Qual Quant

DOI 10.1007/s11135-013-9866-9

Higher education and growth performance of Pakistan:

evidence from multivariate framework

Wasim Qazi · Syed Ali Raza · Syed Tehseen Jawaid

© Springer Science+Business Media Dordrecht 2013

Abstract This study investigates the relationship between higher education and economic
growth in Pakistan by using the annual time series data from the period of 1980–2011.
The ARDL bound testing cointegration approach confirms the valid positive relationship
between higher education development and economic growth in long run as well as in short
run. Results of Granger causality test, Toda and Yamamoto Modified Wald causality test
and variance decomposition test confirm the bidirectional causal relationship between higher
education and economic growth in Pakistan. Results of rolling window estimations suggest
that the contribution of higher education in economic growth is significantly increased after
the formation of higher education commission of Pakistan in 2002. It is clear from our
findings that higher education commission plays an important role in the development of
higher education in Pakistan which leads to enhance economic growth. It is recommended
that policy makers should make policies to strengthen the higher education commission to
ensure continuous and rapid economic growth in Pakistan.

Keywords Economic growth · Higher education · Time series analysis

1 Introduction

The development and accumulation in human capital is a main substance of growth of any
economy. The difference between the living standards of different nations is just because
of difference in their human capital (Lucas 1993). The fortune of economic growth of any

W. Qazi · S. A. Raza (B) · S. T. Jawaid

IQRA University, Karachi 75300, Pakistan
e-mail: [email protected]
W. Qazi
e-mail: [email protected]
S. T. Jawaid
e-mail: [email protected]

123
 W. Qazi et al.

country is mainly based on their human or intellectual capital. The quality and extent of
education is very important for socio-economic development of any society.
We are living in an era where intellectual capital becomes a symbol of a rapid growth
of any economy and higher education is become a tool to shine the human or intellectual
capital. Higher education provides critical thinkers, researchers, scholars, innovators and
responsible citizens to societies. Higher education also provides the opportunities to maintain
social mobility and high living standards. Consequently, the requirement of higher education
is more important for less developed or developing country like Pakistan for their rapid and
sustainable economic growth in future.
In Pakistan, education sector has persistently suffered from the low investment by the
government from the last six decades. The current government spending on education in
2012 is still just 2 % of gross domestic product of country. According to UNDP, Pakistan is
one of the 12 countries in the world, which spend less than 2 % of GDP on education.1 The non
availability of adequate funds over the past years was one of the reasons of the unsuccessful
education programs. It is a fact that Pakistan clearly be deficient in implementing policies to
improve the education system.
In Pakistan, up till 2002, the universities were recognized by the University Grant Com-
mission (UGC). In 2002, the government introduce new ordinance with a name of Higher
Education Commission Ordinance, and since then, Higher Education Commission of Pakistan
(HEC) is responsible for any policy of higher education, assurance of quality in education,
recognition of degree, uplifting the standards of existing institutions and developments of new
institutions in Pakistan. In last ten years, HEC has performed a leading role towards building
a knowledge based economy in Pakistan by producing more than 3,000 PHD scholars.
The main objective of this study is to examine the relationship between higher education
and economic growth in Pakistan. This paper make a unique contribution to the literature
with reference to Pakistan, being a pioneering attempt to investigate the impact of higher
education and economic growth of Pakistan by using the long annual time series data from
1980 to 2011 and by applying more rigorous econometric techniques.
The rest of the paper is organized as follows: following introduction Sect. 2 review some
selected studies, Sect. 3 discuss empirical strategy, Sect. 4 shows estimations and results,
Sect. 5 shows results of rolling window estimation, Sect. 6 discuss the results of cumulative
sum and cumulative sum of square estimations, Sect. 7 shows the results of causal relationship
between higher education and economic growth and the final section conclude the study and
provide some policy implications.

2 Review of literature

Solow (1956, 1957) argues that the growth of any economy could not only be explained by
the increases in capital and labor. Lucas (1988) considered human capital development as a
main factor of production and consider education as a means of human capital accumulation.
He argues that advancement in education level of labor force has a positive impact on the
productivity level that’s leads to better economic performance of an economy.
Barro (1991) proved the human capital development measured by education as a main
determinant of economic growth by using the data of 98 countries from the period of 1960
to 1985. Mankiw et al. (1992) also favor the Solow’s model by analyzing the significant
contribution of accumulation of human capital measured by education level on economic
growth. Barro and Lee (1993) investigate the impact of level of schooling (primary education,
1 Social Policy and Development in Pakistan (2003).

