The relationship between higher education and
economic growth in China
Jilei Cai (10435018)
Bachelor Thesis Economics and Business Specialization: Economics and Finance
Faculty of Economics and Business University of Amsterdam
Abstract:
Previous literatures have researched the relationship between higher education and economic growth, but there are only a few papers about the China. To analysis the correlation between higher education and economic growth in China, this paper uses conintegration test, Granger causality test and variance decomposition based on vector autoregressive model to estimate a modified augmented neoclassical model for time series data over the 1978-2012 period in China. We find that both higher education and physical capital investment have positive and statistically significant effect on the growth of GDP per capita. We also find that higher education not only promotes national income directly but also boosts national income indirectly through enhancing the productivity of labor. The policy implication of our results will also be discussed in this paper.
Table of Contents
1.
Introduction ... 3
2. Literature Review ... 6
3. Methods and data ... 10
3.1.
Econometric Model ... 10
3.2.
Sources and Data ... 13
4.
Econometric Analysis ... 16
4.1.
Stationarity test ... 16
4.2.
VAR model and Cointegration test ... 17
4.3.
Granger causality test ... 21
4.4.
Variance Decomposition ... 22
5.
Conclusions and Recommendations ... 26
1. Introduction
Schultz (1961) theorized that the capability of a nation to use physical capital effectively is a function of its level of human capital and that if its level of human capital does not increase along with its level of physical capital, then human capital will limit this nation’s economic growth. Both macroeconomic literature (Pereria & Aubyu, 2009) and microeconomic literature (Psacharopoulos, 1995) propose a positive correlation between the human capital and education in the economic growth progression. These opinions are accordance with our definition of human capital — education influences the economic growth through improving the productivity of labor. Thus, it is both the quantity of labor and education level embodied in labor influence the level of economic growth.
It is admitted that primary education accounts for the largest part of public expenditure in most developing countries. This situation also happens in China, the Chinese population has an almost universal primary level education because primary education is totally free. However, the neglect of higher education will lead to the low quality of labor force and limit the development of national income (World Bank, 2000). Moreover, McMahon (1998) concludes that additional development in primary education does not provide a good return when universal primary education has been achieved. Chinese national income, on the other hand, always keeps at a high growth rate. This causes the author’s interest to analyze whether higher education is the factor that drives economic growth in China.
The purpose of this paper is to investigate relationship between higher education and economic growth in China in the period from 1978 to 2012. We do so by using the Vector Auto Regressive (VAR) model to estimate an expanded Solow growth model with higher enrolment rates as the proxy of human capital.
Before 1978, China, who’s Real GDP per capita was only 10% of Real GDP per capita of Brazil and 2.5% of Real GDP per capita of the United States, was one of the poorest countries in the world. Since 1978, China moved from centrally planned
economy to market-based economy and achieved a Real GDP per capita growth rate of nearly 8% every year (World Bank, 2002). With this open economy, the GDP of China reached 58786 hundred million in 2010, only lags behind the GDP of the United States. Further, with the development of information and communication technology in recent decades, the development strategy of China has dramatically changed from traditional natural resource-oriented production to knowledge-based production (Ministry of Education of the PRC, 1993). Similar to the case of many other emerging economies, increasing knowledge and skills of workforce is an important strategy to raise the competitiveness of China in the world market. Intellectual human capital is becoming more and more important in this era. Qazi et al. (2013) proposed that ‘Higher education provides intellectual human capital such as critical thinkers, researchers, scholars, innovators and responsible citizens to societies’ (p. 2). Thus, since the number of students attended higher education in China is far lower than other countries have the same level of national income; higher education institutions in China are guided by the Chinese government to expand the enrolment rate of higher education in 1999, the growth rate of higher education enrolment rate increased from 8.5% to about 20% every year. This situation shows the importance of higher education to economic growth has been appreciated among policy makers in China in last 15 years.
This paper attempts to contribute to the literature on the topic since very few empirical studies have been conducted on the correlation between higher education and economic growth in China. Such investigation not only adds to the empirical research on the topic, but it also has policy implications. High-income countries benefit more from higher education (Petrakis and Stamatakis, 2002). If higher education is as important as we referred above, the high increase rate of higher education enrolment rate can explain the GDP rapid growth in China from 1999 to 2010. Thus, this research will provide information for the Chinese governments to decide whether policy efforts should focus on increasing the higher education levels.
previous papers that are relevant to this paper; Section III explores the theoretical foundations for expanded Solow growth model we use and discusses the data sources; Section IV presents and analyzes the statistical results while Section V concludes the paper.
2. Literature Review
An interest in the causal relationship between education and economic growth continues to draw much attention from scholars. It is widely considered that education levels can influence the economic growth directly (Schlottmann, 2010). We therefore discuss some studies that emphasize the importance of education in the progress of economic growth.
