- 1 - / 40
Factors influencing tertiary educational demand: an
empirical study across developing countries
Bin Zu S2152991
Master Thesis for International Economics & Business (2013/2014) Final version, January 2014
Abstract
This paper applies panel regression analysis to discuss the determinants of tertiary educational demand across developing countries. The results indicate that per capita income has a most significantly positive effect on tertiary educational demand (measured by GDP per capita); the poverty ratio and income inequality show a significantly negative influence on tertiary educational demand, which are measured by poverty ratio and GINI index. Furthermore, the unemployment ratio has no significant impact on tertiary educational demand.
Key words: tertiary educational demand, per capita income, poverty ratio, income inequality, unemployment ratio
- 2 - / 40 Table of Context
1.Introduction ... - 3 -
2. Theoretical background and Hypotheses ... - 5 -
2.1 What is tertiary educational demand? ... - 5 -
2.2 Why we choose developing countries as samples? ... - 7 -
2.3 What are the economic factors that influencing tertiary educational demand and how to measure them? ... - 8 -
2.3.1 What is per capita income and how to measure it? ... - 9 -
2.3.2 What is income inequality and how to measure it? ... - 10 -
2.3.3 What is poverty ratio and how to measure it? ... - 11 -
2.3.4 what is unemployment ratio and how to measure it? ... - 12 -
2.3.5 The econometric models ... - 13 -
3.Model and data analysis ... - 14 -
3.1 Sample ... - 14 -
3.2 Dependent Variables ... - 15 -
3.3 Independent variables ... - 15 -
3.4 The econometric model ... - 17 -
- 3 - / 40 1、Introduction
Plenty of researchers have identified the significant impact of education on economic growth (eg., Jorgenson, 1992;Krueger &Maleckova ,2003;Barro &Lee, 2012). "Education‖, as stated by Harbison et al (1965), "is both the seed and the flower of economic development." It is hardly to deny the crucial role of education from the income increasing in cross-country data over a long time (Krueger &Lindahl, 2000). Enabling more people to accept tertiary education is crucial to all the countries, because it has an influence on many social aspects like child mortality, child education level, life expectancy, income inequality etc (Barro &Lee, 1994; de Gregorio &Lee, 2002; Breierova &Duflo, 2004; Cutler et al., 2006). Barro (1991) concluded that one possible way for less developed and developing countries (developing countries) to catch up with developed countries is to improve the their human capital, which asks for a higher demand of tertiary education. Psacharopoulos and Patrionos (2004) have found the return rates of human capital in developing countries are higher than those in physical capital. ―The highest rates were found in the poorest countries.‖
- 4 - / 40 human capital stock) varies among those countries, showed in Appendix 2, it is clearly to conclude that a high labor demand for tertiary educational labor (above than 80% of the employment rate), while the other education level manpower hit a lower employment rate especially in economic crisis.
From above we can find an inharmonious social phenomenon that a high desire of tertiary level educational work force and a relatively lower higher educational attainment rate simultaneous exist. What might be the underneath factors result in this phenomenon? What is the relationship between them? What might be the next step for the government to ameliorate these issues? In this paper, we start from investigating the potential country factors that may affect the tertiary education demand across developing countries and discuss results. For this purpose, our problem statement is as follows:
'What economic factors might affect the tertiary education demand (TED) across developing countries and what are the impacts of them on TED?'
Research questions:
1. What is tertiary educational demand and how to measure it?
2. What are the economic factors that influencing tertiary educational demand and how to measure them?
3. Why we choose developing countries as samples?
4. What is the impact of GDP per capita on tertiary educational demand? 5. What is the impact of poverty ratio on tertiary educational demand？ 6. What is the impact of income inequality on tertiary educational demand?
7. What is the impact of tertiary unemployment ratio on tertiary educational demand?
8. Among those four factors, which one is the most significant factor to affect the tertiary educational demand? And which one is the least important one?
- 5 - / 40 discussed about the other way around, how the other country development indicators may affect the education demand. Second, we only focus on the tertiary education level of discussing general education situation, because nowadays high-skilled workers demand is continuously increasing by rapid technological change, more educated to develop low-skilled workers’ working skill and reading and writing abilities would drive more productive (Education at a glance ,2013; Szirmai,2005). Third, samples are only collected from developing countries, which would cause less bias than those samples are coming from all countries.
The reminder of this paper is organized as follows. In section 2, we discuss the hypothesis of this paper and literature review. In section 3, we illustrate the empirical economic model and data analysis. The descriptive and estimated results for each of the econometric models are in section 4 and conclusions and limitations in the section 5 and section 6 respectively.
2. Theoretical background and Hypotheses
2.1 What is tertiary educational demand?
Before stepping into investigating this research question, an accurate definition of "tertiary education" is necessary. For the level of education, we normally follow the "International Standard Classification of Education" (ISCED) which was designed in 1970 and used as an instrument for illustrating and comparing statistic of education across countries (OECD,2013). Broadly speaking, there are three levels of education: primary, secondary and tertiary. The details of the definition of each level could be found in Appendix 3. There are two types of tertiary education which are Tertiary-type-A education and Tertiary-type-B education.
- 6 - / 40 labor market. In our case, we suggest that tertiary education demand means the population aspiration of tertiary-type A education and tertiary-Type B education.
