頁籤選單縮合
題 名 | 用樣本熵捕捉市場操縱之行為--臺灣股市的實證分析=How Useful is Sample Entropy in Detecting Stock Price Manipulation in Taiwan? |
---|---|
作 者 | 葉偉凡; 林金龍; | 書刊名 | 中國統計學報 |
卷 期 | 51:4 2013.12[民102.12] |
頁 次 | 頁467-499 |
分類號 | 563.54 |
關鍵詞 | 市場操縱; 交易類型的市場操縱; 樣本熵; 隨機度; 臺灣股票市場; 法院判例; 蒙地卡羅分析; 時間序列; Trade-based stock market manipulation; Sample entropy; Approximate entropy; Randomness; Taiwan stock market; Monte Carlo experiment; time Series analysis; Replication principle; Taxonomic diversity; |
語 文 | 中文(Chinese) |
中文摘要 | 市場操縱可概分為三種類型: 資訊類型的市場操縱、行動類型的市 場操縱, 交易類型的市場操縱。每一種市場操縱都會使得股票價格偏離 真實價值, 但尤其是交易類型的市場操縱可以在不需要未公開資訊下賺 取超額的利潤, 犯罪行為的認定上有其不易之處, 故吸引產官學界的重 視。若干既有過去有許多研究指出市場操縱會導致股市價格、交易量的 走勢變得更加規律, 故在上述假設下, 可以利用樣本熵測量序列隨機度 的改變, 進而發現市場操縱的行為, 維護市場交易的公平性。 本文檢驗樣本熵是否適合作為臺灣交易類型市場操縱的警示指標。 除了蒙地卡羅模擬分析外, 研究樣本包含了民國九十一年至民國九十五 年間經台灣各級法院確定判決之案例, 並以各類股中不易發生操縱情事 的個股資料作為對照樣本。模擬分析發現樣本熵不會因母體模型標準差 放大而改變, 卻會因序列的極端值而快速下降。此外, 在不同模型參數 下, 樣本熵的估計仍具有一致性。實証研究則發現序列發生結構轉變時, 也就是操縱期間第一個交易日或最後一個交易日偵測的結果較佳, 但綜 觀所有案例並未產生具一致性的結果, 操縱手法的不同或是股市的消息 面因素皆影響了對市場操縱的判斷。本文模擬與實證分析的結論為, 雖 然不宜將樣本熵作為偵測市場操縱的唯一工具, 但許多成功的案例分析 顯示, 搭配其他資訊, 樣本熵可成為一個良好的輔助偵測系統。 |
英文摘要 | Understanding and detecting market manipulation have long been the most important task for stock market governance and have attracted a great deal of attentions from the academics. Numerous methods and models have been proposed for this purpose. Among them are some recent attempts to use sample entropy (SampEn) to detect stock market manipulation. It is believed that trade-based manipulation introduces more regularity and less randomness into intraday price and volume. SampEn calculates the probability that epochs of window length m that are similar within a tolerance r remain similar at the next point. There are several studies of the usefulness of SampEn in detecting stock market manipulations using data in US and other European countries, and the empirical results are mixed. While some applications show positive evidences but some others don’t. This paper examines the applicability of SampEn in detecting stock market manipulation in Taiwan. We have analyzed numerous cases identified as being manipulated by the government officials and the court. Also included are several stock return series for which manipulations are deemed unlikely and then used as control samples. To better understand the properties of SampEn, we perform some Monte Carlo simulation experiments with underlying processes ranging from white noise, autoregressive process, random walks, and threshold autoregressive process. Our analysis finds significant drop of randomness of stock return at the first and last day of the manipulation period for most cases but there are a few exceptions. News events might temporally lower randomness that makes detecting manipulation more difficult. To conclude, our empirical analysis and Monte Carlo studies confirm some usefulness of SampEn in detectingmanipulation which could become more useful when combining with other methods. |
本系統中英文摘要資訊取自各篇刊載內容。