 1960s there were no significant restrictions on payments or transfers abroad (IMF, Annual Report on Exchange Re- strictions, various issues). In 1960 Honduras joined the Central American Common Market (CACM), so 1961 is chosen as the date of closure. The mean commnon external tariff in the CACM was 40 percent in 1966 (Carnoy, 1972, p. 14). The mean external tariff was 53 percent 1966-86 (World Bank, 1992a, p. 86). An extensive trade re- form between 1990 and 1992 included the elimina- tion of import permits and administrative foreign exchange allocation. Import tariffs were reduced to a range of 5-20 percent (TIDE, p. 214). Hong Kong Always open. Hungary Open since 1990. Source: European Bank for Re- construction and Development (1994). India Open since 1994. Rated as closed before 1991 due to very high tariffs and elaborate quantitative re- strictions dating from the early 1950s (Bhagwati and Desai, 1970). The 1994 dating is based on the start of a trade liberalization program (TIDE, p. 223), and average tariff data in Krishna and Mi- tra (1994, p. 5). 82 Brookings Papers on Economic Activity, 1:1995 Indonesia Open since 1970. Indonesia had a dual exchange rate system that ended April 17, 1970 (Pitt, 1991). The important trade liberalization measures were introduced between May 1966 and July 1967. Im- port licensing was eliminated in October 1966 (Pitt, 1991, p. 181). The median tariff rate had fallen below 40 percent by 1970 (Pitt, 1991, table 5.10, p. 90, which relies on Rosendale, 1981, p. 276). Iran Never open. Iran maintained tight restrictions on imports through the annual publication of import lists by the Ministry of National Economy. The IMF's Annual Report on Exchange Restrictions for 1953 states that the ministry has prohibited a "large number of goods," although there is no quantification. The evidence of numerous admin- istrative controls on foreign trade provided by Amuzegar (1977) leads us to classify Iran as closed. Black market premium data disqualify Iran after 1975. Iraq Never open. Black market premium averaged 230 percent in the 1980s. Import quota system was in place in the early 1950s. All imports that com- peted with Iraqi new industries were on the pro- hibited list. Ireland Open since 1966. Dating is based on membership in the Anglo-Irish Free Trade Area from 1966, and data provided by 0 Grada and O'Rourke (1994, pp. 17, 29). Israel Open since 1985. Israel initially pursued an im- port-substituting industrialization policy in the 1950s. Since then, there has been gradual liberal- ization, so the question is when Israel first quali- fies as open. An import liberalization took place between 1962 and 1965, but Pack (1971, table 4.6) reports rates of nominal protection by sector that Jeffrey D. Sachs and Andrew Warner 85 been extensive trade liberalization, but the black market premium was higher than 20 percent in 1989 and 1990. Source: Husain and Faruqee (1994) and the Economist Intelligence Unit, vari- ous reports. Korea Open since 1968. The exchange rate was unified by the mid-1960s. The black market premium fell below 20 percent in the period 1965-69. A gradual reduction in import tariffs started in the mid- 1960s. Source: Nam (1989, pp. 165-66). By 1968, the average tariff was below 40 percent (Collins and Park, 1989, table 9.12). Lesotho Not rated. Ambiguous case due to membership in the Southern African Customs Union. (See dis- cussion in appendix above.) Liberia Not rated. Insufficient data on trade policy. Luxembourg Open since 1959, when convertibility was estab- lished. The average tariff in the Common Market was less than 40 percent in 1962. Source: Balassa (1965, table 1, p. 580). No major increase in pro- tection 1962-93. Madagascar Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). The ex- port marketing board was still rated as a 4 in 1990. There is no mention of significant reform since (TIDE, p. 294). Malawi Never open. Closed since early 1970s, based on a high black market premium as well as a rating of 4 on its export marketing board (Husain and Far- uqee, 1994, p. 238). Not rated as a recent re- former, based on discussion in TIDE, p. 299. Malaysia Open since independence (1963). The black mar- ket premium has never exceeded 2 percent. In 1965 the IMF states that "most imports are per- 86 Brookings Papers on Economic Activity, 1:1995 mitted freely under open general licenses" (IMF, Annual