3.3.3. Cointegration results
The number of cointegrating vectors was tested based on the
assumption that the series have linear deterministic trend and that there is an
intercept. The null hypothesis that there is no cointegrating vector concerning
the trace statistics could not be rejected since its value was greater than the
5% critical value. Since we failed to reject the null hypothesis with no
Cointegrating equation this indicates that any cointegrating equation has not
been found concerning the trace statistic. However, the maximum eigenvalue
statistics test indicates one (1) cointegrating equation among the variables.
Thus, there was one Cointegrating equation indicating long-run equilibrium
relationship among the variables (Table 3.6).
3.3.4. Vector Error Correction Model (VECM)
The use of the vector error correction model was necessitated
by the fact that the time series were nonstationary in their levels except in
their differences, coupled with the fact that the variables were cointegrated.
In case there was no Cointegration, VECM was not required. The VECM was applied
in order to evaluate the short run properties of the cointegrated series and to
find the real link between the variables. It enables the integration of the
short-run fluctuations. The coefficient of the error correction term must be
negative to report a force towards the long-run equilibrium. The regression
equation of the VECM is express as:
(2)
(3)
(4)
Where Ä is the first difference operator, is the error
correction term lagged one
period, are the short-run coefficients are constant terms,
are coefficient of the vectors and are
the white noise terms.
Table 3.7 Vector error correction estimates
|
Cointegrating Eq:
|
CointEq1
|
|
|
GDP(-1)
|
1.000000
|
|
|
CP(-1)
|
0.765
(0.144)
[ 5.279]
|
|
|
FD(-1)
|
-1.311
(0.206)
[-6.355]
|
|
|
C
|
-25.714
|
|
|
Error Correction:
|
D(GDP)
|
D(CP)
|
D(FD)
|
|
CointEq1
|
-0.061
(0.031)
[-1.935]
|
0.183
(0.089)
[ 2.059]
|
0.283
(0.058)
[ 4.805]
|
|
C
|
0.018
(0.009)
[ 1.858]
|
0.021
(0.027)
[ 0.766]
|
0.035
(0.018)
[ 1.922]
|
|
R-squared
|
0.089
|
0.100
|
0.377
|
|
Adj. R-squared
|
0.065
|
0.076
|
0.361
|
|
Sum sq. resids
|
0.145
|
1.141
|
0.496
|
|
S.E. equation
|
0.062
|
0.173
|
0.114
|
|
F-statistic
|
3.744
|
4.243
|
23.090
|
|
Log likelihood
|
55.560
|
14.374
|
31.019
|
|
Akaike AIC
|
-2.678
|
-0.619
|
-1.451
|
|
Schwarz SC
|
-2.593
|
-0.534
|
-1.366
|
|
Mean dependent
|
0.018
|
0.021
|
0.035
|
|
S.D. dependent
|
0.064
|
0.180
|
0.143
|
|
Determinant resid covariance (dof adj.) Determinant resid
covariance
Log likelihood
Akaike information criterion
Schwarz criterion
|
1.40E-06
|
|
|
1.20E-06
|
|
|
102.433
|
|
|
-4.671
|
|
|
-4.291
|
|
NB: Standard errors in ( ) & t-statistics in [ ]
Short run dynamic
variable coefficient Standard error t-statistic
D(GDP) -0.061* 0.032 -1.935
(*), indicates significant at 10% level
The coefficient should be negative and significant to show
that the long run relationship exists among the variables and that deviation
from equilibrium in the previous year is adjusted back to equilibrium in the
current year. In other words, this indicates a long-run error correction among
the variables. In particular, given that the
coefficient of is -0.061, this means that 6.1% of the
disequilibrium in the previous
year were adjusted back to equilibrium in the current year.
|
Long run dynamic
|
|
|
|
|
Variables
|
Coefficient
|
Standard error
|
t-statistics
|
|
GDP(-1)
CP(-1)
FD(-1)
C
|
1.000000 0.765*** -1.311*** -25.714
|
0.145
0.206
|
5.279
-6.355
|
(*), (**) (***) indicates 10% 5% and 1% significance level,
respectively.
Based on the long run dynamic analysis the relationship between
GDP, CP and FD can be expressed in terms of the coefficients as
(5)
We interpreted the coefficients in terms of elasticity. The
GDP increased by 1.311 percent with an increase of one percent of FD. It had
significant influence on the economic growth of Niger. However, an increase of
one percentage of CP led to a decrease in GDP by 0.765 percent, which confirms
the ambiguity of the sign of CP. With effective allocation of resources, it
will be correlated positively with economic growth; otherwise not, especially
in countries where the financial systems are not well developed. Generally, CP
is expected to have positive effect on investment leading to economic growth;
Demetriades and Hussein, (1996). Contrary, we found negative and significant
impact of CP on economic growth in Niger. This result could be explained by the
huge non performing loans on the private sector between 1970 and 1980. Higher
CP means wider financial sector and higher
financial intermediation. Yet for Niger's case, the CP was
lower; indicating restraint of the financial sector and lower financial
intermediation. Additionally, the attitude of bankers to finance less risky
projects lead to low capital intensity. This does little to improve investment
and may create distortions in the economy. Furthermore, investment in Niger is
weak and unstable leading to unexpected negative returns from projects with
attendant negative impact on economic growth; De Gregorio and Guidotti
(1995).
Chapter 4 Conclusions and Policy Implications
| 
