國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
V. Empirical Results 5.1. The Lerner Index
Table 6 reports the summary statistics of the estimated Lerner index for the sample countries, evaluated by the CSSFM. The estimated values are all non-negative as anticipated. Specifically, the results indicate that banks in these countries generally operate under quite diverse competition environments. Japanese banks have the largest market power among the nine countries, followed by Singaporean, Chinese, Malaysian, and Hong Kong banks. Banks in Taiwan, Indonesia, the Philippines, and Thailand face rigorous competition, since their average Lerner indices are relatively low. It is eminent that the estimated indices have small standard deviations, primarily due to the fact that these indices are estimated by E(𝑢2𝑖𝑡|𝜀2𝑖𝑡)/𝑃𝑖𝑡, on the basis of (5), isolated from the impact of the error term of 𝑣2. Therefore, unlike the conventional Lerner index, our Lerner index is free from random shocks, by and large serving as a robust proxy for market power.
Table 6 Summary statistics of the estimated Lerner Index
MEAN ST.
DEV. MIN MAX
NO. OF TOTAL OBS.
NO. OF NEGATIVE
OBS.
CN 0.3099 0.0853 0.0542 0.5066 261 0
HK 0.2857 0.1390 0.0297 0.6414 83 0
ID 0.1991 0.0735 0.0264 0.6525 611 0
JP 0.3364 0.0697 0.0164 0.9589 1329 0
MY 0.3009 0.1040 0.0248 0.7440 332 0
PH 0.1748 0.0465 0.0106 0.4134 171 0
SG 0.3299 0.1391 0.0208 0.7534 85 0
TW 0.2096 0.0693 0.0362 0.7325 292 0
TH 0.1031 0.0416 0.0290 0.4394 248 0
29
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5.2. The Nexus of NIM and Noninterest Income
Tables 7 and 8 show the country-level empirical results from simultaneous equations system. We consider the interaction of risk factors as extra instrumental variables, as employed by Angbazo (1997) and Maudos and de Guevara (2004), to qualify the Sargan Test. Generally speaking, the signs and the significance of our parameter estimates are as expected. Although not all the explanatory variables for each sample countries demonstrate consistent results, we here hence discuss the general trend of each explanatory variable.
The Lerner index (LER) is significantly and positively correlated with the NIM, except for Indonesia and Thailand. This positive relationship is consistent with Maudos and de Guevara (2004). Thus banks in these countries can enjoy abnormal profits, should they possess strong market power. Conversely, Indonesia and Thailand demonstrates significant and negative nexus, this could be attributed to the fact that banks in these countries are operating under contestable markets15, where firms in low entry-level markets tend to set lower price to avert the entry of new firms. As Carbó and Rodríguez (2007) addressed, banks may charge lower margins to tempt clients and afterwards undertake nontraditional activities to earn extra profits. As a result, market power increases. Such phenomena constitute a negative nexus between LER and NIM while a positive relationship between LER and NII, which is exactly the case of these two countries. As the evidence of Maudos and Solís (2009), Maudos and de Guevara (2004) and Lepetit et al. (2008) indicates, as the operating expenses (MaCo) rise, a bank should manage to charge a higher margin to compensate for this cost increase; therefore a significantly positive relationship is found in Hong Kong, Japan, Malaysia, Singapore, and Taiwan. Although that of Indonesia presents
significantly negative influence, we couldn’t find relevant literatures to explain such a
