Effect of Education on Economic Growth:
Analysis of Indonesia
Name: Suzanne Loeff
Student number: 10179704
Study: Economics and business
Concentration: Economics
Subject thesis: Relationship between education and economic growth in Indonesia
Field: Econometrics
Number of credits: 12
Name of supervisor: R.M. Teulings
Research question: Is there an effect of education on economic growth in Indonesia?
Abstract – This study investigates the relationship between education and economic growth from
1979-2011 in Indonesia. Two Ordinary Least Square regressions have been performed. The results illustrate a significant negative relationship between economic growth and the enrolment of primary education, but this relationship is significant and positive when the enrolment of primary education after five years is used instead. The effect of the enrolment of tertiary education after one year on economic growth is negative and significant, but turns also into a positive and significant effect after five years. The negative effects of primary and tertiary education could be explained by the fact that a change in the enrolment of education is not profitable in the short-term. Both variables however turn into a positive effect after five years and therefore it seems probable that a change in the enrolment of education is only profitable after a couple of years. The negative relationship between tertiary education after one year and economic growth could also be explained by the increase of inequality, since only the elite are able to have tertiary education in Indonesia. The effects of the enrolment of secondary education, public spending on education and the labour force participation rate on economic growth are insignificant. Furthermore, physical capital has a significant positive effect on economic growth and poverty a significant negative effect.
1. Introduction
Throughout the last decades, Indonesia has been a country with a very dynamic economy. The economic growth of Indonesia became, under the regime of Suharto in 1967, the highest priority and since the early 1970s, Indonesia has experienced a rapid economic growth (Suryahadi et al., 2009, p. 110). During the period of 1970 to 2011, the average growth of gross domestic product (GDP) in Indonesia was 4.06% per year. Simultaneously, the average enrolment of primary, secondary and tertiary education increased in this period with respectively 0.40%, 1.54% and 0.63% per year (World Bank Data). The crisis in the 1980s and in the mid-1997s was caused by multiple issues. One example was the competition issues on the international market. Consequently, reforms had to be made within the country (Robertson-Snape, 1999; Soesarto et al., 2005). Indonesia changed from an inward to an outward oriented country, some major monopolies were broken, trade regime improved, power shifted from the government to the private sector and Indonesia became less regulated (Robertson-Snape, 1999; Soesarto et al., 2005).
The variables GDP and enrolment of primary, secondary and tertiary education increased simultaneously in Indonesia. Despite the fact that Indonesia is the third developing country after China and India, there has been little research about the concepts of education and economic growth for Indonesia (Suryahadi, 2009, p. 110). However, the relationship between education and economic growth is widely recognised among researchers (Nelson et al., 1966; Barro, 1991; Glomm et al.,1992; Mankiw et al., 1992; Borensztein et al., 1998; Laszlo, 2008). Therefore, the central question in this paper will be: ‘Is there an effect of education on economic growth in Indonesia?’. To investigate this question, research will be conducted on the effect of the enrolment of different levels of education on the real gross domestic product per capita of Indonesia (Robertson-Snape, 1999; Soesastro et al., 2005).
The relationship between education and economic growth is influenced by multiple variables (Michaelowa, 2000; Moroto, 2000; Pradhan, 2009). A change in education affects health, earnings, the related concept poverty, productivity and the labour force and this has an impact on economic growth (Michaelowa, 2005). Furthermore, investments in education and the economy will also be discussed, since they both have a positive effect on the economy
(Robertson-Snape, 1999; Soesarto et al., 2005; Tilak, 2005).
This thesis is structured as follows. Section 2 contains the illustrative framework. Section 3 will describe the empirical model to calculate the relationship between real GDP (RGDP) per capita and education in Indonesia. Section 4 clarifies the data used in the emprical model.
Section 5 will demonstrate and investigate the results of the two regressions. Section 6 contains the sensitivity analysis. Finally, section 7 draws conclusions from the findings of this research.
2. An illustrative framework
This chapter will contain a theoretical background. Section 2.1 will describe the relation between education and economic growth. Section 2.2 will explain the background of the economic
growth of Indonesia. Section 2.3 will describe the context of education of Indonesia.
2.1 Relationship economic growth and education
It is widely recognised that there is a positive relationship between education and economic growth (Nelson et al., 1966; Barro, 1991; Glomm et al.,1992; Mankiw et al., 1992; Borensztein et al., 1998; Laszlo, 2008). Adam Smith (1776) was one of the first researchers to investigate the relationship between education and economic growth, noting that education is the first step to economic development. He argued that a country cannot be well developed without
investments in education. Following Smith (1776), Theodore Schultz (1961) made a major contribution to this field, establishing that education is not just a consumption activity, but an activity of investment. Expenditures on education will not be profitable in the short-term, but are highly advantageous in the long-term (Schultz, 1961). He also concluded that education leads to the formation of human capital which, consequently, leads to a significant contribution to
economic growth.
However, human capital is not the only factor in the relationship between education and economic growth (Dahlin, 2005, p. 3). Figure 1 demonstrates that health growth, earnings and participation in the labour force are all micro effects of a change in education. The macro effects of a change in education are reduced population growth, better health of the population (and labour force) and an increased labour force. Finally they all affect the central macro variable of growth.
First, the micro effects will be discussed. Micro effects have an impact on a smaller scale than macro effects. This means that they only have impact at a national level, in particular markets, businesses or industries.
The positive relationship between education and earnings is widely recognised (Miller, 1960; Schultz, 1961; Kothari, 1970; Griliches et al., 1972; Tilak, 2005). An increase in education will lead to a more highly educated labour force. Employees who are highly educated are, in
general, better able to process information, perform and evaluate standard tasks, learn how to perform new tasks, adjust in new environments, communicate clearly and coordinate (Nelson et al., 1966; Lau et al., 1991). Since, employers only have a limited knowledge of the skills of their employees, the education of an employee works as a signal. Higher education sends the signal that a worker has more and/or better skills than someone who is less educated (Kerckhoff et al. 2001, p. 2). Therefore, an increase in education will lead to an increase of labour productivity. Furthermore, these more highly educated and productive workers have, in general, higher wages than less educated and productive workers (Miller, 1960). Therefore, it could suggest, that an increase of education will lead to an increase of earnings (Miller, 1960). To conclude, schooling is probably an important aspect of explaining the differences in incomes of a country. Poor people are in general less educated than rich people in an economy (Griliches et al., 1972). To narrow these income differences in a country, the number of college graduates per
Externalities and other indirect effects related to education, health growth
Higher education attainment and achievement of children
Better health and lower mortality of children
Better individual health Lower number of births
Lower population growth and better health of population (and labour force)
Education Increasing earnings
(Higher productivity) Higher growth
Increasing earnings of neighbours Participation in the labour force Increased labour force
Fig. 1 Economic effect of education on economic growth
Reprinted from “Returns to education in low income countries: evidence for Africa”, by Michaelowa, K., 2000 Retrieved from http://www.ipz.uzh.ch/institut/mitarbeitende/staff/michaelowa/publikationen/ Buechern/32-english.pdf
capita and the number of high schools probably has to increase (Miller, 1960).