The Solow growth model, which was the central framework to interpret economic growth, only took exogenous factors (technology and population) as the factors that boost the income growth. Despite the pervasive use of the Solow growth model, Schultz (1961), a famous American economist, noticed that there is a difference between the growth rate of output and the growth rate of physical capital and labor input, proposing that the major explanation for this difference is probably the level of human capital. Then, in 1980s and 1990s, Lucas (1988) introduced the endogenous growth theory and Mankiw, Romer and Weil (1992) proposed the expanded neoclassical growth model to investigate the contribution of human capital to the economic growth. In the literatures of expanding Solow growth model, the human capital is seen as the asset input and hence it predicts that ‘countries with faster growth rate of education will have faster transition growth rates and higher incomes’ (Gyimah-Brempong, Paddison & Mitiku, 2006, p. 8). Mankiw et al. (1992) added a new variable (human capital accumulation) to the Solow growth model and found that the education has positive and strongly significant effect on the economic growth rate in a sample of 88 countries. In the literatures of endogenous growth model, education is regarded as a process that increases the production technology and enhances the knowledge of workers so that they can adapt foreign technology more easily. (Romer, 1990; Barro, 1997). Further, Lucas (1988) and Romer (1986) suggested that education could restrain the reduction of diminishing returns of physical capital inputs to economic growth through both internal effect and spill-out effect. According to the theory of Lucas, ‘the internal effect of education is the effect of individual human capital on productivity and the spill-out effect of education is its ability to increase the
efficiency of socio-economic growth activities’ (Geng, Li & Cai, 2009, p. 1). Hence, it is clear that in both endogenous and expanded Solow growth model, education raises the growth rate of income.
All the papers reviewed above were based on the endogenous growth model and expanded Solow growth model, however, there are papers using different models to research the correlation between human capital and economic growth. The quality of human capital, such as the knowledge, skills and competencies, increased by attending education contributes a lot to the improvement of a country’s income (Barro, 1999; Burja & Burja, 2013). Benhabib and Spirgel (1994) began their study based on a Cobb-Douglass production function rather than a growth equation. Their main result is that education levels do not raise the growth rate of income directly; instead, it raises the income indirectly through facilitating the adoption of advanced technology and promoting the productivity of physical capital. In addition, Barro and Lee (1994) were the first to use average years of schooling among adults whose age 25 and older to represent the stock of human capital. They found that male secondary schooling is positively related to economic growth while female secondary schooling is negatively related to economic growth. The results of Barro and Lee are supported by the study of Sala-i-Martin (1997).
Previous findings verify the positive relationship between education and economic growth, there are also various studies highlight the effect of higher education on national income. Higher education can affect national income positively through developing domestic technology or through adapting foreign technology to local conditions. Greiner and Semmler (2002) find that positive externalities of physical capital inputs exist only when educated human capital is available. Furthermore, Hall and Jones (1999) note that promotion of technology is not able to be dependent on primary and secondary education, but on higher education. Higher education is also able to increase the quality of other inputs. For example, physical exercise, which is an important part of higher education, improves health of human capital that is crucial determinant of economic growth (Nelson and Phelps, 1966). Kimenyi (2011) also
argued that growth of national income has positive effect on higher education in Africa. Moreover, the impact of higher education on national income comes about more through its social role than through its technical role (Gradstein & Justman, 2002). Higher education, Gyimah-Brempong et al. (2006) argue, ‘reduces the cost of enforcing desirable social norms, lessens the potential for ethnic conflicts in ethnically diverse societies, as well as decrease transaction costs by shrinking social distance between individuals in a society’ (p. 8). Furthermore, Agiomirgianaskis, Asteriou and Monasitiriotis (2002) and Voon (2002) find tertiary education has a stronger growth impact than primary education and secondary education in Hong Kong. Gyimah-Brempong, Paddison and Mitiku (2006) suggested that higher education workforces have a large effect on the income growth in a sample of African countries from 1960 to 2000.
Chinese scholars also have done some researches on the relationship between education and economic growth in China. The contribution rate of education to growth rate of the average annual GDP in China increased from 8.84% to 12.66% from 1982 to 2000 (Cui, 2001). The finding of Cui confirms the positive effect of education on economic growth in China. In addition, Cai (1999) used the Feder model to estimate the indirect effect and direct effect of education on economic growth. Cai argued that indirect effects of education, through enhanced health, utilization of physical capital investment and reduction crime rate, have more significant influences on national income than direct effects of education. Yu et al. (2014) proposed that education and economic growth can promote each other in China, meaning that increase of GDP can reciprocally promote the education through allocating more social sources to invest. Chen and Feng (2000) also confirmed that human capital (higher education enrollment rate) is the endogenous growth factor in promoting national income.
Based on the papers discussed above, the role of higher education seems to be a considerable growth factor in China. We might also hypothesis that higher education can promote national income indirectly through enhancing the efficiency of exploiting
3. Methods and Data
This section explains the research method and data collected. Section 3.1 gives a detailed explanation of the variant augmented neoclassical model used. Section 3.2 explains how we choose and collect the data we need and gives a simple analysis of these data.
3.1. Econometric Model
The approach we use to investigate the correlation between the higher education and economic growth in China is to estimate an augmented neoclassical model proposed by Mankiw, Romer and Weil (1992) by using the Vector Auto Regressive (VAR) model for the years 1978-2012.
