Shaun Roache

Recent global financial market volatility and evidence of a slowdown of growth in the United States have rekindled interest in an old question: how are changes in external conditions likely to affect growth in Latin America? The diversity of the region implies that many external factors could have a strong influence, among them external demand, interest rates in advanced countries, investor risk appetite, remittance flows, and commodity prices. This article reviews recent IMF research on how external factors, both macroeconomic and financial, affect economic growth in Latin America.

From the broader literature, a number of stylized facts can be established about the nature of Latin America’s external linkages and how they have changed over time. Linkages seemed to be at their strongest during the 1970s, even though the region was relatively closed to external trade. This was due to a variety of common shocks, including the sharp rises in world oil prices that buffeted the global economy throughout that decade.

During the 1980s, the emerging market debt crisis that afflicted many key countries in the region, together with civil conflict in Central America, weakened these global linkages, causing growth to be determined largely by regional and domestic factors. The story since the mid-1990s has been one of increasing external integration and strengthening linkages. Common shocks are now playing a less important role; trade liberalization, increasing openness, and the globalization of capital markets, including for foreign direct investment, now explain much of the region’s rising sensitivity to external factors.

A number of recent IMF studies have explored external linkages in Latin America and attempted to quantify the effects of external shocks and cycles on regional economic growth. Österholm and Zettelmeyer (2007) assess the effect of a variety of external shocks on the largest six countries in Latin America using a Bayesian vector autoregression (VAR) approach. While VAR models have been used by others to address this topic in the region (e.g., Canova, 2005), the Bayesian approach overcomes some drawbacks of conventional VARs, including a large number of parameters and often short sample periods. This is achieved by imposing “informative priors” on the long-run steady state, on which the forecaster often has an opinion.

In the main specification, external factors are represented by world GDP growth, a trade-weighted index of commodity prices that are relevant for Latin America, U.S. Treasury bill rates, and the high-yield corporate bond spread in the United States to capture investor risk aversion. Also included is the Latin America subcomponent of JP Morgan’s Emerging Market Bond Index (EMBI), which is influenced both by external financing conditions and regional domestic fundamentals. As a measure of Latin American growth, a weighted index for Argentina, Brazil, Chile, Colombia, Mexico, and Peru is used. Imposing the restriction that the external factors were not influenced by developments in Latin America, the key result is that shocks to world GDP growth are passed on to Latin America about one-for-one—that is, a one standard deviation 0.3 percent world growth shock leads to an increase in (four-quarter) Latin American growth by about 0.4 of a percentage point after four quarters. A similar result is obtained when applying a U.S. growth shock. In recent months, the focus has increasingly shifted to the linkages running from financial markets to the real economy, and the evidence from this model suggests that this sensitivity was high, at least over the 1994–2006 sample period. The reaction of Latin American growth to U.S. interest rates is muted, but one standard deviation shocks to the U.S. high-yield bond or Latin American EMBI spreads, interpreted as higher risk aversion, were associated with declines in annual growth of 0.9 and 0.5 percent respectively after three quarters. Commodity prices, a key component of exports for many countries in the region, are also a key transmission mechanism. Overall, the conclusion is that the region may remain sensitive to a number of adverse external scenarios, including sharply tighter financing conditions, particularly when combined with slower world growth or a large and rapid drop in commodity prices, which in turn would have consequences for capital flows to the region.

This approach also yields insights at the country level. Abrego and Österholm (2008) apply this model to Colombia and find sensitivities higher than the one-for-one results from the regional model—in fact, GDP growth would decline by a cumulative 1.4 percentage points over four quarters in response to a 1 percentage point decline in global growth over the same period.

The results for Latin America from the Bayesian VAR approach are broadly similar to those obtained from a more conventional VAR model described in the April 2007 World Economic Outlook, which included three external variables unaffected by developments elsewhere—growth in the United States, Japan, and the euro area—and three country-specific variables, including growth, inflation, and the real effective exchange rate (IMF, 2007). Using quarterly data over 1991–2005, changes in U.S. growth have almost a one-for-one impact on Latin American growth, with spillovers peaking after one quarter and dying out after three to four quarters.

A second, different way of measuring linkages is by identifying the “common factors” that influence economic growth. These factors may not just be one-off shocks from one country to another, but instead recurring influences that determine the periodic fluctuations of the business cycle that may originate in one country or be common to both. A study in the April 2007 World Economic Outlook, closely related to the Bayesian dynamic factor model of Kose, Otrok, and Whiteman (2003), follows this approach and assesses common global and regional factors influencing growth in output, private consumption, and private fixed investment across 93 countries.

Using a variance decomposition measure for 19 Latin American countries, the influence of global and regional factors for the two decades through 2005 appears to have declined compared with the 1960–85 period. This is consistent with other results that suggest that the common global shocks of the 1970s and the influence of the regional debt crisis in the early 1980s prompted external linkages to strengthen sharply. A second key result is that country factors have tended to be much more important in Latin America than in other regions over the later sample—just over 50 percent of output growth variance is explained by country-specific influences, compared with 31 percent in emerging Asia and Japan and 27 percent in Western Europe.

Aiolfi, Catão, and Timmerman (2006) use a different technique to extract common dynamic factors for four large Latin American countries using a comprehensive set of annual economic data that, in some cases, stretch back to 1870. Using a range of indicators, a country business-cycle indicator was constructed and linkages were then assessed by measuring the proportion of time that two cycles are in the same state—for a sample period dominated by the 1990s, results were diverse across country pairings, confirming the heterogeneity of the region throughout this period.

Interesting regional case studies are provided by Central America. These economies are relatively open and geographically close to the United States, with a number of transmission channels through which U.S. cyclical fluctuations could be transmitted, including trade, the financial sector, and migrant worker remittance flows. Roache (2008) explores the strength of Central America’s linkages by focusing on the extent to which these economies share common trends and cycles with each other and the United States. Using the cofeature method of Vahid and Engle (1993), the paper applies the insights of cointegration to the analysis of stationary, or cyclical, economic data. The results indicate that the Central American business cycle, defined as periodic and transient fluctuations in growth, is dominated by the United States. Indeed, these linkages appear to be much stronger than simple regressions of GDP growth rates would imply. So why does output growth sometimes diverge? This model suggests that region-specific long-run shocks, including civil conflicts, terms of trade shocks and poor policy responses, rather than a unique regional business cycle, are responsible for those periods when growth has diverged from the United States. Kose and Rebucci (2005) use a VAR approach to show that external shocks, on average, are estimated to account for around one-third of total real output variance, close to the levels estimated for Mexico. Multi-country versions of this model suggest that these external shocks are in fact less important than regional (i.e., Central American) shocks, which account for around one-half of output variance.

Future work is likely to sharpen the focus on linkages that run from external financial markets to the real economies of the region. Recent work has touched upon this issue, but two fundamental changes likely imply that the nature of these linkages has changed. First, recent external shocks have emerged after an exceptional period of strength both globally and for countries in Latin America, which is helping contain the impact. In particular, compared with the 1990s, improved public and private sector balance sheets, lower and better anchored inflation expectations, and strengthened policy frameworks have made the region more resilient to changes in international financial conditions. Second, financial linkages have become more complex, moving beyond foreign currency sovereign borrowing in international markets to include increasing corporate financial market activity, local currency debt issuance to foreign investors, and the potential role that remittance inflows might play in the financial system.