影響選擇權模型衡量信用風險有效性的公司特性探討

Abstract

精確衡量信用風險已成為近年來風險管理領域中最重要的一環，而在諸多衡量信用風險的方法中，選擇權評價法是目前最受重視的一種方法。惟將其應用在國內上市櫃公司的財務危機預測上，多數實證卻顯示其效果不彰，本文認為主要原因是股價並非隨時都能反映公司所有的資產甚或權益價值，而不同特性公司的股價所反映程度也不盡相同。股價資訊的不真實所導致預測失靈的情形又分兩種：預期違約機率低估或者高估。 本文以修正後的Merton模型為基礎，分別以OLS法與logit法，從公司資訊環境、股價操縱、股價泡沫、與無形資產程度等四個角度，探索容易導致預期違約機率低估及容易導致授信單位錯誤核貸（誤授）的情形。在OLS法下的實證結果發現，在公司資訊環境越差、股價中存在泡沫、及股價被操縱等情境下，較易導致以選擇權模型算出的預期違約機率偏低，但無形資產程度對其影響則不顯著。而在logit法下的實證結果發現，在公司資訊環境越差、股價中存在泡沫、股價被操縱、及無形資產程度越高時，選擇權評價法較易引致誤授。 本文亦另外以相同的方法，從不同的角度來檢測容易導致預期違約機率高估及容易導致授信單位錯誤駁回（誤拒）的情形。在OLS法與logit法下的實證結果皆顯示，當公司資訊環境越差、投資人過度反應於部分以股東角度為出發點的財務比率、及受測對象為營建業時，較易導致以選擇權模型算出的預期違約機率偏高或引致誤拒的情形。

In the Basel 2 environment, to measure credit risk more precisely may be a bank’s core competency. Among many methods of measuring credit risk, Merton model (option pricing) is the prevalent one. However, prior research documents Merton model’s weak predictability of corporate distress as compared with the competing models in Taiwan. In this thesis, we conjecture the major explanation is that stock price may fail to reflect the asset value of one company under certain circumstances. Moreover, the informativeness of the stock price varies among different company. Based on the adjusted Merton model, we use both OLS and logit methods, aiming at identifying the variables underestimation of the expected default frequency or the outcome of type 1 error (default companies being categorized as non-default companies) from four aspects: information environment, stock price manipulation, stock valuation bubble, and the intensiveness of intangible asset. Our OLS analysis results show that companies subject to poor information environment, stock price bubble, and manipulation of stock price, are likely to be with Merton model underestimation of the expected default frequency. Consistently, our logit model tests reveal that observations subject to poor information environment, stock price bubble, manipulation of stock price, and high intensiveness of intangible asset, are more likely to be with the type 1 error. We adopt the same method to identifying the variables contributing to overestimation of the expected default frequency or the type 2 error (non-default companies being categorized as default companies) from different aspects. Both OLS and logit analyses exhibit that when the market overreact to the financial ratios that primarily serve the stockholders, observations with poor information environment, especially those in the construction sector among these companies, are more likely to be subject to Merton model overestimation of the expected default frequency or the type 2 error.