在不同經濟狀況下應用迭代式組合模型於股票市場的風險溢酬與條件波動率之預測

Applying Iterated Combination Approach to Predict Stock Market Risk Premiums and Conditional Volatility under Differen t Economic Conditions

本研究旨在解決過去文獻中提出的兩個主要問題。首先，儘管過去研究顯示許多模型在樣本外達到統計顯著性，但往往無法獲得顯著的已實現效用收益。本研究通過結合大量預測因子並應用迭代式組合迴歸，以預測股票市場風險溢酬與條件波動率。其次，模型的預測能力在不同經濟狀況下存在顯著差異，景氣循環會提高模型預測能力的不穩定性。鑑於此，本研究提出雙重狀態預測迴歸模型，在模型內納入總體經濟指標的虛擬變數與預測因子的交乘項來控制市場變化。實證發現，該模型在不同預測期間下均獲得顯著的樣本外預測力，相較於歷史平均法可獲得超過4%的效用收益，大幅提升資產配置績效，具有顯著的經濟意義。

This study aims to address two main issues raised in the literature. First, while numerous past research has shown that many models can exhibit out-of-sample statistical significance, they often fail to achieve significant realized utility gains. This study combines a large set of predictors and applies an iterated combination approach to predict stock market risk premiums and conditional volatility. Second, the predictive performance of forecasting models varies significantly under different economic conditions, with business cycles increasing the instability of model predictions. To tackle this challenge, we propose a twostate predictive regression model. This model incorporates the interaction terms between macroeconomic indicators and predictors to control for market variations. Empirical results show that the model achieves significant out-of-sample predictive power over different forecasting horizons and obtains utility gains exceeding 4% compared to the historical average model, substantially improving asset allocation performance and demonstrating significant economic value.