CHAPTER III TRADE AND FINANCIAL INTEGRATION (2024)

Table 3.2.

Trade Openness and Saving-Investment Correlations

¹Slope coefficient from a pooled OLS regression of the saving rate (gross domestic saving divided by GDP) on a constant and the investment rate (gross investment divided by GDP), estimated over 1975–99. “—” indicates that the coefficient is not significantly different from zero at the 5 percent level.

²Based on a country’s degree of trade openness relative to the median.

Box 3.1.

Using Prices to Measure Goods Market Integration

Goods market integration is traditionally measured using bilateral trade flows, with larger flows implying greater integration. One important reason for the popularity of this approach is that data on bilateral trade flows are readily available, including from the IMF Direction of Trade Statistics. The determinants of goods market integration are then typically analyzed using a gravity model, which consistently finds that countries that are closer to each other and more similar (in terms of historical and cultural factors) are more integrated (Box 3.3). One potential limitation of the flows-based approach is that trade flows may not be a good proxy for market integration. For example, across countries with similar production structures, even small trade barriers could make it uneconomic to trade certain goods, leading to low trade flows. To check the robustness of the flows-based approach, it is useful to look at the prices of goods across markets, with smaller price differentials implying greater goods market integration.

In one of the first studies to use price dispersion to measure goods market integration across a large number of countries, Parsley and Wei (2001) analyze data on the prices of 95 tradable goods across 83 cities all over the world from 1990 to 2000.1 The goods are highly disaggregated and essentially identical, such as frozen chicken, light bulbs, toilet paper, and tonic water, all standardized by weight or volume. The data set was compiled by a single source, the Economist Intelligence Unit (EIU), ensuring comparability of the goods across locations. Using all tradable products, the authors compute the standard deviation of the price differences for every pair of cities for each year, with a smaller standard deviation indicating greater market integration. The standard deviations are then used in an econometric analysis of the factors underlying goods market integration, including transport costs, tariffs, and currency arrangements.

Reassuringly, the price-based approach reaches similar conclusions about the pattern and determinants of goods market integration as the more traditional flows-based approach.2 Specifically, both approaches suggest the following.

• Goods market integration has increased over the past decade. The figure shows the down-ward trends in the standard deviation of price differences for two city-pairs: Hong Kong SAR and San Jose, Costa Rica; and Hong Kong SAR and Beijing, China.

• Higher transport costs—proxied by distance—lead to lower market integration. In the flows-based gravity model, bilateral distance always has a negative coefficient, indicating that farther-away countries tend to trade less. In the price-based approach, the distance variable consistently has a positive coefficient, indicating that the price dispersion for identical products (i.e., lack of market integration) increases with distance. For example, the figure shows that price differences are lower between Beijing, China, and Hong Kong SAR than between Hong Kong SAR and San Jose, Costa Rica.

• Some regional preferential trading arrangements have a positive and significant impact on goods market integration. One way to characterize the magnitude of these effects is by their equivalent tariff reductions, i.e., how much tariffs would have to decline to achieve the same effect. The paper finds that the North American Free Trade Area and the European Union both have equivalent tariff reductions of about 5 percent. This effect is large compared with the average external tariff rate of industrial countries of about 4 percent.

• Institutionalized currency arrangements (a union or a board) generally increase goods market integration among their member countries.3 However, the authors find that the effects of institutionalized currency arrangements are not all the same. For example, the estimate of the equivalent tariff reduction associated with the Communaute Financiere Africaine (CFA) is small and not significantly different from zero, while the estimate for the euro area is about 2 percent and significant.

• Finally, border effects are significant, even after taking account of common currencies and free trade areas. The authors find that city-pairs within the United States are the most highly integrated. Relative to the U.S. benchmark, the degree of goods market integration across other city-pairs, including those within common currency and free trade areas, still has further to go.

The similarities in the results from the different approaches imply that the two are complementary, each providing insights into the measurement and determinants of goods market integration.

