本篇探討造成預警模型失效的原因,由於模型的預測力主要來自變數資料對違約的解釋能力,因此如果納入變數資料受到扭曲而無法反映公司的真實狀況,將會使模型預測能力降低。本研究藉由提出五大假說,試圖觀察在不同可能扭曲資訊情況下,會計及市場預警模型相對失效的狀況。 本研究所獲得的實證結果發現,不管是會計預警模型或者市場預警模型對公司發生財務危機都有相當顯著的解釋效果。產業效果假說發現市場模型在調整過產業效果之後,其對違約的解釋力相對於會計模型來說有明顯增加;公司治理假說檢驗人為操縱資訊之能力及動機大小對預警模型結果之影響,會計模型的普遍失效顯示會計資訊比起市場資訊更容易遭受到操弄;除此之外,資訊不對稱性高的小公司,其會計資訊雖受規範多,但資訊不透明狀況使得會計模型相對失靈。但市場模型也非萬能,在流動性過高及投資人過於樂觀下仍會失效。此研究結果為預警模型失效提供了可能之原因,並可在未來運用預警模型結果上可做為參考。
he thesis investigates reasons behind failed distressed prediction models. Since variables of models reflect the current status of a company’s operating and financing situation. The quality of inputted information is quite influential for an effective distressed model. We compare our defined best accounting and market distressed forecasting models under five hypotheses to see the relative performance of both models. Our defined best accounting and market distressed prediction models have reached statistical significant in forecasting default. Particularly, the market model gets extra information after adjusting industry effect. On the other hand, accounting distressed forecasting model failed when the management has higher incentive and capability to manipulate information. Moreover, for small firms, the accounting model fails because of high information asymmetry. Nevertheless, market model fails when liquidity is too high and investors are too optimistic toward growth stocks. The thesis provides some empirical reasons for failed distressed prediction models. It also provides some references for people who will use these forecasting results in the future.