Higher education will bring more economic improvement and non-economic benefits (Schuller et al,2001). Firstly, as mentioned before, higher education would affect many social aspects, like child mortality, child education level, life expectancy etc (Barro &Lee, 1994;de Gregorio & Lee,2002; Breierova & Duflo,2004). Secondly, higher-education-level human capital is more popular when people hunting for jobs because of their more productive (Szirmai,2005). Normally, an additional year of education may bring 5% to 15% extra salary to people (Krueger &Lindahl, 1999). Due to that, appropriate for improving the demand of tertiary education level is a necessary and crucial way to a country development. In this paper, we try to figure out the factors may affect a country tertiary education demand.
- 7 - / 40 inequality and unemployment ratio) of a country to tertiary education demand, the number of tertiary school enrollment ratio could better match the demand of data-how the development indicators of a country can affect the personal desire to get a higher education. Due to that, in this paper we use the tertiary school enrollment ratio across countries which also used frequently by researchers theses years mentioned before to quantify the relationship between education and country development indicators.
2.2 Why we choose developing countries as samples?
For academic research purpose, World Economic Situation and Prospects (WESP) organization classifies all countries into three generally categories: developed countries, economic in transition countries and developing countries. We called the economic transition countries and developing countries in this paper as developing countries. Generally speaking, the developed countries or developing countries are measured by many development indicators such as income per capita (GDP per capita), life expectancy etc. In this paper, we refer the countries classification according to the least United Nations definition which also accepted by many other institutions, like the world bank, they are: 1) per capita income (GDP); 2) industrialization and 3) human development index (United Nation,2013). Based on that, we definite the developing countries means countries which have a lower gross national income (GNI), weak industrialization and a relatively low human development index.
- 8 - / 40 because of high GDP per capita and human development resources, even a large income inequality within developed countries exists; the relatively poor people also can afford the education investment. Due to the development indicators we choose, the developing countries are more suitable for our research topic.
2.3 What are the economic factors that influencing tertiary educational demand and how to measure them?
Before we investigating the factors might affect TED, one reasonable theoretical background should put into contact with our outcomes. There are two approaches to find such relationships. One approach treats the decision to enroll to a TED as an investment way, while the others consider this way as a consumption way (Campbell &Siegel,1967). Both of them can be a reasonable theoretical background for this paper. As an investment way, people treats the enrollment to TED as an investment behavior, there are many factors would affect the personal decision ,like personal income, country status, economic level, money cost, investment return etc (Campbell &Siegel,1967). One of the significant factor is the country development level (Gylfason,2001) which is also the main explanatory variable in this paper. According to the book of Economics of development (G.R. Arabsheiban,2011), there are four main indicators as dimensions can represent for a country’s economic development: per capita income, poverty ratio, income inequality and unemployment ratio. Bergh et.al (2006) also ever indicated per capita income; income inequality and unemployment ratio would affect the higher school enrollment. What is a difference, they compare with richer countries with less developed countries by OLS test, but in our paper, we focus on these four elements in relatively poorer countries (developing countries) by panel data to investigate the different influence level of them to TED, which one is most related and which one is the least important one for increasing the educational demand.
- 9 - / 40 education. So the countries development level would be important determinants of those countries tertiary educational demand. As a result, it is meaningful to find out the impact of country development level on TED through the four indicators of country development. In order to figure out the influence factors of TED in the end, some unit of analysis should be discussed firstly. Due to that, this part will interpret the conception and theoretical background of TED and some most related factors mentioned before in our research.
2.3.1 What is per capita income and how to measure it?
Normally when we discuss about educational demand, family income and per capita income are widely used as the explanatory variables to education demand in theoretical and empirical analysis (Barro, 1991;Davis-Kean,2005). Barro (1991) claimed that the educational demand is significant positive related with per capita income and Davis-Kean also claims that the family income also constrained the educational demand. Even there is more information about family income than that of personal income, but there are some problems when we discuss the impact of family income to TED here. The problems include: 1) It is hard to know the exactly proportion of family income to each members within the family. 2) If we use the family income as an explanatory, we should consider about the size and different structures of each family (Lazear &Michael,1981). Due to that, we choose the per capita income in our paper.
- 10 - / 40 Machin,1999,2000). Because relatively low per capita income will limit the expenditure on tertiary education (Blondal, 2002), there is a huge gap of GDP per capita between developed countries and economically developing countries around the world, showed in Appendix 5. From the Appendix 5 we can conclude that income level among different countries are different, the average income per capita in Sub-Saharan Africa (developing only) is only 1440 USD while in Europe the personal income almost six times as high as Sub-Saharan Africa (6821 USD). Based on the relationship mentioned before, different GDP per capita among countries, the tertiary education demand should change through the income level. We suggest that higher GDP per capita will bring higher demand for tertiary education.
Hypothesis 1: Higher GDP per capita leads to a higher tertiary education demand.
2.3.2 What is income inequality and how to measure it?
Besides a large per capita income gap among countries, within a country, there may still exist a huge income inequality, and theses years this gap is increasing (eg.,Park,1996;ning,2010). Sylwester(2000) found that a higher level of income inequality will lead to more individual investment on public education. Not all the level of income will constrain the education demand through capital investment which is one of reasons why we choose the developing countries as our samples. Disability to afford the tuition fees will constrain the tertiary educational demand in many developing countries. The rich enough families tend to pay as high money as super rich families which will lead to an increasing of tuition fees, especially for tertiary education, because tertiary education always brings higher rate of return as a result-higher employment rates and higher earnings premium(OECD at a glance, 2013). Bergh (2006) claimed that GINI coefficient has a negative influence to tertiary educational demand in a richer country.