However, according to Carneiro et al. (2003), family income is also an important factor to evaluate when investigating the relationship between education and economic growth. There are two possible ways to explain this. The short-term explanation is that credit constraints of families affect the resources which are required to finance education. As a result, families with high incomes are able to finance education and are therefore, better educated (Michaelowa, 2000; Carneiro et al., 2003). The long-term explanation is associated with the environment of high income families. Children from affluent families generally develop in environments that encourage academic as well as social skills. Their families have better resources and therefore, their children will go to schools, usually of a higher quality, for a longer period of time (Carneiro et al., 2003, p. 12). As a result, they are financially able to finish university, gain higher paid jobs and mostly marry partners that are of the same high socioeconomic level. Therefore, most of the children from these affluent families go on to create their own affluent family as well (Carneiro et al., 2003, p. 12). For both cases, individual as well as family income, earnings increase, which leads to an increase of total GDP. This results in an increase of economic growth (Kothari, 1970; Carneiro et al., 2003).
Simultaneous to the increase of earnings is a decline in poverty. Therefore, an increase in education will, in general, lead to a decrease in poverty (Tilak, 2005; Chaudhry et al., 2010; Afzal et al., 2012). The effect of an increase in primary education is particularly strong.
Furthermore, primary education also gives the highest private return compared to other
education levels (Michaelowa, 2000; Tilak, 2005; Chaudhry et al., 2010; United Nations, 2010; Afzal et al., 2012). An increase in primary education in developing countries would therefore be a possible solution to decrease poverty in especially developing countries (Tilak, 2005, p. 55).
Another frequently used variable to explain the relationship between education and economic growth is the labour force participation rate (Nelson et al., 1966; Michaelowa, 2000; Morote, 2000; Babatunde et al., 2005; Afzal et al., 2011). As was argued before, a more highly educated worker has greater job opportunities and has a lower chance of unemployment
(Schultz, 1961). Therefore, their participation rate in the labour force is higher. Thus an increase of education will, in general, lead to the increase of the labour force participation rate. Clearly, since the labour force participation rate increases, the total labour force will normally also increase (Schultz, 1961; figure 1).
The last micro effect of an increase in education is related to health growth. According to figure 1, increased education leads to, for example, better individual health as well as increased total health and lower mortality rates among children. A possible explanation for the relationship
between education and individual health is that the knowledge about health increases when individuals are better educated (Michaelowa, 2000). This leads to an increased life expectancy, contributing to the increase of total health (Michaelowa, 2000; Tilak, 2005) Since these
individuals are better educated and are able to lead a healthier life style, they will also have more knowledge about the health of their children and will be financially capable of sustaining them (Michaelowa, 2000; Lutz et al., 2011). Hence, an increase in education will lead to an increase of individual health, the health of children and a decline in the rate of the mortality of children (Lindert, 1980; Michaelowa, 2000; Tilak, 2005; Lutz et al., 2011). This will, in general, also lead to an increase of total health.
Increased education has also an effect on the number of births in a country. Higher educated men and women are, in general, better informed about using birth control and will therefore be able to control the amount of children they have (Lutz et al., 2011). Furthermore, the insurance that children have to financially support their parents when they are retired becomes less important. Higher educated people are able to save money for their retirement, since they usually have higher-paid jobs than less educated people (Lindert, 1980; Nugent, 1985). Those who receive a better and more extensive education when they were younger are also more aware of the costs of having children (Lindert, 1980, p. 6). Therefore, an increase in education will usually lead to a lower number of births and this has a negative impact on population growth (Lindert, 1980; Nugent, 1985; Lutz et al., 2011). However, especially in developing countries, the lower rate of mortality of children and the increase of individual health also leads to an increase in population growth (Hagen, 1959; Lindert, 1980; Michaelowa, 2000; Tilak, 2005; Lutz et al., 2011). Therefore, the negative effect of health on population growth, according to figure 1, is inconclusive.
Figure 1 demonstrates also some macro effects due to the increase of education. The first effect that will be discussed is the increased labour force and the second being the health of the population. Macro effects have an impact on the economy as a whole, on entire economies and industries, not just specific markets.
As explained above, the labour force becomes larger when the labour force participation rate increases. As figure 1 illustrates, the increase of the labour force is a macro effect. This means that the whole economy is affected by a change of the labour force. An increase of the labour force leads to more GDP, since more people earn money. Therefore, GDP will increase and this leads to economic growth (Babatunde et al., 2005; Afzal et al., 2011).
The second macro effect is a change of total health of the population. An increase of individual health as well as increased health and lower mortality rates of children results in a
total increase of health. A healthy employee is both physically and mentally stronger and will therefore be less absent from work, will be more productive and will earn higher wages
(Michaelowa, 2000; Bloom et al., 2004). Therefore, an increase of total health of the population leads, in general, to an increase of GDP and thus economic growth (Bloom et al., 2004).
2.2 Background: economic growth in Indonesia
According to the World Bank (2013), Indonesia is a developing country and belongs to the class of low-middle-income economies, which is the second lowest scale. In former days, the
Indonesian government was dominated by state enterprises, which created many opportunities for corruption. Inflation was running at over 600 percent and there was a large amount of foreign debt (Robertson-Snape, 1999, p. 92-593). The Indonesian economy had an inward focus, an import-substitution-industrialisation, the government was centrally planned, inhospitable to foreign investment and highly regulated, with a restricted trade regime (Robertson-Snape, 1999, p. 592-594). Both import-substitution-industrialisation and restricted trade regimes were
frequently used policies by developing countries. Until the 1970s, developing countries, such as Indonesia, were skeptical about their likelihood of success in exporting manufactured goods (Krugman et al., 2011, p. 256-266). Their new manufacturing industries could not compete with the existing, well-established manufacturing industries of developed countries. Tariffs were used to protect their industries from the already existing industries, until they grew strong enough to meet international competition (Krugman et al., 2011, p. 256-266). However, these protection policies also provided much opportunity for political patronage and corruption, seeing as Indonesia was already the most corrupt country in Asia according to a 1999 survey by Hong-Kong based Political and Economic Risk Consultancy (Robertson-Snape, 1999, p. 589-594).
Figure 2, based on data of the World Bank, illustrates the added value per sector of the Indonesian economy in percentages of GDP from 1962 to 2012. The considered sectors are agriculture, manufacture, industry and services for the period between 1961 and 2010 (see appendix 1 for extensive description). As figure 2 demonstrates, the overall share of the agricultural sector sharply declined by circa 30 percent in total, the overall share of the manufacturing sector increased by circa 15 percent in total, the overall share of the industry sector sharply increased by circa 30 percent in total and the overall share of the services sector did not change. Particular notable is the period of the mid-1960s to 1980s. This period was characterised by both a sharp decrease in the value added to agricultural sector and a sharp
increase in the value added to industry (Soesastro et al., 2005, p.3). During this period of time, there were structural transformations in the economy of Indonesia (Hill, 2000; Suryahadi et al., 2009). After 1966 investment rates rose sharply, resulting in rapid technological progress (Hill, 2000; World Data Bank). This had a positive effect on technological development and labour productivity and resulted in the first, permanent increase in economic growth (Hill, 2000, p. 17-26). Due to these structural changes, the share of the manufacturing sector overtook the agricultural sector and the share of labour force for agricultural felled below 50 percent (figure 2).
