| This study investigates the long memory property in the volatility of the stock market indices of the six largest and least restrictive Latin American markets, as well as that of the United States. The application of the long memory model, associated with fractional integration via the operator (1 − L)d and noninteger d, allows flexible modeling of low-frequency behavior with important implications for the measurement of long memory in the volatility. Evidence of long memory in volatility is found in the stock markets for Argentina, Brazil, Chile, Mexico, Venezuela, and the United States, and not for Colombia. These findings indicate that Colombia's stock market volatility is stationary, implying that, external shocks in these markets have a short-term impact, but little long-term effect, as the volatility reverts to its mean at an exponential rate. On the other hand, the volatility of the stock markets in Argentina, Brazil, Chile, Mexico, Venezuela, and the United States is fractionally integrated, implying that these stock markets volatilities do no return to the previous means after an external shock has been felt.; For the second part of this study, a generalized notion of cointegration, called fractional cointegration, is used to examine the empirical relevance of the long run dynamics between stock market indices of the six Latin American countries and the United States. The empirical work uses weekly data ranging from December 1988 to November 1998 from Argentina, Brazil, Chile, Colombia, Mexico, Venezuela, and the United States. The analysis of fractional cointegration allows the equilibrium error to follow a fractionally integrated process and avoids the stringent I(1) and I (0) distinction maintained in previous empirical work. By allowing for fractionally integrated equilibrium errors, it provides a flexible and parsimonious way to model the low-frequency dynamics of deviation from equilibrium. I find that the stock market indices form a fractionally integrated system. Even though each stock market index is best characterized as a unit root process, the system of stock market indices appears to possess a common fractional nonstationary component. Shocks to the system exhibit significant persistence in the short run, but they eventually dissipate so that an equilibrium relationship is obeyed in the long run. Close examination of the long run behavior of the Latin American Stock markets with and without the United States stock market identify a strong co-movement between these markets.; With respect to stock market volatility, I find that the system of stock market volatilities is not cointegrated. This absence of cointegration implies that each stock market volatility follows its own set of fundamentals, that is, the volatility of each stock market is related to local political and economic events. However, results from the VAR analysis imply that, given the stochastic environment in the region due to the Mexican crisis of December 1994, the degree of co-movement in stock market volatility among the Latin American countries and the United States has increased significantly. |