15 See William A. Brock (1983).
30
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
result. Notwithstanding, we conjecture that Indonesian banks tend to lower NIMs to bring in clients, and conduct nontraditional activities with these clients afterwards while MaCo is high, utilizing such behavior as compensation for high operating expenses. This explanation can be partially supported by the high percentage of noninterest income to total bank income in Indonesian banks in Table 4. The degree of risk aversion (RiAv) is positively related to the NIM in Indonesia, Japan, the
Philippines, and Thailand, similar to Berger (1995), Angbazo (1997) and Rogers and Sinkey Jr. (1999). When the RiAv is increased, a bank tends to charge a higher NIM to compensate for the higher cost of equity financing. A noteworthy country is China, who demonstrates pivotal and negative nexus, opposite to the results of Berger, Hasan and Zhou (2010) and Heffernan and Fu (2010) on Chinese markets. For Chinese banks, pervasive political penetration in turn makes bank operation concern other political factors more than profitability. We surmise that under this condition, as equity capital dilates, banks might instead diminish NIMs to secure market power;
hence a negative nexus is observed. Risk factors (CrRi, InRi, LiRi) pervasively demonstrate significantly negative sway on NIM, indicating that the higher the risk exposure, the higher the margin. Angbazo (1997) and Maudos and de Guevara (2004) derive the same results considering that a bank requires higher NIM to maintain profitability when assuming higher risks. While there are still countries displaying disparate signs from those we anticipated, such as China, Japan, and Singapore. As we can see from Table 6, banks in these countries present stronger market power among our sample countries. Thus we conjecture that banks in these countries are inclined to utilize market power to charge higher NIMs when risk reserves are plentiful, to make up for the opportunity cost of risk reserves. Regarding the opportunity cost of reserves (OpCo), as the volume of non-earning assets increases, the empirical result reveals
31
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 7 Determinants of NIM (Eq. (1))
China Hong Kong Indonesia Japan Malaysia Philippines Singapore Taiwan Thailand Constant 0.018705*** 0.022557*** 0.076498*** -0.042615*** 0.010312** 0.04237* -5.26E-03 5.54E-03 1.16E-03 (3.23E-03) (7.80E-03) (0.011463) (6.54E-03) (4.34E-03) (0.023286) (4.64E-03) (6.37E-03) (8.99E-03) LER 0.068804*** 0.057508*** -0.273778*** 0.041263*** 0.07008*** 0.210141** 0.053078*** 0.01528 -0.062486*
(5.28E-03) (0.014086) (0.034085) (4.35E-03) (9.67E-03) (0.093733) (0.0128) (9.45E-03) (0.035126) MaCo -0.04855 0.31096*** -0.422888*** 1.72813*** 1.1245*** 0.108182 0.38491*** 1.13669*** -0.08261
(0.120899) (0.067943) (0.083661) (0.195259) (0.107629) (0.146475) (0.108446) (0.176064) (0.148646) RiAv -0.022124*** -5.03E-03 0.226142*** 0.034389*** -4.97E-03 0.065495** 0.015411 -4.40E-05 0.075169***
(5.35E-03) (0.012303) (0.020528) (8.81E-03) (0.012032) (0.028009) (0.013461) (0.010694) (0.011411) CrRi -0.02177 -0.065904** -0.01728 -0.01864*** 4.86E-03 -0.04348 0.180419*** 2.04E-03 -0.066585***
(0.016163) (0.030775) (0.016439) (5.73E-03) (0.010934) (0.037481) (0.015264) (0.055945) (0.015588) InRi 0.0000018** 5.06E-04 1.98E-04 -0.000302*** -1.98E-06 0.0049447*** 0.0010148*** -2.97E-04 1.66E-04
(8.89E-07) (6.66E-04) (3.14E-04) (1.00E-04) (3.71E-05) (1.69E-03) (2.15E-04) (3.72E-04) (5.17E-04) LiRi 0.00180678** -3.29E-03 -0.000061*** 0.018617*** -0.011832*** -0.020908*** -0.021558** -0.014807** 4.39E-03
(8.52E-04) (6.27E-03) (1.56E-05) (3.68E-03) (3.96E-03) (4.85E-03) (8.77E-03) (6.80E-03) (9.07E-03) OpCo 0.010568** 0.035841*** 0.113448*** -0.0219 0.016764*** 0.066165*** -0.004873*** -2.26E-03 -2.77E-03 (5.39E-03) (7.60E-03) (9.37E-03) (0.019816) (2.06E-03) (0.011092) (1.57E-03) (9.84E-03) (6.78E-03) ImPa 0.928411*** 0.4881*** 7.18E-03 0.627498*** 0.249455*** 0.400289*** 0.528693*** 0.045592 0.161303*
(0.089413) (0.108153) (5.05E-03) (0.065877) (0.073944) (0.082787) (0.03508) (0.031458) (0.085306) AsSi -0.013156*** -0.006886*** -0.014925*** -0.002928*** -0.004461*** -0.022711*** -0.006228*** 6.98E-04 -2.20E-03 (1.41E-03) (1.50E-03) (4.86E-03) (5.03E-04) (1.62E-03) (7.56E-03) (1.44E-03) (1.07E-03) (2.36E-03)