Report on Exchange Restrictions, 1965, p. 347). Malaysia qualifies on all of our trade indi- cators and there is no evidence of any major pol- icy changes in the 1970s (TIDE, p. 304). Mali Open since 1988. Pursued state-led development between independence in 1960, and 1988. State monopolization of exports (Husain and Faruqee, 1994, p. 238) and extensive import licensing (IMF, Annual Report on Exchange Restrictions, 1965, p. 353). Scores a 3 on the export marketing index in 1990; TIDE, p. 314, dates the reforms as start- ing in 1988. Malta Not rated, due to insufficient data on trade poli- cies. Mauritania Open since 1992. Rated closed during 1970-90 be- cause of a high black market premium, and a 4 on the export monopoly index. TIDE, p. 320, states that 1992 marks the decisive intensification of re- forms. Mauritius Open since independence in 1968. Source: TIDE, p. 324. Mexico Open since 1986. A combination of moderate tar- iffs and extensive import licensing since the early 1950s. In the 1960s, 80 percent of tariff lines were covered by licensing (Bueno, 1971, p. 181). A high black market premium also disqualifies Mexico in the early 1980s. The 1986 dating for the reform is based on TIDE, p. 328. Morocco Open from independence in 1956, to 1964, closed 1964-84, open since 1984. Imports could be made freely from French franc area countries up to 1964 (IMF, Annual Report on Exchange Restrictions, various issues). Introduction of a list of permitted imports, and prohibition of everything else, in 1964 (IMF, Annual Report on Exchange Restric- Jeffrey D. Sachs and Andrew Warner 87 tions, 1965). In 1980 the mean unweighted tariff was 47 percent (IMF, Annual Report on Ex- change Restrictions, 1995, p. 33). The dating of the 1984 liberalization is based on Nsouli and oth- ers (1995, pp. 32-33). By the mid-1980s the quota coverage, mean tariff, and black market premi- ums were all below our thresholds for openness. Mozambique Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). TIDE, p. 351, does not present any evidence of a major recent reform effort. Myanmar Never open. Nepal Open since 1991. Not rated open between 1960 and 1990 because of a high black market exchange rate premium. Not rated open in the 1950s, based on evidence in Shreshtha (1981). The dating of the reform in 1991 is based on the discussion in TIDE, p. 356. Netherlands Open since 1959, based on date of full currency convertibility. Member of the EEC. The average tariff in the Common Market was less than 40 per- cent in 1962. Source: Balassa (1965, table 1, p. 580). No major increase in protection 1962-93. New Zealand Open since 1986. Quantitative trade restrictions covered more than 40 percent of imports in 1981 and 1983, and had fallen below 40 percent by 1986 (Laird and Yeats, 1990, table 4.2). Nicaragua Open 1950-60, closed 1961-90, open since 1991. In the 1950s Nicaragua had import licensing and surcharges for acquiring foreign exchange for im- porting, but the licenses were freely granted and the average surcharge did not exceed 40 percent (IMF, Annual Report on Exchange Restrictions, various issues). The open period ends in the 1960s 90 Brookings Papers on Economic Activity, 1:1995 timetable for removing tariffs, but average tariff levels were low and applied mainly to agricultural products. The postrevolution government in 1974 sought a delay on the tariff reductions but under- took no increase in protection rates (Avillez, Fi- nan, and Josling, 1988). We have found no source that reports average tariff rates for Portugal, but based on our reading of Avillez, Finan, and Jo- sling (1988) and our knowledge of the tariff rates of other EFTA countries, it is very unlikely that Portugal's average tariff exceeded 40 percent. Trade liberalization resumed in 1980. Rwanda Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). Senegal Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). TIDE, p. 437, does not present any evidence of a major recent reform effort. Sierra Leone Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). The state-controlled export monopsony is still in oper- ation. The average black market premium was 408 percent in 1985-90. There is no evidence in TIDE of a major recent reform effort. Singapore Open since independence in 1965. Somalia Never open. Average black market premium ex- ceeded 20 percent in the 1970s and 1980s. Rated as socialist by Kornai (1992). In 