Note: The main authors of this box are David Parsley and Shang-Jin Wei.1 Engel and Rogers (1996) were the first to use price dispersion to measure goods market integration, but they only considered Canada and the United States. Other recent studies include Engel and Rogers (2001), Rogers (2001), and Hufbauer, Wada, and Warren (2002).2Both approaches take account of the effects of a variety of factors—including tariffs—in a single framework.3Reducing nominal exchange rate variability reduces price dispersion and improves goods market integration, but by a smaller order of magnitude than a currency union or currency board.

Box 3.2.

Transport Costs

Transport costs play a central role in explaining trade patterns. Transport costs are still large when compared with tariffs and represent the main impediment to trade in a majority of countries (World Bank, 2002). Transport costs, usually proxied by distance and other geographical variables, are highly significant in the gravity model (Box 3.3) and are important in explaining international vertical specialization—the slicing up of production processes into distinct steps, allowing specialization across countries (Box 3.4). They also help explain the substitution between trade and foreign direct investment (Loungani, Mody, and Razin, 2002).

Transport costs have declined considerably in the past fifty years, but not by as much as during the late nineteenth century, which was also a period of rapid global economic integration. Time series measures of transport costs are difficult to obtain, because no single data set covers all transport costs in a systematic way.1 The costs of shipping a bushel of wheat from New York to Liverpool declined by about two-thirds between 1854 and 1913, and from New York to Chicago by about three-fourths between 1870 and 1913 (Harley, 1980). The figure shows that sea freight costs have not changed much since 1960, while airfreight costs have been steady since 1980. These trends are broadly consistent with more detailed estimates in Hummels (1999).2 They are also consistent with Baier and Bergstrand’s (2001) finding that the effect of declining transport costs on the growth of industrial country trade between the late 1950s and the late 1980s was only about one-third of the impact of tariff reductions.

The less rapid decline in transport costs in the recent era reflects three main factors.

• Innovations in transport technology. The late nineteenth century saw more dramatic improvements in transport technology, including the introduction of steamships and rail-roads, as well as the construction of the Suez and Panama Canals. The most important technological breakthroughs of the past fifty years have been the introduction of containers for sea transport in the late 1960s and the introduction of jet engines and large body air-craft for air transport.3

• Competitive environment. In the late nineteenth century, the disappearance of trading monopolies sharply lowered shipping costs. By contrast, the introduction of containers in the late 1960s raised fixed costs, which increased market concentration, preventing the efficiency gains from being transmitted to users of transport services (Gilman, 1984).4 Partly in response to this concentration, many developing countries adopted policies to ensure that national flag fleets were granted a large share of shipping, which further stifled competition and increased transport costs, especially for developing countries.

• Measurement issues. Comparing the recent decline in transport costs with that in an earlier historical period is complicated by the fact that transport services are not hom*ogenous. Transport services differ along several dimensions, such as speed, reliability, and frequency, and prices are not adjusted for changes in quality. The past fifty years have seen major improvements in quality: Hummels (2000) finds that the decline in shipping time between 1950 and 1998 due to the introduction of air shipping and faster ocean vessels was equivalent to reducing tariffs on manufactured goods from 32 percent to 9 percent. However, because prices are not adjusted for quality, they do not capture this significant improvement in speed.

Notwithstanding the measurement issues, the decline in transport costs—driven by improvements in technology—was likely a more important factor behind the increase in economic integration in the late nineteenth century than globalization in the late twentieth century.

For developing countries, an important issue is that the quality of transport infrastructure is a significant determinant of transport costs. Freight costs as a share of import values vary considerably across countries, ranging from about 4.5 percent in industrial countries to 12 percent in Africa (UNCTAD, 2001). These variations in transport costs are mostly due to differences in geography and transport infrastructure. Limão and Venables (2001) find that transport costs in landlocked countries are about 50 percent higher than average. Clark, Dollar, and Micco (2002) find that improving the efficiency of a seaport from below-average to above-average decreases transport costs by more than 12 percent. Low-quality transport infrastructure is especially relevant for poorer countries and substantially limits their trading potential.