32
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
BuSc 0.01191*** 0.004795*** 0.018322*** 0.0046762*** 0.00282078* 0.016265** 0.0065735*** -9.84E-04 0.00545003**
(1.40E-03) (1.17E-03) (4.82E-03) (5.06E-04) (1.63E-03) (7.32E-03) (1.03E-03) (9.09E-04) (2.21E-03) NonIn -9.89E-05 -1.10E-04 6.71E-07 0.0000039*** -0.000132*** -1.75E-04 -0.000071*** 1.80E-06 0.000059587*
(8.55E-05) (6.96E-05) (3.02E-05) (1.27E-06) (4.55E-05) (3.28E-04) (2.45E-05) (2.74E-06) (3.31E-05) Sargan Test 21.0318 16.4287 15.396 25.3396 18.1969 18.9544 19.7488 8.15932 9.85344
[.177] [.288] [.352] [.189] [.198] [.167] [.138] [.881] [.773]
Sum of squared
residuals 5.57E-03 7.14E-03 0.844781 0.020088 0.02773 0.150063 3.73E-03 0.012879 0.059807 Std. error of
regression 4.72E-03 0.011502 0.03884 3.98E-03 9.21E-03 0.030625 8.47E-03 6.70E-03 0.015561
Table 8 Determinants of NII (Eq. (2))
China Hong Kong Indonesia Japan Malaysia Philippines Singapore Taiwan Thailand Constant 8.03115*** 23.6999*** 0.385553 1.69525 0.691093 -1.11267 3.76025** 3.45464 -1.83973*
(2.53033) (4.55324) (1.05451) (1.23416) (0.628299) (2.55588) (1.70539) (7.26548) (0.966223) LER -0.00316** 2.87003 2.71146 -3.80286** -2.40191*** -3.20056 1.17067 17.5059** 0.380604
(1.35E-03) (3.31496) (2.23001) (1.74546) (0.636155) (4.48377) (1.23375) (8.48874) (1.85403) MaEf -14.6355 95.7799*** 0.105362 7.25908** 16.3995** 2.15872 -26.5675** 33.412 2.10994
(12.5917) (34.0979) (1.68033) (3.59666) (6.87797) (2.75327) (12.0351) (32.5109) (1.46357) RiAv -4.258*** 15.4909 -5.01285** 4.19084 -0.58654 5.27017** 0.808429 42.4889 0.41958
(1.32044) (17.7232) (2.16124) (5.11563) (1.15949) (2.67657) (6.43206) (42.611) (1.19713) CrRi 6.63449 -19.7285 -8.40876*** -1.96548 1.65688 2.97563 8.52977 -21.65 -0.80123
(4.40989) (20.2566) (2.55285) (3.03692) (1.29157) (2.27012) (6.6973) (40.3606) (0.643999) InRi -4.52E-04 0.0214 0.033278 -0.5665*** 0.027199*** -0.03092 0.154513* 0.017334 2.64E-03
33
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(4.51E-04) (0.295646) (0.032189) (0.13229) (9.96E-03) (0.147865) (0.079568) (0.2678) (0.018129) LiRi 0.484846*** 26.8752*** -0.02129 13.5157*** 0.331659 -0.27727 3.94179* 3.20481 0.777331
(0.18855) (4.93867) (0.013254) (2.73836) (0.224097) (0.922872) (2.2649) (6.45476) (0.588285) AsSi -0.50616*** -24.8019*** 1.20121 -8.68327*** 1.27394*** -2.07935 -0.84242 -17.2185 0.09387
(0.161552) (7.66121) (0.838981) (2.14045) (0.417731) (1.89399) (2.27461) (14.0457) (0.274926) LiaSt 0.282938* 21.9006*** -1.33141 8.76411*** -1.19543*** 2.455 0.866958 17.0101 0.125304
(0.15611) (7.50563) (0.863237) (2.22204) (0.407561) (1.85194) (2.30085) (14.5434) (0.287836) FeCom 222.764*** -10.2953 101.967** 191.016*** -2.81545 123.071* 184.281*** 51.74 5.32969
(58.6914) (51.0805) (44.0269) (52.9548) (28.0912) (64.9329) (24.1094) (252.909) (20.5787) Div -27.1069*** -120.584*** 2.5835 20.6757* -12.0879*** -1.04248 -72.7419*** -6.20813 7.62402**
(9.71862) (31.6583) (2.18127) (11.5154) (2.52189) (4.03523) (8.31213) (46.0768) (3.69502) NIM 0.834837 -17.9581*** 2.94822*** -1.1255 -0.24991 -0.26396 -10.383*** -14.5482 1.43663*
(0.743736) (5.19432) (1.11218) (1.10943) (0.548259) (1.72032) (2.18363) (9.15213) (0.772212) Sargan Test 21.0318 16.4287 15.396 25.3396 18.1969 18.9544 19.7488 8.15932 9.85344
[.177] [.288] [.352] [.189] [.198] [.167] [.138] [.881] [.773]
Sum of squared
residuals 522.089 2537.94 58362.8 272432 2213.47 1988.54 2395.47 312424 4017.79
Std. error of
regression 1.44512 6.85557 10.2088 14.6752 2.60173 3.52539 6.78725 32.9937 4.03316
34
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
that a bank would try to price higher to counteract the negative effect in China, Hong Kong, Indonesia, Malaysia, and the Philippines, similar to Angbazo (1997) and Maudos and Solís (2009); whereas Singaporean banks displays unusual negative influence on NIM. Ditto to the empirical evidence of Maudos and de Guevara (2004) and Lin et al. (2012), the implicit interest payment (ImPa) displays a significantly positive influence on NIM. Asset size (AsSi) and business scale (BuSc) display significance and signs as anticipated as well.