The neoclassical model, originally proposed by Solow (1956), assumed an aggregate production function and took the rates of saving and population growth as exogenous variables determined the level of income per capita. Schultz (1961), however, argued that the estimation of the returns of investment in education to economic growth should take both physical capital input and human capital input into account since these two inputs are complementary. Additionally, Mankiw et al.’s (1992) work tested Solow model and argued that effects of saving and population growth on income growth are too large in the Solow model. Mankiw et al. (1992) proposed that human capital accumulation was correlated with saving rates and population growth; this confirmed that estimated coefficients on saving and population growth was biased by ignoring human capital in the model. Thus, Mankiw et al. augmented the Solow model by adding a proxy for human capital as an additional explanatory variable. Specifically, a Cobb-Douglas production function of Mankiw et al. model is assumed in the following form:
Y t = K t αH t β A t L t 1−α−β (1) where Y is aggregate output; K is physical capital; H is the stock of human capital; A
represents the level of technology; L is labor; and t is time. Mankiw et al. assume that L and A grow at constant and exogenous rates n and g, respectively. The exponents α and β measure the elasticity of output to the physical capital and human capital inputs. Assuming decreasing returns to physical and human capital, that is α + β < 1, the steady-state level of income per capita can be derived from the equation (1) as the following form:
lnYL = lnA + gt −1−α−βα+β ln(n + g + δ) +1−α−βα ln(sκ) +1−α−ββ ln(sh) (2) where sk is the fraction of income invested in physical capital; sh is the fraction of income invested in human capital; n, g and δ are all exogenously determined growth rate of population, technology and depreciation rate of capital, respectively. In the neoclassical model, Solow (1956) argued that A term reports not only technology but country-specific resource endowments, climate, institutions, and so on. He assumed that lnA = a + ε, where a is constant and ε is country-specific. Thus, Mankiw et al. (1992) assumed that the error term (ε) is independent of the explanatory variables, which allowing estimation to continue by utilizing ordinary least squares (OLS). In this situation, the problem of endogeneity will not happen as the independent variables are not correlated with the error term.
Following the study of Mankiw et al. (1992), we assume the growth equation like the equation (2). Nevertheless, this paper makes one modification to the augmented neoclassical model: we use the higher education enrolment rate as a measure of human capital as the higher education human capital is the major interest in this thesis. Based on the foregoing, this paper postulates the variant of the augmented neoclassical growth equation. Formally, the equation is written as:
lnqt = a0+ a1lnkt + a2ln(n + g + δ) + a3lnht (3) where qt is the GDP per capita; kt is the gross capital formation as a fraction of GDP; n, g and δ are growth rate of population, technology and depreciation rate of capital, respectively; ht is the higher education enrolment rates. All the variables are
turned to their natural logarithms form for analysis.
The estimation of the higher education enrolment rate is calculated by using the following equation (National Bureau of Statistics of China, 2012):
GHERt =E
t
Pt× 100 (4)
where GHERt = Gross Higher Enrolment Ration in year t; Et = Enrolment for higher level of education in year t (age 18-23); Pt = Population in age-group which officially corresponds to higher level of education in year t (age 18-23). There is no possibility that people older than 23 are enrolled in higher education as people over 23 are not allowed to take the Gaokao which is the higher education entrance examination.
To investigate the relationship between higher education and economic growth, empirical literatures were conducted by using different econometric analysis methods to estimate the augmented neoclassical growth equation. First, we use two unit root tests to identify the integrated order of the variables. If all variables are integrated of order zero in levels, all variables are stationary. Then, we can use OLS directly to do regression to estimate our growth equation. However, if some variables are stationary only in their first difference, using OLS directly to do the regression will cause the problem of spurious regression, meaning that growth equation estimated through OLS cannot truly reflect the relationship between independent variable and explanatory variables. In this situation, we should build Vector Auto Regressive (VAR) model and examine whether non-stationary variables in levels have cointegration relationship (long-run stable relationship) between higher education and economic growth through using Johansen test based on the VAR model we built. The presence of cointegration relationships among the non-stationary variables in levels implies that Granger causality also exists. Last, we employ the variance decomposition to further examine the results of cointegration test and Granger causality test. This paper gets all econometrical results through using econometrical software (Eviews 8).
3.2. Sources and Data
The dependent variable in our model is the GDP per capita (lnqt). The explanatory
variables are investment in physical capital, higher education enrolment rate, the sum of annual growth rate of labor, depreciation, and technology. Following earlier researchers, we measure physical capital investment as the gross capital investment/GDP ratio (lnkt) and measure the growth rate of population as the rate of labor force growth (lnnt) in a period. Following Mankiw et al. (1992), we measure the sum of growth rate of depreciation and technology remains constant, presuming that g + δ = 0.05. The proxy of human capital is the key issue in our model. Different proxies have been used to measure the impact of education on national income (Tsamadias & Pegkas, 2012). For example, some researchers, such as Barro (1999) and Perakis & Stamatakis (2002), use enrolment ratios while others use education expenditure/GDP ratio (McMahon, 1987; Appiah and McMahon, 2002). Pegkas and Tsamadias (2012) argued that the school enrolment rate is just the quantitative measurement of human capital, which means the quality of human capital is not considered. This paper, therefore, uses the higher education enrolment rate as the proxy of human capital. The reason for choosing the year from 1978 to 2012 is that the China experienced ten-year Chinese Cultural Revolution before 1978. During 1967 to 1977, all higher education institutions were not allowed to recruit the students (Barnouin & Yu, 1994).
The time series data were collected over the 1978-2012 period. The data of the GDP per capita, Gross capital formation/GDP ratio, the annual growth rate of GDP per capita, and the annual growth rate of Gross capital formation/GDP ratio were obtained form World Bank’s World Development Indicators (Washington, DC: World Bank, 2013) and were for the years 1978 to 2012. The data of the annual growth rate of population of labor force, the annual higher education enrolment rate and the annual growth rate of higher education enrolment rate were obtained from China Statistical Yearbook, China Education Finance Statistical Yearbook and China Population Statistics Yearbook 1978-2009 (National Bureau of Statistics of China,
1978-2009).