Looking forward, the terrorist attacks of September 11, 2001, will likely increase transport costs in the long run. The fact that transport costs remained generally stable in the months after September 11, during a recession and a period of relatively low oil prices (when transport costs usually fall), suggests that underlying transport costs may have increased. Higher transport costs could result, for example, from the need for customs authorities to examine shipping containers on a systematic basis. Costs of security measures could be on the order of 1-3 percent of the value of trade (OECD, 2002). Checking even a small share of containers will imply an explicit cost for governments and a much higher implicit cost for importers and exporters in terms of delays. Hummels (2000) estimates that a one-day delay is equivalent to a loss of 0.8 percent of the value of a manufactured good. In addition, greater uncertainty about delays will force firms to hold higher levels of inventories.

Note: The main author of this box is Antonio Spilimbergo.1Estimating transport costs is difficult because transport services are not hom*ogenous and available measures are unreliable. A commonly used method is to calculate transport costs as the difference between the value of f.o.b. exports and the value of c.i.f. imports for pairs of countries. However, the original data sources are unreliable and the composition of trade changes over time, making any cross-time comparison problematic (Hummels, 1999).2However, Hummels (1999) reports increasing sea shipping charges between 1970 and 1985.3The share of U.S. import value transported by air rose from 6.2 percent in 1965 to 24.7 percent in 1998.4However, the practice of registering ships under flags of convenience reduced shipboard operating costs by 12–27 percent (Hummels, 1999).

Which Regions Undertrade?

To measure the shortfall in actual trade, a benchmark for expected trade is needed. The gravity model of international trade without explicit policy variables is used to derive measures of the expected volume of trade between trading partners (Box 3.3). Just like the force of gravity between two objects is proportional to their mass and inversely proportional to the distance between them, the gravity model of trade postulates that the magnitude of bilateral trade flows between two countries is positively related to the joint size of the two trading economies and negatively related to the distance between them. Over time, the gravity model of trade has been elaborated to incorporate a wide variety of other factors.

A country is said to “undertrade” if its actual trade across trading partners is, on average, substantially below the level predicted by the gravity model without explicit policy variables. Similarly, a country is said to “overtrade” if its actual bilateral trade is substantially above the average level predicted by the gravity model. As the gravity model accounts for the “natural” causes of trade, under- and overtrading must largely represent above- or below-average “artificial” impediments (Rose, 2002).13 This approach has the benefit of capturing the overall impact of a country’s policy and institutional environment, including a wide variety of artificial impediments and not just trade policies. However, this approach depends on getting the natural causes exactly right—that is, the results are sensitive to the specification of the gravity model.

The analysis of undertrading is based on a conventional gravity model estimated over the period 1995–99, along the lines of Rose (2002). Bilateral merchandise trade data covering 131 industrial and developing countries are taken from the IMF’s Direction of Trade Statistics. The country coverage reflects the availability of data for the explanatory variables in the gravity model.14 The data are averaged over 1995–99 to abstract from cyclical developments. The estimated coefficients of the gravity model are similar in sign, magnitude, and statistical significance to the results in the recent literature.

The results suggest that significant undertrading occurs in certain developing country regions, though of course individual countries sometimes diverge from regional averages. Table 3.3 shows the average differences between actual and predicted trade by region, expressed in logarithms following convention.15 The same rank ordering of regional undertrading is found by Rose (2002). The degree of undertrading is large in the Middle East and North Africa, in line with the results of Al-Atrash and Yousef (2000), and south Asia, though this assessment does not take account of trade in services, which have grown especially rapidly in that region. The degree of undertrading is smaller in the Western Hemisphere. Countries in sub-Saharan Africa trade slightly more than predicted, consistent with Foroutan and Pritchett (1993), Rodrik (1998), and Coe and Hoffmaister (1999). Countries in east Asia are strong traders relative to other developing countries.

