Higher Education and Economic Development in the Balkan Countries: A Panel Data Analysis

Transcription

1 1 Further Education in the Balkan Countries Aristotle University of Thessaloniki Faculty of Philosophy and Education Department of Education October 2008, Konya Turkey Higher Education and Economic Development in the Balkan Countries: A Panel Data Analysis Murat KARAGOZ Inonu University of Malatya-TURKEY Nicholas TSOUNIS Technological Educational Institute of West Macedonia-GREECE Kadir KARAGÖZ Inonu University of Malatya-TURKEY Abstract The nature of the relations between education and growth is far from having been perfectly determined. It is therefore of interest to examine the links between education and economic growth in the Balkan countries. The main question to be addressed is Does higher education play a part in the growth process of this area? Using the notion of causality developed by Granger, the relationship between higher education and economic growth is compared. This study makes use of panel data multivariate causality analysis to examine relationship between higher education and growth in the Balkans using annual time series of major Balkan countries. The panel data analysis results show that higher education has an influence on gross domestic product in the case of Balkan countries. Key words: Higher Education, Economic Development, Causality. Introduction There are some reasons for expecting relationships between education and economic growth. It is intuitively plausible to say that living standards have been raised proportional with education and so there is a link between scientific advance and the way in which education has facilitated the development of knowledge. In fact econometric studies indicate that the incomes individuals can command depend on their level of education. In the same way, the level of output per hour worked in a country depends on the educational attainment of the population. As the expenditures on education delivers returns of some sort, in much the same way as spending on fixed capital, then it is sensible to analyse the process of education as an investment decision.

2 2 An interesting historical data for the link between education and economic performance is provided by Maddison (1991) and reported in Stevens and Weale (2003). The data is presented graphically in Figure 1 and as a regression equation estimation 1 below. The plot is GDP per capita in 1913 against the primary school enrolment rates of It is clear that, high levels of GDP per capita are associated with high levels of primary school enrolment some thirty years earlier. This interval can be considered as a maturing time for investment for education or human capital. Figure 1: Education and GDP per capita The result of regression analysis yields the following result with standard errors in parenthesis: ln GDP per capita= ln Enrolment rate (1) (0.77) (0.12) R2 = 0.59 Thus this suggests that a 1% increase in the enrolment rate raises GDP by 0.35%. In a recent study, the determinants of economic growth and investment were analyzed in a panel of around 100 countries observed from 1960 to With respect to education, growth is positively related to the starting level of average years of school attainment of adult males at the secondary and higher levels (Barro 2002). Contemporary views on the determinants of economic growth place education in centre stage. Yet the way in which education affects growth is not yet well understood. Johnes (2006) begins by surveying the recent literature on the factors that affect growth, paying particular attention to education and then proceeds to

3 3 estimate a comprehensive model of growth, testing its robustness across regions of the world. Policy conclusions are drawn. There are other several recent evidences in the literature. Some countries may not obey the dominant theory whereby education is the cause of growth. However, economic growth increases the number of students. This is consistent with the idea that education is a growth-driven accompanying investment (Jaoul 2004). By analyzing the causality and cointegration between education and gross national product (GDP) for the case of China, the economic development is the cause of higher education and result of primary education (Kui 2007). The some empirical results in the literature show evidence of bidirectional causality between education and growth (Islam et al 2007). The empirical results of a study made for Turkey show that the direction of causality between human capital and economic growth depends upon the selected measures of human capital. These results indicate that economic growth, on the one hand, determines the human capital and, on the other, be determined by the indicators of human capital (Taban and Kar 2003). Model and Methodology Here the objective is to determine the long-run relationship between Higher Education and economic growth for the case of Balkan countries (BCs). In a panel data context other variables effecting the economic growth can easily be absorbed into the cross-section and periodic effects. Specifically, we are testing the education led growth hypothesis in terms of BCs. The six BCs involved in this study supposed to be open economies. The basic type of model can be defined in terms of panel or pooled data econometrics is as follows: Y = α+ X β + γ + δ + ε (2) it it i t it where Y it is the dependent variable, and X it is a vector of k-regressors, and are ε it the unobserved error terms for the cross-sectional units i= 1, 2, K, N over the given time periods t= 1, 2, K, T. The α and β parameters represent the overall constant and slope (vector) in the model, while the γ i and δ t represent cross-section or period specific effects (fixed and random respectively). Panel Unit Root Tests Recent literature approves that panel-based unit root tests have higher power than unit root tests based on individual time series. Panel unit root tests are similar, but not identical, to unit root tests carried out on a single series. Here, we briefly describe a general panel unit root test. Consider a following AR(1) process for panel data:

4 4 Y = ρ Y + X δ + ε (3) it i it 1 it i it The X it represent the exogenous variables in the model, including any fixed effects or individual trends, ρ i are the autoregressive coefficients, and the errors ε it are assumed to be mutually independent idiosyncratic disturbance. If ρ i < 1, Y it is said to be weakly (trend-) stationary. On the other hand, if ρ i = 1 then Y it contains a unit root. i For purposes of testing, there are two natural assumptions that can be made about the ρ. First, one can assume that the persistence parameters are common across crosssections so that ρi = ρ for all i. The Levin, Lin, and Chu (LLC) (2002), Breitung, and Hadri (2000) tests all employ this assumption. Alternatively, one can allow ρ i to vary freely across cross-sections. The Im, Pesaran, and Shin (IPS) (2003), and Fisher-ADF and Fisher-PP tests are of this form. Panel Cointegration Tests The Pedroni (1999) and Kao (1999) tests are based on Engle-Granger (1987) twostep (residual-based) cointegration tests. The Fisher test is a combined Johansen test. Here we give a brief summary of Pedroni (Engle-Granger based) Cointegration Tests. Pedroni (2004) proposes several tests for cointegration that allow for heterogeneous intercepts and trend coefficients across cross-sections. Consider the following regression Y = α + δ t+ β X + K + β X + ε (4) it i i 1i 1it Mi Mit it for m= 1,2, K, M ; where Y and X are assumed to be integrated of order one, e.g. I(1). The parameters α i and δ i are individual and trend effects which may be set to zero if desired. Under the null hypothesis of no cointegration, the residuals will be I(1). The general approach is to obtain residuals from (4) and then to test whether residuals are I(1) by running the auxiliary regressions, or e = e + u (5) it ρi it 1 it p i e = ρ e + ϕ e + υ (6) it i it 1 ij it j it j= 1

5 5 for each cross-section. Pedroni describes various methods of constructing statistics for testing for null hypothesis of no cointegration ( ρ i = 1). There are two alternative hypotheses: the homogenous alternative ( ρ = ρ) < 1 for all i and the heterogeneous alternative, ρ i < 1for all i. The Pedroni panel cointegration statistic is constructed from the residuals of auxiliary regression and the standardized statistic is asymptotically normally distributed. i Data and Empirical Results We have made use of a panel data comprising Bulgaria, Greece, Romania, Slovenia, Slovakia and Turkey for the period The variables included in the study were Gross domestic product (GDP) millions of euro at prices of the previous year and the total number of persons who are enrolled in tertiary education (EDU). The major data source is EUROSTAT. Throughout this study we have used Eviews 6 version for empirical evaluations. For the panel data of two variables GDP and EDU we have conducted panel unit root tests available in the software. The results are reported in Table 1 below. Table 1. Panel Unit Root Test Results. Im, Pesaran and Shin W-stat ADF - Fisher Chisquare Series Levin, Lin and Chu t* Breitung t- stat PP - Fisher Chi-square GDP EDU NOTES: (1) For the first two test The Null: Unit root (assumes common unit root process) (2) For the last three test The Null: Unit root (assumes individual unit root process). (3) Exogenous variables: Individual effects, individual linear trends, (4) Newey-West bandwidth selection using Bartlett kernel. (5) First lines are test statistics and second lines are related probabilities. Table 2. Panel Cointegration Test Results. Statistic Prob. Weighted Statistic Prob. Panel v-statistic Panel rho-statistic Panel PP-Statistic Group rho-statistic Group PP-Statistic NOTES: (1) Null Hypothesis: No cointegration, (2) Trend assumption: No deterministic trend (3) Newey-West bandwidth selection with Bartlett kernel.

6 6 Panel unit root test statistics generally have the implication first order unit root. Panel unit root tests in the first differences not reported here have shown no unit roots. Therefore we conclude that these series are all have stochastic trends. That is, all the series at hand are homogenous of order one. As it is well known, the existence of stochastic trend does not hinder the long run relationship between variables. Therefore we have next carried out the cointegration tests and reported the results in Table 2 above. Cointegration tests in general reject the null hypothesis of no cointegration. Therefore we conclude that there is at least one long run relationship between these variables. The third problem with this analysis is to detect the direction of causality between these two variables. The answer is provided with the Table 3 below. We have performed panel data Granger causality analysis for both directions. As it can be followed from the last column, the causality from education to growth is meaningful, while the reverse is insignificant. This result is a confirmation of Education led Growth hypothesis, in the case of BCs. Table 3. Panel Granger Causality Test Results. Null Hypothesis F-Statistic Prob. LOG(EDU) does not Granger Cause LOG(GDP) LOG(GDP) does not Granger Cause LOG(EDU) NOTES: (1) Method: Panel Least Squares, (2) Periods Included: 12, (3) Cross-Sections Included: 5 (4) Total Panel (Balanced) Observations: 60. As we have determined the type of causality, there remains only one problem that to estimate the long run relationship from education to economic growth. We have several alternatives. The significant models are reported in the Table 4 below. Table 4. Long-run Relationships in Pooled OLS Regression. Model Variables Coefficient Std. Error t-statistic Prob. Pooled OLS CONSTANT LOG(EDU) Fixed Effects CONSTANT LOG(EDU) Redundant CONSTANT Fixed Effects LOG(EDU) Random Effects CONSTANT LOG(EDU) NOTES: (1) Method: Panel Least Squares, (2) Periods Included: 9, (3) Cross-Sections Included: 6 (4) Total Panel (Balanced) Observations: 54. All four models have significant coefficients estimates both for constant term and slope parameter. In these models, as the variables are in terms of log values, the

