本研究使用臺灣證券交易所之日內資料探討上市公司股票投資報酬率與知訊交易機率間的關係,由於頻繁的買賣交易次數會使得在極大化最大概似函數過程中出現浮點異常的現象,估計偏誤可能會影響知訊交易機率對報酬的解釋力。本研究計畫最主要目的有三個,首先是再次檢驗知訊交易機率PIN(Probability of Informed trading )對資產定價能力,探討PIN對股價報酬的解釋力,第二個為探討估計PIN時產生的浮點運算偏誤FPE(float point exception)對資產定價的影響,最後一個為探討估計PIN時產生的浮點運算偏誤與定價因子間的關係。研究結果顯示PIN在台灣交易市場的資產定價能力並不穩定,而估計偏誤對PIN的定價能力是會有影響的,但估計偏誤本身並不具定價能力,而其他定價因子中,只有市場風險因子貝他(β)有受到估計偏誤的影響,本研究推論台灣知訊交易機率估計偏誤對股價報酬率及其他因子無顯著影響,原因在於台灣交易市場成交量較小,因此偏誤不明顯,但我們預期市場會隨時間持續發展,交易量會持續成長,故日後在使用知訊交易機率評價時,都應考慮估計偏誤的影響。
This study uses the intraday data of listed companies on Taiwan Stock Exchange to test the relationship between stock returns and the probability of informed trading (PIN) with and without the bias caused by the float point exception (FPE). During the maximization of the PIN likelihood function, the high trading frequency may cause the FPE, which lead to the underestimated PIN (i.e. FPE bias). This estimated bias may affect the pricing ability of PIN. There are three main objectives in this research. First, this study tests the ability to asset pricing of PIN and its explanatory power for stock returns. Second, this study explores how the FPE bias impacts the asset pricing. Last, this study tests the influences between FPE bias and other pricing factors. The results show that the ability to asset pricing of PIN is not stable in Taiwan market, and that the FPE bias affects the pricing ability for PIN, and that only market beta (βp), one of the pricing factors, is affected by the FPE bias. Although this study shows that PIN is not a stable pricing factor, this study also shows that the FPE bias in PIN estimates little impacts the assets pricing in Taiwan Stock Market. The reason may be that the trading volume is too small in current Taiwan market so the estimated bias is not obvious. We expect the market will continue to develop over time, and transaction volume will continue growing, so the impact of the FPE bias in PIN estimates should be considered if PIN is adopted in the future research.