123
Evidence from multivariate framework

secondary education and higher education) on economic growth by using the data of 129
countries from the period of 1960 to 1985. They concluded that different levels of education
have positive and significant explanatory capacity to explain economic growth.
Benhabib and Spiegel (1994) argue that level of education play a significant role to increase
in the per capita income. Gemmell (1996) confirms that the different level of education and
their growth rates both are very important in explaining the economic growth. Gylfason and
Zoega (2003) analyze the impact of education on economic growth of 87 countries on the
sample period of 1965–1998. They suggest that education plays an important part in growth
of any economy and also play a part in reducing the inequality.
On the other side some researchers are also found weak relationship between education
and economic growth. Bils and Klenow (2000) investigate the impact of schooling education
and economic growth of cross countries data from the period of 1960 to 1990. They suggest
that education does not directly affect the economic growth. Pritchett (2001) argues that
education plays very little role in explaining the variations in the growth rate of cross country
data. Temple (2001) also argues that in developing countries the large investments in education
have yielded a very small pay-offs.
Most of the empirical studies have studied the relationship between basic education as
human capital and economic growth. On the other hand, very few studies have been focused
on particularly the impact of higher education and economic growth. Danacica et al. (2010)
investigate the causal relationship between higher education and economic growth in Roma-
nia by using the time series data from the period of 1980 to 2008. Results of Granger causality
test suggest the unidirectional causality runs from economic growth to higher education in
case of Romania.
Chaudhary et al. (2009) examine the impact of higher education on economic growth of
Pakistan by using the time series data from the period of 1972 to 2005. Johansen cointegration
and Toda and Yamamoto (1995) causality test have been used. Results indicate the existence
of unidirectional causality runs from economic growth to higher education in Pakistan. Huang
et al. (2009) investigate the relationship between higher education enrollment and economic
growth in China by using the sample from the period of 1972 to 2007. Granger causality
test confirms the bidirectional causality between higher education enrollment and economic
growth in China.
Gyimah-Brempong et al. (2006) investigates the impact of higher education on economic
growth of African countries. They used data of 34 African countries from the period of 1960
to 2000. Results indicate the positive and significant impact of higher education on economic
growth. They conclude that the elasticity of higher education is almost twice as large as the
elasticity of physical capital investment.
Khorasgani (2008) analyzes the impact of higher education development on economic
growth of Iran by using the annual data from the period of 1959 to 2005. Autoregressive
distributed lag (ARDL) cointegration and error correction models have been used. Results
suggest the positive and significant impact of higher education on economic growth of Iran
in long run as well as in short run. Katircioglu (2010) investigates the relationship between
higher education enrollment and economic growth of North Cyprus by using the annual data
from the period of 1977 to 2007. Results of ARDL cointegration indicate the positive and
significant relationship between higher education and economic growth.
Many studies suggest the positive effect of higher education on economic growth. Previous
studies concluded that the investment in the education enhance the economic performance.
However, the elasticity and benefits of this positive impact vary by nation to nation, ethnicity,
gender and race.

123
 W. Qazi et al.

3 Empirical framework

After reviewing the theoretical and empirical work, the model to examine the relationship
between higher education and economic growth in Pakistan derived by using the production
function framework. The production function in general form as follows:
GDP = f (L AB, C A P, A) (3.1)
where GDP is the real gross domestic production, LAB is the labor force, CAP is the capital
stock and A is the total factor productivity. It has been assumed that effect of higher education
on economic growth operates through A.2
A = g(H E D) (3.2)
Substituting (3.2) in (3.1)
G D P = f (L AB, C A P, H E D) (3.3)
The empirical models for estimations are developed as follows:
G D Pt = β0 + β1 L ABt + β2 C A Pt + β3 H E Dt + εt (3.4)
where, εt is the error term, GDP is the real gross domestic product, LAB is the total labor
force and HED represents the higher education enrollment. Real gross fixed capital formation
as percentage of GDP has been used as a proxy for capital stock because of unavailability of
data of capital stock.3 The expected signs for labor and capital stock are positive while, the
sign of HED is to be determined. Annual time series data have been used from 1980 to 2011.
All data are gathered from different issues of economic surveys of Pakistan. All variables are
used in logarithm form.