After a first data analysis, we notice that there has been a significant increase in GDP per capita as well as a great increase in higher education enrolment rate during 1978-2012 (Figure 1).
Figure 1. GDP per capita and Higher education enrolment rate (1978-2012).
Table 1. Statistical description of the sample data
Variables Mean value Minimum value Maximum value
GDP 1254 154.97 (year 1978) 6091 (year 2012)
K 39.52% 33.33% (year 1981) 49% (year 2012)
L 63486.8 40152 (year 1978) 76704 (year 2012)
H 10.57% 1.55% (year 1978) 30% (year 2012)
GDP growth rate 0.11 -0.108 (year 1987) 0.2878 (year 2008)
K growth rate 0.018 -0.051 (year 1994) 0.187 (year 1993)
L growth rate 0.02 0.0032 (year 2008) 0.1703 (year 1990)
H growth rate 0.096 -0.093 (year 1982) 0.335 (year 1979)
More specifically, during 1978-2012, GDP per capita of China experienced an
0 5 10 15 20 25 30 35 0 1000 2000 3000 4000 5000 6000 7000
average annual increase of 11% and the highest growth rate of GDP per capita existed in year 2008 and the lowest growth rate appeared in year 1987. During the examined period, the enrolment of higher education has shown a radical increase, with 1.55% in 1978 to 30% in 2012, exhibiting an average annual growth rate of 9.6% (Table 1). Moreover, the maximum value of GDP per capita, Gross capital investment/GDP ratio, population of labor force, and higher education enrolment rate all appear in year 2012 while the minimum value of these indicators all exist in the late 1970s and early 1980s.
4. Econometric Analysis
This section will provide the results and discussion of econometric tests
4.1. Stationarity test
Firstly, in order to check the stationary of the variables (GDP per capita, physical capital investment, population of labor force and higher education enrolment rate), we employ two different unit root tests. The stationarity of the data set is assessed using Augmented Dickey and Fuller (ADF) (1979) and Phillips and Perron (PP) (1988). In this step, these two tests are applied to test the existence of unit roots for each variable in levels and in first differences. The variables are clearly identified by including intercept and including intercept and trend. Schwarz (1978) determined the criteria for the optimal lag length of the ADF unit root test. The PP statistics are acquired by the Bartlett Kernel and the automatic bandwidth parameter approach (Newey and West, 1994). Moreover, for the ADF and PP tests, the null hypothesis predicts that the variable tested has unit root that means the variable is non-stationary.
Table 2. Results of unit root tests
ADF test PP test
Variables
(in levels & first differences) With intercept in equation With intercept and trend in equation With intercept in equation With intercept and trend in equation ln𝐪𝐭 1.879797 -0.347772 2.683128 -0.480893 Δ ln𝐪𝐭 -3.490470** -4.46163*** -3.577764*** -4.345298*** ln𝐤𝐭 -0.579244 -2.288689 -0.579244 -2.288689 Δ ln𝐤𝐭 -4.823228*** -4.829060*** -4.783779*** -4.755812*** ln(𝐧 + 𝐠 + 𝛅)𝐭 -6.006345*** -8.322542*** -5.817465*** -9.122103*** Δ ln(𝐧 + 𝐠 + 𝛅)𝐭 -9.796345*** -9.628440*** -15.46653*** -15.36163*** ln𝐡𝐭 0.164256 -3.896973** -0.485902 -1.974240 Δ ln𝐡𝐭 -3.820118*** -3.753158*** -3.990215*** -3.925223***
Note: ***, ** represents the rejection of the null hypothesis (the variable tested is non-stationary) of ADF test and
PP test at 1% and 5% level of significance respectively. The hypothesis is tested by F-test. If the calculated
F-statistics is lower than the lower bounds critical value given by Mackinnon (1996), the null hypothesis is
rejected.
The results in Table 2 show that GDP per capita (lnqt), higher education (lnht) and physical capital investments (lnkt) of ADF test has unit roots in their levels but they do not have unit costs in their first differences at the 1% level of significance, meaning that these three variables are stationary in their first differences instead of in their levels.. The labor growth rate is integrated of order zero both in its levels and in its first differences. Based on these results, applying OLS to do the regression directly will cause the problem of spurious regression as only labor growth rate is stationary in levels. Hence, we have to employ VAR model to examine the relationship between higher education and economic growth.
4.2. VAR model and Cointegration test
To estimate the relationship between complete endogenous variables, VAR model regress the lagged value of explained variables on the dependent variable in the form of simultaneous equations (Sims, 1980). VAR model, therefore, is always been utilized to explore the dynamic relationship between explained variables and dependent variable of a growth equation (Anwar et al. 2011). However, in order to examine the reliability of VAR model we built, a cointegration test is required.
Herrerias (2010) argued that the first requirement to use VAR model is that explained variables should all be stationary in first difference and the second requirement is that degree of freedom of lag length should maintain reasonable logical. Stationary tests show that GDP per capita, physical capital investment and higher education are stationary in first differences despite the unit root tests show that they follow an I(1) process in levels, which means the first requirement of building VAR model is fulfilled. However, the variable (n+g+δ) is exogenous in our model now as it
is stationary in both levels and first difference. Thus, in order to apply VAR model directly to determine the relationship among three variables used in our model, we should determine the reasonable logical lag length. Seven versions of the system were estimated for identifying the optimal lag length: a one, a two, a three, a four, a five, a six and a seven-lag version. The result of lag length determined by likelihood ratio statistics, Akaike information criterion (AIC), Schwarz criterion and Hannan-Quinn information criteria is shown in Table 3. Among these criterions, AIC is the most common one to be used to define optimal lag length (Pegkas & Tsamadias, 2014; Soytas & Sari, 2009). Table 3 shows that the AIC identified seven lags as optimal lag length.