When Suharto came to power at 1967, economic development became the highest priority of his regime (Robertson-Snape, 1999, p. 592-593). However, in the 1980s, the
economy faced various problems. As an exporter of oil, Indonesia suffered from the detoriating oil prices of the early 1980s. Budget revenues and export earnings reduced significantly and the large decrease in oil prices affected the balance of payments of Indonesia (Soesarto et al., 2005, p. 3). Oil revenues could no longer subsidise the inefficient domestic industries and after the oil price collapsed in 1986, interventionist, inward focused policies became unsustainable
0 5 10 15 20 25 30 35 40 45 50 55 60 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 % of G DP
Fig. 2 % of GDP, value added by sector
(Robertson-Snape, 1999, p. 594). During the period of 1975 to 1980, the economy grew by 7.5 percent per year. However, during the period of 1980 to 1985 this sharply declined to 3.7 percent (Soesarto et al., 2005, p. 3). In response to the weakened economic growth, the
government, under leadership of Suharto, introduced more than twenty-four recovery packages in the period of 1983 to 1995, to increase the economic efficiency. The encouragement of investment, non-oil and gas exports became a high priority. Trade reforms were introduced to increase and improve trade and the Indonesian trade regime became outward, instead of inward focused (Soesarto et al., 2005, p. 3). Power shifted from the government to the private sector, some major monopolies were broken and Indonesia became less regulated. The
deregulation led to debureaucratisation and vice versa (Robertson-Snape, 1999, p. 596). Levels of protection declined and Indonesia became export orientated (Soesarto et al., 2005, p. 3). Due to the decline of protection, Indonesia became a more important player in the international market. This meant the country had to compete with the more effecient, well-established
industries of developed countries (Krugman et al., 2011, p. 256-266). To survive, Indonesia had to encourage trade by fostering private enterprises to be more efficient, cut production costs, and create an industrial base which could compete effectively on world markets (Robertson-Snape, 1999, p. 594). These reforms led to a decrease in the extent of corruption and reduced the high costs of doing business in Indonesia (Robertson-Snape, 1999, p. 594, 596). Finally, at
0 350 700 1050 1400 1750 2100 0 1E+13 2E+13 3E+13 4E+13 5E+13 6E+13 1961 1967 1973 1979 1985 1991 1997 2003 2009 con stan t 20 05 US $
Fig. 3 RGDP per capita of Indonesia and RGDP of the World
the end of 1980, Indonesie got attention on international level for its contuining success of the declining poverty level (Hill, 2000, p. 4).
Despite these improvements and the power shift from government to private sector, the link between government and business remain close. In 1990, an Indonesian civil service wage only covered one-third of an official’s household needs (Robertson-Snape, 1999, p. 590). Robertson-Snape (1999) claims that Hill said that corruption is high when provisions of salaries and welfare are low. The low wages of civil servants helps to explain the existence of corruption in Indonesia (Robertson-Snape, 1999, p. 590).
The economic reforms of the period of 1983 to 1995, were highlighted by the important, historical shift from import to export orientation, particularly in the manufacturing sector. The economy revived and the annual growth rate increased from 3.7 percent between 1980-1985 to 6.3 percent between 1985-1990 (Soesarto et al., 2005, p. 3-4). Figure 3, based on data of the World Bank, illustrates that RGDP per capita of Indonesia was growing faster than RGDP of the world (see appendix 1 for extensive description). This highlights the rapid growth of Indonesia, beginning in 1990.
Despite the reforms of 1983 to 1995, the Indonesian government was not able to stabilise the rupiah, the Indonesian currency. Furthermore, the decline of protection and tariffs did not result in high economic growth (Soesarto et al., 2005, p. 5-7). Some manufacturing products suffered from lack of competitiveness on the international market and Indonesia was faced with issues such as low labour productivity and lack of new investments. As a result, Indonesia could not compete with other low-cost Asian countries and the textile export from Indonesia to the United States and Europe was overtaken by countries such as China, Korea, Thailand and Taiwan (Soesarto et al., 2005, p. 5-7). The demand that once existed for
Indonesian products shifted to other countries and hence, many businesses and corporations were unable to cover the costs (Soesarto et al., 2005; Krugman et al., 2011). This resulted in increased unemployment, a depreciation of the rupiah and eventually, a financial crisis in mid-1997 (Dahnani et al., 2002; Krugman et al., 2011). As figure 3 demonstrates, RGDP per capita started to decrease sharply at the end of 1997 and the gap between RGDP of the world and RGDP per capita of Indonesia widened. The crisis not only resulted in an increased number of people falling below the poverty line, but also a rise in extreme poverty. Consequently,
nutritional and health standards of the poorest people worsened (Dahnani et al., 2002, p. 1220). However, at the beginning of 1999 RGDP per capita increased again (figure 3). Eventually, in 2005, RGDP per capita reached the same level as before the crisis (figure 3). Due to this crisis, the government of Indonesia was forced to ask the International Monetary Fund (IMF) for
support to adopt a programme for economic recovery and trade reforms (Soesarto et al., 2005, p. 4). This led to more financial transparency and was expected to reduce high-level corruption. New regulations were also introduced, a Bank Restructuring Agency was formed to implement reforms and the IMF broke up some monopolies (Robertson-Snape, 1999, p. 600).
In 2001, Japan, the United States and the developing countries of East-Asia became more important for the non-oil export market of Indonesia. Exports from Indonesia to East-Asia significantly increased from 15 percent in 1990 to 26 percent in 2001 (Soesarto et al., 2005, p. 4). The market share of Indonesia on the international market increased between 1995 and 2005 and developing countries became almost as important as OECD countries for the export of manufacturing products of Indonesia. The stronger position of Indonesia on the international market was also caused by the relatively low tariffs of Indonesia (Soesarto et al., 2005, p. 4). Developing countries have, in general, relatively higher tariffs than developed countries. However, Indonesia has lower tariffs than other developing countries and therefore has a comparative advantage, with a relatively strong position in the international market (Krugman et al., 2011; Soesarto et al., 2005).
After the attacks of 11 September 2001, the Indonesian exports slowed down, due to weakened global economic demand (Soesarto et al., 2005, p. 4). However, according to figure 2 and Soesarto (2005), industrial production started to pick up around 2003 and 2004. This was a result of the breakout of Severe Acute Respiratory Syndrome (SARS) in 2003. Mainly Asian countries suffered from this disease, and most comprehensively secured themselves to protect their country against it (Caballero-Anthony, 2005, p. 475-478). However, Indonesia developed their own security version, which covered both domestic as international environment and with particular attention for economic development (Caballero-Anthony, 2005, p. 475-478).
