4. EMPIRICAL RESULTS
4.2. The Findings of the Four-Stage DEA
4.2.2. Stage Two: Quantifying the Effect of the Operating Environment
There are two regression equations from equation (12), one for each input as below.
xsi1
= f1(Ei1
, 1, ui1
) (20)
xsi2 = f2(Ei2, 2, ui2) . (21)
The dependent variables (xsi1and xsi2) are total radial movement plus slack movement based on the first stage DEA results. Here, Ei1
and Ei2
are the vectors of environmental variables for ISF i that may affect the utilization of input.
The four independent variables are VOL for annual brokerage volume in an ISF, which is deeply influenced by Taiwan’s trading volume, DUR for the duration or the years of service in the securities market, ASV for the ISF’s asset value and one dummy variable FHC to show if this ISF is a subsidiary of an FHC. The purpose of the FHC dummy is to investigate whether the FHC would benefit from its ISF subsidiary or not.
A (positive) negative coefficient on these environmental variables suggests that the environment is (un)favorable for a DMU, since it is associated with (greater) less excess use of inputs.
This regression result indicates that the duration of establishment (DUR) has a significantly negative coefficient in two equations, suggesting that it is a favorable environmental variable. It shows that the ISFs with a longer duration are able to draw the customers’ attention, build up customer loyalty, and create a lot of wealth from the brokerage revenue. Experienced ISFs are able to make less discounted
expenditures and utilize branch resources.
The FHC subsidiary variable (FHC) has a significantly positive coefficient in two equations. This suggests that an ISF under FHC is in an unfavorable operating environment. The empirical result in the first stage has shown that it is most of the non-efficient ISFs that are able to join FHCs. Fourteen law-induced FHCs were established through persuasion from Taiwan’s regulatory authority. This might reveal that politics are possibly intertwined with economic activities in Taiwan.
Consequently, the purpose of forming an FHC is not to leverage the synergy among subsidiaries and to improve their efficiency score, but instead an FHC is turned into a negative factor from its securities subsidiary. This result is consistent with the empirical finding for Malaysian banks in 2006. Chong et al. (2006) indicated that the forced merger mechanism destroys shareholders’ value. Contrary to the findings on voluntary mergers in the United States and Europe, Malaysian acquiring banks that had merged the other target banks have a significantly negative cumulated abnormal return under the forced merger scheme. Moreover, FHC’s securities subsidiaries diversify their dedication on the brokerage business in Taiwan due to on-going mergers from FHCs and the cross-selling of banking products. Plus, Taiwanese regulatory authority limited banking branches to not sell securities products to customers directly due to the firewall issue and protecting small-scale securities firms from banks competition. It is another major reason corrupting the one-stop shopping synergy. It also prompts the ISFs under FHCs not to be able to leverage the banking resources and furthermore decrease branches.
The annual sales amount has an insignificant coefficient on two equations in model I of Table 4-3. It shows that ISFs could increase their share of the brokerage
asset value of each firm has also an insignificant coefficient on two equations in model I of Table 4-3. More assets is not proven to be favorable or unfavorable to the securities firms.
The coefficients of annual sales volume variable (VOL) and asset value (ASV) are insignificant and are hence omitted from the equation of slack prediction. Those environmental variables with significant coefficients such as DUR and FHC are included for slack prediction.
TABLE 4-3. Tobit Regression Results
Model I Model II
function -89.3665 -56.1899 -90.2002 -56.2851
Note: 1. Numbers in the parentheses are standard errors;
2. **, *, and†indicate significance at the 1%, 5%, and 10% levels, respectively;
3. The sample size is 54.
The parameter estimates present in model II of Table 4-3 and the following Tobit regression models (22) and (23) shown below are used to adjust the original input data according to equation (13).