Table 3. Vector autoregressive lag order selection criteria
Lag LogL LR FPE AIC SC HQ
0 4.409496 NA 0.000182 -0.100678 0.042058 -0.057042 1 116.5017 192.1580 1.16e-07 -7.464405 -6.893461* -7.289862 2 128.6342 18.19874 9.50e-08* -7.688155 -6.689001 -7.382704 3 134.2825 7.262146 1.29e-07* -7.448750 -6.021388 -7.012391 4 142.5257 8.832026 1.55e-07 -7.394695 -5.539124 -6.827428 5 146.2252 3.170964 2.88e-07 -7.016085 -4.732305 -6.317911 6 152.6239 4.113435 5.27e-07 -6.830276 -4.118288 -6.001194 7 193.6410 17.57877* 1.14e-07 -9.117214* -5.977017 -8.157224*
Note: LR, likelihood ratio; FPE, final prediction error; AIC, Akaike information criterion; SC, Schwarz
criterion; HQ, Hunnan-Quinn information criteria; NA, not applicable. The * indicates the lag length
determined by each criteria.
After defining the optimal lag length, it is possible to build VAR model and check whether GDP per capita, physical capital investment and higher education are cointegrated through cointegration test. Only when these three variables are cointegrated, the VAR model we built to estimate the augmented neoclassical model
will include economical meaning. We use the reduced rank procedure developed by Johansen (1988) to conduct the cointegration test. The Johansen multivariate cointegration approach is utilized to regress the variables at their original levels to examine the long-run relationship between lnqt, lnkt, and lnht. The estimation procedure of cointegration test assumes only an intercept in the VAR estimation (Johansen and Jeselius, 1990). Furthermore, two statistics, the Trace and maximum Eigenvalue test, are recommended to check the long-run relationship between variables (Yu, Zhao, Xu & Wang, 2014). The result of cointegration test is shown in Table 4.
Table 4. Johansen and Juselius cointegration test GDP per capita, physical capital
investments, and higher education enrolment rates: sample 1978-2012
Series: ln𝐪𝐭 ln𝐤𝐭 ln𝐡𝐭 Hypothesised No. of CE(s) Eigenvalue Trace Statistics 5% critical value Max-eigen statistic 5% critical value None* 0.884693 78.96479 29.79707 60.48448 21.13162 At most 1* 0.469630 18.48030 15.49471 17.75707 14.26460 At most 2 0.025499 0.723235 3.841466 0.723235 3.841466
Note: No. of CE(s) indicates the number of cointegrating relationship. *Trace and Max-Eigen tests
indicate 2 cointegrating equation at the 5% level.
Lags interval: 1 to 6
Table 5. Normalized cointegrating coefficients (standard error in parentheses)
LNQ LNK LNH
1.000000 -4.698494 -0.642227
(0.73083) (0.04297)
The null hypothesis in the cointegration test is that there is no cointegration vector, meaning no long-run relationship. The null hypothesis of at most one co-integrated
vector in the Trace test could be rejected at 5% whereas the null hypothesis of at most two cointegration relationships could not be rejected at 5%, which implies that there are two cointegration relationships. The finding of the Trace test was further supported by the results of the maximum Eigenvalue test, in which only the null hypothesis that there are at most two cointegrating vector could be rejected at 5%. Thus, the results (Table 4) lead to the conclusion that the GDP per capita, physical capital investments, and higher education enrolment rates are cointegrated and these variables have a long-run relationship during the period examined. The existence of cointegration among GDP per capita, physical investment and higher education confirms that the VAR model we estimated has economical meaning. In addition, Table 5 shows that lnqt and lnht has opposite sign of their cointegrating
coefficients. As both of the LNQ and LNH are at the same side of the equation, the opposite sign of their cointegrating coefficients means that GDP per capita and higher education are positively related.
One estimated cointegration relationship estimated by VAR model is presented in the following equation (standard error in parentheses):
Lnqt = 0.408959lnkt + 0.138002lnht - 1.183283 (0.52207) (0.30094) (-0.27540)
From the above estimated equation, it is concluded that physical capital investment and higher education have a significant positive long-run effect on economic growth since the coefficients of these two variables are statistically significant at 1% level. The elasticity of GDP per capita regarding higher education, which uses enrolment rate as proxy, is 0.138. This means that a 1% increase in higher education enrolment rates will foster economic growth by about 0.138%. The role of higher education in determining and explaining economic growth seems to be very important. The findings of this paper of the positive long-run relationship between higher education and economic growth are consistent with most of the previous literatures discussed above.
4.3. Granger causality test
This step is to clarify the relation by using Granger causality test, which is based on the VAR model, introduced by Wiener (1956) to judge whether there exists mutual interactions between variables. As this paper has proved that there exists long-term positive correlation between GDP per capita, physical capital investments, and higher education, there must be Granger causality in at least one direction. The Granger causality means if prediction accuracy of current value of y is significantly increased by including information about past value of x, x is said to Granger-cause y (Granger, 1969; Granger 1980). In this paper, if GDP per capita is Granger caused by higher education, it means higher education has positive effect on economic growth. The Granger Causality implemented here employed the seven-lag for all variables. The result is shown in Table 6.
