The contribution of education to economic growth: Empirical analysis in the Middle East and North Africa region

In this paper we will focus on education. Indeed, most theoretical analyzes have confirmed that human capital has a positive and significant effect on growth. The paper aims to examine in time series the causality between human capital and growth in MENA’s region. For this, we carry out our empirical investigation by employing various human capital measures suggested in the literature. The results show that cointegration between education and economic growth exists only in Tunisia, Turkey, Morocco, Iran and Israel. However, in the other countries the causality does not exist because they don’t have effective means to improve their growth.

effect of education (especially higher education) on economic growth coincide with the theoretical results. It is essential to remember the old basic empirical work before presenting recent works. Economists still refer to the old basic models. These will be summarized in a summary table. Finally, we will examine the causal relationship between higher education and growth for the studied countries. We will answer to our problem: Does the effect of higher education on economic growth exist?

2-Empirical investigation: Variables and data
In this study, we chose four indicators of human capital. The first one represents a traditional proxy of human capital; it's the number of graduates in science and engineering (GRD).
Second, we have the openness rate (Trade). In fact, it is the sum of exports and imports of goods and services as a % of GDP. The third indicator of human capital is the secondary school enrolment rate (School), refers to Benhabib and Spiegel (1994) this indicator represents a good proxy of the human capital. Finally, the fourth indicator is used to measure the physical capital. We mean the gross fixed capital formation (GFCF) as a % of GDP.
Concerning the economic growth, the standard literature on the ties between economic growth and human capital generally uses the growth rate of GDP per capita. The data sources are the Word Development Indicators (WDI) of the World Bank. (2011), and all variables are expressed in national currencies. The time span of the variables is 1960-2011. The study focuses on only 9 countries because of the non availability of data.

2-1: Unit Root Testing
This test consists to detect the non-stationary variables and then apply the cointegration test on these variables. If the variable is stationary, it called integrated I(0). Besides, the nonstationary variable is integrated I(1). In the table 1, we find the different indicators of human capital and the proxy of economic growth expressed in their natural logarithm. The results of unit root tests are presented in level and in first difference. The results show that all the variables in level are integrated I(1) except for Algeria and Egypt where the variable School is stationary since the unit root hypothesis is strongly rejected. In addition to this, we note that for Jordan and Israel the variable Trade is I(0) and for Mauritania the variable GFCF is also stationary. When the tests are carried out on the first difference, the hypothesis of unit root is rejected in the case of some countries such as Iran, Egypt, Jordan, Morocco and Turkey.

2-2: Cointegration Testing
The cointegration tests consist to identify the stationarity of the residue of two linear combinations. If the cointegration is demonstrated, so a long-run relationship of equilibrium exists between the two series. In this paragraph we will study the cointegration tests between the different indicators of human capital and the economic growth. The computations are based on the Johanson procedure trace statistic and the null hypothesis (H 0 ) is that there is no cointegration vector; the alternative one (H 1 ) is that there is one cointegrating vector. The hypothesis of non-cointegration is rejected for the GDR for two countries: Israel and Turkey. With the variable GFCF, there are also two cases of cointegration with GDP per capita: Iran and Tunisia. Finally, with the third indicator of human capital Trade, the hypothesis of non-cointegration is rejected in the case of Morocco. For the remaining countries (Algeria, Egypt, Jordan and Mauritania) and for the different proxies of human capital, the hypothesis of the absence of coinegration cannot be rejected. Such an outcome rejects, in these countries, any stable relationship between human capital indicators and economic growth. For the countries where cointegration is detected (Iran, Israel, Morocco, Tunisia and Turkey), a long-run relationship between human capital indicators and growth exist. In other words, the variables are in a long-run equilibrium state. Consequently, the short-run dynamics of the variables are seen as fluctuations around this equilibrium. And the Error Correction Model (ECM) indicates how a system adjusts to converge to its long-run equilibrium state. We note that α 1 represents the adjustment coefficient of the human capital indicators and α 2 is the adjustment coefficient of growth.

Table-3. The adjustment coefficients and the error correction term Countries
The adjustment coefficient The error correction term Vector  β X t-1 X t-1 =y t-1 - 1 (GDR) t-1 - 2 According to table 3, in the cases of Iran, Israel and Morocco α 1 and the error correction term are positives and significant, this means that the effect of human capital on long-run growth is positive. However, for Tunisia, α 1 and the error correction term are negatives and significant, so we have the same conclusion; the effect on growth is positive. Moreover, α 1 is negative and non-significant in the case of Turkey, which excludes any effect of education on long-run growth. For Iran and Tunisia, the effect of growth (α 2 ) on human capital is positive. In contrast, for Morocco and Turkey the effect on education is negative. And there is no effect on education in the case of Israel. To check the robustness of these results, one has to see the dynamic interaction between the cointegrated variables in the long-run and how each one is causing the other.

2-3: Granger causality tests
According to table 4, we have the results of the tests using the Johanson procedure for the determination of the cointegrating vectors. The results show that for Israel the causality tests are in favor of a unidirectional causality between human capital and economic growth.
However, for Turkey, the statistical significance of F-and t-statistics at the 5% level shows that the causation is going in other direction. In other words, the causality tests are in favor of a reverse causation running from economic growth to the human capital. In addition, we note that for Iran, Morocco and Tunisia the evidence is in favor of bidirectional causality between the growth rate of GDP per capita and the proxies of human capital. Indeed, we conclude that in Iran, Morocco and Tunisia t 1 and F 1 statistics are both significant, and in Iran and Turkey t 2 and F 2 statistics are also significant. That means that real growth has two effects on human capital: The first one is coming from the lagged dynamic terms and the second from the error correction term. According to the first effect, each short-term change in the economic growth is responsible to the future change in the growth rate of human capital indicators. For the second effect, given the significance of the error correction term in the second VAR equation, real growth exerts an influence on human capital through the error correction term. We remember that according to the table 2, there are 5 countries where cointegration is detected. For the remaining countries, we applied the causality tests using the first differenced VARs. The evidence presented is not far from the results obtained from the ECMs. The causation turns out to be bidirectional in the case of Israel. Indeed, in Mauritania, Tunisia and Turkey the evidence is in favor of a reverse causation going from economic growth to human capital, with at least one education proxy at 5% level. That is, not only education shows to Granger-cause growth in the short-run (cases of Egypt and Morocco), but also the real growth appears to Granger-cause the education too.

3-Conclusion
This study has examined empirically the causality between human capital and economic growth in a bivariate VAR structure for a sample covering 9 countries of MENA region over the period 1960-2011. Johanson cointegration analysis provides that human capital does not seem to affect positively the long-run economic growth. Indeed, the results of this paper clearly indicate that a strong evidence exist in favor of a reverse causation running from growth to education for 4 countries. For countries where education and economic indicators are not cointegrated, Granger causality tests were carried out with first-differenced VARs to check the causality problem in the short-run. The results display that evidence was found of bidirectional causality between growth and education. The empirical evidence presented above has important implications for the conduct economic policies in these countries