- 11 - / 40 expressed the ratio of population divided by the ratio of income received (Deininger,1996). The Gini coefficient is a number between 0 and 1, where 0 corresponds with perfect equality (where everyone has the same income) and 1 corresponds with perfect inequality (where one person has all the income—and everyone else has zero income).Based on the above, we can make a hypothesis like below:
Hypothesis 2: Higher income disparity leads to a lower tertiary education demand .
2.3.3 What is poverty ratio and how to measure it?
In many developing countries, stock of tertiary school enrollment is always affected by poverty, which means lower income than the poverty line in one country also related to limited credit payment in education (Filmer,1999). Due to the limited credit payment, less capital investment in education community resources is a factor of less quality of education. In the end, poor education quality often leads to a failure in educating country requirement skills and lower enrollment (Brown &Park, 2002).
Before stepping into investigating our research question, an accurate definition of "poverty line" is necessary. There are five dimensions in poverty line in theory according to the researcher by chambers (1988):
'Poverty proper' being a lack of adequate income or assets to generate income;
Physical weakness due to under-nutrition, sickness or disability;
Physical or social isolation due to peripheral location, lack of access to goods and
services, ignorance or illiteracy;
Vulnerability to crisis and the risk of becoming even poorer;
Powerlessness within existing social, economic, political and cultural structures
- 12 - / 40 not significant.
Hypothesis 3: Higher poverty ratio leads to a lower tertiary education demand
2.3.4 what is unemployment ratio and how to measure it?
Plenty researchers have proved that unemployment ratio has a significant influence to TED (Brunello,2001; Kettunen, 1997). As mentioned above, tertiary school enrollment for many people is a way of investment. A low unemployment ratio of tertiary education would encourage more and more people to choose the further education. Because a high return of tertiary education investment is the goal of most of the educational people. Before the early years, Ashenfelter (1979) found that unemployment constrained the choice of work and according to that, additional schooling would reduce such a kind of constrain. Mincer (1991) ever claimed much lower risk of unemployment at TED level, and that is a major advantage of TED, which also can be treated as one way of high return. As for educated people, there are at least three attractive reasons to invest in education according to Mincer: 1) higher wages; 2) greater upward mobility to income and occupation; 3) less risk of unemployment (Mincer,1991). To conclude, unemployment ratio has a negative impact on TED.
In this case, we use tertiary unemployment ratio as a proxy for a country’s unemployment situation matched with TED, which is also widely used by other researchers. According to the definition in the World bank, the tertiary unemployment ratio shows the unemployed by the tertiary educational schooling, as a percentage of the unemployed. The levels of educational school enrollment accord with the International Standard Classification of Education 1997 of the United Nations Educational, cultural and Scientific Organization (UNESCO), which also used in the dependent variable (the World Bank, 2013).
- 13 - / 40 2.3.5 The econometric models
In this paper, we mainly aim to examine the relationship between tertiary education demand and country development level, in specific, whether the per capita income, poverty ratio, income inequality and unemployment ratio of tertiary education will affect the tertiary education demand according to the four hypothesis above. The equation (1) are the completed model using a logarithm because this estimates are therefore more easily interpreted and less sensitive to outliers:
Main Econometric equation :
t it LGDC LUER LPOR LGINI LSERit 1 it2 it 3 it4 (15)
In order to test the relationship between each variable and tertiary educational demand, we also divide all the variables mentioned above to separate regressions.
In the regression model (15), the explanatory variable, represents a proxy for the tertiary school enrollment in country i at the observed time t, defined as the percentage number of people who the highest grade school enrollment level is tertiary education (OECD,2013); The independent variable in equation (1). GDP per capita (GPCit) denotes the gross domestic
product divided by midyear population in country i at time t (OECD,2013).The Gini coefficient (GINIit) is represented by the area between the Lorenz curve and the hypothetical
line of absolute equality, expressed as a percentage of the maximum area under the line during 1985 to 2010 (OECD, 2013). Besides that, the poverty ratio (PORit) is collected by
- 14 - / 40 3. Model and data analysis
3.1 Sample
My data sources are mainly collected from the World Bank data, OECD and CIA. This paper tends to involve a panel dataset of all the developing countries. However, because most of the country data are counted since 1985 and stopped after 2010, so we dropped all the other years before 1985 and 2011,2012 and 2013. Missing data is the most difficult part of collecting sample data, especially for Gini coefficient and poverty ratio. Except the World Bank data, we also collected complemented with data from OECD and CIA. Because some countries are missing most of the Gini coefficient and poverty ratio data from 1985 to 2010, we can only pick the countries with nearly full data of theses four variables to the greatest extent. In the end, there are 56 developing countries in our sample and according to The World Bank’s classification; they can be sorted into 6 regions which are shown below in table 3. The details of the classification could be seen in Appendix 6.
In World Bank Development Data, The tertiary education demand in my model is measured as a ratio of tertiary school enrollment across countries. And also from the same data basement, the independent variable of per capita income in a macro way that we choose the GDP per capita (Lazear, 1981;Gylfason,2001) and calculate the income inequality by using the GINI coefficient method. For developing the poverty effect, we consider a measure of poverty ratio as a proxy of poverty influence and select the proportion of people who is lack of adequate income or assets to generate income according to the researcher chambers (1988). The tertiary unemployment ratio which represented the unemployed by the tertiary educational schooling is based on the World Bank database. (World Bank data, 2013).
