1. Introduction

Urbanization occurs in line with the process of structural economic transformation and development in various countries worldwide. Previous studies have closely investigated this phenomenon, with some focusing on the determinants of urbanization (Pandey, 1977; Firebaugh, 1979; Kentor, 1981; Kasarda & Crenshaw, 1991; Moomaw & Shatter, 1996; Lan & Wei, 2007; Ji & Zhang, 2020). Conversely, others examined the relationship between urbanization and economic development. For example, Chenery and Taylor (1968), Henderson (2003, 2010), and Hasan and Pitoyo (2017) stated that it is strongly correlated to the GDP level per capita. Therefore, managing urbanization is an essential part of maintaining economic growth (Collier & Venables, 2017; Zheng & Walsh, 2019; Khan, 2020).

The impact of urbanization on economic growth has been extensively studied in recent decades. Several previous studies such as Mills and Backer (1986), Moomaw and Shatter (1996), Bertinelli and Strobl (2007), Kolomak (2011), Lewis (2013), Tripathi and Mahey (2017), Chao and Xiaojie (2015), Zi (2017), Liang & Yang (2018), and Long (2020) stated that urbanization drives economic growth. Meanwhile, several others, namely Timberlake and Kentor (1983), Zheng and Walsh (2019), and Lewis (2014) stated that urbanization has a negative effect on economic growth. The World Bank (2016) reported that every 1% growth in urban population correlates with an increase in GDP per capita by 13%, 10%, and 7% in India, China, and Thailand, respectively. However, Indonesia realizes only 4% GDP growth for every 1% increase, based on McCoskey and Kao (1998), the impact of urbanization on growth varies in different countries.

According to Bloom et al. (2008), there is no evidence that urbanization rates affect economic growth. Chen et al. (2014) obtained similar results globally, while Narayan (2014) discovered some empirical evidence showing that the relationship between urbanization and development changes at the developmental stages. This means that urbanization varies between positive, negative, and neutral (Ahmad & Zao, 2018).

However, economic growth is not influenced by urbanization (Henderson, 2003; Turok & McGranahan, 2013), rather, it affects the growth process, through the increased flow of ideas and knowledge caused by agglomeration in cities (Narayan, 2014; Liang & Yang, 2018). Urbanization also increases investment in education (Bartenelli & Black, 2004), human capital accumulation (Bertinelli & Zou, 2008), economic, institutional, and cultural innovation (Nguyen & Nguyen, 2017), infrastructure and institutional conditions, including government policies (Pugh, 1995; Hope, 1998; Friedmann, 2006; Turok & McGranahan, 2013), and the accumulation of physical, knowledge, and human capitals (Liang & Yang, 2018), and decreases poverty rate (Nguyen et al., 2020).

Based on several fundamental arguments, this study investigates the relationship between urbanization and economic growth in Indonesia. First, Indonesia has the second-largest amount of urban land in East Asia, after China. Moreover, between 2000 and 2010, the country experienced an increase in urban land from relatively 8,900 square kilometers to 10,000, equivalent to 1.1% annually. This is the largest increase ever recorded after China (Word Bank, 2019). Second, in 2017, the country reached a medium urbanization rate of approximately 55%. This is higher than other developing countries in East Asia and the Pacific, such as the Philippines, Thailand, and Vietnam, although it is lesser in South American regions, for example, Brazil.

Third, the urbanization rate increased rapidly from 1970 to 2018, at an average of 2.5%. The maximum growth rate occurred in the 1980s and 1990s with an average of approximately 3.2%. In developing countries, particularly those in East Asia, the Pacific, and China, this condition was higher during the same period. The World Bank (2019) described this condition with mega-urbanization, which shows economic growth and prosperity opportunities. However, this description is not absolutely accurate because Indonesia benefits little or nothing from the increase in GDP per capita than the developing countries in East Asia, the Pacific, and China. The most appropriate term for this condition is “over urbanization” because an increase in the urbanization rate causes a decline in GDP per capita. Over urbanization describe cities whose rate of urbanization outpaces their industrial growth and economic development.