3.1 Unit root test

Augmented Dickey Fuller (ADF) and Phillip Perron (PP) unit root test are used to examine
the stationary properties for long run relationship of time series variables. Augmented Dickey
Fuller (ADF)4 test is based on equation given below:

k
ΔYt = α0 + α1 Yt _1 + d j Yt _ j + εt
j=1

where εt is pure white noise error term,  is first difference operator, Yt is a time series, α0
is the constant and k is the optimum numbers of lags of the dependent variable. Augmented
Dickey Fuller (ADF) test determines whether the estimates of coefficients are equal to zero.
ADF test provides cumulative distribution of ADF statistics. The variable is said to stationary,
if the value of the coefficient δ is less than critical values from fuller table. Phillip and Perron
(PP)5 unit root test equation is given below:
ΔYt = α + ρ ∗ Yt _1 + εt
The Phillip and Perron unit root test is also based on t-statistics that is associated with
estimated coefficients ofρ ∗ .
2 See, Kohpaiboon (2003), Jawaid and Waheed (2011), Jawaid and Raza (2012).
3 See, Balasubramanyam et al. (1996), Barro (1999) and Kohpaiboon (2003).
4 See, Dicky and Fuller (1979).
5 See, Phillips and Perron (1988).

123
Evidence from multivariate framework

3.2 ARDL bound testing approach

The Auto Regressive Distributed Lag (ARDL) method of cointegration developed by Pesaran
and Pesaran (1997), Pesaran and Shin (1999), Pesaran et al. (2000, 2001) has been used with
the help of unrestricted vector error correction model to investigate the long run relationship
between higher education and economic growth. The ARDL approach has several advan-
tages upon other cointegraion methods. ARDL approach may applies irrespective of whether
underlying variables are purely I (0), I (1) or mutually co-integrated.6 ARDL approach has
estimated better small sample properties.7 In ARDL procedure the estimations of results is
even possible if the explanatory variable are endogenous.8 The ARDL model is developed
for estimations as follow:

p

p

p
GDPt = ψ0 + ψ1 GDPt−1 + ψ2 LABt−1 +ψ3 CAPt−1
i=1 i=1 i=1


p
+ψ4 HEDt−1 + γ1 GDPt−1 + γ2 LABt−1 + γ3 CAPt _1 + γ4 HEDt−1 + μt
i=1

where ψ0 is constant and μt is white noise error term, the error correction dynamics is denoted
by summation sign while the second part of the equation corresponds to long run relationship.
Schwarz Bayesian Criteria (SBC) has been used to identify the optimum lag of model and each
series. In ARDL model we first estimate the F-statistics value by using the appropriate ARDL
models. Secondly, the Wald (F-statistics) test is used to investigate the long run relationship
among the series. The null hypothesis of the cointegration is (H0 = γ1 = γ2 = γ3 = γ4 = 0).
The null hypothesis of no cointegration is rejected if the calculated F-test statistics exceeds
the upper critical bound (UCB) value. The results are said to be inconclusive if the F-test
statistics falls between the upper and lower critical bound. Lastly, the null hypothesis of no
cointegration is accepted if the F-statistics is below the lower critical bound. If long run
relationship between higher education and economic growth is found then we estimate the
long run coefficients. The following model will be use to estimate the long run coefficients:

p

p

p

p
GDPt = ζ0 + ζ1 GDPt−1 + ζ2 LABt−1 +ζ3 CAPt−1 + ζ4 HEDt−1 + μt
i=1 i=1 i=1 i=1
If we find evidence of long run relationship between higher education and economic
growth then we estimate the short run coefficients by employing the following model:

p

p

p
GDPt = ϕ0 + ϕ1 GDPt−1 + ϕ2 LABt−1 +ϕ3 CAPt−1
i=1 i=1 i=1

p
+ϕ4 HEDt−1 + nECMt−1 + μt
i=1

The error correction model shows the speed of adjustment needed to restore the long run
equilibrium following a short run shock. The n is the coefficient of error correction term in
the model that indicates the speed of adjustment.

6 Pesaran and Shin (1999).

7 Haug (2002).
8 Pesaran and Shin (1999) and Pesaran et al. (2001).

123
 W. Qazi et al.

Table 1 Stationary test results

Variables Augmented Dickey–Fuller Phillips–Perron

I(0) I(1) I(0) I(1)

C C&T C C&T C C&T C C&T

GDP 2.29 0.96 −3.90 −5.02 −1.97 −1.13 −3.90 −5.17

LAB −1.83 −1.00 −5.01 −5.35 −1.32 −0.72 −5.01 −5.38
CAP −0.84 −2.23 −4.29 −4.20 −0.91 −2.07 −4.29 −4.20
HED −0.51 −2.18 −6.05 −6.25 −1.75 −2.41 −4.34 −5.86
The critical values for ADF and PP tests with constant (c) and with constant & trend (C&T) 1, 5 and 10 %
level of significance are −3.711, −2.981, −2.629 and −4.394, −3.612, −3.243, respectively
Source Authors’ estimation