- 15 - / 40 3.2 Dependent Variables
The dependent variable in the model we chose the tertiary school enrollment across countries. Various indicators are used to measure different aspects of education across countries. The tertiary school enrollment ratio is chosen as a proxy for the tertiary education demand in this paper.
Tertiary educational enrollment ratio (SERit): the percentage number of people who the
highest grade enrollment level is tertiary education in country i at time t, computed by dividing the total number of all levels of education people among a country at the same time of a same age group. The formula is as follow:
SEAit=SENit/Popit
SERit, the dependent variable, is the number of people who enrolled the highest tertiary
education level in one country at the observed time t. SENit istotal enrollment in tertiary
education. Popit represents the total population in one country at time t, measured in millions.
The complete tertiary educational enrollment ratio data are available in the World Bank, represented in English.
3.3 Independent variables
In this paper, we mainly discuss the relationship between tertiary educational demand and country development situation. From the model, we choose to represent the independent variables with some most related country development indicators. They are per capita income, poverty, and income inequality and unemployment ratio.
- 16 - / 40 we use the Gini coefficient, according to the United Nations (UN), theWorld Bank, the US Central Intelligence Agency (CIA), and the Organisation for Economic Co-operation and Development(OECD), to define the income inequality indicator within a country. The Gini coefficient is a number between 0 and 100, where 0 corresponds with perfect equality (where everyone has the same income) and 100 corresponds with perfect inequality (where one person has all the income—and everyone else has zero income). Poverty in this paper we use the poverty ratio which means a rate of people who is lack of adequate income or assets to sustain a family in terms of food, housing, cloths, education and so on during time t (OECD,2013). Normally, the poverty ratio is the population who is lower than poverty line in one country which is divided by total population at time t.The data is sourcing from the World Bank (1985-2010).We use the tertiary educational level of unemployment ratio on refers to the unemployment ratio. The data are sourced from word bank development department during 1985 to 2010.
GDP per capita (GPCit) denotes the gross domestic product divided by midyear population
in country i at time t. GDPit collected the sum of gross value added by residents in one
country among all producers in the economy plus any product taxes and minus any subsidies not included in the value of the products (word bank data,2013). The formula is as follow:
it it it GDP Pop
GPC /
Gini coefficient (GINIit) devotes the degree of inequality of family income within countries.
- 17 - / 40 Poverty ratio (PORit) devotes the percentage of people whose income is lower than poverty
line, which divided by total population within a country i at time t. PVPit represented the total
poverty people in one country, and POPit is the total number of people in one country at time
t.
it it
it PVP POP
POR /
The data are taken from the World Bank data (1985-2010).
Tertiary unemployment ratio (UERit) Based on the unemployment ratio definition used in
Mincer (1991), unemployment rate could be defined as the fraction of time lost by all members of labor force who enter a tertiary education within a unit period, the formula is as follows:
UERit=UETit/LFRit
the LFRit is the number of workers in the labor force with tertiary educational background, UETit devotes the number of workers who experienced unemployment with tertiary educational background.
The data are taken from the World Bank (1985-2010). The data description is in Appendix 7.
3.4 The econometric model
Normally, there are many types of data, like cross-section data, time series data and panel data (Adkins,2008). In panel data, observations of the same cross-sectional unit are observed over time, and if each individual variable has the same observations, we call it balanced panel data, if each individual variable has the different observations, we name it unbalanced panel data (Wooldridge,2002). One of the advantages of panel data is allowed to control for variables you cannot observe or measure like culture, politics across countries (Hsiao,2003). In this case, we observe the educational demand influence factors across data during years, panel data will be the best choice.
- 18 - / 40 influence to tertiary school enrollment in one country at time t.
Before us running the regression to get a result, some diagnostic checks would be done to get better results. With the response to the panel data in the research setting regression equations, we can estimate it through random or fixed effects which are test by Hausman test. A Jarque-Bera test used here to test the normality of variables by checking values of skewness and Kurtosis (Hill et al, 2007), generally, a normal distribution the Kurtosis value is around 3. If the test results is larger than 3, we would choose logarithm one to better estimate the relationship among them. Endogeneity is a problem that occurs when one of the independent variable is correlated with the error term in a regression model. One main source that arise endogeneity is simultaneity, which also refers to reverse causality. We also use Hausman test to check whether endogeneity exists in our model and use IV regression to avoid it.
3.5 Descriptive statistics
Before we start the data analysis, outliers should be first to recognized. There are many ways to inspect the outliers. In this case, 1% and 99% percentile data are treated as the outliers by stata summarize. An outlier is an observation that outside of the main patten, which will affect the whole result and not fit the model, it is better to remove them first, and then we would do two tests with or without the outliers and to compare whether the outliers here would affect the final results (in section 5).
Table 1-Descriptive Statistics
Variable Obs Mean Std. Dev. Min Max ser 1078 27.00508 18.23228 .9767 78.1285 gini 631 42.75472 10.09664 22.76 63.1 por 595 12.58261 15.50347 .04 68.51
uer 545 16.37537 12.87652 .3 73.3 gdc 1373 2515.916 2477.82 144.149 14053.8 Source: Author’s own calculation
- 19 - / 40 final observations among the dependent and independent variables are quite different, there is a large amount of data of gdc (1373) while the least one uer is only 545. Because such a big numbers of sample difference, sampling bias may exists. Sampling bias would potentially lead to an over- or underrepresentation of the corresponding parameter. However, it is hard to perfectly avoid the sampling bias in a realist test, which would be a limitation. In our case, all data are picked up randomly, even in the smaller groups- gini and por (the World Bank, 2013), which means although the some samples are missing, but the results also can generally reflect relationships among those items.