This research attempts to reaffirm the relationship between urbanization and economic growth. It is generally an accepted fact that economic growth promotes the expansion of modern industry and increases the urban population. Consequently, urbanization also drives economic growth to some extent. This shows that there is a causal relationship between urbanization and economic growth, which tends to occur in either one or two directions. Furthermore, Watson (2009) stated that urbanization played an irrelevant role in driving economic growth till the 21st century. Conversely, economic growth does not significantly increase the urbanization rate. He and Sim (2015) reported a one-way relationship in which growth drives urbanization. On the contrary, Turok and McGranahan (2013) stated that it is a one-way relationship in which urbanization drives economic growth. This study examines the aspect of the relationship which involves the causality that occurs between urbanization and unemployment.

This study also investigates the link between urbanization and education. Choy and Li (2017) stated that there is a causal relationship between them, where education affects the urbanization rate. This implies that one year of schooling increases urbanization by some percentage points. Moreover, according to Choy and Li (2017), education also plays an important role in the urbanization of villages to cities. It was empirically discovered that cities with a higher level of education attract students due to their infrastructure. Wu and Xie (2003) reported that economic growth and education are also related, with an increase in per capita income capable of affecting education. Based on this background description, the purpose of this study is to examine and analyze the causal relationship between urbanization, education, and economic growth, from short and long-term perspectives.

2. Literature Review

2.1. Urbanization and Per capita Income Growth

According to Zhao and Wang (2015), Megeri and Kengnal (2016), and Khan (2020), urbanization occupies a disorganized position in terms of development and growth. On the contrary, it is recognized as a fundamental multidimensional structural transformation experienced by low-income rural communities to be modernized and join the middle and high-income regions. Urbanization refers to the population shift from rural to urban areas, the decrease in the proportion of people living in rural areas, and the ways in which societies adapt to this change. Historically, the factors affecting the transformation of rural areas are explained by using the theory developed by Lee (1966), who stated that the movements of residents from rural areas to cities are influenced by factors related to the origin, and destination areas, the barriers between them, as well as personal determinants. In each region, some factors act to either reject or attract individuals to a particular area. Furthermore, factors that do not affect each of these areas, and everyone tends to feel less concerned about. These factors influence population, displacement, and an individual’s decision differently, for example, a good school system has a positive value for parents with school-age children. Unfortunately, it has a negative value for homeowners without schoolage children because of the cost of housing taxes incurred. Meanwhile, from the perspective of unmarried individuals without property, these factors do not influence their decisions to migrate to urban areas. Also, each of these factors possesses certain obstacles that affect decisionmaking, however, in some other instances, they are difficult to overcome. The most frequently studied obstacle is the distance, while the real or physical ones include the Berlin Wall or migration laws that restrict movement (Heiland, 2003). People are affected differently by the same set of obstacles, for example, certain factors, such as transportation cost and household goods, which are regarded as trivial by some individuals, are considered an obstacle by some others (Lin, 1982; Annez & Buckley, 2009).

Edwards (2007) stated that urbanization is an advantage closely related to urban areas that are inseparable from the industrial sector. It provides benefits that reflect Marshall’s externalities, namely, Access to Specialized Labor, Access to Special Resources, and Technological Abundance. Urbanization started in the industrial revolution era, which was marked by the increasing number of people living in urban areas. It also involves a series of changes from rural to urban areas in terms of industrial structure, employment, living conditions, and public services (Chen et al., 2014).

The relationship between urbanization and growth is traced to Arthur Lewis’s theory (The Lewis Theory of Development) (Lewis, 1984; Ranis, 2004). Based on this theory, the economy comprises traditional and modern economic sectors located in rural areas and cities and comprising industries. The traditional economy, irrespective of its excess population, is characterized by low labor productivity. This situation is interpreted by surplus labor, which allows a gradual shift to the modern economic sector with high labor productivity. Therefore, this model mainly focuses on the labor transfer process and the growth of output and employment in the modern sector. This Lewis theory constitutes the main developmental process in developing countries, and it is also applied when studying economic growth.

Previous studies carried out in ASEAN stated that urbanization positively impacts economic growth. The positive relationship between urbanization and per capita income is one of the most obvious and striking development facts. Generally, the more developed a country is in terms of per capita income, the greater the population living in urban areas (Todaro & Smith, 2008). According to Arouri et al. (2017), urbanization influence economic growth through several pathways. First, cities play an essential role in the economy or society. Second, urbanization causes agglomeration of both individuals and companies. Third, it is the crucial factor of entrepreneurship, where there are spillover effects or positive externalities from the development of urban and rural areas.

