以Cubist迴歸樹建構公司情境特質之股權評價模型

Abstract

本研究旨在考量公司情境特質下，以Cubist迴歸樹模型建構各所屬之股權評價模型，歸納股權評價之門檻攸關因素，並論證公司情境特質對股權評價之重要性。股權評價是一複雜的評估內容，非單純的線性估計所能解決，故導入Cubist迴歸樹模型，以資料特性去探勘不同情境特質下之股權評價模型較佳。實證結果經由樣本外預測之準確性，發現Cubist迴歸樹模型之MAPE值為36.54%，小於OLS估計方式的MAPE值62.26%，顯示Cubist迴歸樹模型優於Ohlson (1995)股權評價模型之原始估計方式。進一步與以往研究及傳統簡單模型比較發現，考量公司之公司治理機制與公司特性等情境特質因素，較能準確刻劃出公司的股權評價模型，且財務變數較公司治理因素與公司特性因素重要；再者，Cubist迴歸樹模型相較傳統簡單模型，Cubist迴歸樹模型較為準確、較具效率性且有助於提昇決策正確性。此說明應考慮個別公司情境特質因素，建構其所屬最適之股權評價模型，可使股權評價模型預測更加準確。本研究實證結果可建議管理當局於公司所處之情境特質下，應如何強化其公司價值，並供管制機構推動公司治理相關政策之參考。

This study constructs an equity evaluation model by considering a firm’s conditions, characteristics, and corporate governance. The model generalizes the key factors affecting the threshold of equity evaluation, and validates the importance of corporate conditions and characteristics on equity evaluations. Equity evaluation is very complicated and the application of OLS estimation to it may be overly simplified. Thus, this study adopts the Cubist regression tree model, which uses data characteristics to probe threshold variables and threshold values, to determine each equity evaluation model. The empirical results reveal that, according to the accuracy of the out-of-sample forecast, the MAPE value of the Cubist regression tree model is 36.54%, which is smaller than that value of 62.26% obtained by the OLS method. This result shows that the Cubist regression tree model is superior to the original estimation method for the equity evaluation model proposed by Ohlson (1995). A further comparison of existing and traditional models finds that considering corporate conditions and characteristics, as well as corporate governance mechanisms, one can obtain a more accurate construction of the equity evaluation model. In addition, financial variables seem to be more important than corporate governance factors and other firm characteristic factors. Moreover, as compared with traditional models, the Cubist regression tree model is found to be more accurate, more efficient, and more conducive in enhancing the quality of decision-making. The individual corporate conditions and characteristics are suggested to be incorporated into the construction of the equity evaluation model, which can lead to more accurate forecast results. The empirical findings may provide indication to management for what to be focused to strengthen corporate values and offer reference to regulatory bodies for formulating and mandating corporate governance policies.