風險、報酬的抵換關係與波動性預測-臺灣股市的實證研究

Revisiting the Risk-Return Trade-off and Volatility Forecast─Evidence from Taiwan Stock Market

喬慕恩(Mu-En Chiao)

指導教授 ： 林金龍

國立臺灣大學/社會科學院/經濟學研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.10360

摘要

本研究應用 Ghysels et al. (2004)的 MIDAS 模型分析台灣的股市。分析分為兩個部分，第一部分以日報酬，以及使用日內高頻資料的估算的已實現波動度檢驗台灣股市的預期風險與預期報酬之間是否存在正向的抵換關係；第二部分則比較幾個使用高頻資料的已實現波動 (Realized Volatility) 估計子，對於未來一周波動率的預測能力。MIDAS模型的優勢，在於可用涵蓋更多資訊的高頻率資料來預測頻率較低、期間較長的目標，這對於估計一周到一個月內的風險報酬替換關係，以及一周以上的波動度預測，都比傳統必須使用相同頻率資料的計量模型擁有更好的估計能力。
使用2002年一月到2010年三月的台灣加權股價指數日資料，以及每分鐘紀錄一次的高頻率日內交易資料，迴歸結果顯示，使用日內高頻資料的估計子，發現台股存在風險與報酬的替換關係，使用日資料的結果則欠缺統計顯著性。此外截距項普遍顯著不為零的結果，顯示對彌補風險的要求，並非影響台股預期報酬的唯一動力。第二部分的研究顯示30分鐘的已實現波動估計子具有最好的波動預測能力，其次為針對市場微結構噪音做調整的已實現波動估計子，這個結果隱含台灣股市的微結構噪音可能並非IID，值得採取進一步的分析。

關鍵字

MIDAS ； risk-return trade-off ； realized volatility

並列摘要

This paper applies the MIDAS framework developed in Ghysels et al. (2004) to analyze the Taiwan Stock Exchange (TAIEX) data. The research project is divided into two main parts: First, daily returns and estimates of realized volatility calculated from high frequency intraday data are analyzed to see if there exists a risk-return tradeoff. Second, several estimators of realized volatility calculated from high frequency data are compared on their one-week ahead volatility forecast precision. It is the advantage of the MIDAS framework to use time series data of varying frequencies to estimate a variable with lower frequency or longer period. This property makes MIDAS better than traditional models for estimating one week to one month risk-return tradeoffs and one week or longer volatilities.

The data used are daily returns and one-minute high frequency price series of TAIEX index spanning January 2002 to March 2010. Using the high frequency intraday data, the regression results show that Taiwanese stock market do in fact exhibit a risk-return tradeoff. The daily data however fail to give a statistically significant result. In addition, the intercept of the former regression is statistically non-zero, suggesting that risk compensation is not the only driver of Taiwanese stock returns. The results of the second part show that the 30 minute realized volatility estimator has the greatest predictive power while the noise-corrected realized volatility estimators was second best. This may imply that the microstructure noise in the Taiwanese market is not IID and warrant further investigation.

並列關鍵字

MIDAS ； risk-return trade-off ； realized volatility

參考文獻

Andersen, T. and T. Bollerslev (1998). "Answering the skeptics: Yes, standard volatility models do provide accurate forecasts." International Economic Review 39: 885-905.

Andersen, T. G., T. Bollerslev, et al. (2001). "The distribution of realized stock return volatility." Journal of Financial Economics 61.

Andersen, T., T. Bollerslev, et al. (2003). "Modeling and forecasting realized volatility." Econometrica 71: 529-626.

Baillie, R. and R. P. DeGennaro (1990). "Stock returns and volatility." Journal of Financial and Quantitative Analysis 25: 203-214.

Bandi, F. M. and J. R. Russell (2006). "Separating market microstructure noise from volatility." Journal of Financial Economics 79: 655-692.