Forecasting emerging market volatility in crisis period: Comparing traditional GARCH with high-frequency based models

Abstract

This chapter discusses the topic of modeling and forecasting volatility in emerging market and presents the strength and weakness of the several highfrequency based approaches available in the literature. We compare the forecasting performance of traditional GARCH with high-frequency based models namely, HAR-RV, HAR-RV-J, and HAR-RV-CJ under the financial crisis and non-financial crisis periods. We extend our study scope by focusing not only on general market index BIST-30, but also on each constituent of market index. Our empirical results indicate that the global financial crisis does not affect the forecasting performance of the models in emerging markets. All high-frequency based volatility forecasting models perform better than the traditional ARCH-class models in both non-crisis and crisis periods. We conclude our paper with the statement that high-frequency based models do not affect the structural break in the underlying process. The best outperforming model among the high-frequency based volatility models for both stable and turmoil period is HAR-RV-CJ model. The empirical findings for the individual stocks are consistent with the general market index ISE-30.