標題: 財務危機預測模型比較:考慮總體經濟變數
A study of financial distress prediction models:considering macroeconomic variables
作者: 李國樞
Li, Gau-Shu
周雨田
Chou, Ray Yeu-Tien
經營管理研究所
關鍵字: 離散存活;財務危機;最大概似法;discrete-time survival;financial distress;maximum likelihood method
公開日期: 2009
摘要: 在本文中,我們收集國內股票上市公司的產業效應變數(industry effects; Chava and Jarrow, 2004)、市場導出變數(market-driven variables; Shumway, 2001)、財務比率(financial ratios)、以及總體經濟變數(macroeconomic variables),分別應用至離散倖存模型(discrete-time survival model)、離散卡氏模型(discrete-time Cox model)、羅吉特模型(Logit model)、以及機率模型(Probit model)。並且使用最大概似法(maximum likelihood method)去估計模型的參數。   Shumway在其實證研究中將離散倖存模型定義為為多期Logit模型,然而,其所定義之多期Logit模型之概似函數忽略了樣本公司在 時仍存活之機率。本研究將對此進行修正,並分別應用上述變數去比較各模型對公司發生財務危機之預測能力,研究在何種模型下,使用何種變數組合及考慮總體經濟情況是否可以得到公司發生財務危機之最佳預測效果。   實證結果顯示,動態模型會比靜態模型有較佳的表現,除了在Zmijewski變數組合下。而在使用離散倖存模型去分析各變數組合及總體經濟變數組合時發現,除了在Altman變數組合及總體經濟變數組合情況下,其他變數組合的預測能力皆有提升。不幸地,三種變數組合的型一誤差皆比原始變數組合高出些許。
In this paper, we use the discrete-time survival model, discrete-time Cox model, Logit model, and Probit model to predict the probability of financial distress for each firm. We use a set of variables, including market-driven variables, accounting ratios, and macroeconomic variables to predict financial distress. Shumway estimates a multi-period Logit model that can be defined as discrete-time survival model. The maximum likelihood method is used to determine the probability of the model. Then we want to modify the discrete-time survival model’s likelihood function because it ignores the probability of surviving at time t for all firms. Empirical studies demonstrate that the dynamic models can yield more accurate out-of-sample forecasts than static models in general. All these macroeconomic variables are significant. Unfortunately, three sets of variables’ type-one-error rate are higher than original sets of variables.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079737556
http://hdl.handle.net/11536/45616
Appears in Collections:Thesis