Turning to the noninterest income, as we can see in the last regressor in Table 7, we do not observe consistent relationship between NIM and NII across the sample countries, nor does it demonstrate a regular pattern in accordance with the undertaking volume of nontraditional activities across country. What is noteworthy is that
countries presenting significant relationship between the two variables are those with less engagement in nontraditional activities among our sample countries (see Table 4).
We attempt to explore the interrelationship further in the full sample regression (see Appendix Table 1) to examine the comprehensive results and discover significantly positive relationship between these two variables, which is consistent with Heffernan and Fu (2012) and Lin et al. (2012), illustrating that banks in these countries overall benefit from more engagement in nontraditional
activities.16
Pertaining to the determinants of NII, we derive similar results to that of Nguyen (2012). The signs of the explanatory variables in this regression in general match our expectation while the levels of significance are not as ideal as anticipated. And, same as the determinants of NIM, since not all the explanatory variables for each sample
16 Considering the capacity of our study, we provide the empirical results of full sample regression in Appendix Table 1. In general, the results are similar to the country-level regressions, except for the significant and specific relationship between NIM and NII. Here we thus supplement the results to provide another avenue to explore the interrelationship.
35
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
countries demonstrate consistent results, we manage to explore the general trend of each explanatory variable.17
Table 8 shows that our Lerner index displays a significantly negative influence on NII merely in three countries, indicating that these Asian banks generally alter their operation policy towards specialization when possessing stronger market power.
There are only banks in Taiwan presenting significantly positive nexus, heightening the possibility of loss-leader behavior in this market, while the concern of loss-leader behavior seems to be unnecessary to the remaining sample countries. The proxy for liquidity risk (LiRi) presents significantly positive influence on NII in China, Hong Kong, Japan, and Singapore, indicating opposite result as we expected. It illustrates that when a bank possesses more liquid assets, these Asian banks are apt to allocate their resources towards nontraditional activities to make up for the opportunity cost of liquidity reserves. As for the proxy for net fees and commissions (FeCom), since fee and commission incomes are the main contributors for noninterest income, a
significantly positive relationship is verified in five out of the eight countries. The variable Div, proxied by the ratio of other earning assets to total assets, exhibits a significant and negative relationship with NII, indicating that noninterest incomes for these countries probably are predominantly fees and commissions incomes. Since as the volume of other earning assets increases, resources and efforts are afterwards diverted from the more lucrative fees and commissions activities, resulting in lower NIMs. Identically, from our country-level regressions, we do not observe consistent relationship and significance between NII and NIM, neither does the full sample regression, same as the conclusion we draw from the empirical evidence of equation 1.
17 With a simultaneous system, we manage to deal with the endogeneity issue and explore the simultaneity between NIM and NII. Our main focus is the nexus of NIM and NII and the determinants of NIM. As a consequence, we generally examine the determinants of NII instead of thoroughly discussing the significance and signs by each of all the explanatory variables in equation 2.
36
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Yet Singapore and Thailand are the two countries worth thoroughly examining the nexus of NIM and NII. From Table 7 and Table 8 we can observe that the nexus exhibits mutually significantly negative influence, which leads to the conclusion that it might be inappropriate for banks in these two countries to engage too much in nontraditional activities.
With an eye to comparing the robustness and estimation efficiency of the two market power indices, Lerner Index and HHI, we substitute HHI for the Lerner Index and re-run the simultaneous regressions. The empirical results are provided in
Appendix Table 2, which are in general similar to those in Tables 7 and 8. However, since the results in Appendix Table 2 present higher sum of squared residuals and deteriorated significance of all explanatory variables, the Lerner index appears to be a better proxy for market power and a better explanatory variable than HHI, consistent with Maudos and de Guevara (2004). The number of explanatory variables which present the expected signs and significance is 75 of the former versus 67 of the latter.
Inter alia, the exploration of the relationship between NIM and market power displays distinct results. The significantly positive nexus between the renovated Lerner Index and NIM is consistent with quite a few relevant literatures; whereas estimation with HHI as a regressor demonstrates a significantly negative relationship, which lacks evidence from academia. In sum, the renovated Lerner Index does contribute its robustness and estimation efficiency to our regressions.
37
‧
國立 政 治 大 學