台灣上市櫃公司隱含權益資金成本之估計及其效度之評估

Evaluating the Validity of Estimates of Implied Cost of Capital: Evidence from Taiwan Listed Firms

"為了克服分析師預測資料涵蓋範圍不足之限制，本研究以Hou, van Dijk, and Zhang (2012)橫斷面盈餘預測模式為基礎估計個別公司之隱含權益資金成本（ICC），提供了台灣非金融業上市櫃公司隱含權益資金成本之長期時間序列（1994-2013）與橫斷面大樣本之敘述性統計。本研究亦提供七種ICC估計橫斷面與時間序列相對效度之證據。依據實證結果評估，本研究建議，研究者若欲探討特定事件（如制度變革）對權益資金成本之影響，可選擇時間序列效度最佳之GLS模式（Gebhardt, Lee, and Swaminathan, 2001）與GG模式（Gordon and Gordon, 1997）估計權益資金成本；若旨在探討橫斷面特性（如資訊環境與治理結構）與權益資金成本之關聯性，可選擇橫斷面效度最佳之GLS模式或將GLS與效度次佳之CT (Claus and Thomas, 2001)、GG與PEG (Easton, 2004)等模式所估計ICC彙總為一項綜合ICC，以降低個別ICC估計之衡量誤差。本研究所估計之多項隱含權益資金成本對於估計預期報酬率具有良好的效度，且能涵括長期時間序列與橫斷面大樣本分析，可提供台灣評價實務與學術研究之參考。"

"This study examines the relative validity of alternative estimates of firm-specific implied cost of capital (ICC) and provides descriptive statistics of those ICC estimates for a large longitudinal time-series and cross-sectional sample of Taiwan listed firms over 1994-2013. To address the problem of limited coverage of analysts' earnings forecast data, I use earnings forecasts estimated by the cross-sectional model developed by Hou, van Dijk, and Zhang (2012) to proxy for market earnings expectations. A series of tests have been conducted to assess the validity of seven ICC estimates based on their cross-sectional and time-series performance. Based on the empirical results, researchers are advised to select the model with the best time-series validity including the GLS (Gebhardt, Lee, and Swaminathan, 2001) and the GG model (Gordon and Gordon, 1997) to estimate the ICC when they attempt to identify the effect of a certain event (e.g., institutional reforms) on the cost of capital. In addition, if the researchers aim to identify the association between certain cross-sectional characteristics (e.g., information environment and governance structure) and the cost of capital, they can select the GLS model which has the best cross-sectional validity, or alternatively estimate a ICC composite by combining the GLS's ICC estimate with three second-best ICCs estimated by the CT (Claus and Thomas, 2001), GG, and PEG (Easton, 2004) model in order to mitigate the measurement errors by individual ICC estimates. The evidence documented in this study suggests that several ICC estimates can serve as valid proxies for expected return and are useful for the valuation practice and academic research in Taiwan."