近年來大陸經濟蓬勃發展,企業數量成長快速,對投資者而言,財務危機的影響更顯重要。但過去研究因模式的假設或無法有效說明分析結果而有所受限,新進統計學習上的核函數正規化(Kernel Regularized)運用彈性的模式設定(specification),可較深入的說明變數影響之形式,有助於傳統模式的改善。本研究應用核函數正規化最小平方模式(Kernel Regularized Least Squares Model, KRLS)分析大陸企業財務危機的影響因素,藉以提供投資者更正確的資訊。 本研究結果顯示財務危機發生機率確實很複雜,模式設定對傳統Logit模式推論結果差異很大,在眾多主要影響變數下,前10項顯著設定項目中,單獨顯現的變數較少,多以交互形式出現。例如「總資產凈利潤率」及「留存收益資產比」確實為重要影響變數,但「總資產凈利潤率」與「成本費用利潤率」及「速動比率」與「負債與權益市價比率」的交互項亦為非常重要的因素。此結果充分顯示應用KRLS的配適能力與Logit模式的簡易解釋性可進一步探索影響財務危機發生變數間的複雜形式。
The analysis of financial distress can help investors to make right decisions. But most past analysis models are either limited by the assumptions or cannot be used for effective explanation. The advance kernel regularized model from statistical learning can apply more flexible model specification and explores the affect forms of variables. Recently, the flourish of China economics has increased large amount of enterprises, this highlights the influence of financial distress effect. Therefore, this work applies Kernel Regularized Least Squares (KRLS) model to analyze the factors for Chinese enterprises’ financial distress to provide more accurate information for investors. The results of the work demonstrate the integration of fitting capability of KRLS and clear interpretation of Logit models can further explore the forms of factors for the financial distress analysis, such as the interactions between “return on assets” and “ratio of profits to cost”, or “quick ratio” and “debt-to-equity ratio”.