Undertrading is generally less pervasive in intraregional than in extraregional trade. This result is not obvious, because geographical factors that tend to boost intraregional trade are already taken into account in the gravity model. One possible reason for this finding is regional preferential trading arrangements, such as MERCOSUR. If dummy variables representing regional preferential trade agreements are included in the gravity model, the extent of intraregional overtrading in the Western Hemisphere is noticeably reduced, though intraregional overtrading in other regions is not affected much. Also, extraregional over- or undertrading remains roughly unchanged, suggesting that regional trade agreements do not divert trade in developing country regions.16

Over the past twenty years, the degrees of regional undertrading have changed, reflecting developments in artificial barriers to trade (Table 3.4). Two regions became relatively weaker traders: sub-Saharan Africa and—especially—the Middle East and North Africa, which went from slight overtrading to large undertrading. The weakening of sub-Saharan Africa’s trade performance (which is only partly related to a secular decline in non-oil commodity prices) is consistent with the concerns noted in the May 2001 World Economic Outlook about the marginalization of this region within the world trading system. The other regions—east Asia, south Asia, South America, and especially the Caribbean and central America—became relatively stronger traders. Changes in regional preferential trade agreements cannot account for changes in intraregional overtrading during the 1980s and 1990s, except in South America where MERCOSUR led to an increase in intrabloc trade during the 1990s.17

In summary, the analysis suggests that undertrading remains a serious problem in many developing countries, especially in the Middle East and North Africa, and south Asia. Undertrading reflects above-average artificial barriers to trade in all aspects of a country’s policy and institutional environment.

What Is the Impact of Trade and Balance of Payments Restrictions?

While undertrading is a measure of overall artificial barriers to trade, it is not directly connected to any specific policies. As a result, this measure cannot be used to assess the impact of trade or balance of payments liberalization on trade flows. To answer this type of question, the gravity model was reestimated over 1995–99 with two measures of policy restrictiveness as explanatory variables: (1) the IMF’s index of overall trade regime restrictiveness, which is based on average import tariffs and nontariff barriers (IMF, 1998); and (2) an index of balance of payments restrictiveness, which ranks the overall restrictiveness of current and capital account restrictions (Mody and Murshid, 2002).18

Both indices suggest that trade and balance of payments policies are generally less restrictive in sub-Saharan Africa, east Asia, and the Western Hemisphere than in other developing country regions, though there are important measurement problems (Table 3.5).19 In other words, the marginalization of sub-Saharan Africa within the global trading system noted above is not primarily due to measured trade and balance of payments restrictiveness, but rather to other aspects of the region’s policy and institutional environment. Similarly, measured trade and balance of payments restrictions are relatively small in the Western Hemisphere, although the region’s undertrading suggests that overall artificial barriers to trade are higher than average.

The two policy variables have significantly negative effects on bilateral trade flows (Table 3.6). The magnitudes of the coefficients measure the effects on trade of changes in policy restrictiveness: one-point increases in both trade and balance of payments restrictiveness reduce trade volumes by about 5 percent. The coefficients on the other variables are generally similar to those obtained for the gravity model without explicit policy measures, and remain comparable in sign, magnitude, and statistical significance to the results reported in the recent literature. Interestingly, the coefficient on the product of per capita incomes becomes somewhat smaller once the policy variables are included, indicating that policy restrictiveness is inversely related to the level of economic development. In other words, trade and balance of payments policies tend to be more restrictive in poorer countries, which presumably reflects in part the adverse effect of policy restrictiveness on growth.

A striking result is that balance of payments restrictiveness has a significant and large adverse effect on trade, consistent with the idea that financial frictions can help to explain the segmentation of global goods markets. This idea parallels the view that trade frictions are a factor behind the segmentation of international financial markets (Obstfeld and Rogoff, 2000a). There is a growing literature on the role of international financial frictions in dampening trade. Tamirisa (1999) finds that exchange and financial controls represent a significant barrier to trade. Rose (2000) shows that belonging to a currency union more than triples a country’s trade with the other members of the union, with no evidence of trade diversion.20 Rose and Spiegel (2002) find that sovereign defaults also tend to have adverse trade effects.21

Multilateral liberalization of trade and balance of payments restrictiveness would have large effects on trade. Table 3.7 presents illustrative calculations of the impact of policy liberalizations in industrial countries, developing countries and all countries, respectively.22 If industrial countries reduced their trade restrictiveness to the lowest possible level, trade between industrial and developing countries (North-South trade) would increase by about 14 percent. The trade effects of balance of payments liberalization are generally smaller, given the already low levels of restrictiveness in industrial countries. The full liberalization of both trade and balance of payments policies in all countries would increase trade between industrial countries (North-North trade) by about 40 percent, North- South trade by about 63 percent, and trade between developing countries (South-South trade) by about 94 percent.