Table 2 Lag length selection and

Lags order AIC HQ SBC F test statistics
bound testing for cointegration
0 40.245 40.303 40.435 28.664*
1 29.543 29.834 30.633
2 28.919* 29.443* 30.494*
* 1 % level of significant

Rolling window estimation method has been used to analyze the stability of coefficients
of long run model. Cumulative sum and cumulative sum of square estimations have been
used to analyze the stability of coefficients of short run model. This study also employs
the three different techniques of causality analysis namely, Granger causality analysis, Toda
and Yamamoto modified Wald test causality analysis and variance decomposition method to
analyze the robustness of causal relationship between higher education and economic growth.

4 Estimations and results

To check the stationary properties we use Augmented Dickey Fuller (ADF) and Phillip Perron
(PP) unit root tests. Table 1 represents the results of stationary tests. First, these tests are
applied on level of variables then on their first difference.
Results of Table 1 show that all variables are stationary and integrated at first difference.
This implies that the series of variables may exhibit a valid long run relationship.
Autoregressive distributed lag method for cointegration is used to estimate the long run
relationship between higher education and economic growth. The first step is to determine
the optimal lag length of the variables. The order of optimal lag length is decided by using
the Schwarz Bayesian Criterion. Table 2 shows the results of ARDL cointegration method.
The ARDL results suggest the rejection of null hypothesis of no cointegration in model
because the value of the F- statistics is greater than upper bound critical value at 1 % level
of significance in favor of alternative hypothesis that the valid long run relationship is exist
between higher education and economic growth in Pakistan.
Now we estimate the lag length order of the all variables through unrestricted vector auto
regression method. The decision criterion is based on minimum value of Schwarz Bayesian
Criterion.
Table 3 represents the results of lag length order of all variables. Results of Schwarz
Bayesian Criterion indicate that the gross domestic product and labor force should be include

123
Evidence from multivariate framework

Table 3 Lags defined through VAR of variables

Lag 0 1 2 Selected lags

SBC SBC SBC SBC

GDP 9.326 3.475a 3.942 1

LAB 7.662 1.185a 1.298 1
CAP 5.375 2.633 2.291a 2
HED 27.858 23.641 23.555a 2
a Indicate minimum SBC values

Table 4 Long run results using

Variables Coeff. t-stats Prob.
ARDL approach
C −0.080 −0.541 0.595
GDP(−1) 0.940 5.241 0.000
LAB 0.362 6.388 0.000
LAB(−1) 0.296 0.628 0.538
CAP 0.245 4.132 0.000
CAP(−1) −0.084 −0.785 0.443
CAP(−2) −0.033 −0.340 0.738
HED 0.054 4.855 0.000
HED(−1) −0.030 −1.515 0.147
HED(−2) 0.014 1.105 0.284
Adj. R2 0.968
D.W stats 2.455
F-stats (Prob.) 267.667 (0.000)
Source Authors’ estimation

in model at 1st lag while capital and higher education should be include in model at 2nd
lag. After having the valid evidence of long run relationship between higher education and
economic growth now we applied the ARDL method to estimate the long run and short run
coefficients. The model for long run coefficients as follow:

p

p

p

p
GDPt = ζ0 + ζ1 GDPt−1 + ζ2 LABt + ζ3 LABt−1 + ζ4 CAPt
i=1 i=1 i=1 i=1

p

p

p

p
+ζ5 C APt−1 +ζ6 CAPt−2 + ζ7 HEDt + ζ8 HEDt−1
i=1 i=1 i=1 i=1
p
+ζ9 HEDt−2 + μt
i=1

Table 4 shows the results of long run ARDL estimations. The results of (LAB) and (CAP)
are having expected positive sign and are highly significant. Results indicate the positive
and significant impact of higher education on economic growth in Pakistan. The findings are
consistent with Gyimah-Brempong et al. (2006), Khorasgani (2008) and Katircioglu (2010).
The coefficient of higher education showing that the 1 % increase in higher education causes