The 2018 Revised World Urbanization Prospects produced by the Population Division of the United Nations Department of Economic and Social Affairs (UN DESA, 2018) reported that urban transition and economic growth are linked in parts because both drive urbanization. Besides, people are usually attracted to cities that offer various opportunities for education and employment, especially in the industrial and service sectors. Urbanization has generally become a positive force for economic growth, unemployment, poverty, and human resource development.

Based on this review, urbanization and per capita income have different relationships in various countries. Spence et al. (2009) stated that this relationship is striking, and it offers some insight into the evolution of urbanization and per capita income over time. The United States experienced urbanization and per capita income simultaneously until 1940, when urbanization reached approximately 60%, while per capita income was much higher. In the initial stage, both variables increased at roughly the same rate, and this reflected a shift in resources from low-productivity rural activities. In the following phases, the rapid gains in productivity reflect a large aspect of the increase in industrial services. China is also following the United States path, with rapid growth in its per capita income at an urbanization rate of relatively half of the United States.

Subsequently, another previous study in Punjab (India) stated a positive correlation between urbanization and economic growth. Some cities in India, such as Punjab, have undergone a transformation from an agriculturalbased economy to the industrial service sectors. This is also evidenced by the increasing number of urbanization in Punjab. The study further stated that the potentials for urbanization through urban investment in infrastructure are essential to unlocking India’s potential for economic growth (Tripathi & Mahey, 2017). Explosive growth stems in part from urbanization in terms of economic restructuring and spatial agglomeration. However, other direct aspects play a more vital role in this evolution, such as reform and openness of policies, institutional transitions, or educational development (Chen et al., 2014). A more recent study carried out in China found that urbanization stimulates economic growth through physical knowledge and human capital (Liang & Yang, 2018).

2.2. Urbanization and Education

One of the driving factors for urbanization is education. This study focuses on the economic growth and aspects of education affecting urbanization. According to Todaro and Smith (2008), the migration of population movement from rural to urban areas exceeds job opportunities in these regions, thereby requiring an allotment (quota) to select new employees. Employers tend to use educational attainment or the number of years spent completing ones’ qualifications. They tend to hire more educated employees irrespective of the fact that additional education does not necessarily contribute to better performance outcomes.

The phenomenon where companies prefer properly educated employees certainly promotes migration from villages to cities. Therefore, education is a driving factor for individuals to live a better life. This is similar to one of the studies, which stated that sociologically, education functions to promote social mobility and certification. This social certification is a prerequisite for getting a job, and it tends to inspire people to obtain professional qualifications to achieve higher results in their workplaces (Ren, 2017). This research is in line with previous studies, which stated that urbanization and education positively affect labor employment (Choy & Li, 2017; Wu & Xie, 2003).

Some other research further stated that urbanization and education were positively correlated. Hofmann and Wan (2013) carried out research using the OLS (Ordinary Least Square) method, stated that industrialization and education have a positive and significant correlation on the urbanization rate. This is due to the occurrence of industrial agglomeration that leads to the movement of the labor market. Meanwhile, according to research carried out by Hofmann and Wan (2013), the positive causality effect of education on urbanization rate shows that on average, one year of schooling increases it by 2 percentage points. The impact of education on urbanization is due to the increasingly advanced science, and the effect of knowledge on industry becomes a determining factor for agglomeration, especially with advanced technology. The existence of a high-tech industry depicts an educated workforce, and the outcome complements each other as well as drives urbanization. Education also plays an important role in the urbanization of villages, besides, it was empirically discovered that cities with a higher level of education attract students that even intend to settle there due to infrastructural reasons (Choy & Li, 2016). Education and development are also related, especially the growth of per capita income. Similar research also stated that a higher income tends to affect the education levels (Wu & Xie, 2003). Similarly, Ren (2017) further stated that urbanization has an impact on education.