How Important Is Policy Relative to Other Factors in Explaining Trade?

To put the role of policies into perspective, this subsection compares the effects of trade and balance of payments restrictiveness on trade with those of other key factors, including economic size, level of economic development, and geography. For this comparison, North-North trade is taken as a benchmark, because industrial countries generally have less restrictive policies, especially for manufactured goods, which account for the bulk of trade between industrial countries. 23 This analysis is based on the gravity model that includes policy variables.

The most important reason why the absolute level of trade between industrial countries is much larger than trade between developing countries is that average industrial country GDP is twice as large as average developing country GDP. Using the coefficients from the gravity model, differences in economic size account for 80 percent of the difference in average bilateral trade flows. As economic size has such an overwhelming impact on bilateral trade flows, the results below are adjusted for economic size. Even after this adjustment, bilateral trade flows involving developing countries are smaller than those among industrial countries.

Overall, the results suggest that trade and balance of payments restrictiveness play a significant role in explaining why developing countries trade less per unit of GDP than industrial countries, though economic development and geography are even more important (Figure 3.4). Lower income per capita is the single most important reason why adjusted South-South trade is smaller than adjusted North-North trade, and accounts for about one-fifth of the difference between adjusted North-South trade and adjusted North-North trade. Differences in income per capita matter a lot for trade because richer consumers tend to have a higher demand for product variety, while the production of differentiated goods remains specialized, leading to intra-industry trade. Geography, especially distance, is the single most important impediment to North-South trade, and accounts for two-fifths off the shortfall in South-South trade.

More restrictive trade and balance of payments policies account for about 10-20 percent of the shortfall in adjusted bilateral trade flows. Restrictive policies hurt South-South trade more than North-South trade because developing countries have on average greater restrictions than industrial countries, so their adverse effect is doubled as both trading partners have higher restrictions. Other determinants, like linguistic or historical factors, are much less important because the differences between industrial and developing countries are on average small. Finally, unexplained differences account for only a small part of the shortfall in trading involving developing countries.

The gravity model is less successful at explaining why developing countries in east Asia trade more per unit of GDP than countries in other regions (Figure 3.5). In contrast to the results discussed above, unexplained differences account for much of the excess in east Asia’s trade relative to other developing country regions. In terms of the explained differences, east Asian countries have on average higher income per capita and less restrictive policies (both of which tend to increase trade) but are relatively distant from their trading partners (which tends to reduce trade).

The large unexplained difference between trade volumes in east Asia and other developing country regions may be related to increasing vertical specialization in global production (Box 3.4). In recent years, the further slicing up of the production chain has accompanied the substantial expansion of trade. With the expansion of international vertical specialization, trade flows per unit of GDP rise even when all other factors remain unchanged. Indeed, the contribution of increased intra-industry trade to total trade growth, which partly reflects greater vertical specialization, has been higher in east Asia than in other developing country regions (Table 3.8). Also, in the gravity model, the difference between actual and expected trade in east Asia falls once the share of intra-industry trade is taken into account.24 Thus far, trade flows between industrial countries, and flows with and among east Asian economies, have been most affected by outsourcing, but it is increasingly assuming a global dimension.

The increasing role of vertical specialization in east Asia is consistent with the region’s focus on labor-intensive production. Figure 3.6 shows the factor composition of net exports—that is, the embodiment of capital, labor, land, natural resources, and technology—across developing country regions.25 East Asia is a net exporter of labor-intensive manufactures; sub-Saharan Africa and Latin America are net exporters of agricultural products; and sub-Saharan Africa, Latin America, and the Middle East and North Africa are net exporters of raw materials and fuels. While the share of manufactured goods in total trade has increased over time in many developing countries, the most significant change in the pattern of factor specialization has occurred in east Asia, where net exports of labor-intensive manufacturing products have increased while net exports of agricultural products and raw materials have fallen.

The analysis of developing countries’ trading patterns suggests three main points.

• Undertrading, which reflects the overall impact of artificial barriers to trade in a country’s policy and institutional environment, remains a serious problem in many developing countries, especially in the Middle East and North Africa, and south Asia.

