The first essay examines which volatility estimation model best explains KOSPI200-realized volatility in the Korean stock market, which has both heteroscedasticity and jump risk. The sample covers from July 1, 2010 to July 31, 2014, which is a low-vol...
The first essay examines which volatility estimation model best explains KOSPI200-realized volatility in the Korean stock market, which has both heteroscedasticity and jump risk. The sample covers from July 1, 2010 to July 31, 2014, which is a low-volatility period in Korean stock market by which time the effects of the global crisis had almost vanished. We use the intra-day return of KOSPI200, which has been measured by 5-minute intervals. This study finds GARCH-family models are efficient estimators compared to historical volatility and EWMA. Also, among the GARCH-family models, Jump-Diffusion GARCH has shown comparatively good results. Especially this study finds that VKOSPI200 is the most efficient model with the largest adj. R2 and the smallest evaluation statistics during the sample period. Meanwhile, it seems to be necessary to consider jump risk when we estimate volatility in Korean stock market.
The second essay examines whether REIT has heteroscedasticity and jump risk, which are observed in general financial time series data, and the relationship between REIT volatility and the stock market. To achieve these objectives, we have used the Component-Jump GARCH model. Using the FTSE NAREIT All Equity REIT Index from January 4, 1999 to August 31, 2015, this paper finds that REIT has both heteroscedasticity and jump risk, and heteroscedasticity is decomposed into permanent component and transitory component. Jump risk, particularly, comes every 19 trading days in the REIT market, with the estimated jump frequency parameter being 0.0519. This implies that REIT could have market jump or crash risk like common stock. Moreover, REIT total volatility is well explained by the S&P1500 permanent volatility component and approximately 10% of REIT jump component is explained by the S&P1500 jump component.
Using monthly return data from 143 actively managed U.S. domestic real estate mutual funds over a recent 20-year period (January 1996 – December 2015), the third essay investigates whether real estate mutual fund managers possess market-timing abilities. To examine whether there is a difference between survivor funds and non-survivor funds regarding market-timing abilities, we categorize according to the existence of a fund as of December 2015 and analyze the market-timing abilities at the portfolio and individual levels. Based on a comprehensive data set, we use a bootstrap methodology to distinguish between ‘skill’ and ‘luck’ in market-timing abilities for individual funds. We find that, on average, U.S. domestic real estate mutual funds do not have market-timing abilities at the portfolio level. However, a small number of U.S. domestic real estate mutual funds have positive market-timing abilities at the individual level. According to the results of the cross-sectional luck distribution, we find that market-timing abilities in regard to U.S. domestic real estate mutual funds are from 'skill' rather than 'luck'.