本篇論文根據Sudipto(2001)這篇文獻提出影響附加贖回條款的五個因子進行實證研究並比較Sudipto(2001)的發現和我們的實證結果。研究期間為1996年至1997年前二個月在美國發行的所有公司債。公司債發行基本條件來自Bloomberg，公司財務資訊來自Compustat以及相關的權益市值資料來自CRSP。 我們使用平均數檢定和邏輯迴歸進行債券附加贖回條款組成份子的實證研究。實證結果發現，在移動平均模式下，負債契約條款假說不成立，但支持高票面利率的債券，其附加贖回條款的可能性愈高和破產成本愈低的公司，傾向發行可贖回債券的論點。然而，在移動平均模式下，證據顯示不支持利率水準假說和到期期限假說。對於在最大概似估計模式下，結果顯示不符合負債契約條款假說和租稅利率假說。但支持到期期限假說，高票息債券為何大多附加贖回條款和高公司風險的公司偏愛發行可贖回債券的觀點。總的來說，我們的結果大多不支持Sudipto(2001)的發現。 此外，我們使用一個更嚴謹的方法去估計公司價值波動度，即Duan(1994)這篇文獻中所提出的最大概似估計法，並與原先的方法進行比較。結果顯示兩種方法對債券附加贖回條款可能性間的影響有明顯地不同。使用移動平均模型估計而得的公司價值波動度與債券附加贖回條款可能性間呈負相關但不顯著，利用最大概似估計法估計而得的公司價值波動度與可贖回債券發行之可能性間卻有正向且顯著地影響。 The paper examines various hypotheses concerning bond adding a call provision in the light of determinants of call feature of Sudipto (2001) advanced based on the same periods and compares the findings of Sudipto (2001) with ours. We gather details of all corporate bonds issued in the U.S. in 1996 and the first two months of 1997 from Bloomberg, related financial information from Compustat and affiliated equity value data from CRSP. We use the means test and logistic regression to engage in the empirical study of determinants of call features. Empirical investigation shows that the results are at variance with debt restrictive covenant hypothesis and in concert with the argument that there is a good possibility of callable bonds are high coupon rate bonds and firms with less bankruptcy cost tend to issue callable bonds under the moving average model. However, the evidence provides no support for interest rate level hypothesis and maturity hypothesis under moving average model. For the maximum likelihood estimation model, our results disagree with debt restrictive covenant hypothesis and tax advantage hypothesis while in unanimity with the maturity hypothesis and the point that the higher the coupon rate, the greater the likelihood of bonds with a call feature. Furthermore, firms with high firm risk prefer to issue callable bonds. Overall, most evidence provide no support the findings of Sudipto (2001). Additionally, we adopt a religious method which Duan (1994) addressed to estimate volatility of firm value and compare the difference between the old method which refers to the moving average model and new method which means to maximum likelihood estimation. The results indicate that two estimation methods have a different effect on bonds with call feature. Volatility of firm value estimated by the moving average model have a negative but insignificant effect on the probability of callable bonds but volatility of firm value implied by maximum likelihood estimation method have a positive and significant effect on the bond attaching call provision.