• Trade and balance of payments restrictiveness are important reasons why developing countries trade less than industrial countries, though economic development and geography matter even more.

• International vertical specialization has played a growing role in east Asia, where less restrictive trade policies helped create a favorable environment. Vertical specialization is likely to become increasingly important for other developing countries with open trading regimes, abundant labor, and flexible economies.

Box 3.4.

Vertical Specialization in the Global Economy

The growth of world trade has been accompanied by international vertical specialization. Vertical specialization refers to the slicing up of the production process into distinct steps, allowing specialization across locations (locational decentralization) and firms (outsourcing). Historically, international vertical specialization can be traced to the mid-1960s, when electronics components began to be assembled in Hong Kong SAR, Thailand, Malaysia, and Singapore, and apparel and leather goods in the Dominican Republic and the Philippines. Thus, international specialization has become two-dimensional, with countries specializing both vertically in certain stages of the production processes and horizontally in the production of some categories of final goods.

Rising vertical specialization implies an increasing ratio of international trade to value added, as parts and components are shipped back and forth across national borders for further processing along the production chain. The table shows that the ratio of merchandise trade to merchandise value-added rose sharply between 1980 and 2000 in both industrial countries and emerging market economies, especially in Asia and Mexico. While the increase in the trade-to-value added ratio could also be due to greater trade in final goods, other empirical evidence confirms the growing role of international vertical specialization in international trade.1 Hummels, Ishii, and Yi (2001) find that increased vertical specialization accounted for one-third of world trade growth between 1970 and 1990, after taking account of inter- and intra-industry linkages.2 Yeats (2001) finds that the share of components and parts in trade in machinery and transportation equipment among industrial countries rose from 26 percent in 1978 to 30 percent in 1995. Asian exports and imports of components and parts grew even faster than those of most industrial countries between the mid-1980s and the mid-1990s (Ng and Yeats, 1999).

What are the implications of vertical specialization? First, increasing vertical specialization tends to accelerate the global propagation of shocks, as industry-specific shocks are immediately transmitted to countries along the production chain. By contrast, with horizontal specialization, industry-specific shocks tend to initially affect a more limited number of countries. The rising sensitivity of east Asian economies to cyclical developments in the global information technology industry was discussed in Chapter III of the October 2001 World Economic Outlook. Second, vertical specialization allows countries to specialize in the stages of production that best fit their relative factor endowments. Specifically, labor-abundant developing countries can increase their role in global manufacturing by specializing in the production of labor-intensive parts and components or labor-intensive assembly processes. Vertical specialization also likely facilitates the convergence of wages for similar types of labor and other factor prices across countries. Finally, vertical specialization underlines the importance of reducing trade barriers further, as the back-and-forth shipping of goods across borders compounds the effects of even relatively low trade barriers.

Vertical specialization is driven by three main factors.

• Improving service links. Service links between production locations or producers, including activities like transport, telecommunication, insurance, coordination, and supervision, are critical to the success of vertical specialization. Innovations in transport and communications technology, as well as the deregulation of service provision, have made service links more reliable and less costly (Jones and Kierzkowski, 2001). Economic growth has also helped, as greater demand offsets the fixed set-up costs of services links between locations.

• Increased customization. Technological change has increased the scope for the inexpensive customization of generally standardized components and parts, allowing the exploitation of increasing returns to scale in the production of parts and components.3 The greater attractiveness of outsourcing relative to small scale in-house production has led to the “commoditization” of some manufactures, especially parts and components for electronics and information technology goods (e.g., memory chips and disk drives).

• Falling trade barriers. With back-and-forth shipping across borders, the effects of trade barriers are compounded, so reductions in trade barriers have a more than proportional positive impact on vertical specialization. It is not surprising that vertical specialization has sometimes occurred in the context of special tariff or quota provisions for offshore assembly or regional preferential trade agreements.4

Looking forward, vertical specialization is likely to become even more important. The negative impact of the September 11, 2001, terrorist attacks on transport costs (Box 3.2) will probably be more than offset by further innovations in telecommunications technology (which will strengthen services links), expansion of inexpensive customization (especially in the electronics and information technology industries), and trade liberalization (which will reduce the adverse effects of trade barriers more than proportionately). Vertical specialization has so far mostly affected trade in industrial countries, east Asia, and the Caribbean and Central America, but it is increasingly assuming a global dimension.

