動態因子波動度模型與股票預期報酬：建基在無跡卡爾曼濾波分析法與自我相關條件異質變異模型

Dynamic Factors Volatility and Expected Stock Return Based on the Unscented Kalman Filter Method and GARCH Model

王睦舜(Wang, Mu-Shun) ； 林基煌(Chihuang Lin)

《經濟研究》 53卷1期 (2017/01) Pp. 129-179

摘要

本文提出一個動態多因子波動度模型並分別測驗此模型在橫斷面上和時間變異上對價格的影響。動態因子的產生係經由無跡卡爾曼濾波法則(unscented Kalman filter)及一般化自我相關條件異質變異模型，進而討論組合波動度與報酬之間的關係。相對於靜態模型，動態因子係以Fama and French (2015)的五因子模型來進行研究，此一特色包括了市場組合、公司規模、價值效果、營運利潤與投資類型所計算出來的預期報酬。我們提供了事後因子與落後事後因子實證研究，發現在橫斷面迴歸上有助於增進解釋力並提高在樣本內與樣本外預測對組合報酬的預測能力。我們的分析指出公司規模與帳面對市值比的分類組合會產生正面且顯著的風險溢酬，但是以市值來預期的動態因子對組合報酬的影響卻是顯著的負面效果。我們也再執行向前一期的樣本外預測並與其他模型進行配對以比較預測的正確性，此比較採用Diebold-Mariano測試的方法。我們發現動態因子對營運利潤溢酬和投資獲利溢酬會產生顯著的正面效果，此結果更加支持Bali and Cakici (2008)的理論可測性。而且我們的實證結果也進行了穩健性測試，相較於其他不同的模型設定和估計結果都要穩妥。

關鍵字

公司特有風險 ； 樣本外預測 ； 無跡卡爾曼濾波 ； 定價偏誤 ； 動態因子波動度

並列摘要

This paper proposes a dynamic factors volatility model and examines its cross-sectional and time-varying impacts on assets prices. The dynamic factors were extracted by using both the unscented Kalman filtering method and the GARCH model. We then investigated the relationship between the volatility of the portfolio dynamic factors and the portfolio returns. In contrast to the static model, the dynamic factors incorporate features of Fama and French 5-factor models, which capture the characteristics of market portfolio, size, value effect, operating profitability, and investment patterns in average returns. We provide evidence that the ex post and lagged ex post factors can improve cross-sectional explanatory power and can increase the predictability of portfolio returns both in-sample and out-of-sample. Our analyses demonstrate that the size and book-to-market sorted portfolios earn a significant positive variance risk premium. However, the dynamic factors predicted by market value have significant negative effects on portfolio returns. We also perform the one-prior-period out-of-sample Diebold-Mariano test of forecasting accuracy paired with other models and find that the dynamic factors have significant effects on the risk premia of both operating profitability and investment policies. This finding supports the theoretical prediction of Bali and Cakici (2008). Finally, our evidence is robust to various specifications and estimation results.

並列關鍵字

Idiosyncratic Volatility ； Out-of-Sample Predictability ； Unscented Kalman Filter ； Pricing Bias ； Dynamic Factors Volatility

參考文獻

Kan, R. and G. Zhou (1999), “A Critique of the Stochastic Discount Factor Methodology,” The Journal of Finance, 54:4, 1221-1248.

陳家彬、劉映興、楊踐為Chen, Chia-Pin, Ying-Sing Liu and Jack J. W. Yang (2010)，「三因子模型在多頭月及空頭月之條件異質性與雙重時間變動貝他值」“Conditional Heteroscedasticity, Dual Time-Varying Betas in Bull and Bear Months of the Three-Factor Model”，證券市場發展季刊 Review of Securities & Futures Markets，22:1，1-27。(in Chinese with English abstract)

Adrian, T. and J. Rosenberg (2008), “Stock Returns and Volatility: Pricing the ShortRun and LongRun Components of Market Risk,” The Journal of Finance, 63:6, 2997-3030.

Aharoni, G., B. Grundy and Q. Zeng (2013), “Stock Returns and the Miller Modigliani Valuation Formula: Revisiting the Fama French Analysis,” Journal of Financial Economics, 110:2, 347-357.

Ang, A. and G. Bekaert (2007), “Stock Return Predictability: Is It There?” The Review of Financial Studies, 20:3, 651-707.