Based on the description of previous studies, it is evident that the most widely used methods in analyzing the relationship between urbanization and economic growth are VECM (Solarin & Shahbaz, 2013; Lewis, 2014; Zhao & Wang, 2015), GMM (Nguyen & Nguyen, 2017), FLOMS (Asif et al., 2015; Ali et al., 2020), OLS (Hofmann & Wan, 2013), VAR (Li, 2017), GIS (Chen et al., 2014), ARDL (Raggad, 2018), and Toda-Yamamoto methods (Saliminezhad & Bahramian, 2019), as well as 2SLS / 3SLS used by Liang and Yang (2018). This study applied the VECM method.

3. Research Methods and Materials

This study uses a quantitative approach with secondary data analyzed in the form of Time Series from 1990 to 2018. The research object was obtained from Indonesia. The data on GDP Per Capita (based on constant 2020 prices transformed into natural or lnycap logarithms) and urbanization (urbanization growth in percent or urban form), was obtained from World Bank, while the average duration of schooling in years is obtained from United Nations Development Program (Educ). The determination of Time Series analysis is based on the use of potentially related variables. Therefore, it is important to use a detailed analysis in accordance with the characteristics of the time series variables involved (Lutkepohl & Kratzig, 2004). The VECM model was used because it is also considered a VAR method with a cointegration stage (Zou, 2018). One of the advantages of using the VAR / VECM method is based on the fact that it is a simple model. There is no need to worry about determining both endogenous and exogenous variables. The forecasts generated using the multiple case methods were discovered to be better than those derived by the complex simultaneous equations (Gujarati & Porter, 1999).

3.1. Stationarity and Optimal Lag Test

The Augmented Dickey-Fuller (ADF) was used to test whether a given Time series is stationary or not. However, when the value of absolute t-statistic is smaller than the critical value in the MacKinnon table at various levels of confidence (1%, 5%, and 10%), it simply indicates that the data is not stationary. Furthermore, this is also evident when the value of probability is greater than 0.05, which means that the data is not stationary. Conversely, when the ADF value is greater than the critical value at various confidence levels (1%, 5%, and 10%), then there is a lack of unit root or stationary data. The ADF was formed to obtain the autoregressive equation, which was stated as follows:

\(\Delta Y_{t}=\beta_{1}+\beta_{2} t+\delta Y_{t-1}+\sum_{t=1}^{m} \alpha_{i} \Delta Y_{t-i}+\varepsilon_{t}\)       (1)

Where ε_t is error and \(\Delta Y_{t-1}=\left(\Delta Y_{t-1}-\Delta Y_{t-2}\right), \Delta Y_{t-2}=\left(\Delta Y_{t-2}-\Delta Y_{t-3}\right)\) is the lag difference that is often determined empirically.

Also, VAR estimation is sensitive to the lag length used. Moreover, when the lag in judicial stationarity is too small, the regression residuals are unable to show the process of white noise. Therefore, the models are unable to estimate the actual error accurately, thereby, making it impossible to estimate the standard. However, supposing the lag is deeply incorporated, it declines the ability to reject H0 as additional parameters, thereby reducing the degree of freedom. The testing lag also benefits from eliminating the autocorrelation problem, therefore, this issue is no longer expected to arise.

3.2. Cointegration Test

The existence of non-stationary variables causes the most likely long-term relationship between the variables. Therefore, it is important to carry out a cointegration test. This is aimed at determining the long and short-term relationships between variables. In this study, Johansen cointegration tests were applied to determine the existence of cointegration between the variables. The test developed by Johansen is also used to determine the cointegration number of variables (vectors). Johansen test is expressed with the autoregressive models as follows:

\(\Delta y_{t}=\Pi y_{t-1}+\sum_{i-1}^{p-1} \Gamma_{i} \Delta y_{t-1}+B \pi_{t}+\varepsilon_{t}\)       (2)

Where

\(\Pi=\sum_{t-1}^{p} A_{i}-I, \Gamma_{i}=-\sum_{j-i+1}^{p} A_{j}\)       (3)

3.3. Vector Error Correction Model (VECM) Granger Causality

VECM model is applied when the studied variables have been proven to have a cointegration relationship. Furthermore, assuming the cointegration test results show a long-term equilibrium relationship, the outcome of the relevant dynamic regression estimation is determined with the VECM model. It also serves as a method used to determine the effect caused by the variable in the long and short term. VECM is formulated as follows:

\(\begin{aligned} \Delta \operatorname{lnycap}_{t}=& \alpha_{1}+\sum_{i=1}^{k-1} \beta_{i} \Delta \operatorname{lnycap}_{t-i}+\sum_{j=1}^{k-1} \varphi_{j} \Delta \text { urban }_{t-j} \\ &+\sum_{m=1}^{k-1} \varphi_{m} \Delta \mathrm{edu}_{t-m}+\lambda_{1} \mathrm{ECT}_{t-1}+u_{1 t} \end{aligned}\)       (4)

\(\begin{aligned} \Delta \text { urban }_{t}=& \alpha_{2}+\sum_{i=1}^{k-1} \beta_{i} \Delta \ln \text { cap }_{t-i}+\sum_{j=1}^{k-1} \varphi_{j} \Delta \text { urban }_{t-j} \\ &+\sum_{m=1}^{k-1} \varphi_{m} \Delta \mathrm{edu}_{t-m}+\lambda_{2} \mathrm{ECT}_{t-1}+u_{2 t} \end{aligned}\)       (5)

\(\begin{aligned} \Delta \mathrm{edu}_{t}=& \alpha_{3}+\sum_{i=1}^{k-1} \beta_{i} \Delta \operatorname{lnycap}_{t-i}+\sum_{j=1}^{k-1} \varphi_{j} \Delta \mathrm{urban}_{t-j} \\ &+\sum_{m=1}^{k-1} \varphi_{m} \Delta \mathrm{edu}_{t-m}+\lambda_{3} \mathrm{ECT}_{t-1}+u_{3 t} \end{aligned}\)       (6)

4. Results and Discussion

4.1. Stationarity, Optimum Lag, and Cointegration Test

The data stationarity testing results in Table 1 were obtained using the Augmented Dickey-Fuller (ADF) method. Besides, Gross Domestic Product per Capita (lnycap), Urban Population (urban), and Education (edu) shows that data is not stationary at this level. However, after testing the stationarity at the First Different I (1) level, the three variables were declared stationary by examining the probability value of each variable, which were all discovered to be less than 5%, namely 0.0072, 0.0003, and 0.0025 for lnycap, urban, and edu, respectively.

Table 1: Stationarity Test and Determination of the Optimum Lag

Furthermore, optimum lag determination and stationary testing were carried out before the cointegration test (Suharsono et al., 2017). Based on Table 1, the results obtained at this stage stated that the optimum lag is determined at lag 4.

Meanwhile, the cointegration test results related to the Johansen procedure in Table 2 show cointegration between the variables (trace statistical value > critical value 5%). The trace statistic value shows that there is a significant cointegration rank at α = 5%, which is indicated by the asterisk sign (*), therefore it has a long-term balance between the variables.

Table 2: Cointegration Test (Trace Statixtoc)

4.2. Vector Error Correction Model (VECM)

Table 3 shows that the level of per capita income (lnycap) in the long term has a significant effect on the urbanization rates (urban).

Table 3: Long-Term and Short-Term VECM Results

These results were obtained by examining the t-statistic of the ycap variable, which is greater than the t-table value. Fortunately, an increase in the economic growth rate causes an increase in urbanization. In line with economic growth, education also has a significant impact on urbanization. Conversely, an increase in education increases the urbanization rate. Meanwhile, for the short-term estimation results, lags 3 and 4 have a significant effect on urbanization. On the contrary, lags 2 and 3 in respect to education also showed a significant effect. This indicates that in the third and fourth years, GDP per capita economic growth encourages urbanization. However, the effect of education on urbanization occurred in the second and third years.

4.3. Impulse Response and Variance Decomposition

This study’s Impulse response test focuses on the urbanization response to the shock caused by the per capita income and education. The Impulse Response graph shows that the horizontal axis is a period indicator (one period represents a year). The vertical axis on the graph shows the change due to the shock caused by a specific variable expressed in Standard Deviation (SD) units) (see Table 4).

Table 4: Long-Term and Short-Term VECM Results

Figure 1 as a whole shows the Impulse Response of urbanization to the shock of the fluctuating economic growth variable. It decreased from the first to the third period, and rose quite sharply, in the fourth period, and the trend returned afterward. Meanwhile, the response of urbanization towards the shock of the education variable also fluctuates. Moreover, the urbanization rate did not respond to shock from the education variable from the first to the third period. Furthermore, it rose till the fifth period, and the trend went down again afterward.

