財務危機預警模型實證研究─考慮表外風險因素

Abstract

本研究以台灣上市櫃紡織業公司2011到2017年的資料為樣本，並以羅吉斯迴歸為基礎建立財務危機預警模型，探討額外加入表外風險因素後是否能增進模型的預測力。本研究設立兩種衡量表外風險的因子：屬量變數「衍生性商品合約金額比」(DER)和屬質變數「是否有操作衍生性商品」(HED)，分別建立以下三種模型，並以2011到2016年的資料為基礎進行樣本內測試、2017年的資料為基礎進行樣本外測試： 模型一：財務變數 ＋ 公司治理變數 ＋ 總體經濟變數 模型二：財務變數 ＋ 公司治理變數 ＋ 總體經濟變數 ＋ DER 模型三：財務變數 ＋ 公司治理變數 ＋ 總體經濟變數 ＋ HED 實證結果顯示，雖然模型一已經可達到99%的樣本內預測準確率和98%的樣本外預測準確率，額外加入表外風險因素的模型二和模型三卻都無法再另外提高準確率。因此本研究推論：表外風險因子並不能增進預警模型對台灣紡織業公司未來是否會發生財務危機的預測力。

This paper uses the data of listed companies or over-the-counter companies in Taiwan’s textile industry from year 2011 to year 2017 to build financial distress prediction model, and tries to examine whether adding the off-balance sheet factor into the prediction model can improve the prediction ability or not. This paper set two kinds of factor to evaluate off-balance sheet risk: “derivative contracts ratio” (DER), which is a quantitative variable, and “use derivatives or not” (HED), which is a qualitative variable or dummy variable, to build three different kinds of model. After that, this paper uses the data from year 2011 to year 2016 to perform the in-sample test, and use the data in year 2017 to perform the out-of-sample test: Model I: finance variables ＋ corporate governance variables ＋ macroeconomics variables Model II: finance variables ＋ corporate governance variables ＋ macroeconomics variables ＋ DER Model III: finance variables ＋ corporate governance variables ＋ macroeconomics variables ＋ HED According to the empirical study, although Model I has 99% prediction accuracy of in-sample test, and 98% prediction accuracy of out-of-sample test, Model II and Model III, which consider the off-balance sheet risk, cannot improve the prediction ability anymore. So we can infer that off-balance sheet risk factor is not the key point to improve the prediction ability of financial distress prediction model of Taiwan’s textile industry.