To be more specific, the average level of ser is only 27.00508%, the largest ratio of school enrollment showed by Venezuel in 2009 is 78.1285%. Compared with other countries, this is a relatively high ratio. What is more, from 1985 to 2010, the tertiary school enrollment ratios in Venezuel are constantly increasing and until 2008, it reached 78%, which means 78% of population who graduated from secondary educational level would choose a further education. Within the 649 observations of gini index, there is a relatively shorter gap between them, because the completely inequality income country showed by 100 in gini index and the completely equality income country with 0. The most income equality country is Belarus in 1988 which the gini index is only 22.76%, while the least income equality country is 63.1% in South Africa (2011). Belarus has a lowest income inequality in 1988, but it increases years until now, that is may mainly resulted by income policies in Belarus: avoided mass
privatization and old-style Soviet social security (Yemelyanau,2009). Ukraine (1992) further outperforms other countries with the lowest percentage of poverty among countries (0.03%), whereas the Zambia (2006) showed the weakest poverty ratio around the world with 68.51%. But about the poverty ratio, in our paper we choose the population below $1.25 per day as a poverty line, that is reasonable for some countries has a little people who live below poverty ratio but in some poorest country like Zambia with a number reached 68.51%. Due to that, we can conclude that there is still a large gap of poverty situation all around the world, also the same as other variables presented in our case. For unemployment ratio of tertiary
- 20 - / 40 start from 2007, Thailand has a lowest unemployment ratio compared with other countries. The main reasons for that is because ―they are starting to get a lot of investment that used to go to China and would have gone to China‖ which is said by Haley.G.T (2012).
According to the various changing numbers of explanatory variables, we can expect that the difference status of economic development across countries and times, will have virtual influences to total tertiary educational demand, and in this paper, we try to figure out what is the relationship between them, and what may be the most influence factor and what is the least one through different combination of variables in regression. The details of the results will be showed below.
4. Diagnostic checks
In this section, many diagnostic checks are conducted step by step to check the relationship between dependent and independent variables.
Firstly, a Jarque-Bera test normally be used to test the normality of variables by checking values of skewness and Kurtosis (Hill et al, 2007). Due to the present context in book Hill et al, ―skewnessrefers to how symmetric the residuals are around zero. Perfectly symmetric residuals will have a skewness of zero.‖ And Kurtosis usually refers to the ―peakedness‖ of the distribution (Hill et al, 2007). Generally, a normal distribution the Kurtosis value is around 3. In our regressions, dependent and independent variables are larger than 3, a
probably way is to change the dependent variable into natural logarithm, at last the extremely high or low variable will be more fittable to a natural distribution.
- 21 - / 40 variables in our regression and that multi-collinearity exists (UCLA, 2010). From the Appendix 7, all the variables of VIF are lower than 5, which suggest the multi-collinearity does not occur in our regression by VIF.
In this model, one potential problem in the model is the existence of heteroskedasticity among all the variables in the model. White test is conducted in our model and results showed heteroskedasticity exist in all models. According to the nature of heteroskedasticity, there existed biased coefficient estimates of the variances and standard errors. Furthermore, the significant tests for the variables in the right-hand would be too high or too low, which will affect the empirical results of factors that may affect the tertiary educational demand. Due to that, we choose robust variance-covariance estimator (vce) to avoid that.
- 22 - / 40 5. Empirical results
5.1 Regressions
Based on the tests above, first we compare the main regression (14) with or without outliers by fixed panel data regression (one year time-lag) to show the difference, see table 2.
Table 2 : Results of main regression with or without outliers variables Equation(1):With outliers
FE
Equation(2): without outliers FE Lgini_1 .5947652*** (.21) .4209624***(.22) Lpor_1 -.0635936*** (.023) -.0598591*** (.022) Luer_1 .0338 (.028) -.0100555 (.027) Lgdc_1 .3617615*** (.044) .3524636*** (.041) Constant -1.646809* (.69) -.8122139 (.79) N 245 236 R-square 50.37 56.12 Prob > chi2 = 0.0000 panel variable: i (unbalanced) time variable: t, 1985 to 2010
Notes: Figures in parentheses are standard deviations. ***, **, and * indicate significance at 1%, 5%, and 10% statistical levels, respectively.
Source: collection from the World Bank data, CIA and OECD, time various from 1985 to 2010.
- 25 - / 40 (.03) (.04) (.04) Constant -3.228241*** (.58) .5796889** (.32) -.8122139 (.79) N 452 238 236 R-square 57.17 55.33 56.12 Prob > F 0.0000 0.0000 0.0000
Notes: Figures in parentheses are standard deviations. ***, **, and * indicate significance at 1%, 5%, and 10% statistical levels, respectively.
Source: collection from the World Bank data, CIA and OECD, time various from 1985 to 2010.
The correlations between each independent variables and tertiary educational demand are shown in model 1 to model 4, which illustrate influence of per capita income, poverty, unemployment ratio and income inequality to tertiary educational demand in each country. We have already known that there is endogeneity existing among those four variables so fixed instrument-variables regression (times lag) for panel data is chosen to solve this issue. In model 1 to model 4, only one variable has a relatively strong explanatory power to explain a relationship with high R-square, GDP per capita (uer) with 50.18% respectively in model 4. That is mainly because of a large sample base (1011).And the R-squires in model 1, 2 and 3 are 5.66%, 9.24% and 1.87%, quite low, which means any one of those three variables alone cannot be a strong explanatory to tertiary educational demand. 5.66% of variation in regression tertiary educational demand is explained by the including explanatory variable- GINI index in the model 1; 9.24% of variation in regression tertiary educational demand is explained by the poverty ratio in the model 2; only 1.87% of variation in regression tertiary educational demand is explained by the unemployment ratio in the model 3.