7 7 slope parameter represents the elasticity. Having slope parameter estimates around 1, we conclude that there is an almost unit elasticity between education and economic growth. Conclusion The nature of the relations between education and growth is of interest to examine. Here the objective was to determine the long-run relationship between Higher Education and economic growth for the case of Balkan countries (BCs). We have employed panel causality analysis to examine relationship between higher education and growth in the Balkans using annual time series of major Balkan countries. We have made use of a panel data comprising Bulgaria, Greece, Romania, Slovenia, Slovakia and Turkey for the period The variables included in the study were Gross domestic product (GDP) millions of euro at prices of the previous year and the total number of persons who are enrolled in tertiary education (EDU). Panel unit root test statistics generally have the implication of first order unit root. Cointegration tests in general reject the null hypothesis of no cointegration. Therefore we conclude that there is at least one long run relationship between these variables. The panel data analysis results show that higher education has an influence on gross domestic product in the case of BCs. This result is a confirmation of Education led Growth hypothesis, in the case of BCs. The results for BCs support the historical data for the link between education and economic performance provided by Maddison (1991) in terms of regression analysis. However our findings are not consistent with the idea that education is a growthdriven investment (Jaoul 2004). That is, in the case of BCs, there is a unidirectional causality between education and growth, contrary to the findings of (Islam et al 2007). Four types of panel models have significant coefficients estimates both for constant term and slope parameter. Having slope parameter estimates around 1, we conclude that there is an almost unit elasticity between education and economic growth. The findings here can be further calibrated with an estimation of a comprehensive multivariate model allowing other factors that affect growth, paying particular attention to education and then proceed, testing its robustness across region. Policy conclusions then would be more accurate and beneficial. References Barro, R. J. (2002) Education and Economic Growth, dataoecd/5/49/ pdf,

8 8 Hadri, Kaddour (2000). Testing for Stationarity in Heterogeneous Panel Data, Econometric Journal, 3, Im, K. S., M. H. Pesaran, and Y. Shin (2003). Testing for Unit Roots in Heterogeneous Panels, Journal of Econometrics, 115, Islam, T. S., Md A. Wadud, and Q. Islam, (2007) "Relationship Between Education and GDP Growth: A Mutivariate Causality Analysis for Bangladesh." Economics Bulletin, Vol. 3, No. 35 pp Jaoul, M. (2004) Higher education, causality and growth: a comparison of France and Germany before the Second World War Compare, Volume 34, Number 1, pp (17). Johnes, G. (2006) Education and Economic Growth, written for presentation as the twelfth lecture in the Eric John Hanson Memorial Lecture Series at the Department of Economics, University of Alberta, Edmonton. Kao, C. (1999). Spurious Regression and Residual-Based Tests for Cointegration in Panel Data, Journal of Econometrics, 90, Kui, L. (2007) The Interactive Causality between Education and Economic Growth in China final paper for the course Development Economics. Department of Political Economics, School of Marxism, Peking University. Levin, A., C. F. Lin, and C. Chu (2002). Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties, Journal of Econometrics, 108, Maddison, A. (1991), Dynamic Forces of Capitalist Development, Oxford University Press, Oxford. Pedroni, P. (1999). Critical Values for Cointegration Tests in Heterogeneous Panels with Multiple Regressors, Oxford Bulletin of Economics and Statistics, 61, Pedroni, P. (2004). Panel Cointegration; Asymptotic and Finite Sample Properties of Pooled Time Series Tests with an Application to the PPP Hypothesis, Econometric Theory, 20, Stevens, P. and M. Weale (2003), Education and Economic Growth, National Institute of Economic and Social Research, London. Taban, S. and Kar, M. (2006) Human Capital and Economic Growth: Causality

9 Analysis, Anadolu University, Journal of Social Sciences, 6(1),