建構技術分析危機預警條件預測股市泡沫與均數復歸研究

回顧383年金融泡沫史，每次泡沫破裂都造成金融風暴和嚴重的經濟衰退，本文試圖運用技術分析的指標過熱且正乖離過大，建構股市泡沫的危機預警條件：隨機指標(KD)的K值和相對強弱勢指標(RSI)的6RSI值>90且乖離率位於5%～15%，作為股市泡沫之預測工具，來預測股市泡沫，期望在泡沫破裂前提早發現，降低金融風暴的衝擊。現以已開發市場中的美國、日本和德國股市，以及新興市場中的中國、巴西和南韓股市(1995-2019年)為實證對象，用技術指標的危機預警條件預測泡沫和均數復歸的時間分布。從實證結果發現，已開發市場的波動性較小，調降預警條件可提高泡沫預測準確率，其泡沫破裂時間較長約6個月而新興市場的波動性較大，調升預警條件可提高其泡沫預測準確率，其泡沫破裂的時間較短約3個月內；不論基準泡沫預警條件或調整後較佳的泡沫預警條件，都能發揮泡沫的預警作用。2020年初美股引發全球股市泡沫，在泡沫破裂前，那斯達克和標普500指數皆符合本研究的泡沫預警條件，即時且準確地發揮了預警功能，成功避開股市泡沫。另外，我們從全球主要股市驗證結果得知，泡沫破裂的時間幾乎等於均數復歸的時間，全球41個主要股市中有高達31個股市的均數復歸時間為0週，比例高達75.61%。由驗證結果推論得知，均數復歸的時間有兩種：泡沫破裂點到均數復歸的時間約為0～1週。泡沫預警點到均數復歸的時間約5週至10週之內。

關鍵字

泡沫；技術指標；乖離；移動平均線；均數復歸

並列摘要

Looking back on the history of the financial bubbles in 383 years, each bubble burst caused a financial turmoil and severe economic recession. This paper attempts to use technical analysis indicators to construct a crisis warning condition of the stock market bubble: The K value of the Stochastics Oscillator (KD) and the 6RSI value of the Relative Strength Index (RSI) are both greater than 90, and the Bias is between 5% and 15%. It is hoped that it can be used as a tool for predicting financial bubbles in the stock market to reduce the impact of financial turmoil. The U.S., Japan and Germany stock markets in the developed markets, and China, Brazil and South Korea stock markets in the emerging markets (1995-2019) are the empirical objects. The crisis warning conditions of the technical indicators are applied to predict the financial bubbles, and the time distribution of the mean reversion. From the empirical results, it is found that the volatilities of the developed markets are smaller, and thus lowering the warning conditions can improve the accuracy of bubble prediction, and the bubble burst time is about 6 months. The emerging markets are more volatile, and raising the warning conditions can improve the accuracy of bubble prediction, and the bubble burst time is shorter in about 3 months. Regardless of the baseline bubble warning conditions or the adjusted bubble warning conditions, the bubble warning function can be useful. At the beginning of 2020, US stocks triggered a global stock market bubble. Before the bubble burst, the Nasdaq and S&P 500 index both met the bubble warning conditions of this study. Therefore, the warning conditions immediately and accurately predicted this financial bubble. In addition, we know from the empirical results of the world's major stock markets that the bubble burst time is almost equal to the mean reversion time. Among the 41 major stock markets in the world, the average reversion time of up to 31 stock markets is 0 weeks, with a proportion as high as 75.61%. It is inferred from the empirical results that there are two kinds of the mean reversion: The time from the bubble burst point to the mean return is approximately 0 to 1 week. The time from the bubble warning point to the mean return is about 5 weeks to 10 weeks.