對於財務危機公司的預測，一般研究皆以傳統的財務指標為主，但一般的財務指標皆著重於對本業經營不善或資本結構有嚴重偏差的公司進行預測，對於本業經營良好而過度使用財務操作的公司則較無預測能力。另外，許多經營不善的公司亦透過子公司囤積其無法銷售的存貨，或利用子公司虛增其業績，這些從原始的財務報表都無法預知公司的體質是否還屬健全。因此本文除了採用傳統的財務比率外，亦增加了大股東質押比，子公司購回母公司股票，及短期投資比例三項財務操作變數，並考慮合併報表與一般報表的差異性。 綜合本文的實證結果，發現公司在加入了財務操作指標後，在類神經網路預測模式下整體公司的正確預測率由90.625%提升到93.75%，另外以合併報表取代一般報表後，在logit預測模式下整體公司的正確預測率也由68.42%提高到76.31%，兩類報表間相同的解釋變數為負債比率，由此結果顯示，目前上市公司不當的財務操作確實存在，本文希望能藉由探討發生財務危機的各種可能情況的原因著手，以期能真正做到防患於未然，而使金融機構與投資人所需付出的社會成本降到最低。 Researches on the prediction of financial distress companies focus on traditional financial indicators. Common financial indicators have provided good predicting ability over firms whose core businesses are operated inefficiently, or having extremely high leverage. For those whose core businesses remain profitable, but craving for financial manipulation, the predictability is insufficient. Besides, many under-performed companies varnish their financial statements by selling inventories to subsidiaries; this phenomenon will not be discovered through the conventional model. To improve the predictability of the financial distress model, we include the pledge ratio of major shareholders, the percentage of subsidiaries purchasing parent companies’ shares, and the ratio of short-term investment as explanatory variables, and examine the differences between consolidated statement and ordinary statement. Based on our empirical results, we find that incorporating financial manipulation indicators helps to increase predictability (90.625%-93.75%). In addition, consulting consolidated statements instead of unconsolidated statements boosts the predictability from 68% to 76%.
