VaR of SSE returns based on Bayesian markov-switching GARCH approach

Abstract

© 2019 Association for Computing Machinery. This study compares the accuracy of the single-regime and two-regime Bayesian Markov Switching GARCH models, in the forecasting the Value-at-Risk (VaR) of Shanghai Stock Exchange (SSE). The research addresses the question of whether considering the structural change for stock markets with high volatility improves the accuracy of the forecasting VaR. To take account of regime changes in stock market, we employ Markov-switching model with GARCH model. Regarding to DIC model selection, two-regime GJR model with Student-t distribution is chosen indicating that it is the best-fitted to the data. The estimated results confirm that the two-regime switching models beat the single regime switching model in forecasting VaR of SSE. Thus, the Markov switching model with GARCH model appears to improve the VaR forecasting of SSE.