美國航空業破產預測模型：二元分量迴歸之應用

Abstract

根據Morrell(2007)，航空業相較其他產業而言，更容易受到總體環境與人為因素影響，進而波及公司之營運與獲利。除此之外，航空業對一國經濟極為重要，因此一破產預測模型之建立將有助於公司經理人、投資人、債權人與政府機關等團體監督其營運績效。本研究以1990年至2011年之美國航空業資料為樣本，採用Kordas(2000)所提出之二元分量迴歸(Binary Quantile Regression)以及 DRIES & DIRK (2012)所提出之貝氏二元分量迴歸(Bayesian Quantile Regression)建構破產預測模型，並與Logit模型同時進行比較，以累積正確率曲線(Cumulative Accuracy Profile)所計算出之準確率(Accuracy Ratio)與白氏得分(Brier Score)兩項指標為比較準則。實證結果顯示，貝氏二元分量迴歸具有較佳之預測能力。

According to Morrell (2007), macroeconomic and artificial factors have greater impact to the aviation industry, comparing to others, thus further influence business operation and profitability. In addition, aviation industry accounts for a significant proportion of most economies, hence an early warning model of bankruptcy is critical and beneficial to interest groups, such as managers, creditors, investors, and governments, to monitor the performance of operation. This study utilizes data of U.S. airline companies, ranging from 1990 to 2011. Binary Quantile Regression, Bayesian Quantile Regression, and Logit models are employed for construction of bankruptcy prediction. Two indices, Accuracy Ratio calculated from Cumulative Accuracy Profile and Brier Score, are used as standards of comparison. Empirical result demonstrates that Bayesian Binary Quantile Regression has the best performance in terms of prediction accuracy.