xŝ1 = 5.35987 – 0.31405 DUR +1.8928 FHC (22)
xŝ2 = 1.18623 – 0.090371 DUR +0.0643657 FHC (23)
Table 4-4 summarizes the predicted slacks and maximum predicated slacks for all inputs based on equation (14). The adjusted data control influences the external operating environment. In 2002, one year before most FHCs’ establishment, the result that Taiwan, Sinopac and Grand Cathay securities firms under FHCs contributed to the maximum predicted slack reveals an unfavourable external environment under FHCs. In 2003 and 2004, the maximum predicted slacks are from Fuhwa and Mega securities firms, which own the least favourable external environment including the shortest duration in the securities industry and a subsidiary of an FHC. The maximum predicted slacks are contributed by Fuhwa securities firms in 2004 and Mega securities firms in 2005, respectively which are all securities subsidiaries in FHC. This predicted slack result is also consistent with the result of the parameter estimates above.
TABLE 4-4. Predicted Slacks and Maximum Predicted Slacks
4. Yuanta Core Pacific 41 0 -7.516 -2.519
5. Capital 14 0 0.963 -0.079
6. President 14 0 0.963 -0.079
7. Polaris 14 0 0.963 -0.079
8. MasterLink 13 0 1.277 0.011
9. SinoPacf 14 1 2.856 0.565
10. Grand Cathayf 14 1 2.856 0.565
11. Jih Sunf 41 1 -5.623 -1.875
2002
12. Taiwan International 14 0 0.963 -0.079
Maximum predicted slack [Maxk{xŝik}] 2.856 0.565
1. Fubonf 20 1 0.972 0.022
2. Taiwanf 15 1 2.542 0.474
3. KGI 15 0 0.649 -0.169
4. Yuanta Core Pacific 42 0 -7.830 -2.609
5. Capital 15 0 0.649 -0.169
6. President 15 0 0.649 -0.169
7. Polaris 15 0 0.649 -0.169
8. MasterLink 14 0 0.963 -0.079
9. SinoPacf 15 1 2.542 0.474
10. Grand Cathayf 15 1 2.542 0.474
11. Jih Sunf 42 1 -5.937 -1.966
12. Taiwan International 15 0 0.649 -0.169
13. Fuhwaf 7 1 5.054 1.197
2003
14. Megaf 14 1 2.856 0.565
Maximum predicted slack [Maxk{xŝik}] 5.504 1.197
1. Fubonf 21 1 0.658 -0.068
2. Taiwanf 16 1 2.228 0.384
3. KGI 16 0 0.335 -0.260
4. Yuanta Core Pacific 43 0 -8.144 -2.700
5. Capital 16 0 0.335 -0.260
6. President 16 0 0.335 -0.260
7. Polaris 16 0 0.335 -0.260
8. MasterLink 15 0 0.649 -0.169
9. SinoPacf 16 1 2.228 0.384
10. Grand Cathayf 16 1 2.228 0.384
11. Jih Sunf 43 1 -6.251 -2.056
12. Taiwan International 16 0 0.335 -0.260
13. Fuhwaf 8 1 4.740 1.107
2004
14. Megaf 15 1 2.542 0.474
Maximum predicted slack [Maxk{xŝik}] 4.740 1.107
1. Fubonf 22 1 1.914 0.294
2. Taiwanf 17 1 1.914 0.294
3. KGI 17 0 -8.458 -2.790
4. Yuanta Core Pacific 44 0 0.021 -0.350
5. Capital 17 0 0.021 -0.350
6. President 17 0 0.021 -0.350
7. Polaris 17 0 0.335 -0.260
8. MasterLink 16 0 0.021 -0.350
9. SinoPacf 17 1 1.914 0.294
10. Grand Cathayf 17 1 -6.566 -2.146
11. Jih Sunf 44 1 1.914 0.294
12. Taiwan International 17 0 2.533 0.373
13. Fuhwaf 9 1 2.228 0.384
2005
14. Megaf 16 1 6.939 1.740
Maximum predicted slack [Maxk{xŝik}] 6.939 1.740