| 摘要: | 本研究旨在探討美國股票市場本益比（Price-to-Earnings Ratio, P/E）結構性上升的潛在驅動因子，並進一步利用該趨勢建構具備預測與操作邏輯的量化交易策略。近年來，標普500指數（S&P 500）的整體本益比水準已持續高於歷史平均，儘管企業盈餘增長未必同步提升，仍未見估值水準大幅回落的現象。此一現象引發市場對於「估值新常態」的廣泛討論，亦成為本研究之切入點。
為解釋本益比持續上行的邏輯，本研究首先回顧並整合多篇文獻，並提出三項主要假設作為研究核心：一、在低利率環境下，折現率下降將推升資產估值；二、穩定的通膨率能降低市場的不確定性與風險溢酬，使估值上行具備更高的可持續性；三、標普500指數成分股的汰換機制存在偏好高本益比產業的趨勢，導致整體指數估值結構逐步抬升。為驗證上述假設，本文採用2011年5月至2025年4月的月度資料進行實證分析。
進入到主要模型，本文設計兩種預測模型以捕捉本益比之趨勢：未加權模型為一個加入MA結構處理時間序列的自我相關性的回歸模型，以時間變數t解釋P/E的長期變動趨勢；加權模型則在未加權模型基礎上引入VIX波動率指數作為市場恐慌程度的權重調整因子，以反映情緒面對於估值的扭曲影響。兩套模型皆進行共計168期（14年）之訓練與預測，並繪製趨勢線與上下界。
進一步地，本文根據預測值與實際P/E之間的偏離程度，設定多組不同標準差倍數(k)作為上下界，並採用了「確認機制」的訊號邏輯，當實際P/E突破上下界並反轉再次穿越上下界即產生交易訊號，並依據固定持有期間策略進行多空回測，評估模型與策略之整體報酬表現。
結果顯示：第一，兩種模型中時間變數對P/E皆呈顯著正向關聯，支持估值長期上升趨勢的存在；第二，回測期間內加權模型在趨勢預測之精準度與穩定性表現優於未加權模型；第三，雖然整體樣本數有限，但部分參數組合已呈現出明確的報酬潛力，證明本研究設計具備可行性與延伸價值。
本研究之貢獻在於將總體經濟背景與市場行為理論整合至估值變動與交易策略設計中，亦為後續應用日資料、結合技術指標與風控機制發展實用型策略提供重要基礎。
;This study aims to explore the potential drivers behind the structural upward trend in the Price-to-Earnings Ratio (P/E) of the U.S. stock market and to develop a quantitative trading strategy based on this trend with predictive and operational logic. In recent years, the overall P/E ratio of the S&P 500 Index has remained well above its historical average. Despite the fact that corporate earnings growth has not necessarily kept pace, valuation levels have not significantly reverted. This phenomenon has sparked widespread discussion about a possible “new normal” in valuation, which serves as the point of departure for this research.
To explain the persistent upward movement of the P/E ratio, this study reviews and integrates relevant literature and proposes three core hypotheses: (1) In a low-interest-rate environment, lower discount rates drive asset valuations higher; (2) Stable inflation reduces uncertainty and risk premiums, supporting a more sustainable rise in valuations; and (3) The S&P 500’s component rebalancing mechanism tends to favor high-P/E growth industries, gradually lifting the overall index valuation. To validate these hypotheses, we conduct an empirical analysis using monthly data from May 2011 to April 2025.
The core of this research involves constructing two forecasting models to capture the P/E trend. The first, an unweighted model, uses a regression framework with a moving average (MA) structure to address autocorrelation, explaining long-term movements in P/E via a time variable. The second, a weighted model, incorporates the VIX index as a sentiment-based weighting factor to adjust the forecasted trend, thereby capturing distortions in valuation arising from market fear. Both models are trained and tested over 168 monthly periods (14 years), generating trend lines and prediction bands.
Based on the deviation between the predicted and actual P/E, we set multiple standard deviation multipliers (k) to define upper and lower bounds and employ a “confirmation mechanism” to generate trading signals: a signal is triggered when the P/E crosses and reverses across a bound. A fixed holding period strategy is then applied to backtest both long and short trades and assess overall model performance.
The results indicate the following: (1) In both models, the time variable has a significantly positive relationship with P/E, supporting the existence of a long-term upward trend in valuation; (2) During the back testing period, the weighted model outperforms the unweighted model in terms of trend forecasting accuracy and stability; (3) Although the sample size is limited, certain parameter combinations demonstrate strong return potential, proving the feasibility and extensibility of the model design.
The key contribution of this study lies in integrating macroeconomic and behavioral finance theories into valuation modeling and trading strategy design. Furthermore, it provides a practical foundation for future research incorporating daily data, technical indicators, and risk management systems to develop more robust investment strategies. |