Table 6. Granger causality test
Null hypothesis Lag length F-Statistics Probability Conclusion
LNK does not Granger cause LNH 7 1.29877 0.3242 Accepted LNH does not Granger cause LNK 7 2.53127 0.0704 Refused LNQ does not Granger cause LNH 7 0.66777 0.6960 Accepted LNH does not Granger cause LNQ 7 0.50897 0.8122 Accepted LNQ does not Granger cause LNK 7 1.77724 0.1758 Accepted LNK does not Granger cause LNQ 7 0.30828 0.9375 Accepted
Granger causes of economic growth while economic growth does not Granger cause physical capital investment and higher education at the seventh lag order. Furthermore, higher education is the Granger cause of physical capital investment even if physical capital investment is not the Granger cause of higher education. This result is consistent with the paper mentioned above that higher education can improve the productivity of labor to use physical capital investment. However, the Granger causality test does not show any Granger causal interaction between economic growth and higher education, meaning that higher education and economic growth do not have any relationship. On the other hand, the result of cointegration test shows that higher education is important in economic growth. In conclusion, as there is a difference between result of cointegration test and result of Granger causality test, we employ variance decomposition under VAR system in next section to further gauge the relationship between economic growth and higher education.
4.4. Variance Decomposition
Variance decomposition divides fluctuation of each variable into proportions to attribute shocks to individual variables through different time period to analyze the relative importance between independent variables (Masih and Masih, 1997). Variance decompositions are also constructed from VAR model and can directly address the contribution of independent variables to forecasting dependent variable (Litterman and Weiss, 1985). In many papers, it is argued that variance decomposition is the most plausible way to identify causal relationship between considered variables based on economic theory (Lorde et al., 2010; Hye, 2012; Raza and Jawaid, 2013). The results of variance decomposition of lnqt, lnkt and lnht are shown in Table 7, Table 8 and Table 9.
Table 7. Variance decomposition of lnqt
𝐥𝐧𝐪𝐭
1 0.100445 71.46655 3.750836 24.78216 2 0.163219 61.14688 5.545117 33.30800 3 0.194368 51.29020 6.618379 42.09142 4 0.227889 44.09545 8.948426 46.95612 5 0.264557 40.18416 10.37454 49.44130 6 0.281447 38.15113 11.12389 50.72497 7 0.295453 37.38681 10.79793 51.81526 8 0.308318 35.08524 10.18365 54.73111 9 0.318273 33.76510 9.558862 56.67603 10 0.332877 32.21747 8.738523 59.04401
Table 8. Variance decomposition of lnkt
𝐥𝐧𝐤𝐭 Periods S.E. lnqt lnkt lnht 1 0.046735 0.000000 35.07359 64.92641 2 0.061708 19.12031 27.14098 53.73871 3 0.062750 20.26702 27.52908 52.20390 4 0.064223 19.54868 26.28247 54.16986 5 0.072038 15.84163 22.20441 61.95395 6 0.073226 15.38951 21.62711 62.98338 7 0.079517 18.45460 18.40667 63.13873 8 0.083792 17.92530 17.63932 64.43538 9 0.084899 19.92560 17.29310 62.78130 10 0.086729 22.29611 17.52944 60.17444
Table 9. Variance decomposition of lnht
𝐥𝐧𝐡𝐭
Periods S.E. lnqt lnkt lnht
2 0.121528 4.526725 1.883613 93.58966 3 0.172423 7.606181 1.072267 91.32155 4 0.213534 6.914769 1.275989 91.80924 5 0.252009 8.110373 1.887734 90.00189 6 0.293540 8.483441 2.088816 89.42774 7 0.315127 7.384770 2.055981 90.55925 8 0.341041 9.234812 1.757013 89.00818 9 0.372147 10.31930 1.505058 88.17565 10 0.407493 11.59256 1.276800 87.13064
When we use the Eviews to do the variance decomposition, the results will be divided into ten periods automatically. This division of time period can help us to better analyze the furcating effect between different variables. Results of Table 7 show that in the first period 71.47% of the variance of economic growth is explained by its own innovations while higher education only accounts for about 24.78%. In the second period, 61.15% explained by own innovation and 33.31% by higher education, which increased almost 10 percent from period 1. Then, the impact of higher education on economic growth begins to experience a stable increase from 42.09% in the third period to 59.04% in the tenth period whereas the change of GDP explained by own innovation decreased from 51.29% to 32.22% over the same period. Thus, higher education has a long-run positive effect on economic growth. Furthermore, the impact of physical capital investment made up only 3.75% in the first period. Despite the shock of physical capital investment rose to a higher value (8.95%) in the fourth period, it fluctuated at an average level of 10% within a limited range. We can conclude that physical capital investment has a slight shock on economic growth over ten periods. Higher education, however, had much more significant influence on economic growth. All these confirmed that the higher education is a major impetus to accelerate economic growth in China.
capital investment is predominantly explained by higher education, maintaining about 60% during ten periods. So these results reaffirm that higher education had a long-term positive effect on physical capital investments.