1992 the United Nations declared Somalia a country without a government (Famighetti, 1993, p. 808). South Africa Open since 1991. Source: Lachman and Bercuson (1992, pp. 32-37). South Africa has traditionally Jeffrey D. Sachs and Andrew Warner 91 followed an import-substitution and inward-look- ing development strategy. This was reinforced by externally imposed trade and financial sanctions in 1985. The United States and several other countries began lifting trade sanctions in the sum- mer of 1991. Thus although it is hard to put a pre- cise date on qualification as open, 1991 seems a reasonable assumption. Spain Open since 1959. In July 1959 Spain unified its ex- change rate, liberalized imports, and made its cur- rency convertible with the currencies of other Or- ganization for European Economic Community (OEEC) countries (IMF, Annual Report on Ex- change Restrictions, 1960, p. 284). The sum of tariffs and indirect taxes on imports averaged 18.1 percent in 1961, and fell gradually for the next 27 years (Gamir, 1990). Sri Lanka Open 1950-56, closed 1956-77, open 1977-83, closed 1983-91, open since 1991. The dating is based on Cuthbertson and Athokorala (1991). The dating for 1983 is based on the annual black mar- ket premium data in Cowitt (1986). Swaziland Not rated. Inherently ambiguous case due to membership in the Southern African Customs Union. (See discussion in appendix above.) Sweden Open since 1960, based on the date of full cur- rency convertibility following membership in the European Free Trade Association. No black mar- ket exchange rate premium; average tariff less than 40 percent in 1962. Source: Balassa (1965, ta- ble 1, p. 580). No major increase in protection 1962-93 (IMF, Annual Report on Exchange Re- strictions, various issues). Switzerland Always open. Full currency convertibility. Mem- bership in the European Free Trade Association 92 Brookings Papers on Economic Activity, 1:1995 since 1960. No black market exchange rate pre- mium, and low tariffs since at least 1950. Syrian Arab Open 1950-65, closed since 1965. The dating of Republic the initial phase of liberalization is based on the IMF's Annual Report on Exchange Restrictions, 1950-66, which does not report any significant im- port barriers. This open era ended in 1965, when a state trading company, SIMEX, was granted a monopoly on the purchase of imports. In 1965 SIMEX purchases represented 55 percent of all imports (IMF, Annual Report on Exchange Re- strictions, 1966, p. 521). Classified as closed dur- ing the 1980s due to a high quota coverage and a high black market premium. The country had multiple exchange rates for everything in 1993, and current account restrictions, according to the IMF's Annual Report on Exchange Restrictions, 1993. The average effective import tariff was 27 percent in the mid-1980s. In 1980s the quota cov- erage was above 40 percent and the black market premium was well above 20 percent. There is no evidence of recent reform. Taiwan Open since 1963, based on Lin (1993). Tanzania Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). Tanza- nia is not rated as a reformer since 1990, due to the discussion in TIDE, p. 475. Thailand Always open. Source: Phongpaichit (1992). Togo Never open. Not rated as open before 1990 be- cause it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). Togo is not rated as a reformer since 1990 due to the dis- cussion in TIDE, p. 486. Trinidad and Never open. Based on our indicators and the dis- Tobago cussion in TIDE, p. 490. Jeffrey D. Sachs and Andrew Warner 95 Zambia Open since 1993. Not rated as open before 1990 because it has a score of 4 on its export marketing board (Husain and Faruqee, 1994, p. 238). The 1993 date is based on TIDE, p. 538. Zimbabwe Never open. The Federation of Rhodesia (Zim- babwe) and Nyasaland, established in 1953, had very high rates of protection. This economic iso- lation was intensified with the imposition of United Nations sanctions in 1966. Source: Ndlovu (1994, pp. 10, 59). Rhodesia, and after its independence in 1980, Zimbabwe, had a black market premium that averaged above 90 percent in the 1970s and 1980s. It is also rated as closed because it is on Kornai's (1992) list of socialist economies. It is not rated as a recent reformer, due to the discussion in TIDE, p. 553. Comments and Discussion Anders Aslund: Jeffrey Sachs and Andrew Warner have written a broad and