The results show that the insured firms have less and significant volatility in ROA, but more volatility in EPS and short-term investment. This implies firms might have more opportunistic behavior in EPS and short-term investment after D&O purchase. However, the differences between firms with and without D&O insurance are not significant in ROE and debt-asset ratio. Consequently, it is suspicious that insured firms have more volatility in returns and investments. The result of descriptive analysis is reported in Table 7.
Table 7 Comparison between firms without and with D&O insurance (3)
Firms without D&O Firms with D&O Total
Mean N Mean N Mean N
S_ROE 3.5135 3409 3.1864 4880 3.3209 8289
S_ROA* 1.4555 3409 1.2971 4880 1.3622 8289
S_EPS* .5750 3407 .6496 4880 .6189 8287
S_DAratio 34.6412 3409 63.6736 4880 51.7335 8289
S_Sti* 153349.33 3409 553883.44 4880 389156.60 8289
* indicates that the difference is significant in independent sample test.
110 See Seunghan Nam, The Impact of Non-audit Services on Capital Markets 18 (2006), available at http://ssrn.com/abstract=693422.
111 Similarly, Standard deviation of ROE is also often used as a proxy variable for the risk of insurers. See J.
David Cummins & Gregory P. Nini, Optimal Capital Utilization by Financial Firms: Evidence from the Property-Liability Insurance Industry, 21. 1(2)J.FIN.SERV.RES.15,23(2002).
26 2. Regression analysis
As the previous tests, binary variable insurance purchase is used in the first panel, and numeric variable nature logarithm of coverage is used in the second penal. Standard deviations of ROE, ROA, EPS, debt-asset ratio and short-term investment are used as dependent variables in respective regressions.
In the first panel, D&O insurance purchase is not positively significant in these regressions except when dependent variable is standard deviation on short-term investment. This implies no statistically significant evidence proving that the purchase of D&O insurance will increase opportunistic behavior. Moreover, D&O insurance purchase is negatively significant when dependent variable is standard deviation on ROA, implying insurance purchase correlates to even less volatility. In the second panel, insurance coverage is negatively significant when dependent variable is standard deviation of ROE and ROA, and not significant in other specifications. Clearly, it does not support the concern that firms with more coverage may have some intention to conduct opportunistic behavior, and implies less volatility in ROE and ROA. Also, the test results of corporate governance hypothesis are consistent. For litigation risk, it is generally and positively significant in many specifications. This supports the concern that firms with more litigation risk usually have more volatility in investment.
In conclusion, all these results demonstrate that the D&O increase purchase and coverage is not positively and significantly correlated to the variance in earnings and investment behavior. In other words, no consistent evidence is found to prove a positive correlation between D&O insurance and the opportunistic behavior and moral hazard of firms. The detailed results are reported from Table 8 and 9.
D. Summary
In this section, this study empirically tests whether D&O insurance is correlated to opportunistic behavior and moral hazard. Empirical evidence shows that the purchase of D&O insurance and its coverage are not significantly and negatively correlated to variances of earnings and investments. In consequence, even though some scholars argue that insurance may cause opportunistic behavior and moral hazard or even damage firms eventually, the empirical work does not find such significant evidence. Hence, the previous tests about the monitoring and signal effect of D&O insurance may not be affected by opportunistic behavior and moral hazard in Taiwan. And the previous proposal to limit D&O insurance may need more evidence to be supported and justified in Taiwan.