123
 W. Qazi et al.

Table 5 Short run results using

Variables Coeff. t-stats Prob.
ARDL approach
C 0.004 0.178 0.861
GDP(−1) 1.022 0.562 0.582
LAB 0.221 2.869 0.011
LAB(−1) 0.213 0.538 0.598
CAP 0.259 0.868 0.398
CAP(−1) −0.101 −1.247 0.230
CAP(−2) −0.093 −1.313 0.208
HED 0.024 2.201 0.043
HED(−1) 0.018 1.778 0.094
HED(−2) 0.007 0.301 0.767
ECM(−1) −0.363 −3.408 0.004
Adj. R2 0.589
D.W stats 1.994
Source Authors’ estimation F-stats (Prob.) 22.826 (0.000)

the increases in the economic growth by 0.054 %. It is concluded that the higher education is
a better leading indicator for economic growth in Pakistan. Following model is used to check
the short run relationship among the considered variables with the different lag length.


p

p

p
GDPt = ϕ0 + ϕ1 GDPt−1 + ϕ2 LABt +ϕ3 LABt−1
i=1 i=1 i=1

p

p

p

p
+ϕ4 CAPt +ϕ5 CAPt−1 + ϕ6 CAPt−2 + ϕ7 HEDt
i=1 i=1 i=1 i=1
p

p
+ϕ8 HEDt−1 + ϕ9 HEDt−2 + nECMt−1 + μt
i=1 i=1

Table 5 represents the short run relationship between higher education and economic
growth. Results indicate the lagged error correction term for the estimated economic growth
equation is both negative and statistically significant. This confirms a valid short run rela-
tionship between higher education and economic growth in Pakistan. The coefficient of error
term is −0.36 suggest that about 36 % of disequilibrium is corrected in the current year.
Results indicate the positive and significant effect of higher education on economic growth
in short run as well in Pakistan.

5 Stability of long run model: a rolling window analysis

The stability of coefficients of the long run model in the sample size is evaluated by using
the rolling window estimation method. Figure 1 and Table 6 represent the results of rolling
window regression method of higher education. Two standard deviation bands show the upper
and lower bounds. Results indicate that the higher education is having positive coefficients
throughout the sample period. The results of Fig. 1 shows that the coefficient of higher
education is remains steady with minor changes from 1984 to 2001 but from 2002 to 2010 the

123
Evidence from multivariate framework

Fig. 1 Coefficient of HED and its two SE bands based on rolling OLS (dependent variable: GDP)

Table 6 Long run coefficients of

Year Coeff. Year Coeff.
HED
1981 0.016 1996 0.014
1982 0.016 1997 0.014
1983 0.015 1998 0.014
1984 0.016 1999 0.015
1985 0.016 2000 0.016
1986 0.016 2001 0.027
1987 0.015 2002 0.040
1988 0.015 2003 0.047
1989 0.015 2004 0.053
1990 0.013 2005 0.055
1991 0.012 2006 0.058
1992 0.012 2007 0.066
1993 0.013 2008 0.072
1994 0.013 2009 0.080
Source Authors’ estimation 1995 0.013 2010 0.082

coefficient of higher education contribution in economic growth is continuously increasing

very sharply. These findings suggest that the contribution of higher education in economic
growth is significantly increased after the formation of higher education commission of
Pakistan in 2002. It is clear from our findings that higher education commission plays an
important role in the development of higher education in Pakistan which leads to economic
growth.

6 Stability of short run model

The stability of short run model in the sample size is evaluated by using the cumulative
sum (CUSUM) and CUSUM of square test on the recursive residuals. CUSUM test detects
systematic changes from the coefficients of regression, while, CUSUM of square test is able

123
 W. Qazi et al.

15

10

-5

-10

-15
1992 1994 1996 1998 2000 2002 2004 2006 2008

CUSUM 5% Significance

Fig. 2 Plot of cumulative sum of recursive residuals. The straight lines represent critical bounds at 5 %
significance level

1.6

1.2

0.8

0.4

0.0

-0.4
1992 1994 1996 1998 2000 2002 2004 2006 2008

CUSUM of Squares 5% Significance

Fig. 3 Plot of cumulative sum of squares of recursive residuals. The straight lines represent critical bounds
at 5 % significance level

to detects the sudden changes from constancy of regression coefficients (Brown et al. 1975).
Figures 2 and 3 represents the results of CUSUM and CUSUM of square tests, respectively.
Results indicate that the statistics of both CUSUM and CUSUM of square test are lie within
the interval bands at 5 % confidence interval. Results suggest that there is no structural
instability in the residuals of equation of economic growth.

7 Causality analysis

In this section three different techniques of causality analysis namely, Granger causality
analysis,9 Toda and Yamamoto modified Wald test causality analysis10 and variance decom-
9 See, Granger (1969).
10 See, Toda and Yamamoto (1995).