most stimulating paper. They have included a large number of countries and formulated a clear hypothesis which can be statistically tested, thanks to a strict categorization. Their main conclusion is that reform works and that there is no invin- cible poverty trap, which is easy to agree with. Even if countries experi- ence falling GDP for years, they can catch up by adopting the right eco- nomic policies. A second conclusion is that openness to global integration is the crucial criterion of good economic policies, and the rest follows. That is less obvious. A third, less elaborated, theme is why some countries adopt the right policies at certain times, and others do not. Here the reasoning is neither complete nor stringent. While the denial of an invincible poverty trap seems convincing, there are several factors to consider. The authors bring up a great many, but there are others; for instance, migration and various forms of inter- national intervention. If we ignore racism and look upon all factors of production as transferable, it is indeed difficult to accept that a poverty trap is given once and for all. However I feel uneasy with the word convergence, because it sug- gests that there is one ideal that everyone can learn; that the leaders can do no wrong or unlearn this ideal, and that their challengers can do no better. Coming from Sweden, I am firmly convinced that good economic policies and institutions can be unlearned and abolished. Argentina is a country with a longer record of economic unlearning. We are consider- ing very long periods and this convergence may be temporary, lasting only a few decades. Sachs and Warner have simply defined their criteria for an open econ- omy, found the statistics for their categorization, and tested their hypotheses. In this fashion, they have largely avoided the question: 96 Jeffrey D. Sachs and Andrew Warner 97 Which are the necessary prerequisites for successful reform? They ar- gue that relatively free foreign trade and a reasonably convertible cur- rency are sufficient conditions for the success of economic reform; they assume, for instance, that macroeconomic stabilization follows. The correlation between openness to foreign trade and the ability of poorer countries to catch up is convincing, but the causality needs to be proven further. What roles do other factors play? Usually, a government adopts a sensible policy covering many fields, and trade liberalization is only one aspect. For instance, liberalization and macroeconomic stabili- zation are usually introduced in parallel, in one package. In recent years Sachs has presented a number of alternative lists of the four to six factors essential to the success of economic reform. ' All of them seem quite sensible. It would be useful to test these factors as far as possible to find the truly crucial preconditions of success. Common suggestions have been: openness to foreign trade, domestic liberaliza- tion, convertibility, macroeconomic stabilization, international finan- cing, a pegged exchange rate, mass privatization, a social safety net, and certain political criteria (strong leader, insightful political elite, civil so- ciety, manageable interest groups, political pluralism, public educa- tion). Apart from international financing, all of these criteria are institu- tional, which makes testing more complicated. The liberalization of foreign trade and the introduction of a convert- ible currency are hardly sufficient conditions for economic growth. A country with a very open economy can have bad incentives in the form of excessive taxes and public expenditures, and stay at a suboptimal equilibrium for decades. Sweden is an obvious example. In a recent pa- per Sachs has written about the entitlement trap in eastern Europe, par- ticularly in Hungary.2 Similarly, the importance of a social safety net has been exaggerated in the discussion of former socialist countries. I am struck by its absence in discussions of East Asia. It is difficult to understand why a social safety net is enormously important in eastern Europe and of no conse- quence in East Asia. Clearly the social safety net has a bigger place in political rhetoric than in sound economic analysis. Even Russia had so- cial expenditures of 21 percent of GDP in 1994.3 No country at this level 1. See, for example, Sachs (1992, 1995d, 1995e). 2. Sachs (1995b, 1995c). 3. Shapiro (1994).