Table 8 Result of regressions with panel data (3)
(1) (2) (3) (4) (5)
S_ROE S_ROA S_EPS S_DAratio S_Sti
Purchase -0.0800 -0.0752** 0.0554 61.66 67885.2***
(-1.03) (-2.59) (1.44) (1.21) (3.66)
Industry 0.00604 -0.00350 -0.000162 7.516* 486.7
(0.70) (-0.63) (-0.09) (2.41) (0.14)
27
ROE -0.0598 -0.00980 0.000661 0.693 5864.6***
(-1.87) (-1.64) (0.19) (0.46) (4.30)
Remuneration -0.0151 -0.000582 0.00188 3.299 13731.1*
(-1.32) (-0.23) (1.87) (1.55) (2.34)
Indptdirector -0.255*** -0.0174 0.0231* -40.24 79795.7***
(-5.86) (-0.85) (1.99) (-1.37) (3.98)
Auditcomitee -0.635* -0.146 0.00691 -57.84*** 1400111.4***
(-2.02) (-1.82) (0.08) (-3.41) (4.07)
Ctrldirector -0.241*** -0.0943*** -0.0298*** -12.87** 12475.1
(-5.52) (-6.72) (-13.68) (-3.00) (1.12)
Sdirector 0.0192** 0.00192 0.000759 -0.333 -7383.2**
(2.80) (0.95) (1.73) (-1.04) (-2.99)
Dual 0.619* 0.208** -0.00792 -32.32 -230547.1***
(2.43) (2.70) (-0.68) (-0.89) (-6.88)
DAratio 0.00617*** -0.0000514 0.0000619* 0.410*** 2076.7*
(9.78) (-0.71) (2.47) (5.83) (2.30)
Litigation 0.313* 0.0883* 0.0491*** -8.582* 89696.8**
(2.36) (2.46) (3.33) (-2.31) (2.81)
Constant 3.140*** 1.598*** 0.576*** -15.48 -132515.5
(13.21) (8.60) (18.66) (-0.37) (-1.85)
N 6401 6401 6398 6401 6401
R2 0.1379 0.0282 0.0121 0.0037 0.0735
Hausman test 0.0000 0.0000 0.0000 0.9711 0.0000
Model xtscc xtscc xtscc re xtscc
Mean VIF 1.96 1.96 1.96 1.96 1.96
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
28
Table 9 Result of regressions with panel data (4)
(1) (2) (3) (4) (5)
S_ROE S_ROA S_EPS S_DAratio S_Sti
Coverage -0.000336* -0.000114** 0.0000582 0.103 616.4
(-2.14) (-3.22) (1.12) (1.12) (1.86)
Industry 0.00434 -0.00436 -0.000242 7.542* 1102.2
(0.47) (-0.79) (-0.14) (2.44) (0.34)
ROE -0.0342 -0.00913 0.00105 5.328 5324.5***
(-0.95) (-1.52) (0.31) (1.33) (4.20)
Remuneration -0.0194 -0.000439 0.00193 2.025 12039.1*
(-1.29) (-0.17) (1.82) (1.54) (2.41)
Indptdirector -0.294*** -0.0258 0.0223 -40.21 70191.5***
(-7.29) (-1.08) (1.60) (-1.52) (4.63)
Auditcomitee -0.557 -0.128 0.0100 -88.47*** 1150799.4**
(-1.55) (-1.65) (0.12) (-3.98) (3.15)
Ctrldirector -0.271*** -0.0942*** -0.0305*** -19.42** 15236.4
(-6.59) (-6.72) (-12.39) (-3.10) (1.79)
Sdirector 0.0167** 0.00216 0.000681 -0.747 -7144.8**
(2.75) (1.06) (1.32) (-1.28) (-3.01)
Dual 0.673** 0.191* 0.0229 -10.47 -196490.5***
(2.67) (2.45) (1.21) (-0.34) (-7.12)
DAratio 0.00629*** -0.0000471 0.0000617* 0.424*** 2052.2*
(10.02) (-0.65) (2.23) (5.39) (2.26)
Litigation 0.307* 0.0820* 0.0490*** -3.215 83863.8**
(2.30) (2.28) (3.81) (-0.63) (3.03)
Constant 3.297*** 1.590*** 0.599*** 27.95 -166103.0**
(11.31) (8.56) (18.57) (0.85) (-2.97)
N 6625 6625 6622 6625 6625
R2 0.1218 0.0275 0.0096 0.0030 0.0771
Hausman test 0.0000 0.0000 0.0009 0.9626 0.0000
Model xtscc xtscc xtscc re xtscc
Mean VIF 1.70 1.70 1.70 1.70 1.70
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
29
IV. Legal implications for D&O insurance in Taiwan A. Improving D&O insurance and regulation
The empirical result in this paper indicates that there is a relationship between the purchase of D&O insurance and the quality of corporate governance in Taiwan. If the monitoring function of insurers is expected to develop more in the future, in the long-term the litigation systems and relevant regulations should be improved first. Many details in the regulations and insurance policies in Taiwan are not as complete as the United States. For example, some inconsistencies and conflicts in insurance policies also cause confusion.112 These not only influence the compensation for directors and corporations, but also the incentives to purchase D&O insurance. Also, the signal effect of D&O insurance implies that investors in the market may rely on such information.
Thus, the underwriting of insurers and the relevant regulation would be more important.
Compared with the United States, the authorities concerned in Taiwan have less experience in supervising the insurer due to the relatively short history of the development of D&O insurance. Even though the evidence proves that no adverse selection problem exists in the current market, relevant regulation will be necessary if these problems occur.