123
Evidence from multivariate framework

Table 7 Results of Granger

Variables F-statistic Probability
causality test
GDP does not Granger cause LAB 0.454 0.506
LAB does not Granger cause GDP 8.060 0.009
GDP does not Granger cause CAP 11.766 0.002
CAP does not Granger cause GDP 8.609 0.007
GDP does not Granger cause HED 3.915 0.059
The lag length is 1
Source Authors’ estimations HED does not Granger cause GDP 5.724 0.024

position method11 have been used to analyze the robustness of causal relationship between
higher education and economic growth.

7.1 Granger causality analysis

The direction of causality between dependent and independent variables is analyzed by

Granger (1969) causality test. We determine the causality analysis of our economic growth
model on lag one. Jones (1989) favors the ad hoc selection method for lag length in Granger
causality test over some of other statistical method to determine optimal lag. The equation
of Granger causality model is given below:


t 
t
Y = αi X t _i + βi Yt _i + μ
i=1 i=1
t t
X = λi X t _i + δi Yt _i + v
i=1 i=1

It is assumed that μ and v are uncorrelated. There are two variables and dealt with bilateral
causality. Above equation states that Y is related to its lag values and X is related to its lag
values.
The results of Granger causality test are reported in Table 7. Results show the bidirectional
causal relationship between higher education development and economic growth.

7.2 Toda and Yamamoto modified Wald test causality analysis

The direction of causality between dependent and independent variables is analyzed by using
the causality test based on Toda and Yamamoto (1995) procedure. This test use a modified
Wald (MWALD) test which can be applies irrespective of the of whether underlying variables
are purely I (0), I (1) or mutually co-integrated. Toda and Yamamoto (1995) augmented
Granger causality test uses the Seemingly Unrelated Regression (SUR) technique through
estimating a two equation system. The Wald test improves efficiency when SUR models are
used in the estimation. So, the model can be specified as follows:


k+d 
k+d
Yt = α1 + γ1i Yt _i + γ2i X t _i + ε yt
i=1 t−i

11 The variance decomposition method is estimated through VAR framework, it shows the proportion contri-
bution in one variable caused by the shocks in other variables, Pesaran and Shin (1998).

123
 W. Qazi et al.

Table 8 Results of Toda and Yamamoto causality test

Dependent variable Modified Wald—statistics

GDP LAB CAP HED

GDP – 21.622 (0.000) 3.324 (0.079) 17.071 (0.000)

HED 6.488 (0.017) 23.892 (0.000) 0.044 (0.836) –

The lag length for GDP is 1, LAB is 1, CAP is 2 and HED is 2 as per Schwartz Bayesian Criteria (SBC)
Source Authors’ estimations


k+d 
k+d
X t = α2 + δ1i Yt _i + δ2i X t _i + εxt
i=1 t−i

where k is the optimal lag order, d is the maximum order of integration of the series in the
system, and ε yt and εxt are error terms that are assumed to be white noise. Usual Wald tests
are then applied to the first k coefficient matrices using the standard χ 2 – statistics.
The results of Toda and Yamamoto (1995) procedure based causality test are reported in
Table 8 Results indicate the bidirectional causal relationship between higher education and
economic growth.

7.3 Variance decomposition analysis

Generalized forecast error variance decomposition method under vector autoregressive

(VAR) system has been used to analyze the strength of the causal relationship of higher
education and economic growth. The variance decomposition method provides the magni-
tude of the predicted error variance for a series accounted for by innovations from each of
the independent variable over different time period. Wong (2010), Hye (2012), Shahbaz et al.
(2012) and Raza and Jawaid (2013) have used this approach to find causal relationship among
considered variables. Table 9 represents the results of variance decomposition analysis.
Results of Table 9 show that in the first round the change in economic growth explain
completely 100 % by its own innovations. In the second period 94.01 % explain by own
innovation, 1.93 % by labor, 3.87 % by capital and 0.19 % by higher education development.
In period five the shocks in economic growth explain 62.03 % by own innovation, 16.02 % by
innovations of labor, 14.99 % by innovations of capital and 6.96 % by innovations of higher
education development. In tenth period the shocks in economic growth explain 23.08 % by
own shocks, while, 33.37 % explain by innovations of labor, 15.04 % explain by innovations
of capital and 28.51 % explain by innovations of higher education development.
The shocks in higher education explain 82.79, 39.20, 26.94 and 24.00 % by its own inno-
vations in period 1, 2, 5 and 10, respectively. The shocks in higher education explain 13.25,
32.662, 50.43 and 45.21 % by innovation of economic growth in period 1, 2, 5 and 10, respec-
tively. These findings suggest the bidirectional causal relationship between higher education
and economic growth in Pakistan.