Therefore, the demand for, in particular, garments of developed countries shifted from other East-Asian countries to Indonesia (Caballero-Anthony, 2005; Soesarto et al., 2005).
2.3 Education in Indonesia
As mentioned in section 2.1, education and economic growth should result in a positive relationship (Nelson et al., 1966; Barro, 1991; Glomm et al.,1992; Mankiw et al., 1992; Borensztein et al., 1998; Laszlo, 2008). The level and amount of education received by the population is one of the most important factors contributing the process of development (Laszlo, 2008; Michaelowa, 2000; Raja, 2005; Schultz, 1961; Smith, 1776). An increase in education increases productivity, increases earnings and reduces poverty, stimulates economic growth and has an important role in building human capabilities through skills and knowledge.
Therefore, a country with a high education level is also more attractive to investors (Raja, 2005). Education leads to the formation of human capital which has a significant positive effect on economic growth (Schultz, 1961). However, many developing countries suffer from lack of skilled labour. This is one of the reasons that increasing economic growth of developing
countries is difficult to maintain (Kinda, 2010; Krugman et al., 2011). In this section, the quantity and quality of education in Indonesia will be discussed.
Figure 4, based on data of the World Bank, compares the different education levels in Indonesia (see appendix 1 for extensive description). This figure only presents the enrolment rates of primary, secondary or tertiary school, excluding the graduation rate. The primary
enrolment ratio is above one hundred percent and decreases after 1985. This is possibly due to the inclusion of over-aged and under-aged student and grade repetition (see appendix 1 for extensive description).
According to the World Bank data, primary school starting age is at seven years of age and the duration of primary school is six years. Since the 1970s, the central government has played a dominant role in primary education by building new public schools, supplying teachers and providing teaching materials (James et al., 1996, p. 388). In 2009 the Indonesian enrolment rate for primary education was 97 percent, and the completion rate 98 percent in 2009 (Unicef). The reason for this high enrolment rate is that primary school is compulsory and the cost of
0 10 20 30 40 50 60 70 80 90 100 110 120 130 1970 1975 1980 1985 1990 1995 2000 2005 2010 % enrollmen t o f t h e to tal p o p u latio n
Fig. 4 School enrolment ratio in Indonesia
primary school is covered by the government (Frederick et al., 2011, p. 33).
According to the World Bank data, secondary school starting age is at thirteen years of age and the duration is also six years. These six years are divided by three years in junior secondary school, which is compulsory, and three years in senior school, which is not compulsory (Unicef). Nevertheless, although during the period of 1970 until 2010 the overall secondary enrolment ratio increased, 20 percent of the children still did not attend secondary school in 2010 (figure 4). While a large amount of children finish primary education, they do not start or, if they do start, do not finish secondary school (World Bank data). The explanation for this drop-out rate is that children attain in some form of child labour and are thus not able to study full-time or part-time (Unicef).
The enrolment rate of tertiary education increased during the same period as secondary education. However, the education levels of primary and secondary education have a relatively higher enrolment ratio compared to tertiary education (figure 5). Two reasons for this difference is that the college fee for tertiary education is high for most families and that tertiary school is compulsory (Unicef; Welch, 2007). Furthermore, students who did not finish secondary school are not allowed to start tertiary school (Welch, 2007, p. 669). Consequently, the enrolment rate of tertiary education was not even 30 percent in 2010 (figure 4).
It is not always the case that a high enrolment rate implies higher economic growth. Developing countries that are well focused on higher education instead of primary education are more likely to have a positive relationship between the enrolment of tertiary education and future inequality, noting future inequality is negative related to economic growth (Forbes, 2000; Gruber et al., 2014). Therefore, whether the relationship between tertiary education and economic growth is positive or negative, depends on the economic situation and education focus of the country (Forbes, 2000; Gruber et al., 2014).
Particularly within primary and secondary schools, the enrolment rates of Indonesia are relatively high compared to other developing countries (World Bank data). However, most of these children are still not well-educated, because the quality of education, particularly public education, is very low. Despite this low quality, circa 15 percent of the children are still not able to finish their first or second year in one year (Unicef). Furthermore, the low quality of education contributes to the high rate of illiteracy. In 2005, 16 percent of the Indonesian population was not able to read or write (Raja, 2005, p. 1; Frederick et al., 2011, p. 33). To receive a higher quality education, students must go to private schools. These private schools are relatively expensive and are therefore, only accessible for the rich elite who are able to pay the high college fees (Welch, 2007).
Developing countries have frequently expanded their educational systems and made large investments in education. In many cases, their rates of growth on educational systems even exceeded the rates of economic growth (Tilak, 2005, p. 21). Indonesia has also made large investments in education. Figure 5, based on World Bank Data, demonstrates that the overall public spending on education increased during the period of 1989-2011 (see appendix 1 for extensive description). This increase is large compared to, for example, health expenditure. In the period from 1975 to 1995, the Indonesia’s education policy resulted in an increased numbers of schools and school placement possibilities, and the cost of education per student declined (James et al., 1996, p. 388). As a result, the enrolment ratios of education increased. But despite these large investments in education, the quality of education in Indonesia is still very low (James et al., 1996, p. 388). Therefore, since quality of education is more important then quantity, the Indonesian government should focus on investments in education to improve quality instead of quantity. It is an increase in their quality of education which would lead to a highly educated Indonesian labour force (Barro 2001, p. 17).
0 5 10 15 20 25 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 % of go v er n men t ex p end it u re
Fig. 5 Total public spending on education in Indonesia
3. Empirical model
In chapter 2, different related factors of education, economic growth and/or the relationship between these two variables are discussed. Section 2.1, figure 1 demonstrates the macro and micro effects which are important for the relationship between education and economic growth. An investigation of this figure illustrates that the most important factors of this figure are health, earnings and/or poverty and the labour force. Multiple issues are discussed in section 2.2 and 2.3, including the importance of investments in the economy and education as well as the explanation of the concepts in primary, secondary and tertiary education. This section will introduce the empirical model used to calculate the relationship between education and economic growth.
The empirical model will include most of the factors that were discussed above. The variables primary, secondary and tertiary education represent the concept of education. The variables poverty, labour force and investments in education and the economy are also added to investigate whether they have a significant effect on economic growth. Furthermore, a dummy, to correct for the financial crisis of 1997 and 1998 (see section 2.2), will be included in the empirical model. However, some of these variables do not contain much observations. Therefore, two time series regressions, using the Ordinary Least Square (OLS) method, will be performed.
The first regression which will be performed is
with t=1,2 , ...,n
This regression includes the dependent variable RGDP per capita. This is used as a
measurement of economic growth. Since RGDP often has a unit root and to correct for serial correlation, the first difference of RGDP, , is created and is used as independent variable in regression one. RGDPt is the original value of RGDP per capita at time t
and RGDPt-1 is the first lag of RGDP per capita.
The independent variables primary (PRIM), secondary (SEC) and tertiary (TER)
education are added to the regression as a measure of education. However, it takes a couple of years before an increase or decrease in PRIM, SEC or TER has an effect on education.