According to the variance decomposition of lnht in Table 9, higher education
suffers a shock from its own innovation by 100, 93.6, 89.43 and 87.13% in period 1, 2, 6 and 10, respectively. Higher education suffers a shock from economic growth by 0, 4.53, 8.48 and 11.59% in period 1, 2, 6 and 10, respectively. These findings suggest that economic growth has an unapparent effect on higher education, meaning that the bidirectional causal relationship between higher education and economic growth does not exist in China.
5. Conclusions and Recommendations
This paper provides an investigation of the relationship between higher education and economic growth in China by using the time series data from 1978 to 2012. Firstly, this paper presents a review of empirical literatures that have identified the relationship between economic growth and higher education. Empirical studies show that higher education can directly promote economic as well as indirectly stimulate the economic through increasing the utilization of physical capital investment. Secondly, in order to examine the relationship between higher education and economic growth in China during the period 1978-2012, this paper uses the VAR model to estimate a variant augmented neoclassical growth equation and employs the higher education enrolment rate as the proxy for human capital. The empirical econometric analysis shows that GDP per capita, physical capital investments and higher education are cointegrated when GDP per capita is the dependent variable. The Johansen test identifies that the higher education has long-term positive effect on higher education in China, 1% increase in higher education enrolment rates will foster economic growth by about 0.138%. The Granger causality test, however, shows that there is no Granger causal relationship between higher education and economic growth, meaning that higher education does not have any effect on economic growth. But there is a directional Granger causality from higher education to physical capital investment, which confirms the evidence that higher education enhance the utility of human capital to utilize the physical capital investments in China. Finally, we employ the variance decomposition to further gauge the correlation between higher education and economic growth. The variance decomposition supports the result of cointegration test that higher education is a major impetus for economic growth and the result of Granger causality that higher education is important for utilization of physical investment. On the other hand, the results of variance decomposition show that the effect of economic growth on higher education is very subtle in China. Our main conclusion is that the role of higher education on promoting economic growth seems to be very crucial in China during the period 1978-2012.
We note that our findings of the positive effect of higher education on economic growth is likely to be biased because the growth rate of labor, which is stationary in levels, is excluded from our econometric analysis. In order to better estimate relationship between higher education and economic growth, it is recommended that other variables such as higher education investment should be included in the growth equation. Our future research will consider investigating the dynamic correlation between higher education investment and higher education.
Implications for policy can also be drawn from our results. As the insufficient higher education investment and emigration of higher education human capital, the shortage of higher education has become a challenge for the economic growth of China. Hence, it is recommended that Chinese policy makers should pay attention to higher education human resources and make policies increase the higher education investment to build and attract outstanding university faculty. Our results recommend Chinese higher education institutions to take actions to prevent this massive ‘brain strain’. In addition, they should make a great effort to attract these emigrant experts back home to help them to build excellent higher education institutions. In this way, the enrolment rates of higher education will increase, meaning that economic growth will be further boosted. To sum up, in the development of national income of China, the value of higher education should be recognized.
Bibliography
Agiomirgianakis, G., Asteriou, D., & Monastiriotis, V. (2002). Human capital and economic growth revisited: A dynamic panel data study. International Advances
in Economic Research, 8(3), 177-187.
Anwar, S., Shabir, G., & Hussain, Z. (2011). Relationship between financial sector development and sustainable economic development: time series analysis from Pakistan. International Journal of Economics and Finance, 3(1), 262-270.
Appiah, E. N., & McMahon, W. (2002). The social outcomes of education and feedbacks on growth in Africa. Journal of Development Studies, 38(4), 27-68. Barnouin, B., & Yu, C. (1994). Ten years of turbulence: the Chinese cultural
revolution. International Affairs, 70(1), 159.
Barro, R., & Lee, J. W. (1994). Sources of economic growth. Carnegie-Rochester
Conference Series on Public Policy, 40, 1-46.
Barro, R. J. (1997). Determinants of economic growth: A cross-country study (Cambridge, MA: MIT Press).
Barro, R. J. (1999). Human capital and growth in cross-country regressions. Swedish
Economic Policy Review, 6(2), 237-277.
Benhabib, J., & Spiegal, M. (1994). The role of human capital in economic development: evidence from aggregate cross-country data. Journal of Monetary
Economics, 34, 143-173.
Berthelemy, J., Pissarides, C., & Varoudakis, A. (2000). Human capital and growth: the cost of rent socking activities, in M.S Oosterbam, T.R. van Steveninck, and N. van der Windt (eds.). Economic Growth and its Determinants (Dordrecht: Kluwer Academic Press).
Burja, C., & Burja, V. (2013). Education’s contribution to sustainable economic growth in Romania. Procedia – Social and Behavioral Science, 81, 147-151. Cai, Z. (1999). An econometric analysis of the contribution rate of education to
economic growth. Economic Research, 2, 39-48.
Chen, B., & Feng, Y. (2000). Determinants of economic growth in China: Private enterprise, education, and openness. China Economic Review, 11(1), 1-15.
Cui, Y. (2001). On the contribution of China’s education to the economic growth rate.
Journal of Beijing Normal Universitiy (Social Science), (1), 76-81.
European Commission. (2011). Recent developments in European higher education systems. Commission Staff Working Document SEC, 1063 final.
Felsenstein, J. (1981). Evolutionary Trees from DNA Sequences: A maximum likihood approach. Journal of Molecular Ecolution, 17, 368-376.
Geng, C., Li, N., & Cai, T. (2009). Quantitative analysis on the contribution of Jiangsu’s education investment to its economic growth. Grey Systems and Intelligent Services, 2009, 1003-1007.
