Cointegration Test

Under the unit root test, the results show that the variables of financial intermediation,
money (MMB) and output are stationary at the first difference [I(1)]. Continuously, all
the variables will be tested in cointegration test, by using the Johansen test statistics,
imply that if financial intermediation and output are cointegrated, so there is a long term
equilibrium relationship between these variables.

Vector Error Correction Model (VECMs)

The cointegration test is used to prove the cointegrating relationship between financial
intensity and output in the long run. However, they offer no information about the
adjustment speeds of the variables to deviations from their common stochastic trend.
Indeed, in the short run there may be disequalibrium. Therefore, we need the error term
as the ‘equilibrium error’ which can corrected gradually the deviations from long run
equilibrium through a series of partial short run adjustment. Here, the vector error
correction mechanism will corrects for the disequilibrium. A VECMs is a restricted VAR
that has cointegration restriction built into the spesification. The VECMs can be written
as follow:
k-1 k-1
 x1,t =  1 +   1,i  x1,t-i +   1,i  x2,t-i +  1(  x1,t-1 +  x2,t-1 ) ................ (3)
i=1 i=1

where x1 is output (GDP) and x2 is a measure of the financial intensity or money. The last
component of eq. (3) is the error correction term (ECT), which is formed with the
elements of the cointegrating vector and enters the model at a single lag. Since the
cointegration test have proved the long run relationship between financial intensity and
output, therefore, the sign and size of the coefficient on the ECT reflect the direction and
speed of adjustments in the dependent variable to temporary deviations from this
relationship. A negative loading on the intermediation measure in the cointegrating
vector coupled with negative and significant coefficients on the ECT equation in eq. (3)
would imply that output rises in response to fluctuations that depress the value of the
 2

stationary combination. The sign of the loading would also be consistent with a role for
increased intermediary activity as a source of these negative deviations.

VECMs Causality Test

Even though the cointegration and VECMs test have identified the cointegrating
relationship between financial intensity and output, but these tests do not imply the
causation. Therefore, we will establish the causality test by using VECMs causality test
to develop a leading role for the variable in the fluctuations of another. Under this
causality test, we can examine the existance of unidirectional causality or bilateral
(feedback) causality from financial intermediation and money to output and vice versa.

Example

The parameter estimates of the cointegrating model are reported in Table 3. The
Johansen test reject the null hypothesis at 5% which proves the existence of cointegrating
relationship between financial intermediation and output, and also money and output, in
the long term. However, the Johansen test result for LOAN also rejects the null
hypothesis that at least exist one cointegrating vector between LOAN and GDP.
Therefore, this result indicates two cointegrating equations at 5% significant level for
LOAN variable.

Insert Table 3

To examine the short-term adjustment process to deviations from the long-term

equilibrium relationship, the parameters of the error correction models of equation (3)
are estimated and presented in Table 4. The results of cointegrating vector show that
short run changes in financial intermediation and money have significant effects on GDP.
Thus, it implies the existance of cointegration between the variables in the short run.
Given the negative sign on the financial intensity measure in the cointegrating vector,
 3

increases in this measure could provide the impulses that drive the upward adjustments in
output, though downward movements in past output would also reduce the value of the
stationary combination. However, the negative and significant coefficient on the ECT in
the output equation indicates a rapid response of output to deviations from its long run
relationship with financial intensity measure and money. In particular, negative
deviations from the stationary relationship are ‘corrected’ by increases in output. In
general, the findings of VECMs test suggest that there is cointegrating relationship in
short term between financial intermediation and money with GDP.

Insert Table 4

Finally, Table 5 presents the result of VECMs test for each variable. From the table, we
can prove the existance of bilateral causality from financial intermediation to GDP and
from GDP to financial intermediation. Similarly, the results also indicates the existance
of bilateral causality from money (MMB) to GDP and from GDP to money. Overall, the
findings from causality test suggest that the financial intermediation and money will
affect the output fluctuation, and later, output will significant in explaining and affecting
the fluctuation of financial intensity and money.

Insert Table 5

Table 3
Johansen Test Statistics For Cointegration Between Log GDP and Measure of Money and Financial
Intermediation

Malaysia Trace Statistics (Likelihood Ratio) Eigen Value Normalized

1959-2000 r=0 r1 r=0 r  1 Cointegration Vector
LCBA 26.3953* 6.5258 0.3915* 0.1505 1, -1.4514
LBANKA 26.2644* 6.8008 0.3853* 0.1564 1, -1.4419
LFIA 24.5643* 6.7191 0.3599* 0.1546 1, -1.3662
LMMB 29.8209* 6.2471 0.4453* 0.1446 1, -1.4958
LLOAN 32.1244* 14.1662* 0.3690* 0.3046* 1, -1.4339
Note: The lag length for this analysis is 1 lags. The columns labelled r=0 test a null of
cointegration, while the r  1 columns test a null of at most one cointegrating vector. All
 4

variables were tested in lag form. The normalized cointegration vector refers to cointegration
vector of GDP and financial intensity.
* denote rejections of the null at the 5%

Table 4
Vector Error Correction Model Test For Difference and Log GDP, Financial Intermediation and Money

Cointegrating Vector Error Correction Term (ECT) R2

D(LCBA) 1, -1.4514 -0.0635 -0.0341

(-2.3047)*

D(LBANKA) 1, -1.4419 -0.0655 -0.0361

(-2.1809)*

D(LFIA) 1, -1.3662 -0.0236 -0.0063

(-1.7223)**

D(LMMB) 1, -1.4958 -0.1211 -0.0515

(-2.0832)*

D(LLOAN) 1, -1.4339 -0.0162 -0.1558

(-2.7834)*
Note: In the ECT column, the t-statistics is in parentheses, with significant level.
* denote rejections of the null at the 5%.
** denote rejection of the null at the 10%
All variables were tested in lag and difference form.

Table 5
VECMs Causality Test For Difference and Lag GDP, Financial Intermediation and Money
            
Cointegration Vector Null Hypothesis ECT Decision
D(LGDP) does not Granger cause D(LCBA) -0.1454 Reject
(-4.8646)*
1, -1.4514
D(LCBA) does not Granger cause D(LGDP) -0.0635 Reject
(-2.3047)*
            
D(LGDP) does not Granger cause -0.1475 Reject
D(LBANKA)
(-4.8042)*
1, -1.4419
D(LBANKA) does not Granger cause -0.0655 Reject
D(LGDP)
(-2.1809)*
 5

            
D(LGDP) does not Granger cause D(LFIA) -0.0573 Reject
(-4.5270)*
1, -1.3662
D(LFIA) does not Granger cause D(LGDP) -0.0236 Reject
(-1.7223)**
            
D(LGDP) does not Granger cause D(LMMB) -0.4235 Reject
(-5.3759)*
1, -1.4958
D(LMMB) does not Granger cause D(LGDP) -0.1211 Reject
(-2.0832)*
            
D(LGDP) does not Granger cause -0.1158 Reject
D(LLOAN)
(-3.4171)*
1, -1.4339
D(LLOAN) does not Granger cause -0.0162 Reject
D(LGDP)
(-2.7834)*
            
Note: The lag length for this analysis is 3 lags.
* denote rejection of the null at 5% level.
** denote rejection of the null at 10% level.
All variables were tested in lag and difference form.