- 26 - / 40 According to the literature Bergh et.al (2006), they claimed that a lower income inequality a richer country has, the higher tertiary school enrollment it has. However, there are still some reasons why our results are different. One is that we only focus on developing countries but Bergh concluded it for richer countries, and second is that Bergh chose to omit the country difference by OLS test, and we try to deal with data by panel and handle the endogenity with time lag. Those two reasons may drive to a different result and the positive relationship only happens in developing countries. Furthermore, because of significant influence between TED and income inequality of one country, more and more countries’ governments tend to spend more for tertiary education to reduce the impact of income inequality on TED, which also would stimulate more to enter the tertiary education level (Gregorio &Lee, 2002).
LPOR is statistically significant at 1% level in model 2. Also, it is suggested a same relationship as mentioned in hypothesis 3, that poverty ratio has a significant negative influence to tertiary school enrollment. In another word, with an additional increase in poverty ratio, the tertiary school enrollment would decrease by 0.1314687% as other variables would keep constant.
The coefficient of LUER_1 is significant at 1% level in model 3, if all the other variables are constant, one more increase in unemployment ratio in one country, 0.818117% increase of tertiary school enrollment. It is unmatched with hypothesis 4 that a higher unemployment ratio on tertiary education leads to a lower demand on tertiary education. The reason why the unemployment ratio doesn’t show a negative coefficient to tertiary school enrollment might be that we didn’t include enough control variables to constrain the unemployment. Mincer (1991) showed the negative influence based on a higher wages, parents educational level etc. Those factors might affect the final result.
- 27 - / 40 demand would increase by 0.5737205%. It is supported by hypothesis 1.
It is for sure that LGDC has a significant influence to tertiary educational demand if we ignore other variables. For example, we can find that in all the models which LGDC variable included in, LGDC has a significant at 1% level with a largest coefficient, quite larger than other variables, and the other variables even has no significant influence, such as the LUER and LPOR in the model 14. There are two reasons for this: 1) LGDC is the most virtual variable which can determine the tertiary educational demand in those four variables. 2) It is biased because of the missing data. LUER and LPOR have a relatively less valuable data in this sample compared with LGDC, which is the LGDC can be countered more observable.
How about the relationship if we combined more variables into one regression? Which one is more important among those variables, which one is less virtual even or unimportant? How the coefficient changed? Based on those questions, we added one more variable into the regression and see what is the new relationships between them (model 5 to model 10). In model 11 to model 15, two or more variables until all variables are added into a same regression to check the new findings.
Firstly, from the model 5 to model 14, the most significant change is the R-square, from the average 1.87% explanatory power sharply increase to 56.12% (model 15). It means average 56.12% of variation in regression tertiary educational demand is explained by the including explanatory variable- GINI index, poverty ratio, unemployment ratio and GDP per capita in the model 15. Compared with the model 15, we can find that in the model 7 and model 13, there is a higher R-square even the explanatory variables are less than the model 14. It may because of the observation number (N) in the model 7 and 13 are much more (N=488/452) than that in the model 14 (N=236), and larger number can be more precise.
- 28 - / 40 enrollment would increase by 0.420964%. If LPOR dropped one unit, the tertiary school enrollment should increase 0.0598591%, while a 0.3524636% increase in tertiary school enrollment with 1 unit increase in LGDC. However, the LUER is not significant in this model when we added the other three variables together, which means compared with other three variables, the unemployment ratio has not significant impact to the tertiary school enrollment.
6．Conclusion
This paper is seeking to shed some light upon the factors that may affect th e tertiary educational demand in one country, especially considering the influence by country development. To be specific, this paper aims to determine whether the country development status would affect the tertiary educational demand measured by four country development indicators- per capita income, poverty, income inequality and unemployment ratio. Compared with former studies, we mainly focus on developing countries and explores how the development indicators may affect the tertiary educational demand, which in most of the case s, researchers are paid much more attention on the way around. Overall, the results show the presence of a positive link between GINI index, GDP per capita and unemployment ratio. To be more specific, income inequality, income per capita and unemployment ratio of tertiary educational background could stimulate t he tertiary educational demand, simply because more income means more possibility devote more energy and money into higher education, and income inequality would stimulate poorer people to improve their life standard through education. However, the unemployment ratio is insignificant when we add the other three indicators into the regression, which may be caused by poor data availability. The poverty level would also affect the tertiary educational demand in a negative way. It is obvious that lack of enough income could constrain people to pay more tuition fees in tertiary education.
7．Limitation
- 29 - / 40 educational demand and four development indicators and how the four variables could influence the final tertiary education decision. Nevertheless, there are still some limitations. First of all, there is a lot missing data for GINI index and unemployment ratio, which would cause sample bias and negatively affect the final results. This issue can be reduced through the time passing by, because more and - 32 -more authority institutes already noticed that those indicators are important and reported them every year. Secondly, in this paper we only consider the quantity of tertiary educational demand through tertiary school enrollment. However, we know that quality of tertiary education demand is also a virtual aspect; step further to this area would contribute more in future researches’. Thirdly, it is better to discuss more indicators that would affect the tertiary educational demand, not only limited in country development indicators, not only four indicators. Fourthly, endogeneity could also be a limitation of the results, though we already use the one-year leg to reduce endogeneity, but it still exists. Human capital is always an important determinant for a country’s development. The better we know how to improve the human capital, the better further of a country. How to improve this limitation could also be the future option in this area.
