以追蹤資料分量迴歸方法衡量臺灣股市預期報酬與風險關係

Abstract

本文根據跨期資本資產定價模型(Intertemporal Capital Asset Pricing Model, ICAPM)探討臺灣股票市場預期報酬與風險之跨期關係。在實證模型中使用臺灣50指數報酬、其各成份股報酬以及影響投資人未來投資機會集合的狀態變數(State Variable)間隨時間改變的共變異數。我們考慮的狀態變數有vix指數、期間利差及融資流動性風險利差。在研究方法上，我們先使用動態條件相關模型(Dynamic Conditional Correlation, DCC)估計各項條件共變異數作為解釋變數，再由各成份股預期報酬作為被解釋變數以追蹤資料分量迴歸(Panel Quantile Regression)得出相對風險趨避係數估計值。實證結果顯示，在預期報酬條件分配的右尾，風險與預期報酬呈現正向關係；預期報酬條件分配的左尾，風險與預期報酬為負向關係；而在預期報酬條件中位數附近，則顯示風險與報酬無顯著關係。強韌性檢驗(robustness checks)顯示估計結果並不會因為解釋變數不同、風險替代變數相異及使用計量方法不一而有太大的差異。

This paper explores the Taiwan stock market’s intertemporal relation between risks and expected returns in the context of Intertemporal Capital Asset Pricing Model (ICAPM). Our panel data models mainly rely on the time-varying conditional covariances among the return of Taiwan 50 Index (market portfolio), each corresponding component stocks returns, and the state variables including VIX, term spread, and funding liquidity spread. Specifically, the following two-stage econometric procedure are implemented: we first estimate the time-varying conditional covariances by Dynamic Conditional Correlation (DCC) models, and then treat the estimates as explanatory variables in the second-stage panel quantile regression (PQR) methods to explore the shape of conditional distribution of excess returns. The risk coefficients estimated via PQR are positive over the upper right tail of the conditional distribution of excess returns; the estimation results signify negative risk coefficients over the lower left tail of conditional distribution of excess returns. No significant intertemproal relation between risk and return are identified over the neighborhood of conditional median of excess returns. Robustness checks indicate that our empirical results are robust to the choice of proxies of risk, explanatory variables, and econometric methodologies.