The microeconomic evidence on contagion, capital controls, and capital flows.

• Author: Forbes, Kristin J.
• Position: Research Summaries

On December 19, 2006, Thailand implemented restrictive capital controls on foreign investment and the Thai stock index fell by over 15 percent before trading was halted. Investors quickly began to draw comparisons to the summer of 1997, asking if events in Thailand would spark contagion and be the first in a series of crises around the globe. My research over the past few years has attempted to provide the framework and analysis to understand these types of events--covering topics from financial contagion and global linkages to capital controls and capital flows.

Many researchers--including other NBER affiliates--have made important contributions to these topics. Much of my own research has tried to take a different approach by using firm-level data to understand the macroeconomic movements in exchange rates, stock indices, investment, and growth. This "microeconomic approach" to answering questions in international finance has the benefit of using the wealth of information incorporated in firm behavior--information that is lost in the aggregation process used to create macroeconomic statistics. Heterogeneity across firms can be an important tool in identifying the impact of various macroeconomic events. This strategy of using disaggregated and firm-level data in international economics recently has become popular in the international trade literature (1) as well as in international finance. (2)

Contagion, Currency Crises, and the Cross-Country Transmission of Shocks

A series of financial crises--Mexico, Asia, Russia, Argentina--motivated an academic literature on "contagion" and the international transmission of crises. (3) The term contagion is generally used to refer to the spread of negative shocks--although the definition has evolved over time. (4) The last decade has dearly shown that crises that originate in relatively small economies (such as Thailand) can quickly affect markets of very different sizes and structures located around the world, including markets that appear unrelated to the country where the crisis originated.

Early analyses of contagion tested for increased comovement between countries after a crisis (in variables such as their stock returns, bond spreads, exchange rates, or capital inflows). One complication with this approach, however, is that the correlation coefficients underlying this analysis depend on market volatility and can be biased. My work with Roberto Rigobon shows how this bias can significantly affect estimates of contagion. (5) We develop a correction for this bias and show that most recent crises were transmitted to other countries through linkages that exist in all states of the world--and not through special transmission channels that only occur during crises. Other authors have since used different idenrifying assumptions in order to adjust for this bias in tests for contagion. (6)

This approach of testing how crises affect cross-country co-movements, however, has a major shortcoming: it cannot explain why these macroeconomic variables co-move or exactly how shocks are transmitted internationally. For investors interested in how a crisis spreads, and especially for governments and policymakers that would like to contain and prevent crises, understanding exactly how shocks spread is of critical importance. Do crises spread mainly through "real" linkages, such as trade and banking flows? Or, do they spread through investor behavior, driven by portfolio balancing or informational asymmetries or herding or irrationality? Sorting out these various explanations is further complicated by the fact that many cross-country linkages are highly correlated, so it is difficult to identify these various mechanisms in empirical work.

In order to differentiate between these various mechanisms, my research moved from the macroeconomic to the microeconomic level. Within each country there is a large variation in how different companies are affected by shocks. By using this...