Using error-correction mechanism (ECM) test, the study has re-confirmed that no long run relationship of the output between foreign banks and domestic banks in Malaysia. This finding re-confirms the previous work by Tang (2001). Some policy implications have been noted in this study.
The study aims to re-examine the presence of a long run relationship between foreign banks and domestic banks in Malaysia, using ECM (error-correction mechanism) test that developed by Banerjee, et al. (1998). The ECM test accounts for the number of regressors (k) and the number of observations (T) used in cointegration analysis. However, previous study (Tang. 2001) does not account these issues by using the EngleGranger ADF test, and the results may be bias. Banerjee el al. (1998) have showed that the ECM test provided a more reliable test of cointegration as well as an unbiased estimate of the long run relation when the explanatory variables are weekly exogenous of the parameters of interest. The implication of the finding is to investigate the effectiveness of the Central Bank of Malaysia's policies to develop domestic banks by reducing the dominance of foreign bank activities.1
ECM Test (Banerjee, et al.,(1998) and the Empirical Results
Banerjee, et al. (1998) have proposed a new test for cointegration in as a single-equation framework which is based on the coefficient of the lagged dependent variable in an autoregressive distributed lag (ADL) model with leads of the regressors. The procedure depends upon the significance of the lagged dependent variable since this is equivalent to testing the significant of the error-correction terms in the ECM reparameterizaton of the model. Banerjee et al, (1998) recommend estimating the following (unrestricted) ECM regression by OLS:
[gamma](L)[Delta]Y^sub t^ = [alpha](L)'[Delta]X^sub t^ + [beta]Y^sub t-l^ + [theta]'X^sub t-l^ + [Sigma]^sup s^^sub l^a'^sub j^[Delta]X^sub t+j^ + [epsilon]^sub t^, where [gamma](L) and [alpha](L) are polynomials in the lag operator, L. When [beta] (or its t ratio) exceeds the critical values (provided in Banerjee et al. 1998), the null hypothesis of non-cointegration is rejected.
The output variable (loans plus deposits) is in aggregate level and is used as proxy for bank performance; LnY^sub f^ for output of foreign bank and LnY^sub d^ for output of domestic banks (Tang, 2001). Annual data from 1964 to 1996 have been used for cointegration analysis.2 The data are obtained from Monthly Statistical Bulletin (various issue, previous format), Central Bank of Malaysia. The both series are found to be nonstationry, or I(1) (see Tang, 2001, p. 45)
The computed test statistics (ECM test) are reported in Table 1. We use a lead of one year for unrestricted ECM regression since it is not feasible given 33 annual observations. Up to 3 years lag structure is used considering the use of annual data. The computed t-statistic of coefficient of lagged level dependent variable (LnYf^sub t-1^; LnYd^sub t-1^) lies below the 10 per cent critical value (-2.95), thus, the null of non cointegration is rejected for every lag structure. The finding indicates no long run relationship between foreign banks' output and domestic banks' output in Malaysia.
Using the cointegration concept, the present study has re-investigated the long run linkages among the output of foreign and domestic banks in Malaysia. …