Note: The main author of this box is Thomas Helbling.1It is often difficult to distinguish between final and intermediate goods in trade data, partly because the distinction sometimes depends on circ*mstances.2Using a similar methodology, Campa and Goldberg (1997) find that the share of imported intermediate inputs in total manufacturing production increased by about one-fourth in Canada and doubled in the United States between 1974 and 1993.3See Grossman and Helpman (2002) for a discussion of customization and outsourcing.4See Graziani (2001) on tariff and quota provisions in the context of offshore assembly of textiles and clothing.

Trade Integration and Vulnerability to Financial Crises

What is the impact of greater trade integration on a country’s vulnerability to external financial crises? Two types of external financial crises are considered: debt defaults and currency crashes. A debt default is defined as occurring if there are external arrears to commercial creditors of more than 5 percent of total commercial debt outstanding or if there is a rescheduling or restructuring agreement with commercial creditors, based on Detragiache and Spilimbergo (2001). A currency crisis is defined as an exchange rate depreciation vis-à-vis the U.S. dollar of at least 25 percent and at least double the rate of depreciation in the previous year, as long as the latter is less than 40 percent—to exclude hyperinflationary episodes (Milesi-Ferretti and Razin, 1998).

The upper panel of Figure 3.9. shows the frequency of debt defaults and currency crises by five-year periods between 1975 and 1999. The two types of external financial crises have similar time profiles throughout the entire period, including peaks during the early 1980s.

External financial crises have been more frequent in countries that are less integrated into the global trading system (middle panel of Figure 3.9). Over the past quarter century, less integrated countries have been on average about one-fifth more likely to suffer a debt default, and one-third more likely to have a currency crisis than the average developing country. Across developing country regions, the Western Hemisphere has been significantly more vulnerable to episodes of financial turmoil than any other region (lower panel of Figure 3.9), as highlighted in Chapter II of the April 2002 World Economic Outlook. The benefits also turn out to be largest for countries that already have open capital markets.

The inverse relationship between trade integration and external financial crises remains statistically significant in a multivariate econo-metric framework. Sgherri (2002) shows that the relationship is robust to alternative definitions of external financial crises and after taking account of the conventional determinants of crises, including a country’s economic fundamentals, its solvency position, foreign exchange reserves, and external macroeconomic conditions. In addition, the result remains significant after incorporating—in a bivariate probit framework—two crucial supplementary relationships: the fact that trade openness may be related to the same factors that affect the frequency of financial crises, and the fact that the frequency of external financial crises and trade openness may both be affected by factors not included in the empirical framework.27 This result is consistent with other analyses of the determinants of crisis frequency, including Klein and Marion (1997) and Milesi-Ferretti and Razin (1998), although Detragiache and Spilimbergo (2001) find that recently trade openness has tended to increase the likelihood of financial crises.28

Trade integration reduces a country’s financial fragility by increasing both the ability and willingness to service external obligations. A higher ratio of exports to GDP implies that exchange rate depreciation will provide a greater boost to a country’s ability to earn foreign exchange, which is essential to service foreign currency-denominated debt. In this way, a greater export ratio decreases the likelihood of a sharp reversals of capital flows, as the country is considered to be more able to service its foreign currency-denominated debt. This reassurance is especially important in developing countries, where domestic financial markets are shallow and economic and policy prospects are generally more uncertain than in industrial countries.29 In addition, trade openness serves as an incentive to meet external obligations by making a country more vulnerable to creditors’ sanctions in case of default (Bulow and Rogoff, 1989).30

In summary, higher trade integration tends to reduce the frequency of external financial crises. One interpretation of this result is that trade integration provides an important buffer for the inherent volatility associated with financial integration. A possible implication is that countries that are already financially open could see a decline in the frequency of external financial crises by increasing trade openness.

