針對新上市上櫃公司股票(initial public offerings, IPOs)長期報酬的異常現象,以往學者大多偏重在探討其發生的原因,而忽略了衡量方法對於異常報酬的影響,Chen (2001)指出在各種不同衡量模型下,IPOs之長期績效會有不同的結果,而建議必須審慎選定衡量模型。國外研究均發現IPOs確有長期績效不佳的現象,但國內的研究則沒有明顯長期績效不佳的現象,而造成這樣結果一種可能的原因是:衡量模型本身的檢定力不足。以往研究在IPO預期報酬(benchmark)的衡量上並未去除IPO樣本,如此一來容易造成IPO異常報酬被其預期報酬所抵消,而造成沒有異常報酬存在。本文以Fama-French的三因子模型再加上股價之動能(Momentum)組合成四因子模型,並將衡量資產報酬因子模型中的因子組合(factor portfolio or factor mimicking portfolio)去除IPOs樣本,以淨化因子組合(purified factor portfolio),藉此來強化衡量模型的檢定力。實證結果發現,淨化因子組合模型確實能有較強之檢定力,而台灣IPOs並沒有長期績效低落的現象,符合市場效率性。
The long-run underperformance of initial public offerings (IPOs) is commonly documented in prior literature. Previous studies focus on the causes for the anomalies, while few ever examine the validity of the measurement models. Chen (2001) argues that IPO long-run performance significantly depends on the measurement model. This paper employs the Fama-French three-factor model plus momentum factor to investigate the IPO performance, especially focusing on the power of the measurement models. In the factor model, previous studies form factor mimicking portfolios with the entire sample including the IPO firms. However, forming mimicking factor portfolios with the entire sample tend to mitigate the abnormal returns of IPOs. In this paper, we purify factor portfolios which are the mimicking portfolios without IPO samples in the portfolios to reinforce the power of the measurement model. Our results show that purified factor model increases the power of the measurement model and that IPOs do not experience poor long-run performance, which is consistent with market efficiency.