台灣的金融市場近兩年來因衍生性金融商品的蓬勃發展，而產生巨幅
的改變，當中以認股權證因能提供投資人以少量資金獲取無限利潤的特性
而廣受青睞。然本土型認股權證自民國86年12月上市至今，尚未有一套明
確的定價模式，故本研究擬藉由文獻的參考，找出合適的訂價模型，評估
這（些）模型的預測績效，以做為投資人做投資決策時的依據。 合
理的定價影響選擇權市場的發展至巨，發行者依此決定發行價格、投資人
據此判斷投資與否，因此，慎選合適的定價模型實為要務。在參閱相關文
獻後發現，連續型的二次逼近法及離散型的二項式模型，為評價美式選擇
權的最佳模式。鑑於台灣的認購權證屬於一種美式選擇權，故本研究擬採
用這兩種評價模式來預測認股權證的價格，並比較這兩種模型的預測績效
。另外，在研究過程中亦可發現，採用隱涵標準差與歷史平均標準差的模
式，在預測績效上明顯不同。 本研究對象為最早發行的前六家個股型
認股權證，即大華01、02、03、寶來03、京華01、金鼎01，期間涵蓋自民
國87年1月9日至87年4月30日。研究結果得到如下結論 ： 1. 不論是
採用歷史標準差或隱涵標準差做為模型中變數的估計值，大致上說來，二
項式模型在價外認證的評價上所產生的絕對誤差（市價減去模型價格的絕
對值） 誤差比率（市價減去模型價格再除以市價）均較二次逼近
法小。相反地，在評價 價內認證時，二次逼近法的績效較二項式
模型佳。 2. 在納入隱涵變異數後，兩模式的預測績效均顯著改善了
。 3. 兩模型在S/E<1.05 一項中的誤差比率，相對於在S/E>=1.05中
的誤差比率要大得 多，這大概是因為前者在作為誤差比率運算式
中的分母-------市價，比後者的 市價要為小得多之故。 4.
兩模式均低估認股權證的價格。

The financial market in Taiwanhas changed a lot for the latest
two years mostly because of the spring up of derivative
financial instruments. Among these newly developed financial
commodities, warrants shock investors most because they provide
excellent mechanisms for controlling risk and locking up
resources at a relevant minimal fee. So this study tries to find
a more proper pricing model to value warrant prices so as to
provide a guide for the investors when making investment
decision. After reviewing relevant literature on option pricing
theory, we find that the quadratic approximation method and the
binomial model are two better models valuing American options.
Since warrants in Taiwan are American-style call options, this
study decides to adopt these two models as pricing tolls. The
target warrants are six first-issued single-stock ones. Study
period covers from 1,9,''98 to 4,30,''98 Dividing the warrants
into five groups according to the extent of moneyness, we
measure the absolute pricing error and percentage pricing error
between market prices and model prices. Then we use t-test to
exam if these errors are significant and compare the performance
of the two models. In addition, we adopt two methods to estimate
the volatility, simple historical s.d. and simple implied s.d..
One can observe the performance of the two methods
simultaneously. The experimental results are: 1. No
matter applying historical volatility or implied volatility ,
the binomial model generates smaller average absolute
pricing and percentage pricing error than the quadratic
approximation when valuing out-of-the-money warrants. But
when pricing in-the-money warrants, the condition reverses.
2. The average absolute error and percentage pricing error of
all the warrants drops dramatically after applying implied
s.d. as the estimator of the volatility. 3. After dividing
market price - model price by market price, the percentage
pricing error for both models are much higher when dealing with
valuing warrants with S/E <1.05 than that of warrants with
S/E >=1.05. This is mostly because that the market prices
of warrants with S/E<1.05 are much lower that those of
warrants with S/E>=1.05. 4. Both models obviously underprice
the value of warrants in Taiwan.