Reference
Anand, Sudhir, and R. S. M. Kanbur. 1993. "Inequality and Development: A Critique."
Journal of Development Economics 41(l):19-43.
Ammermüller, A., Kuckulenz, A., & Zwick, T. (2009). Aggregate unemployment decreases individual returns to education. Economics of Education Review,28(2), 217-226.
Ashenfelter, O., & Ham, J. (1979). Education, unemployment, and earnings.The Journal of Political Economy, S99-S116.
- 30 - / 40 Barro, R. J. (1991). Economic growth in a cross section of countries. The quarterly journal of economics, 106(2), 407-443.
Barro, R. and J.W. Lee. 1993, ―Sources of Economic Growth,‖ Carnegie Conference Series
on Public Policy, 40: 1.
Barro, R. J., & Lee, J. W. (2012). A new data set of educational attainment in the world, 1950–2010. Journal of Development Economics.
Bergh, Andreas, and Günther Fink. "Higher Education: Does Public Expenditure Increase Enrollment." Ratio Working Papers 84 (2006).
Baltagi, B., U. Blien (1998): The Germanwage curve - evidence from the IAB employment sample, Economics Letters, 61, 135-142.
Blondal, S., Field, S., & Girouard, N. (2002). Investment in human capital through upper-secondary and tertiary education. OECD Economic Studies,34(1), 52-111.
Breierova, L. and E. Duflo. (2004).―The Impact ofEducation on Fertility and Child Mortality: Do Fathers Really Matter Less than Mothers?‖ NBER Working Paper No. 10513.
Brown, P. H., & Park, A. (2002). Education and poverty in rural China.Economics of
Education Review, 21(6), 523-541.
Brunello, G., & Medio, A. (2001). An explanation of international differences in education and workplace training. European economic review, 45(2), 307-322.
- 31 - / 40 Chambers, R. 1988. Poverty inIndia: Concepts, Research and Reality. Discussion Paper 241, Institute of Development Studies, University of Sussex.
Cutler, D., A. Deaton, A. Lleras-Muney. (2006). ―The Determinants of Mortality,‖ The
Journal of Economic Perspectives, 20(3): 97-120.
Davis-Kean, P. E. (2005). The influence of parent education and family income on child achievement: the indirect role of parental expectations and the home environment. Journal of Family Psychology, 19(2), 294.
Deininger, K., & Squire, L. (1996). A new data set measuring income inequality.The World Bank Economic Review, 10(3), 565-591.
Education at a Glance, OECD, Paris, 2002, Glossary
Filmer, D. (1999). The structure of social disparities in education: Gender and wealth. World
Bank, Development Research Group, Poverty and Human Resources.
Glewwe, P., & Jacoby, H. G. (2004). Economic growth and the demand for education: is there a wealth effect?. Journal of Development Economics, 74(1), 33-51.
G.R.Arabsheibani (2011). Economics of development. The London school of economics and political science,EC3044, 2790044
Gregorio, J. D., & Lee, J. W. (2002). Education and income inequality: new evidence from cross‐country data. Review of Income and Wealth, 48(3), 395-416.
- 32 - / 40 Harbison, F. H., & Myers, C. A. (1965). Manpower and Education: country studies in economic development. New York: McGraw-Hill.
Haley, G. T., Haley, U. C., & Tan, C. T. (2012). New Asian emperors: The business strategies of the Overseas Chinese. John Wiley & Sons.
Hill. R.C., Griffiths, W.E. and Lim G.C (2007). Principles of Econometrics, 3rd Edition, New York: John Wiley & Sons, Inc
Hsiao, C. (2007). Panel data analysis—advantages and challenges. Test, 16(1), 1-22.
Jorgenson, D. W., & Fraumeni, B. M. (1992). Investment in education and US economic growth. The Scandinavian Journal of Economics, S51-S70.
Krueger, A. B., & Lindahl, M. (2000). Education for growth: why and for whom?(No. w7591). National Bureau of Economic Research.
Krueger, A. B., & Maleckova, J. (2003). Education, poverty and terrorism: Is there a causal connection?. The Journal of Economic Perspectives, 17(4), 119-144.
Kettunen, J. (1997). Education and unemployment duration. Economics of Education Review, 16(2), 163-170.
KRUEGER, A. and LINDAHL, M. (1999),―Education for Growth in Sweden and the World‖,
NBER Working Paper No. 7190.
Lazear, E. P., & Michael, R. T. (1981). Family size and the distribution of per capita income.
- 33 - / 40 Mincer, J. (1991). Education and unemployment (No. w3838). National Bureau of Economic Research.
Ning, G. (2010). Can educational expansion improve income inequality in China? Evidences from the CHNS 1997 and 2006 data.
Yemelyanau,M (2009). "Belarusian Economic Research and Outreach Center Outreach Center.".
Oakes, J. (1990). Multiplying Inequalities: The Effects of Race, Social Class, and Tracking on Opportunities to Learn Mathematics and Science.
OECD (2013), Education at a Glance 2013: OECD Indicators, OECD Publishing. http://dx.doi.org/10.1787/eag-2013-en
Park, K. H. (1996). Educational expansion and educational inequality on income distribution. Economics of education review, 15(1), 51-58.
