本研究旨在探討臺灣上市公司績效之預測與股價之反應。首先利用多變量區別分析、Logit分析及Probit分析等方法建立區別模式，再以市場模式驗證績效良好及不好等兩群公司的超常報酬情形，最後則以人員訪設及問卷調查方式探討股票投資人的投資決策依據。研究結果發現： (1) 多變量區別模式對原始樣本的區別能力最佳，幾乎都在70%以上，達5%顯著水準，此意味著：前期44家樣本公司中，至少有31家可被正確分類；然而，該模式對後期樣本的預測能力卻未達60%。此外，Logit模式及Probit 模式兩者效果皆不佳。 (2) 在15個月的觀察期間中，模式預測為績效良好的公司群，其累積超常報酬率有10個月為正值，而預測為績效不好的公司群則有11個月為負值；惟大部分與零並無顯著差異； (3) 由9份法人機構投資者及53份小額投資人的回函中可看出，股市投資人對上市公司財務資料及盈虧皆相當重視，惟小額投資人顯然不諳績效預測方法。

This paper applies Multiple discriminant analysis (MDA), Logit analysis, Probit analysis, Market model, and questionnaire to analyze the performance prediction and stock price reaction in Taiwan's listed firms. We find that the discriminant abilities of MDA are most above 70%, when the accounting information comes from original samples; but the ex ante predictive power is insignificant when the data come from the second period samples. We also find that the CAR of the firms that predicted good performance is better than the firms that predicted bad performance, and most investors in stock market understand the importance of accounting information and earnings, but some of them are not good at the techniques of performance prediction.