Therefore, the second lag of primary education (PRIMt-2) and the first and fifth lag of tertiary
education (TERt-1 and TERt-5) are, based on a general to specific method, added to the first
the dummy (D97/98) is added, to correct for the financial crisis of 1997 and 1998. εt is the error
term. To correct for heteroskedasticity, white noise standard errors will be used (White,1980). This will lead to valid errors and thus valid t-values, p-values and confidence intervals (Stock et al., 2011, p. 198-201).
In general, every created lag of a variable leads to a loss of one observation. However, in this case there are more available observations of the variables RGDP, PRIM and TER than regression one suggests. Therefore, the variable GAPt and the lags of PRIM and TER do not
lead to a loss of observations for the first regression.
To perform a correct OLS regression, the regression must meet a number of
assumptions. The first assumption for this regression is that the error term εt has a conditional
mean of zero. So for each t, the error εt has an expected value of zero, given the independent
variables for all time periods. The second assumption is that there is no serial correlation. Conditional on the independent variables, the errors in two different time periods are uncorrelated. The third assumption is that the variable GAPt does not have a unit root. The
fourth assumption is that large outliers are unlikely, meaning all variables have nonzero finite fourth moments. The fifth assumption is that the independent variables are not perfectly multicollinear. In the sample, no independent variable is constant nor a perfect linear combination of the others. Furthermore, heteroskedasticity is allowed. Therefore, the error εt does not always have the same variance for all t given any value of the independent
variables.
The second regression which will be performed is
with t=1,2 , ...,n
Regression two also contains the dependent variable GAPt and the independent variables
PRIM, SEC, PC and D97/98. The dependent variable GAPt, the first difference of RGDP, of
regression two is created for the same reason as for regression one. Since education has a long-term effect on economic growth, the original value of tertiary education (TERt) is subtracted
and the first lag of this variable (TERt-1) is added to the second regression. These two changes
are based on a general to specific method. To investigate the effect of other variables on economic growth, the independent variables poverty (POV), public spending on education (PSE) and labour force (LF) are added to regression two. The error term in this regression is μt.
To correct for heteroskedasticity, white noise standard errors will be used (White,1980).
regression two suggests. Therefore, the variable GAPt and the lag of TER do not lead to a loss
of observations. Furthermore, for regression two, the same assumptions made for regression one are upheld and heteroskedasticity is allowed.
4. Data
In this section, the variables used for the two regressions will be explained in more detail. The World Bank provides time series data of Indonesia with a period ranging from 1961 to 2012.
Due to the fact that economic growth is a broad concept and is therefore complicated to analyse, RGDP per capita will be used instead. The variables RGDP and RGDP per capita are frequently used by researchers to measure economic growth (Barro, 1991; Borensztein et al., 1998; Morote, 2000; Soubbotina et al., 2000; Tilak, 2005; Afzal et al. 2011; Afzal et al., 2012; Afzal et al., 2013). Since the population of Indonesia increased, according to the World Bank data, RGDP has to be corrected for the increase in population. Therefore, the dependent variable RGDP per capita will be used for both regressions. This variable ranges data over the period 1961 until 2012 and is expressed in U.S. dollars (USD). However, to use RGDP per capita as dependent variable for both regressions, the variable is converted from USD to index numbers with a base year of one hundred percent in 2005 (see appendix 1 for extensive description).
The enrolment ratios of the independent variables primary, secondary and tertiary education are included in both regressions and ranges data over the period 1970 until 2011. The enrolment ratios of these variables are measured by dividing the actual enrolment of the education level by the population of the education age (primary, secondary or tertiary), expressed as percentages (see appendix 1 for extensive description). The enrolment ratio of primary school exceeds one hundred percent due to grade repetition and the inclusion of students who are over- or under-aged (World Bank Data; figure 5). However, the enrolment ratios of primary, secondary and tertiary school have respectively two, three and one missing value(s). To replace these missing values (MV), the next formula will be used:
with t=1,2 , ...,n
The random error is added to correct for any mistakes of the calculated missing value.
The independent variable physical capital is added to both regressions as a measure of investments in the economy. Physical capital reflects values of the time period 1979 until 2012
and is calculated by the gross fixed capital formation. This includes land improvements and the construction of schools, railways, roads, etc. (see appendix 1 for extensive description).
Furthermore, this variable is converted from USD to index numbers with a base year equals one hundred percent in 2005.
The independent variable poverty is calculated by the World Bank, ranges values over the period 1984 to 2011 and is expressed as the percentage of the population who lives of less than 2.00 USD per day (see appendix 1 for extensive description). Since this variable does not have much observations, poverty is only added to second regression. This variable also contains some missing values. To solve this problem, the same method will be used as explained for the variables PRIM, SEC and TER.
The independent variable public spending on education is calculated by the World Bank and is expressed as the percentage of total government expenditure on public education (see appendix 1 for extensive description). Since this variable only contains data over the period of 1989 until 2011, public spending on education is only added to the second regression.
Furthermore, to replace the missing values, the same method will be used as explained for the variables PRIM, SEC and TER.
The independent variable labour force contains data for the period of 1990 to 2012 and is expressed as the percentage of the population who is economically active and aged of 15 to 64 years. This includes everyone who supplies labour for the production of services and goods (see appendix 1 for extensive description). Because of the small amount of observations of this variable, labour force is only added to the second regression.
The financial crisis of 1997 and 1998 not only had an impact on the economy of Indonesia, but also on almost all variables used for the analysis of the relationship between education and economic growth (World Bank Data). To correct for this financial crisis, the dummy (D97/98) is added to both regressions.
Regression one and two have periods ranging respectively from 1979 to 2011 and 1990 to 2011. The differences in these time periods can be explained by the amount of variables. Regression one contains less variables with a small number of observations than regression two and therefore, contains more observations.
5. Empirical results
In order to examine the relationship between education enrolment and RGDP, two regressions, based on the OLS method, were performed. This section will present and discuss the results of these two regressions. The first and second paragraph will contain the results of regression one and regression two respectively.
5.1 Regression 1
Figure 6 illustrates the results of regression one. The R-squared is 72.86 percent. This means that the independent variables of regression one explain about 72.86 percent of the variation in the first difference of RGDP per capita for this sample. Furthermore, the F-statistic of 8.35 demonstrates that for a significance level of 5%, the independent variables jointly affect the first difference in RGDP per capita, GAP.
Fig. 6 Regression of GAPt, PRIMt, PRIMt-2, SECt, TERt-1, TERt, PCt and D97/98
* White noise standard errors
The independent variables PRIM and PRIMt-2 are significant for a 5% significance level.
