5.1. Results of Ordinary Least Square Regression
The results of ordinary least square regression are presented in Table Ⅳ. We can see that the coefficient of discretionary current accruals is positive and significant. Firms with higher discretionary current accruals will have higher DEBT3. It presents that the behavior of earnings management will affect the choice of debt maturity structure later.
As we can see, Table Ⅳ shows that the Firm Size is positive related to debt maturity structure significantly, and consistent with the results of Barclay and Smith (1995) and Smith and Warner (1979) that small firms will tend to issue more short-term debts to eliminate the conflicts between stockholders and bondholders. The sign of coefficient of tax rate is also positive and significant. It indicates that firms with high tax rate will tend to issue long-term debt. This result is consistent with Brick and Ravid (1985). The coefficient of M/B ratio which is the proxy of growth option is significantly positive and inconsistent with the result of Barclay’s (1995), but is consistent with the results of Datta, Iskandar-Datta, and Raman (2005) and the original finding of Stohs and Mauer (1995). As our prediction, the coefficient of Assets Maturity is positive but insignificant. However, inconsistent with our prediction, the coefficients of Term Structure and Regulation Dummy are negative, though they are insignificant. The coefficient of Abnormal Earnings is unexpected positive although it is insignificant.
5.2. Results of Two-stage Least Square Regression
Table Ⅴ shows the results of two-stage least square regression. Panel A presents the regression with DEBT3 as the dependent variable. From the empirical results in Table Ⅴ, we can find that the coefficient of discretionary current accruals (DCA) which is the proxy of earnings management is positive and significant at the 1 %, 5% and the 10% level. It supports
our hypothesis that firms with aggressive earnings management will have incentive to issue more debts with longer maturity.
The coefficients of Log of Firm Value and the square of Log of Firm Value are still positive and significant, and consistent with Barclay and Smith (1995). As our expectation, large firms will prefer to issue more long-term debt rather than small firms. The coefficient of Market-to-Book ratio is still positive but insignificant in this situation. We also find that the coefficient of Assets Maturity is still positive as our prediction but insignificant. The Abnormal Earnings is positive related to debt maturity structure and is inconsistent with our prediction but still insignificant. However, the coefficients of Term Structure and Regulation Dummy are both negative and inconsistent with previous prediction, although they are insignificant. Finally we can see that Leverage is positive related with debt maturity but still insignificant. Panel B shows the results of the second-stage regression with Leverage as the dependent variable.
5.3. Results of Long-term Performance of Buy-and-Hold Abnormal Returns (BHAR)
Table Ⅵ reports the results of 5-year long-term performance of our sample firms by using the buy-and-hold abnormal returns method. Panel A shows the equally-weighted results.
We can see that the abnormal returns of holding period from1-year to 5-year are all negative in firms with higher discretionary current accruals (defined as Aggressive firms). On the other hand, firms with lower discretionary current accruals (defined as Conservative Firms) also face positive long-term performance after debt issuing. Panel B shows the value-weighted results. Similar to equally-weighted results, we find that the abnormal returns of holding period from1-year to 5-year in aggressive sample firms are also negative, and the conservative firms have positive long-term performance in first two years then become negative from the third year. However, we can still find that the long-term performances of aggressive firms are
worse than conservative firms. Finally, panel C reports the results of abnormal returns using size and book-to-market matching firms as benchmarks. The result shows that almost all abnormal returns of each holding period are negative in aggressive firms, and conservative firms face positive long-term performance in first two years then become negative from the third year. As our prediction, the empirical results show that after firms with aggressive earnings management issue debt, they will have poor long-term performances in 5 years. This result is also consistent with the analysis of Teoh, Welch, and Wong’s (1998a).
5.4. Results of Long-term Performance of Fama and French Three-factor Model
Table Ⅶ reports the results of 5-year long-term performance of our sample firms by using Fama and French’s Three-factor model. Panel A presents the equally-weighted results.
From holding period 1-year to 5-year, we can see that all the abnormal returns of aggressive firms are negative. The long-term performances of conservative firms are also negative;
however, they are still better than the long-term performances of aggressive firms. Panel B presents the value-weighted results. We can find that there are also negative abnormal returns of each holding period similar to equally-weighted results; however, the abnormal returns of conservative firms are negative in the first year after debt issuing and turn positive from the second year. The results of Fama and French’s Three-factor model also support our prediction that firms with aggressive earnings management will have poor long-term performances in 5 years after firms issuing debt.
From the empirical results of Fama and French’s three factor model, we can find that the abnormal return of value-weighted results is less significant than the equally-weighted results.
According to Loughran and Ritter (1995), when the significantly abnormal return concentrate on small firms, the portfolio based on value-weighted may not reflect the real situation. Thus, we suppose that large firms may dominate the results of abnormal returns in our sample, and
the results of value-weighted three-factor model are less significant.