8 Conclusion and recommendations

This study investigates the relationship between higher education and economic growth in
Pakistan by using the annual time series data from the period of 1980 to 2011. The ARDL

123
Evidence from multivariate framework

Table 9 Results of variance decomposition approach

Period GDP LAB CAP HED

Variance decomposition of GDP

1 100.000 0.000 0.000 0.000
2 94.009 1.933 3.874 0.185
3 80.306 5.640 13.742 0.312
4 67.574 11.347 11.419 9.660
5 62.031 16.020 14.990 6.960
6 49.813 18.641 24.058 7.488
7 37.828 20.727 28.356 13.088
8 24.112 23.939 18.112 33.837
9 27.700 28.776 16.206 27.319
10 23.076 33.370 15.044 28.510
Variance decomposition of HED
1 13.248 3.866 0.101 82.785
2 32.662 5.338 22.800 39.200
3 33.316 3.946 18.879 43.858
4 44.373 4.603 14.925 36.099
5 50.432 9.928 12.702 26.937
6 53.756 11.848 15.918 18.477
7 60.809 17.809 3.996 17.387
8 54.509 19.138 7.436 18.917
9 48.436 21.600 9.536 20.428
10 45.211 23.030 7.756 24.003
Source Authors’ estimation

bound testing cointegration approach confirms the valid positive relationship between higher
education development and economic growth in long run as well as in short run. Results
of Granger causality test, Toda and Yamamoto Modified Wald causality test and variance
decomposition test confirm the bidirectional causal relationship between higher education
and economic growth in Pakistan. Results of rolling window estimations suggest that the
contribution of higher education in economic growth is significantly increased after the
formation of higher education commission of Pakistan in 2002. It is clear from our findings
that higher education commission plays an important role in the development of higher
education in Pakistan which leads to economic growth. It is recommended that policy makers
should make policies to strengthen the higher education commission to ensure continuous
and rapid economic growth in Pakistan.

References

Balasubramanyam, V. N., Salisu, M. A., Sapsford, D.: Foreign direct investment and growth in EP and IS
countries. Econ. J. 106(434), 92–105 (1996)
Barro, R.J.: Economic growth in a cross-section of countries. Q. J. Econ. 106(2), 407–443 (1991)
Barro, R.J.: Human capital and growth in cross-country regressions. Swed. Econ. Pol. Rev. 6(2), 237–277
(1999)