An increase of the enrolment of primary education of one unit (ceteris paribus) will lead to a
_cons 12.50749 9.31967 1.34 0.192 -6.68673 31.70171 D -6.791595 4.532104 -1.50 0.147 -16.12564 2.542448 PC .0400855 .0123664 3.24 0.003 .0146165 .0655546 TER5 1.33349 .6099242 2.19 0.038 .0773276 2.589652 TER1 -1.663187 .6777626 -2.45 0.021 -3.059065 -.2673092 SEC .0248081 .0986014 0.25 0.803 -.1782653 .2278815 PRIM2 .4403355 .1354928 3.25 0.003 .1612827 .7193882 PRIM -.5011744 .1410298 -3.55 0.002 -.7916307 -.2107181 GAP Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = 2.0118 R-squared = 0.7286 Prob > F = 0.0000 F( 7, 25) = 8.35 Linear regression Number of obs = 33 . regress GAP PRIM PRIM2 SEC TER1 TER5 PC D, robust
decrease in economic growth of 0.501. However, one unit increase of this variable after two years (ceteris paribus) leads to an increase in economic growth of 0.440. The effect of primary education has therefore first, a negative effect on economic growth, and then turns into a positive effect after two years. The positive coefficient PRIMt-2 compensates approximately for
the negative coefficient PRIM.
The explanation for the difference in the effects of the negative coefficient PRIM and positive PRIMt-2 are discussed in section 2.1. Investment in education will not be profitable in
the short-term, but is highly advantageous in the long-term (Schultz, 1961). It is therefore likely that the enrolment ratios of primary education also do not have a positive short-term effect on economic growth. Since PRIM has a negative effect on economic growth and PRIMt-2 has a
positive effect, this statement seems probable.
The independent variables TERt-1 and TERt-5 are significant for a 5% significance level.
A single unit increase of the enrolment of tertiary education after one year (ceteris paribus) will lead to a decrease in economic growth of 1.663. However, one unit increase of this variable after five years (ceteris paribus) lead to an increase in economic growth of 1.333. The negative coefficient of TERt-1 could be explained by the same reason aforementioned for primary
education. Another explanation has to do with inequality. When a country is highly focused on tertiary education and the enrolment of this education level increases, this could lead to higher future inequality. Consequently, this higher inequality will lead to a decrease of economic growth (Gruber et al., 2014). Inequality being an issue for Indonesia is explained in section 2.3. Most Indonesian students do not finish secondary school. However, the poor students who do, are not able to start tertiary school, since the college fees are too high. Therefore, almost only the elite are able to start tertiary school (Welch, 2007). As a result, an increase of the enrolment of tertiary school is probably only an increase of the enrolment of students from high-income families, who are able to afford the college fee. Consequently, the income gap between the rich and the poor will become wider, which leads to an increase of inequality and this will, according to Gruber et al. (2014), have a negative effect on economic growth.
The independent variable PC is significant for a 5% significance level. A unit increase of physical capital will lead to an increase in economic growth of 0.040. This result confirms the theory in section 2.2 and 2.3 about the importance of investments in the economy. An increase in physical capital reflects an increase in the investment by the Indonesian government in the economy. The government made large investments in the economy during the period ranging of 1983 to 1995. According to the first regression, this had a positive effect on the Indonesian economy.
The independent variable SEC and the dummy are both insignificant. Therefore, it is not possible to draw any conclusions from their effects on economic growth.
The coefficients of PRIM and PRIMt-2 are both around ±0.5 and the coefficient of TERt-1
and TERt-5 are both around ±1.5. According to section 2.3 this result is unexpected. Figure 4
demonstrates that the total number of students enrolled in tertiary education is much lower than primary education (figure 4). This can be explained by high college fees for tertiary education and the fact that many students are not even allowed to start tertiary education. This is because they did not finish secondary school, due to for example many of them commencing work in some form of labour.
The variable PC has a smaller effect on the economic growth with a coefficient of circa 0.05. Therefore, the effect of the enrolment of primary education on economic growth is smaller than the effect of tertiary education. However, the three variables all have a small effect on economic growth. This is possibly due to the time period of regression. The economy will only benefit from these students once they have graduated. Therefore, education will also have an effect after ten or fifteen years. However, since the first regression only has data for the period 1979 to 2011, these lags are too large. Therefore the real effect of education on economic growth might be larger than illustrated in figure 6. The small time period could also explain the small coefficient of physical capital. For this variable, no lags are included. However,
investments in the economy might not be profitable in the short-term. Hence, the effect of physical capital on economic growth might also be larger than illustrated in figure 6. However, it should also be taken into consideration that there are far more variables than just education and physical capital that affect the economy. As such, large coefficients are highly unlikely.
5.2 Regression 2
Figure 7 demonstrates the results of regression two. The R-squared is 80.58 percent. This means that the independent variables of regression two together explain about 80.58 percent of the variation in the first difference of RGDP per capita for this sample. Furthermore, the F-statistic of 6.73 proves that for a significance level of 5%, the independent variables jointly affect GAP, the first difference in RGDP per capita.
The independent variables TERt-1 is significant for a 5% significance level. A single unit
increase of the enrolment of tertiary education after one year (ceteris paribus) will lead to a decrease in economic growth of 1.972. The coefficient of TERt-1 of both regressions are
between -1.6 and -2.0 and therefore, the differences in the effect of the enrolment of tertiary education after one year in regression one and regression two are nearly the same. The
negative effect of tertiary education could be due to its short-term effect and the impact of inequality, which are already explained in section 5.1.
The independent variable poverty is significant for a 5% significance level. An increase in the poverty level of one unit (ceteris paribus) will lead to a decrease in economic growth of 0.174. Therefore, a decrease of the amount of people who are living off less than $2.00 will actually have a positive effect on RDGP. In section 2.1 the concept poverty was discussed. A decrease of poverty is related to an increase in earnings and this leads to an increase in
economic growth. Therefore, the results of the variable poverty in the second regression confirm
Fig. 7 Regression of GAPt, PRIMt, SECt, TERt-1, PSEt, POVt, PCt, LFt and D97/98
* White noise standard errors
the expectation according to the theory explained in section 2.1.
The independent variable PC is insignificant for a 5% significance level, but is significant for a 10% significance level. Therefore, a unit increase of physical capital will lead to an
increase in economic growth of 0.040. Therefore, the results also confirm the theory in section 2.2 and 2.3 (explained in section 5.1) for the second regression. However, since physical capital is only significant for a 10% level instead of 5%, strong conclusions cannot be drawn.
_cons 36.8073 75.68845 0.49 0.635 -126.7077 200.3223 D -7.015236 4.646596 -1.51 0.155 -17.0536 3.023124 LF 1.277714 1.079892 1.18 0.258 -1.055251 3.61068 PC .0361302 .0167958 2.15 0.051 -.0001549 .0724154 POV -.1744152 .0465792 -3.74 0.002 -.2750434 -.073787 PSE -.1594535 .2425802 -0.66 0.522 -.6835162 .3646093 TER1 -1.972053 .8906422 -2.21 0.045 -3.896169 -.0479375 SEC .2391746 .2848408 0.84 0.416 -.3761865 .8545358 PRIM -.8667115 .5129477 -1.69 0.115 -1.974868 .2414446 GAP Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = 2.2886 R-squared = 0.8058 Prob > F = 0.0014 F( 8, 13) = 6.73 Linear regression Number of obs = 22 . regress GAP PRIM SEC TER1 PSE POV PC LF D, robust
The independent variables PRIM, SEC, PSE, LF and the dummy are insignificant. Therefore, it is not possible to draw any conclusions about their effects on economic growth.
