Research Data and Results of the Equity Holder Model
6.3 Analysis and Discussions of the Results
6.3.1 Empirical Results
The existence of monitoring by shareholders was tested by regressing bank-risk variables and indicators of general economic conditions on stock returns. Table 6.5 presents the results from the panel least square regression of the monitoring model, as described in Chapter 3. The same monitoring model was regressed for subsamples of private and state owned banks. The first column presents regression results for all samples during the period Quarter I of 2003 to Quarter IV of 2011. The second and third columns provide regression results for two sub-samples:
private banks and state banks. The regression on private and state banks is aimed at examining the difference, if any, of shareholder sensitivity on these types of banks.
This study initially estimated a model of market discipline by equity holders employing a GMM estimation procedure, similar to the approach adopted to investigate market discipline by depositors (Chapter 4) and debt holders (Chapter 5). However, the regression results failed to meet the reliability and validity conditions and were deemed unsatisfactory. The results presented in the current chapter are that of a random walk mode, as introduced by Burton Malkiel in 1973. Random walk theory argues that changes in stock prices are independent of each other, therefore the movement or trend of a series of prices or markets in the past could not be used to forecast future stock price movements (Malkiel, 2007). Based on this theory, the lag of equity return might not be a relevant determinant of equity return since successive price changes are independent (Fama, 1970). Besides, initial data analysis could not find any trends.
181 Table 6.5 Regression Results: Equity Return Variable
All Banks Private Banks State Banks VARIABLES Lag
Expected Signs
(1) EQ_RT
(2) EQ_RT
(3) EQ_RT
CAR L1 (+) 0.361902** 0.407999** 0.177284
(0.180629) 0.189474 (0.827451)
NPL L1 (-) 0.5083 0.512063 0.69699
(0.407665) 0.445096 (1.084707)
ROA L1 (+) 0.2083 0.962173 -2.180225
(0.833335) 1.032343 (3.572922)
OPEX L1 (-) 0.0265 0.374642 -0.271875
(0.24083) 0.353739 (0.360709)
NIM L1 (+) 0.9772 1.143825 2.142938
(0.634494) 0.790827 (2.839165)
LDR L1 (-) -0.241045* -0.301137** -0.37267
(0.134689) 0.14504 (0.721251)
BANK_SIZE 0.0030 0.007636 0.021281
(0.007497) 0.009732 (0.085646)
GDP_RT -1.0418 0.26885 -3.972123
(1.506191) 1.762807 (3.159119)
INF_RT -1.099258*** -0.799935** -1.82333**
(0.347016) 0.405623 0.774833
EXC_RT -0.387578** -0.359038* -0.490702
(0.178384) 0.207603 0.377876
Constant 3.64981** 3.004988 4.709858
(1.653292) 1.93766 3.619173
R-squared 0.0862 0.0856 0.1267
Adj. R-squared 0.0617 0.0516 0.0338
Prob(F-statistic) 0.000 0.006 0.209
Number of Observations 385 280 105
Number of bank 11 8 3
This table presents the results from the panel least square regression. Standard errors are presented in parentheses.
The estimation uses quarterly observations over the period 2003–2011. The dependent variable is EQ_RT (equity return). The independent variables include: CAR, NPL, ROA, OPEX, NIM, LDR, and BANK_SIZE (total asset of banks). Control variables for general macroeconomic conditions include: GDP_GR (the growth rate in GDP);
INF_RT (inflation rate); and EXC_RT (the annual average of exchange rate IDR/USD scaled in IDR000). All variables are transformed using the natural logarithmic transformation.
* Indicates statistical significance at the 10% level (2-tailed)
** Indicates statistical significance at the 5% level (2-tailed)
*** Indicates statistical significance at the 1% level (2-tailed)
182 The present study, therefore, used the static panel data model to investigate the relationship between equity returns and bank fundamentals. Within the static panel analysis, the impacts of the determining factors have been simultaneously estimated under a pooled least squares- regression approach. The dependent variable of the model is equity returns calculated as the change in stock price over the previous period53. As can be seen in Table 6.5, the independent variables could jointly explain changes in the dependent variable with p-values less than 0.01 for the whole sample (column 1) and the private banks (column 2), whereas the estimates for state banks is not significant (column 3). However, the model could only explain approximately 8.6%
of the total variation in entire equity returns, with figures for private and state banks at 8.6% and 12.7% respectively. These figures indicate that, apart from the six CAMEL ratios and the three macroeconomic variables that are used in these models, there are other variables that could exert a stronger influence on the movement of equity returns. These figures reflect a common state in developing economies where market risk factors have a tendency to be more dominant and overshadow the risk of individual firms (Levy-Yeyati et al., 2004a).
As shown in the table, of the six CAMEL indicators used in this study, only CAR and LDR had significant impacts on equity returns. The CAR, as expected, had a positive impact in the whole sample at p = 0.05, while LD, had a significant negative impact at p = 0.10. The regression results on the macroeconomic variables indicate that inflation rate has a negative impact at p = 0.01 and exchange rate a negative impact at p= 0.05.
The regression results for the private banks are presented in column 2 of Table 6.5. As shown, both the CAR and LDR variables of the private banks have significant impact at p = 0.05 and, similar to the whole sample, inflation and exchange rates have statistically significant impact on the equity returns of the private banks at p-values of 0.05 and 0.10 respectively. For state banks, as can be seen in column 3, none of the six CAMEL variables have a significant impact on equity returns, and inflation rate is the only macroeconomic variable that has a significant impact on equity returns, at p = 0.05.
53 As alternative dependent variables, this study uses equity return minus the central bank benchmark rate and equity return minus the growth of market index. However, the regression results using these dependent variables were unsatisfactory.
183 6.3.2 Robustness Check
In order to check the robustness of the estimates, regressions were estimated excluding several independent variables that were strongly correlated. As stated in Section 6.2.3 , ROA and OPEX were strongly correlated. To test the possible impacts of multi-collinearity, the study compared the regression results with and without OPEX and ROA. The results of these reduced models were consistent with the regression results presented in Table 6.5, where amongst the CAMEL variables, only CAR and LDR had a statistically significant influence on equity returns and, among the macroeconomic variables, only inflation and exchange rate had a significant impact on equity returns (see Appendix C.1 and C.2 for more detail).
Tests were conducted to verify the presence of fixed effects and random effects in the panel data, as suggested by Baltagi (2008). Fixed effects are tested by an F-test, while random effects are tested by Breusch and Pagan’s Lagrange Multiplier (LM) test. The former compares a fixed effect model and OLS (ordinary least squares) to see how much improvement could be achieved by the fixed effect model, while the latter contrasts a random effect model with OLS. Using a fixed effect method, the goodness-of-fit was increased from 8.62% to 12.01%. However, the null hypothesis of the cross-section F-test was rejected, implying that the fixed effect method does not add any significant improvement to the estimated model. Under this cross-section fixed effect model, the estimation results indicate that only CAR and inflation rate had a significant impact on equity returns. The results of the random effect model did not produce any significant changes to the original one.