123
 W. Qazi et al.

Barro, R. J., and Lee, J. W.: International comparisons of educational attainment. J. Monetary. Econ. 32,
363–394 (1993)
Barro, R.J., Lee, J.W.: International data on educational attainment: updates and implications. Oxf. Econ. Pap.
3(3), 541–563 (2001)
Benhabib, J., Spiegel, M.M.: The role of human capital in economic development. J. Monet. Econ. 34(2),
143–174 (1994)
Bils, M., Klenow, P.: Does schooling cause growth? Am. Econ. Rev. 90, 1160–1183 (2000)
Brown R.L., Durbin, J., Ewans J.M.: Techniques for testing the constance of regression relations overtime.
J. R. Stat. Soc. 37, 149–172 (1975)
Chaudhary, A.R., Iqbal, A., Gillani, S.Y.M.: The nexus between higher education and economic growth: an
empirical investigation for Pakistan. Pakistan. J. Commer. Soc. Sci. 3, 1–9 (2009)
Danacica, D., Belascu, L., Ilie, L.: The interactive causality between higher education and economic growth
in Romania. Int. Rev. Bus. Res. Pap. 6(4), 491–500 (2010)
Dicky, D.A., Fuller, W.A.: Distribution of the estimators for autoregressive time series with a unit root. J. Am.
Stat. Assoc. 74, 427–431 (1979)
Gemmell, N.: Evaluating the impacts of human capital stocks and accumulation on economic growth: some
new evidence. Oxf. Bull. Econ. Stat. 58, 9–28 (1996)
Granger, C.W.J.: Investigating causal relations by econometric models and cross-spectral methods. Econo-
metrica 37(3), 424–438 (1969)
Gyimah-Brempong, K., Paddison, O., Mitiku, W.: Higher education and economic growth in Africa. J. Dev.
Stud. 42(3), 509–529 (2006)
Gylfason, T., Zoega, G.: Education social equality and economic growth: a view of the landscape. CESifo
Econ. Stud. 49(4), 557–579 (2003)
Haug, A.: Temporal aggregation and the power of co-integration tests: a Monte Carlo study. Oxf. Bull. Econ.
Stat. 64, 399–412 (2002)
Huang, F., Jin, L., Sun, X.: Relationship between scale of higher education and economic growth in China,
Asian. Soc. Sci. 5(11), 55–60 (2009)
Hye, Q.A.M.: Long term effect of trade openness on economic growth in case of Pakistan. Qual. Quant. 46(4),
1137–1149 (2012)
Jawaid, S.T., Waheed, A.: Effects of terms of trade and its volatility on economic growth: a cross country
empirical investigation. Transit. Stud. Rev. 18(2), 217–229 (2011)
Jawaid, S.T., Raza, S.A.: Workers’ remittances and economic growth in China and Korea: an empirical analysis.
J. Chinese. Econ. Foreign. Trade. Stud. 5(3), 185–193 (2012)
Jones J.D.: A comparison of lag length selection techniques in tests of Granger causality between money
growth and inflation: evidence from the US, 1959–86. Appl. Econ. 21(6):809–822 (1989)
Katircioglu, S.: International tourism, higher education, and economic growth: the case of North Cyprus.
World. Econ. 33(12), 1955–1972 (2010)
Khorasgani, M.F.: Higher education development and economic growth in Iran. Educ. Bus. Soc. Contemp.
Middle. Eastern. Issues. 1, 162–174 (2008)
Kohpaiboon, A.: Foreign trade regimes and the FDI-growth nexus: a case study of Thailand. J. Dev. Stud.
40(2), 55–69 (2003)
Lucas, R.: On the mechanics of economic development. J. Monet. Econ. 22, 3–42 (1988)
Lucas, R.E.: Making a miracle. Econometrica 61, 251–272 (1993)
Mankiw, N.G., Romer, D., Weil, D.: A contribution to the empirics of economic growth. Q. J. Econ. 107,
407–437 (1992)
Pesaran, M.H., Pesaran, B.: Working with Microfit 4.0: Interactive Econometric Analysis. Oxford University
Press, Oxford (1997)
Pesaran, M.H., Shin, Y.: Generalized impulse response analysis in linear multivariate models. Econ. Lett.
58(1), 17–29 (1998)
Pesaran, M.H., Shin, Y.: An autoregressive distributed lag modeling approach to co-integration analysis. In:
Strom, S. (ed.) Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial
Symposium. Cambridge University Press, Cambridge (1999)
Pesaran, M.H., Shin, Y., Smith, R.J.: Structural analysis of vector error correction models with exogenous I(1)
variables. J. Econom. 97, 293–343 (2000)
Pesaran, M.H., Shin, Y., Smith, R.J.: Bounds testing approaches to the analysis of level relationships. J. Appl.
Econ. 16, 289–326 (2001)
Phillips, P.B., Perron, P.: Testing for a unit root in time series regression. Biometrika 75, 335–346 (1988)
Pritchett, L.: Where has all the education gone? World Bank Econ. Rev. 15(3), 367–391 (2001)
Raza, S.A., Jawaid, S.T.: Terrorism and Tourism: a conjunction and ramification in Pakistan. Econ. Model.
33, 65–70 (2013)

123
Evidence from multivariate framework

Shahbaz, M., Islam, F., Aamir, N.: Is devaluation contractionary? Empirical evidence for Pakistan. Econ.
Chang. Restruct. 45(4), 299–316 (2012)
Social Policy and Development in Pakistan: The State of Education, Annual Review 2002–2003, p. 4. Social
Policy and Development Centre, Karachi (2003)
Solow, R.M.: A contribution to the theory of economic growth. Q. J. Econ. 70, 65–94 (1956)
Solow, R.: Technical change and the aggregate production function. Rev. Econ. Stat. 39, 312–320 (1957)
Temple, J.R.W.: Generalizations that aren’t? Evidence on education and growth. Eur. Econ. Rev. 45(4–6),
905–918 (2001)
Toda, H.Y., Yamamoto, T.: Statistical inference in vector autoregressions with possibly integrated processes.
J. Econom. 66, 225–250 (1995)
Wong, H.T.: Terms of trade and economic growth in Malaysia. Labuan Bull. Int. Bus. Financ. 2(2), 105–122
(2004)
Wong, H.T.: Terms of trade and economic growth in Japan and Korea: an empirical analysis. Empir. Econ. 38,
139–158 (2010)

123