6. Sensitivity analysis
In this section the assumptions from section 3 will be tested and/or discussed with respect to both regressions and the results will be critically evaluated. An important aspect of the empirical model is that the data of both regressions only account for a small time period. Specifically for regression two, it is possible that due to the low amount of observations, the regression will suffer from serial correlation and will therefore demonstrate significant t-values, while they are actually not significant (Wooldridge, 2012, p. 412-418). To test for this serial correlation, the Breusch-Godfrey test will be used. Therefore, the following regression will be performed:
with t= 1, .... ,n
Then a test will be performed, where H0:γ8 = 0 and H1: γ8 ≠ 0.
Fig. 8 Breusch-Godfrey test to test for serial correlation in regression one
According to figure 8, the p-value is 0.2921. This is larger than the 1% and 5% significance level. Therefore, the null hypothesis cannot be rejected. There is approximately no serial correlation in regression one.
To test for serial correlation in the second regression, this regression has to be performed:
with t=1 , .... ,n
Then a test will be performed, where H0: ω9 = 0 and H1: ω9 ≠ 0. H0: no serial correlation
1 1.110 1 0.2921 lags(p) chi2 df Prob > chi2 Breusch-Godfrey LM test for autocorrelation
Fig. 9 Breusch-Godfrey test to test for serial correlation in regression two
According to figure 9, the p-value is 0.0242. This is larger than the 1% significance level, but smaller than the 5% significance level. Therefore, the null hypothesis can only be rejected for a 5% significance level, but not for 1%. Therefore, it could be possible that the second regression suffers from serial correlation. The results could therefore illustrate significant t-values, while they are actually not significant. To construct a valid regression, lags of the variable GAPt should be added. However, since data of the second regression ranges over a
short time period, there would not be enough degrees of freedom, making it impossible to construct a valid regression.
It is also necessary to mention that simultaneous causality of the variables of education and RGDP will give some problems (Borensztein et al., 1998; Morote, 2000; Barro, 2001; Tilak, 2005; Pradhan, 2009; Afzal et al., 2010; Afzal et al., 2011). Therefore, instruments could be used in both regressions, to replace the independent variables of the enrolments of primary, secondary and tertiary education. However, the results including these instruments are often similar to the results without the instruments (Barro, 2001). It should be noted that there may not be an ideal instrument available, which will again produce an imperfect estimation (Borensztein et al., 1998). Therefore, instrumental variable estimation for the two regressions will not be performed.
To test if both regressions are also valid by dropping the third assumption of the unit root, a Dickey-Fuller test on GAP will be performed (Stock et al., 2011, p. 578-595).
The Dickey-Fuller test on RGDP will first test, due to H0 : δ=1 against H1 : δ<1, whether RGDPt
is nonstationary and contains a unit root (regression a). When H0 is rejected, RGDP has an
autoregressive root of 1, which means that the regression is nonstationary (Stock et al., 2011, p. 578-595). Thereafter, the Dickey-Fuller test will test H0 : δ=0 against H1 : δ<0 (regression b).
When H0 is rejected, then ΔRGDPt is nonstationary. To test if it was necessary to use GAP H0: no serial correlation
1 5.078 1 0.0242 lags(p) chi2 df Prob > chi2 Breusch-Godfrey LM test for autocorrelation
instead of RGDP as a dependent variable in both regressions, a Dickey-Fuller test on RGDP will also be performed.
In this case, H0 : =1 against H1 : <1.
First, the unit root of GAP and RGDP of the first regression will be tested. As figure 10
Fig. 10 Dickey-Fuller test of regression one on RGDP to test for a unit root
Fig. 11 Dickey-Fuller test of regression one on GAP to test for a unit root
demonstrates, RGDP of the first regression has a unit root for 1%, 5% and 10% significance level. However, GAP does not have a unit root for 1%, 5% and 10% significance level (figure 11). Therefore, regression one does not suffer from a unit root, due to the replacement of RGDP by GAP.
Now, the unit root of GAP and RGDP of the second regression will be tested. As figure MacKinnon approximate p-value for Z(t) = 0.9911
Z(t) 0.769 -3.702 -2.980 -2.622 Statistic Value Value Value Test 1% Critical 5% Critical 10% Critical Interpolated Dickey-Fuller Dickey-Fuller test for unit root Number of obs = 32 . dfuller RGDP
MacKinnon approximate p-value for Z(t) = 0.0015
Z(t) -3.985 -3.702 -2.980 -2.622 Statistic Value Value Value Test 1% Critical 5% Critical 10% Critical Interpolated Dickey-Fuller Dickey-Fuller test for unit root Number of obs = 32 . dfuller GAP
12 illustrates, RGDP of the first regression has a unit root for 1%, 5% and 10% significance level. However, GAP does not have a unit root for 5% and 10% significance level (figure 13). Therefore, for these significance levels regression one does not suffer from a unit root, due to the replacement of RGDP by GAP.
Fig. 12 Dickey-Fuller test of regression two on RGDP to test for a unit root
Fig. 13 Dickey-Fuller test of regression two on GAP to test for a unit root
The results of the variable enrolment of primary education for regression one (PRIMt), as
explained in section 5.1, remain unexpected, following the explanation in section 2.1 and 2.3. One of the Millennium Development Goals is to achieve universal primary education (United Nations, 2010). However, if primary education would also have a negative effect on economic growth, the United Nations might waste a lot of effort and money. Since, the data of the
enrolment of primary education exceeds one hundred percent, which is theoretically impossible, it seems more reliable to use instead, other data to measure primary education (see figure 5). Therefore, the data of the enrolment of primary education of Indonesia is not used, but data of the developing countries of East Asia and the Pacific (World Bank Data). This data ranges over a period of 1986 until 2011. Therefore, the first regression period will be shortened to 1986 until
MacKinnon approximate p-value for Z(t) = 0.9769
Z(t) 0.290 -3.750 -3.000 -2.630 Statistic Value Value Value Test 1% Critical 5% Critical 10% Critical Interpolated Dickey-Fuller Dickey-Fuller test for unit root Number of obs = 21 . dfuller RGDP
MacKinnon approximate p-value for Z(t) = 0.0200
Z(t) -3.200 -3.750 -3.000 -2.630 Statistic Value Value Value Test 1% Critical 5% Critical 10% Critical Interpolated Dickey-Fuller Dickey-Fuller test for unit root Number of obs = 21 . dfuller GAP
2011. The data of the enrolment of primary education of the developing countries of East Asia and the Pacific will be used to perform the first regression of section 3. For the adjusted
regression one, which includes data of the developing countries of East Asia and the Pacific, the same assumptions are valid as for regression one (see section 3).
Fig. 14 Adjusted regression 1
*data of the enrolment of primary education of the developing countries of East Asia and the Pacific is used ** White noise standard errors
As demonstrated in figure 14, most variables become insignificant, except TERt-1 and
PC, while in the original regression one, the variables PRIMt-1, PRIMt-5 and TERt-5 were also
significant. This is probably due to the shortened time period. The original regression included 33 observations instead of 24. Since the variables of primary education in the adjusted
regression one are insignificant, no conclusions can be drawn from the effect of the enrolment of primary education on the economic growth of Indonesia.
