本研究探討何種公司特性與投資人行為會產生股票報酬變異數的緩長記憶。從行為財務學的觀點來看，如果市場上雜訊交易者的數量大於理性投資者，股價將明顯遠離基本價值，並導致股價報酬變異數有持續性的現象。投資人傾向低估公開資訊、高估私有訊息，甚至是在樂觀時交易過多的股票，這些行為都將產生報酬變異數有緩長記憶的現象。實證結果顯示大公司、高機構投資人持股比例、低營業活動現金流量分散性、歷史較久的公司、較多分析師研究的公司擁有較高程度的緩長記憶。公司擁有較低的套利風險也會有較高程度的緩長記憶。本研究也檢驗何種公司特性與投資人行為會產生制度上的轉換。結果顯示公司擁有較低的套利風險、高交易量的公司、大型公司、歷史較久的公司、高數量的機構投資人、與高機構投資人持股比例的公司，其兩種狀態的轉換機率較大。由結果可知，導致股價報酬變異數緩長記憶的公司特性、投資人行為，與導致轉換機率較高的公司特性、投資人行為一致。本研究與Diebold和Inoue在2001年的結果一致。 In this paper, we examine the cross-sectional determinants of firm characteristics that explain the long memory in stock volatility. From the view of behavioral finance, the number of noise traders is larger than rational arbitrageurs; the price diverges significantly from fundamental values then drives the volatility persistence. Investors tend to underreact to public information and overreact to private information and trade more when they are optimistic. All of those behaviors drive the phenomena of long memory in return volatility. We find that firms with larger market capitalization, higher percentage of institutional ownership, lower dispersion in cash flow from operations, older history, and wider analyst coverage, have higher degrees of long memory in return volatility. Firms with lower arbitrage risk also have higher degrees of long memory. We also examine regime switching model to see what firm characteristics and investor behavior could force the long memory in volatility. And we find that lower arbitrage risk, lower bid-ask spread, lower stock price, higher trading volume, larger firm, higher number of institution owners, higher percentage of institutional ownership, and older history drive higher transition probabilities in both states. This result is consistent with that of Diebold and Inoue (2001) where when both p00 and p11 are large, it appears that the parameter d of GPH model is away from zero, in other words, drive the volatility persistence.