Psacharopoulos, G., & Patrinos*, H. A. (2004). Returns to investment in education: a further update. Education economics, 12(2), 111-134.
Schuller,T., Bynner,J.,Green,A.,Blackwell,L.,Hammond,C. And Perston,J.(2001), ―Modelling and Measuring the Wider Benefits of Learning: An Initial Synthesis‖, Centre for Research on
the Wider Benefits of Learning Institute of Education/Birkbeck College.
Sylwester, K. (2000). Income inequality, education expenditures, and growth.Journal of Development Economics, 63(2), 379-398.
- 34 - / 40 World Economic Situation and Prospects 2013
Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. The MIT press.
Appendix 1
Educational highest level school enrollment (% of total Population, 2010)
country tertiary country tertiary
Armenia 51.53 Honduras 20.62
Belarus 77.85 Jamaica 26
China 25.95 Mexico 28.03
Indonesia 23.12 Panama 45.75
Jordan 37.74 Paraguay 34.56
Kyrgyz Republic 42.14 Peru 42.99
Malaysia 42.28 Uruguay 63.20 Tajikistan 24.18 Egypt 32.37 Thailand 46.17 Guinea 11.04 Turkey 55.42 Mauritania 4.36 Vietnam 22.29 Morocco 14.13 Georgia 28.25 Albania 38.99 Argentina 74.80 Bulgaria 56.86 Chile 66.12 Croatia 54.13 Colombia 39.13 Estonia 64.27 El Salvador 23.44 Georgia 28.25 Poland 72.35 Hungary 60.65 Romania 58.84 Latvia 60.10 Ukraine 79.47 Lithuania 73.98
- 35 - / 40 Mean country school
enrollment ratio
43.69925
Source from: World Bank data 2010
Appendix 2
Employment rates among 25-64 year-olds, by educational attainment (2011)
Difference in employment rates (in percentage points) between tertiary-educated adults and those with only lower secondary education.Countries are ranked in descending order of the employment rate of tertiary-educated 25-64 year-olds.
Source from: OECD
Appendix 3
Level Description Principal characteristics 1 Primary education or first stage
of basic education
Primary education usually begins at ages five, six or seven and lasts for four to six years.
2 Lower secondary education or second stage of basic education
- 36 - / 40 schooling.
Upper secondary education Upper secondary education corresponds to the final stage of secondary education. The entrance age to this level is typically 15 or 16 years.
Post-secondary non-tertiary education
Post-secondary non-tertiary education straddles the boundary between upper secondary and post-secondary education from an international point of view. The students tend to be older than those enrolled at the upper secondary level.
3
Tertiary-type A education Tertiary-type A programmes are largely theory-based and normally they have a minimum cumulative theoretical duration (at tertiary level) of three years’ full-time equivalent, although they typically last four or more years.
Tertiary-Type B education Tertiary-type B are typically shorter than those of tertiary-type A and focus on practical, technical or occupational skills for direct entry into the labour market.
Source from: Education at a Glance, OECD, Paris, 2002, Glossary
Appendix 4
- 37 - / 40 Source from: GDP per capita in ppp of the years 2011, World Bank data, CIA and IMF data was used.
Appendix 5
The difference income level of countries (year of 2011)
Appendix 6
Developing regions (2013)
East Asia & Pacific 7 countries: Albania, China, Indonesia, Malaysia, Philippines, Thailand, Vietnam,
- 38 - / 40 Chile, Colombia, Costa Rica, Dominican Republic,
Ecuador, El Salvador, Guatemala, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, Venezuela RB,
Middle East & North Africa 3 countries: Egypt, Jordan, Morocco
Europe & Central Asia 15 countries: Belarus, Bulgaria, Croatia, Estonia, Georgia, Kazakhstan, Kyrgyz Republic, Latvia, Lithuania, Moldova, Romania, Russia, Tajikistan, Turkey, Ukraine,
South Asia 2 countries: Bangladesh, Pakistan
Sub-Saharan Africa 7 countries: Cote d'Ivoire, Ghana, Guinea, Mauritania, South Africa, Uganda, Zambia
Source from: the World Bank data, 2013
Appendix 7 Data description
Variable Description Source
Tertiary education demand(SERit)
The percentage number of people who the highest grade school enrollment level is tertiary education in country i at time t, computed by dividing the total number of education people (TENit) among a country
at the same time of a same age group(Popit)
it it it TEN Pop
TEA /
World Bank data, 1985-2010
GDP per capita (GPCit)
The gross domestic product divided by midyear population in country i at time t. GDPit collected the
- 39 - / 40 sum of gross value added by residents in one country
among all producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. it it it GDP Pop GPC / Gini coefficient (GINIit)
Degree of inequality of family income within countries. he Gini coefficient is a number between 0 and 1, where 0 corresponds with perfect equality (where everyone has the same income) and 1 corresponds with perfect inequality (where one person has all the income—and everyone else has zero income)
World Bank data, 1985-2010
Poverty ratio (PORit)
The percentage of people whose income is lower than poverty line, which divided by total population within a country i at time t.
it it
it PVP POP
POR /
World Bank data, 1985-2010 Tertiary educational level of unemployment ratio (GERit)
Unemployment rate could be defined as the fraction of time lost by all members of labor force within a unit period, the formula is as fellows:
it it
it UET POP
UER /
The popit is the number of workers in the labor force, UETitdevotes the number of workers who experienced unemployment.
- 40 - / 40 Appendix 8
The VIF table