7. Conclusion
The positive relationship between education and economic growth is widely recognised.
However, little research has been done concerning this relationship in Indonesia, even though it
_cons 74.99465 41.51949 1.81 0.090 -13.02273 163.012 D -7.648319 4.891013 -1.56 0.137 -18.0168 2.720166 PC .0454447 .0155873 2.92 0.010 .0124011 .0784882 TER5 1.18104 .7131283 1.66 0.117 -.3307242 2.692805 TER1 -1.605891 .8120511 -1.98 0.065 -3.327363 .1155801 SEC -.0342001 .1061022 -0.32 0.751 -.2591266 .1907265 PRIM2 -.3363197 .3138026 -1.07 0.300 -1.001552 .3289122 PRIM -.3521211 .3050377 -1.15 0.265 -.9987721 .2945298 GAP Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = 2.4307 R-squared = 0.7315 Prob > F = 0.0002 F( 7, 16) = 8.34 Linear regression Number of obs = 24
is experiencing a sharp increase in economic growth compared to other developing countries. Simultaneously, the enrolment of secondary and tertiary education increased sharply (figure 5). An increase of the enrolment of education has a direct and positive effect on the individual earnings, earnings of neighbours, productivity and participation rates of labour force and a negative effect on poverty (figure 1).
The variables public spending on education, poverty level and labour force were added to the second regression, to investigate their effect on economic growth. The results of section 5.2 demonstrated that the variables public spending on education and the effect of the labour force participation rate were insignificant. The poverty level had, as expected according to section 2.1, a significant negative effect on economic growth. An increase in poverty will lead to a decrease in economic growth in Indonesia.
The variable physical capital in regression one was significant and had a positive effect on economic growth. In regression two, this variable was only significant at a 10% significance level and also had a positive effect on economic growth. Although the coefficient is relatively small compared to other significant variables in regression one and two, an increase in the investment in the economy will still lead to a small positive effect on economic growth in Indonesia.
The variable of the enrolment of primary education in regression one had first a significant and negative effect on economic growth, but turns into a significant positive effect after five years. Therefore, an increase in the enrolment of primary education will probably lead to a total increase of economic growth, since the expectation is that the effect of primary school on education will continue as time goes by. However, in regression two the variable of primary education was insignificant. This is probably due to the short time period of regression two. The second regression also suffers from serial correlation. The t-values can illustrate an insignificant result when the result is actually significant. Since the first effect of primary education on
economic growth is negative and data of this variable exceeds one hundred percent, data of primary education in section 6 was replaced by data of this variable of developing countries of East Asia and the Pacific. However, the variables of this regression were mostly insignificant. Therefore, conclusions could not be drawn.
The results of secondary education on economic growth were in both regressions, insignificant. Therefore, it seems reasonable to state that secondary school has a smaller (or no) effect on economic growth compared to primary and tertiary education. However, it could also be possible that the enrolment of secondary education only has a long-term effect on economic growth in Indonesia. However, since the time periods of the regressions are short due
to few observations, this effect cannot be calculated with this model and the World Bank Data. More research is necessary to determine if there is an actual, long-term effect of the enrolment of secondary education on economic growth.
The results of the variable tertiary education of both regressions have significant effects on economic growth. The enrolment of tertiary education after one year had a negative effect on economic growth, according to both regression one and two. However, the enrolment of tertiary education turns into a positive effect after five years, according to regression one. The negative effect could be explained by the theory in section 2.1. A possible explanation is that a change in the enrolment of education is not profitable in the short-term, but only in the long-term. Another possible explanation is that an increase of the enrolment of tertiary education leads to an increase in inequality (Gruber et al., 2014). With respect to Indonesia, almost only the elite are able to have tertiary schooling (Welch, 2007). As a result, an increase of the enrolment of tertiary education is probably only an increase of the enrolment of rich students, who are able to afford the college fees. Consequently, the income gap between rich and poor will become wider and this will have a negative effect on economic growth, according to Gruber et al. (2014). This theory would suggest that the government of Indonesia should first focus on child labour so that the making of secondary education compulsory would be more effective. Investments in the improvements of tertiary education should then be made. If tertiary education becomes accessible for students of all social classes, the long-term effect of this increase of the enrolment of tertiary education on economic growth will continue to be positive.
Appendix 1: Variable definitions (World Bank, n.d.)
Agriculture, value added (% of GDP). Agriculture corresponds to ISIC divisions 1-5 and includes forestry, hunting, and fishing, as well as cultivation of crops and livestock production. For an extensive description of value added, see appendix 1, variable value added (United Nations Statistics Division, 2008)
Health expenditure, public (% of government expenditure). Public health expenditure consists of recurrent and capital spending from government (central and local) budgets, external
borrowings and grants (including donations from international agencies and nongovernmental organisations), and social (or compulsory) health insurance funds.
Industry, value added (% of GDP). Industry corresponds to ISIC divisions 10-45 and includes
manufacturing (ISIC divisions 15-37) (United Nations Statistics Division, 2008). It comprises value added in mining, manufacturing (also reported as a separate subgroup), construction, electricity, water, and gas. For an extensive description of value added, see appendix 1, variable value added.
Labour force participation rate, total (% of total population ages 15-64). Labour force participation rate is the proportion of the population ages 15-64 that is economically active: all people who supply labour for the production of goods and services during a specified period.
Manufacturing, value added (% of GDP). Manufacturing refers to industries belonging to ISIC divisions 15-37 (United Nations Statistics Division, 2008). For an extensive description of value added, see appendix 1, variable value added.
Physical capital formation (current US$). Gross fixed capital formation (formerly gross domestic fixed investment) includes land improvements (fences, ditches, drains, and so on); plant, machinery, and equipment purchases; and the construction of roads, railways, and the like, including schools, offices, hospitals, private residential dwellings, and commercial and industrial buildings. According to the 1993 SNA, net acquisitions of valuables are also considered capital formation. Data are in current USD.
Poverty headcount ratio at $2 a day (PPP) (% of population). Population below $2 a day is the percentage of the population living on less than $2.00 a day at 2005 international prices. As a result of revisions in PPP exchange rates, poverty rates for individual countries cannot be compared with poverty rates reported in earlier editions.
Public spending on education, total (% of government expenditure). Public expenditure on education as % of total government expenditure is the total public education expenditure (current and capital) expressed as a percentage of total government expenditure for all sectors in a given
financial year. Public education expenditure includes government spending on educational institutions (both public and private), education administration, and subsidies for private entities (students/households and other privates entities).
Real gross domestic product per capita. GDP per capita is gross domestic product divided by midyear population. GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in constant 2005 USD.
School enrolment ratio, primary (% gross). Total is the total enrolment in primary education, regardless of age, expressed as a percentage of the population of official primary education age. GER can exceed 100% due to the inclusion of over-aged and under-aged students because of early or late school entrance and grade repetition.