Gradstein, M., & Justman, M. (2002). Education social cohesion, and economic growth. America Economic Review, 92(4), 1192-1204.
Granger, C. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica, 37(3), 424-438.
Granger, C. (1980). Testing for causality: a personal viewpoint. Journal of Economic
Dynamics and Control, 2, 329-352.
Greiner, A., & Semmler, W. (2002). Externalities of investment, education and economic growth, Economic Modeling, 10(5), 709-724.
Gyimah-Brempong, K., Paddison, O., & Mitiku, W. (2006). Higher education and economic growth in Africa. Journal of Development Studies, 42(3), 509-529. Hall, R. E. and Jones, C. (1999). Why do some countries produce so much more
output per worker than others?. Quarterly Journal of Economics, 8(1), 52-86. Herrerias, M. (2010). The causal relationship between equipment investment and
infrastructures on economic growth in China. Frontiers of Economics in China, 5(4), 509-526.
Hye, Q. A. M. (2012). Long term effect of trade openness on economic growth in case of Pakistan. Quality & Quantity, 46(4), 1137-1149.
Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic
Dynamic and Control, 12, 231-254.
Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration witj applications to the demand for money. Oxford Bulletin of
Economics and Statistics, 52, 169-210.
Kimenyi, M. S. (2011). Contribution of higher education to Economic Development: A survey of international evidence. Journal of African Economies, 20, III14-III49.
Litterman, R. B., & Weiss, L. M. (1985). Money, real interests, and output: a reinterpretation of postwar U.S. data. Econometrica, 53, 129-156.
Lorde, T., Waithe, K., & Francis, B. (2010). The importance of electrical energy for economic growth in Barbados. Energy Economics, 32(6), 1411-1420.
Luca, R. (1988). On the mechanics of economic development. Journal of Monetary
Ecoomics, 22, 3-42.
Mankiw, G., Romer, D., & Weil, D. (1992). A contribution to the empirics of economic growth. Quarterly Journal of Economics, 107(2), 407-437.
Masih, A. M. M., & Maish, R. (1997). On the temporal causal relationship between energy consumption, real income and prices: some new evidence from Asian-energy dependent NICs based on a multivariate cointegration/vector error correction approach. Journal of Policy Modeling, 19(4), 417-440.
McMahon, W. W. (1987). The relation of education and R&D to productivity growth in the developing countries of Africa, Economics of Education Review, 6(2), 183-194.
McMahon, W. W. (1998). Education and growth in East Asia. Economics of Education Review, 17(2), 159-172.
Nelson, R., & Phelps, E. (1966). Investment in humans, technological diffusion and economic growth. American Economic Review, 61, 69-75.
Newey, W. K., & West, K. D. (1994) Automatic lag selection in covariance matrix estimation. Review of Economic Studies, 61(4), 631-654.
Pegkas, P., & Tsamadias, C. (2014). Does higher education affect economic growth? The case of Greece. International Economic Journal, 2014, 1-20.
Pereira, J., & Aubyn, M. (2009). What level of education matters most for growth?: Evidence from Portugal. Economics of Education Review, 28(1), 67-73.
comparative analysis. Economics of Education Review, 21(5), 513-521.
Philips, P. C., & Perron, P. (1988). Testing for a unit root in time series regression. Biometrika, 57(2), 335-346.
Pissarides, C. A. (2000). Human Capital and Economic Growth: a synthesis report. OECD Development Centre Technical Papers No.168, Paris: OECD.
Psacharopoulos, G (1995). The profitability of investment in education: Concepts and methods. Human capital development and operations policy working paper No.76. Washington, DC: World Bank.
Raza, S. A., & Jawaid, S. T. (2013). Terrorism and Tourism: a conjunction and ramification in Pakistan. Economic Moedlling, 33, 65-70.
Romer, P. (1986). Increasing returns and long-run growth. Journal of political
Economy, 94(5), 1002-1037.
Romer, P. (1990). Endogenous technological change. Journal of Political Economy, 98(5), part 2, 71-102.
Sala-I-Martin, X. (1997). I just ran two million regressions. American Economic
Review, 87(2), 178-183.
Schultz, T. (1961). Investment in human capital. American Economic Review, 51(1), 1-17.
Schwarz, G. E. (1978). Estimating the dimension of a model. Annuals of Statistics, 6(2), 461-464.
Sims CA. (1980). Macroeconomics and reality. Econometrica, 48(1), 1-48.
Solow, R. (1956). A contribution to the theory of economic growth. Quarterly Journal
of Economics, 70(1), 65-94.
Solow, R. (1957). Technical change and the aggregate production function. Review of
Economics and Statistics, 39(3), 312-320.
Soytas, U., & Sari, R. (2009). Energy consumption, economic growth, and carbon emissions: Challenges faced by an EU candidate member. Ecological Economics, 68(6), 1667-1675.
Voon, J. P. (2001). Measuring social returns to higher education investment in Hong Kong: Production function approach. Economics of Education Review, 20(5),
503-510.
Wiener, N. (1956). The theory of prediction. Modern Mathmatics for Engineers, 1, 125-139.
Yu, H., Zhao, S., Xu, X., & Wang, Y. (2014) An empirical study on the dynamic relationship between higher educational investment an economic growth using VAR model. Systems Research and Behavioral Science, 31, 461-470.
