Page 10 - Azerbaijan State University of Economics

THE JOURNAL OF ECONOMIC SCIENCES: THEORY AND PRACTICE, V.77, # 1, 2020, pp. 4-18

Table 4: Cointegration test
Model Case Pedroni Cointegration Test
Panel v Panel rho Panel PP Panel ADF Group rho Group PP Group
ADF
Model 1 Ind. interc. -1.98 -3.89*** -11.21*** -11.25*** -0.22 -9.03*** -7.69***
Ind. interc. -5.29 -1.44* -12.92*** -12.50*** 1.96 -11.38*** -8.17***
and trend
Model 2 Ind. interc. -2.15 -6.61*** -25.72*** -16.45*** -2.96*** -23.26*** -16.92***
Ind. interc. -5.53 -3.27*** -33.69*** -17.73*** 0.57 -30.81*** -16.62***
and trend
Model 3 Ind. interc. -0.95 -2.33** -8.40*** -8.87*** 0.26 -9.15*** -9.16***
Ind. interc. -4.34 0.53 -8.24*** -8.64*** 2.90 -9.62*** -7.83***
and trend
Kao Cointegration Test
ADF Residual variance HAC variance
Model 1 Ind. interc. -3.77*** 0.004 0.000
Model 2 Ind. interc. -8.31*** 0.020 0.003
Model 3 Ind. interc. -2.70*** 0.001 0.000
Johansen Fisher Cointegration Test
None At most 1 At most 2
from trace t max-eigen t from trace t max-eigen t from trace max-eigen
t t
Model 1 Ind. interc. 488.7*** 382.4*** 194.0*** 147.1*** 158.2*** 158.2***
and trend
Model 2 Ind. interc. 488.7*** 382.4*** 194.0*** 147.1*** 158.2*** 158.2***
and trend
Model 3 Ind. interc. 488.7*** 382.4*** 194.0*** 147.1*** 158.2*** 158.2***
and trend

*, **, and *** are respectively 10%, 5% and 1% significance levels. Bartlett Kernel method was used
and Bandwidth width was determined by Newey-West method.

It is also seen in Table 4 that the cointegration test findings include that there is a long-
term relationship between all three variables. As a matter of fact, according to the
findings of Pedroni cointegration test, hypothesis was rejected in 10 of 14 tests that
0
measured individual intercept, and individual intercept and trending situations for model
one. An alternative hypothesis showing that there is cointegration is accepted. Likewise,
11 tests for model two and 9 tests for model three show that there is a cointegration
relationship. Kao cointegration test findings, it is seen that in the model one where
education is considered as a dependent variable, it is associated with long-term
education finance and growth. Alternative hypothesis is accepted in model two, where
financing is considered as the dependent variable, and in model three, where growth is
dependent. In the findings of the Johansen Fisher cointegration test, the last of the
cointegration tests, it was concluded that education, finance and growth are in mutual
relationship for all three models.

After determining that there is a long-term relationship between the variables, the
direction and values of this relationship can be estimated. For this, DOLS and
FMOLS methods are used. The results obtained are presented in Table 5.

10