Figure 1: Response of Each Variable to Shock

Then, forecast error variance decomposition is a VAR model tool that separates the estimated variations of a shock component or an innovation variable, assuming they fail to correct each other. Variance decomposition provides information regarding the proportion or magnitude of the shock on a variable, both present and future periods. This study focuses on the shock contribution of the per capita GDP growth variables and education to urbanization. Variance decomposition results show that each variable in the first period did not contribute to the variance of changes in the urbanization rate. This was 100% illustrated by the urbanization rate, from the subsequent to the last periods, each variable started to contribute to the variance of changes. In the second period, each variable started to contribute shocks to the urbanization rate at a different level. The highest shock contribution to the economic growth variable’s urbanization rate was obtained in the fourth period. Besides, the education variable gave the highest shock contribution in the tenth period. The results are stated as follows.

4.4. Granger Causality Test

The Granger Causality testing was simultaneously carried out using the VECM model. Each equation contained in the VECM was tested using the Wald Chi-square.

Each endogenous variable is exchanged to exogenous to test the causality relationship. Furthermore, the Granger Causality for each variable was determined by analyzing the probability value in the significance level (5%). In accordance with the causality test carried out on the three variables, it was discovered that the one with the strongest causality was in the second model, which stated that urbanization was influenced by growth and education variables. These findings are consistent with the VECM test results, which also stated that economic growth and education tend to influence the urbanization rate.

The Granger Causality estimation results (see Table 5) stated that economic growth, which is proxied by GDP per capita growth, is a driving factor for the urbanization rate development in Indonesia. This means that it has a (+) value or an attraction in the push and pull urbanization theory. This relationship is in line with the research carried out by Todaro and Smith (2008), which stated that when there is a positive relationship between urbanization and GDP per capita growth, more people migrate to urban areas. The results of this research are consistent with previous studies carried out in various nations, such as Nguyen and Nguyen (2017) in ASEAN countries, Ali et al. (2020) in Nigeria, Li (2017) in China. Meanwhile, this study is also in with the research carried out by Lewis (2014) in Indonesia.

Table 5: VEC Granger Causality / Block Exogeneity Wald Test

The positive correlation between GDP per capita growth and urbanization rate occurs both in the long and short terms. According to World Bank (2019), urbanization is associated with a shift in employment from the agricultural sector, while the industrial service sectors that have higher total output do not experience transition. This has motivated the population to move to urban areas, especially metropolitan regions and an area that has a higher per capita income than rural nonmetro areas. These metropolitan areas offer stable jobs and higher wages in the industrial service sectors, which contributes to an enormous average income. Fortunately, urban areas also provide more facilities for health, education, and sanitation services.

This study explores the reasons behind the strong correlation between GDP per capita growth and urbanization in Indonesia. Firstly, the increasingly modern economic development has led to a shift from agriculture to industrialization and urbanization. This transition has led to an increase in the income of groups in the industrial service sector, which also plays a relevant role (Sulemana et al., 2019). It even promotes economic growth by 50% of GDP from all APEC member states.

Second, this is because people living in rural areas are poorly educated and are mostly relatively unskilled compared to those in urban regions. They are trapped in poverty because of limited economic opportunities and other obstacles encountered after moving from rural to urban areas. This population displacement causes several experiences. World Bank (2012) reported that in the case of Indonesia, urban areas are 1) population density of 5,000 people per square kilometer, 2) 25 percent or less of their households work in the agricultural sector, and 3) there are eight or more types of certain urban facilities, including primary, junior and senior high schools or its equivalents, cinemas, hospitals, maternity or mother-child hospitals, maintenance center spare parts equipment.

Third, the rapidly increasing urban population growth in Indonesian cities is inseparable from the urbanization element, which continues to increase. Java Island stills dominate the high urbanization rate. In general, cities’ growth on this island further shows a spatial pattern in the form of corridors, such as the Semarang-Jakarta corridor through Cirebon Semarang-Yogyakarta Surabaya-Malang. These cities largely contribute to the migration flows (Wilonoyudho et al. 2017). Although the pace of urbanization has slowed since the 1980s and 1990s, cities in Indonesia have been increasing in number at an extremely rapid rate. Currently, approximately 151 million people live in cities, and it is estimated that by 2045, exactly one hundred years of Indonesia’s independence, relatively 220 million people or more than 70% of the population is expected to migrate to cities. The urban context is quite a dilemma because it is divided into “suburbs” and “core” cities of the economy (Roberts et al., 2019).

