Time horizon volatility forecasting of Malaysian property stocks

Reliable and accurate forecasts can provide important input for fund manager and policymakers to make an informed decision. However, volatility forecast research is still bound by several weaknesses such as scarcity in volatility forecasting literature and the lack of knowledge on the contributing factors to poor forecast, i.e. time-varying series characteristic or model specification. As a result of inaccuracy in forecasting, fund managers could face catastrophic consequences. The first contribution is to prove there are ‘parameter changes (time-varying) in Generalized autoregressive conditional heteroskedasticity (GARCH) model before and during the GFC in Malaysian property stocks. News impact curve (NIC) is adopted to show how the good news and bad news impact (news shock) on the next period’s volatility forecast in these two periods. Findings show that parameters and NICs are changes in both periods, this may incur poor forecast. To further validate the parameter changes in different periods. Second contribution adopted and adapted news impact curve (NIC) for different models in different periods. Adaptive asymmetric Smooth Transition Exponential Smoothing (STES) is reported to be more pragmatic and superior to symmetric model in volatility forecasting. Overall, NIC for the symmetric GARCH model shows the news shock on next volatility estimates during crisis is the highest. NICs for the asymmetric GJR GARCH model and STES-E+AE indicate bad news has higher impact on next period’s volatility forecast during crisis period. The study furthered on the volatility forecasting of STES method (the models are STES-E, STES-SE, STES-ESE, STES-AbsE and STES-E+AE) as compared with other models (total thirteen models) in short-time horizon. The third contribution is to study the performance of STES methods in forecasting the Malaysian property stocks volatility compared to various forecasting methods before, during and after global financial crisis (GFC). Surprisingly, the performance of STES is very encouraging. A model performs well in short-time horizon data may not perform well in long-time horizon data. The fourth contribution is to further investigate the performance of STES method in the long-time horizon. Compared to 18 months data used in the previous section, study employed 2000 daily returns (8 years data) of 33 Malaysian property stocks in this study. The result shows that STES method is still the best method as compared with its competitors such as GARCH family models. Hence, study concludes that STES method outperforms other forecasting methods in forecasting the short and long-time horizon volatility of Malaysian property stocks. Time series data often sampled at a different frequency. It is a dilemma (regression must be at the same frequency) faced by many researchers. MIDAS methods enable different frequency data being used to estimate together. The fifth contribution is investigating the relationship between the house price index (HPI) volatility (quarterly data) and property stock index (PI) volatility (daily data) using MIDAS approach. Modelling and forecasting performance of MIDAS with different weighting functions. The results show there is a negative relationship between HPI volatility and PI volatility indicating that investors can reduce their portfolio’s risk by pairing these assets.

Text SPE 2020 6 IR.pdf Download (1MB)

Actions (login required)

View Item