Financial Integration and Output Volatility

What is the impact of openness to foreign direct investment and portfolio investment on output volatility? Output volatility is defined as the unconditional standard deviation of the growth rate of real GDP per capita over the period 1975–99. The upper panel of Figure 3.10 shows that output volatility in developing countries is roughly double that in industrial countries, and that—among developing countries—output volatility in small countries (those with populations between ½ million and 1½ million) is about one-third greater than average. Developing countries that are relatively more integrated into global capital markets tend to have lower output volatility on average than financially closed countries (lower panel of Figure 3.10). This is true for the period as a whole, but not in the early 1980s and late 1990s, when global financial shocks were especially large.

In particular, the output volatility among financially open small developing countries is about one-third lower than among their financially closed counterparts.31 This result is consistent with the empirical literature on the macroeconomic performance of small economies, which suggests that the benefits from financial deepening may be substantial (Easterly and Kraay, 2000).

Financial openness appears to be associated with lower output volatility through two channels: the magnitude of inflation and exchange rate shocks is lower, and the impact of all shocks on output is dampened. Financially open countries—which are also more open to trade and have slightly higher debt ratios—experience larger external shocks, as measured by the volatility of the terms of trade, trade flows, and financial flows (Table 3.9). However, financially open countries have somewhat more stable real exchange rates and much more stable inflation rates (the volatility of fiscal balances is similar in financially open and closed countries). The lower volatility of inflation and exchange rates in financially open countries may reflect a disciplining effect of international financial markets or a facilitation of the transfer of international best practices in macroeconomic policymaking.32

However, it may also reflect causality running in the opposite direction—from sounder macroeconomic policymaking to greater integration.

Financial integration also seems to be associated with lower output volatility because the impact of shocks on output is dampened. Output volatility is significantly correlated with inflation rate volatility and fiscal balance volatility in financially closed economies, but not in financially open countries (Table 3.10). Similarly, the correlations between output volatility and the volatility of external shocks—including terms of trade volatility, trade flow volatility, and real exchange rate volatility—are significantly lower in financially open economies. Strikingly, although financially closed economies experience less volatile capital flows, this lower volatility is significantly correlated with output volatility. Moreover, in countries that are more financially closed than average, greater financial openness is associated with lower output volatility. In other words, it appears that financial integration helps to smooth the effects of shocks on output, not only by comparing financially open to financially closed countries, but also among financially closed countries.

It is important to distinguish the association between greater openness to foreign direct investment and portfolio investment and lower output volatility from that of external debt. Financially open countries have somewhat higher external debt ratios, and external debt ratios are positively correlated with output volatility in both financially open and financially closed countries. External debt may exacerbate output volatility as balance sheet effects magnify the impact of shocks, especially if domestic financial systems are not yet well developed. However, on average, the indirect effect of financial openness in raising output volatility (through a higher external debt ratio) appears to be smaller than the direct impact in lowering output volatility.

The contribution of financial openness to reducing output volatility remains significant in a more sophisticated econometric framework, which accounts for the endogeneity of financial openness and the heteroscedasticity of shocks across countries (Sgherri, 2002). These results are broadly consistent with the theoretical and empirical literatures. Theoretical models of the international business cycle suggest that global financial diversification may be the right response to terms of trade shocks (Razin and Rose, 1994; Heathcote and Perri, 2002; and Kose, Prasad, and Terrones, forthcoming).33 Empirical work by Bekaert, Harvey, and Lundblad (2002) finds that equity market liberalization is associated with lower volatility of output and consumption. Kim, Kose, and Plummer (forthcoming) show that the amplitude of economic fluctuations in east Asia has fallen over time as countries have become more open. Kraay and Ventura (2001) find that countries use foreign assets as a buffer stock as they try to smooth consumption and investment.

In summary, greater international financial integration is associated with lower output volatility, though this association is generally only realized over the longer term. Output volatility also depends on other factors, including macroeconomic policy stability, domestic financial development, and institutional quality. Greater financial integration appears to be associated with smaller inflation and exchange rate shocks, and a reduced impact of all shocks on output. A possible implication is that economies with less diversified production structures, including small economies, which tend to be very open to trade, could see a decline in output volatility by opening up to financial flows.