Page 9 - Azerbaijan State University of Economics

Rza Mammadov, Erdal Gümüş: Secondary Education, Economic Growth and Finance

this study, tests [For detailed information on conitegration tests, see: Baltagi, B. H.
(2013). Econometric analysis of panel data. Chichester: John Wiley and Sons.] of
Pedroni (2004), Kao (1999) and Johansen Fisher (1988) were performed to
determine cointegration and increase reliability.

After determining that the variables are stationary and cointegration, the coefficient
of the relationship between these variables can be estimated. DOLS (2001) and
FMOLS (2000) coefficient estimators [For coefficient estimation methods, see:
Baltagi, B. H. (2013). Econometric analysis of panel data. Chichester: John Wiley
and Sons] developed by Pedroni were used to determine the relational coefficient in
the study. DOLS reveals dynamic effects, FMOLS reveals modified effects.
Therefore, both methods are preferred.

4. FINDINGS
In this study, which aims to investigate education and economic growth on a 'triple
variable basis'; The variables of 'education', 'finance' and 'growth' were made
stationary at first. Unit root tests were applied to ensure stationarity. The findings are
included in Table 3.

Table 3: Unit root test
Variable
Test Case lnSEC lnSFIN lnPGDP
Ind. interc. Ind. interc. Ind. Ind. interc. Ind. Ind. interc.
and trend interc. and trend interc. and trend
PP I(0) 52.65 40.57 78.47* 62.36 138.25*** 74.40*
I(I) 368.67*** 351.30*** 168.38*** 172.26*** 200.24*** 183.44***
ADF I(0) 49.52 36.23 61.68 37.89 87.42** 50.27
I(I) 199.80*** 229.87*** 96.81*** 112.21*** 173.21*** 130.65***
LLC I(0) -2.22** -1.70** -3.88*** 0.02 -6.41*** -3.39***
I(I) -11.54*** -15.35*** -3.25*** -6.33*** -11.48*** -11.43***
IPS I(0) 1.62 3.16 0.54 3.10 -1.13 0.72
I(I) -7.84*** -11.10*** -3.21*** -3.82*** -7.92*** -4.80***

*, **, and *** are respectively 10%, 5% and 1% significance levels. The lag length was determined
by the Modified Schwarz Information criterion. Bartlett Kernel method was used in LLC and PP tests
and Bandwidth width was determined by Newey-West method.

According to Table 3 with unit root test results, there is no stationarity in the level
values [I (0)] for lnPGDP, lnSEC and lnSFIN. Stationarity was achieved by taking
the first difference [I(I)] of the variables. In order to reveal the relational state of the
variables with the values obtained by the first-degree difference, models 1, 2, and 3
were established and cointegration test was performed. The findings are included in
Table 4.

9