Fourth, the positive relationship between urbanization rates and per capita income is partly due to differences between the economic structures of urban and rural areas. More urbanized regional economies rely on the industrial service sectors than in rural areas. However, approximately 70% of the total output in the “metro-core” is derived from this sector, which is more than double the share from rural areas (Roberts et al., 2019).

Big cities are generally more productive and attract other externalities such as the spatially unbalanced economic growth distribution (Nurlanova et al., 2018). In response to this, the government launched the Master Plan for the Acceleration and Expansion of Indonesian Economic Development (MP3EI), decorating economic centers, namely the Special Economic Zones (KEK), and Free Trade Zones (FTZ), which were developed to increase investment.

Based on the second result, education also influences the urbanization rate in Indonesia. In accordance with the push and pull theory of urbanization, Todaro and Smith (2008) stated that the migration of population movements that exceeds job opportunities in urban areas requires a quota in the new employees’ selection. Employers tend to use educational attainment or the number of years spent in realizing ones’ academic qualifications. Apart from being a selection criterion, educational qualification also shows the income level of a person. This is consistent with the studies carried out by Hofmann and Wan (2013) and Choy and Li (2016).

World Bank (2019) reported that access to education in Indonesia is still not evenly distributed, and in general, it is easily accessed by urban areas. In Realization of the Potential Urbanization in Indonesia, The World Bank (2019) cited an example using Jakarta, the capital and a megapolitan city where almost all the inhabitants have easy access to education. These factors influence a person’s decision to migrate from the village to the city. Higher and prospective education tends to attract rural residents to urbanize with the expectation of developing their productivity (Choy & Li, 2016). Economic growth in urban “megapolitan” areas based on the industrial service sector also leads to the educational sector’s growth. This is because the growth of the industrial sector tends to affect technical majors in higher education. Likewise, growth in the service sector also affects business and tourism majors (Li & Prescott, 2009), which women prefer (Cukier et al., 2017).

According to The World Bank (2012), the quality and extent of infrastructure play an essential role in helping cities and urban areas take advantage of agglomeration and urbanization’s economic opportunities. This is carried out using the public sector investment in infrastructure systems (roads, water supply, wastewater treatment, electricity, education, solid waste management, and other services). This system promotes the formation of industrial clusters by increasing the level of activities in various companies. Special Economic Zones (SEZs) or industrial estates are designed to attract complementary business developments that take advantage of the abundance of knowledge and reduce the costs of procuring intermediate inputs. There are apparent advantages to developing suitable educational services designed to train a skilled workforce to increase productivity by focusing on the required skills.

5. Conclusion

Based on the estimation and discussion, it was concluded that the effect of economic growth and education on urbanization has the strongest causality in the VECM model. This shows that GDP per capita and education are still attracting factors for urbanization in Indonesia. Meanwhile, economic growth and education in the VECM model significantly affect urbanization in the long and short terms. Based on these results, some suggestions concerning policy implications were stated, and they include (1) forming area-based urbanization, where cities within one area are integrated, to get the impact of an agglomeration economy (2) infrastructure and public facilities are proven to be one of the driving factors for urbanization in Indonesia. Therefore, the government needs to accelerate the distribution of infrastructure and public facilities in various regions to avoid population density in one area due to urbanization, (3) education has been proven to have a positive effect on urbanization. Therefore, the government needs to pay attention to easier access to education and more equitable ones in various regions, and (4) on the contrary, after education is evenly distributed in all regions, the government needs to pay attention to transportation access and infrastructure. This tends to greatly affect the mobility of the labor market in a region, as well as its productivity.

Meanwhile, based on research limitations, the following are suggestions made (1) the need to carry out similar research covering all major cities in Indonesia, (2) using more factors in the occurrence of urbanization, both social, economic, and availability of public infrastructure and facilities, as well as (3) using more accurate estimation methods, including ARDL, Toda-Yamamoto, etc.