As the economic and financial characteristics of countries change, so would be their betas and correlations of their investment returns with that of the U.S. Such changes are expected to be particularly significant for emerging market nations as they strive for rapid industrialization and modernization. OLS estimator for the beta coefficient would not be the Best Linear Unbiased Estimator (BLUE) if beta is non‐stationary or changes from period to period. This paper proposes a special type of time weighted least square method (TWLS), which assigns greater weights on the regression errors in more recent periods, for estimating the current beta. This TWLS approach can tackle the problem of intertemporal heteroscedasticity and thus yields a beta that is more efficient. The breakthrough lies on the viability of the method without a‐priori knowledge or estimation of the values of the weights. This yields a significant practical advantage since the weights are unobservable in the real world. Since the Time Weighted Method estimator is the coefficient estimator of beta value for the latest period in the sample, statisticians who base their forecasts on the beta estimates derived from the Time Weighted Least Square can expect to outperform those relying on beta values obtained from conventional estimation. We use a sample of daily returns of thirty‐one emerging markets stock over the period of January 1, 2000 through December 31, 2002. We find that most of the tstatistics for the variances are significant at the 95 per cent level, indicating that the Var(s)’s are not zero for nearly every emerging‐markets. This implies that the betas for these markets do shift over time.
CitationDownload as .RIS
Emerald Group Publishing Limited
Copyright © 2004, Emerald Group Publishing Limited