Keywords

JEL Classification

9.1 Introduction

Studies on inequality in developing countries have actively been conducted by the World Bank, the Asian Development Bank (ADB), and the United Nations University-World Institute for Development Economics Research (UNU-WIDER); the studies published include that of the World Bank (2005), Ravallion (2016), ADB (2012), Zhuang (2010), and Kanbur and Venables (2005). Most of these studies analyzed inequality in developing economies, based on either cross-country data or aggregate data, which are easy to obtain and process.

However, there are only a limited number of studies on spatial inequality in a certain country—with a focus on location groups (e.g., urban and rural areas)—using time series and nationwide household survey data, which are not easy to obtain and process. Among them are Yusuf et al. (2014), Balisacan and Fuwa (2004), Cain et al. (2008), and Gustafsson et al. (2008).

As distinguished development economists such as Lewis (1954), Harris and Todaro (1970), and Kuznets (1955) pointed out, as the social and economic inequalities observed between urban and rural areas have been very large, it may be a critical challenge for developing countries to solve this problem, from the perspective of ensuring fairness across population groups. ADB (2012) emphasizes that the widening income disparity between urban and rural areas in Asian countries in recent years should be addressed directly. However, even though spatial inequality is such an important issue, this field has not been much studied until recently, due to types of data and/or decomposition methods.

Considering this gap in the research, the current study aims to investigate the spatial socio-economic inequalities in developing countries. This study selects Indonesia, the Philippines, and India as its research target countries, based on the following criteria: (1) these countries each have a large population size, which is helpful for the effective use of research results; (2) these countries have adequate spatial expanses, considering that this is suitable for a spatial analysis; and (3) there is the availability of household survey data, which makes the execution of the study possible. Using time series and nationwide household survey data of these three countries, the study focuses on inequality between urban and rural areas and examines the role of education in urban–rural disparity, with reference to my previous studies such as Hayashi et al. (2014, 2015), and Hayashi and Kalirajan (2018).

Indonesia, the Philippines, and India have achieved steady economic development, with an average annual growth rate—in real GDP per capita—of 3.8%, 2.8%, and 5.1%, respectively, for 2000–2010, despite the outbreak of the Lehman shock. Among Asian countries, Indonesia, the Philippines, and India are large countries, with their populations being approximately 240 million, 94 million, and 1.2 billion, respectively, in 2010.

This study attempts to clarify how much influence education has had on the urban–rural gap in consumption expenditure in each of the above three countries. Specifically, the Blinder–Oaxaca decomposition method is adopted to analyze the impact of various household characteristics—such as the number of family members, gender, age, education, and production sector—on inequality between urban and rural areas, in terms of per capita household consumption expenditure, using nationwide household survey data in the three countries at two-time points.

The remainder of this study is organized as follows. Section 9.2 describes the data and methods used in this study, while Sect. 9.3 gives an overview of urban–rural disparity in Indonesia, the Philippines, and Indonesia. In Sect. 9.4, the role of education in urban–rural inequality is analyzed, using the Blinder–Oaxaca decomposition method. Finally, Sect. 9.5 presents the main findings, along with some policy implications.

9.2 Data and Method

This study explored the determinants of the urban–rural disparity in per capita consumption expenditure in Indonesia, the Philippines, and India, respectively; then, the determinants in those three countries were compared. To accomplish these aims, the study used household survey data of each country, as indicated in Table 9.1, decomposed consumption expenditure data into several determinants, and then compared these determinants.

Table 9.1 Household survey data in Indonesia, The Philippines, and India
Full size table

Considering Indonesia, this study used the Susenas (the Survei Sosial Ekonomi Nasional or the National Socio–Economic Survey) panel data on expenditure in 2008 and 2010, compiled by the Central Bureau of Statistics in Indonesia (BPS). The Susenas panel dataset includes approximately 61,000 households, of which 23,700 are in urban areas and 37,300 are in rural areas. The share of urban households—estimated using sampling weights—was around 47% in 2008, a proportion that remains constant during the study period.Footnote 1 To adjust for spatial differences in prices at a point in time and spatial differences in inflation rates, this study converted current price expenditures into expenditures at 2008 constant prices by using current price provincial urban and rural poverty lines in 2008 and 2010.Footnote 2

With regard to the Philippines, we used household data from the Family Income and Expenditure Survey (FIES) in 1997 and 2006, compiled by Philippine Statistics Authority (PSA).Footnote 3 The Family Income and Expenditure Survey datasets in those two years include around 38,500–39,600 households, of which 45–60% are in urban areas and 40–55% are in rural areas. To analyze inequality changes in real terms, the study converted current price expenditures into expenditures at 2000 constant prices by using the 2000 consumer price index (CPI) calculated by the PSA.

In the case of India, our study used household consumption expenditure data from the National Sample Survey (NSS), collected and compiled by the National Sample Survey Office (NSSO) under the Indian Ministry of Statistics and Programme Implementation.Footnote 4 Specifically, the study used the NSS 55th Round Survey (July 1999–June 2000) and 68th Round Survey (July 2011–June 2012). The NSO datasets in those two rounds include around 102,000 to 120,000 households, of which approximately 40% are in urban areas and 60% are in rural areas. The shares of urban households estimated using sampling weights are roughly 27% in 1999/2000 and 31% in 2011/12.

To adjust for spatial differences in prices at a point in time and spatial differences in inflation rates, the study converted current price expenditures into constant price expenditures by using current price provincial urban and rural poverty lines in 1999/2000 and 2011/12. Based on the poverty line in the urban area of Delhi (National Capital Territory) in 2011/12, per capita household expenditure at 2011/12 constant prices with consideration of spatial differences was calculated and utilized.Footnote 5

To explore the determinants of the urban–rural disparity in mean per capita consumption expenditure, this study used household survey data for the three countries and performed a Blinder–Oaxaca decomposition analysis, which was popularized by Blinder (1973) and Oaxaca (1973).Footnote 6

In the equation below, let YU and YR be the natural log of per capita expenditure of urban and rural households, respectively. Given the linear regression model,

$$Y_{k} = {{\varvec{X}}_{k}}^{\prime } {\varvec{\beta}}_{k} + e_{k} \quad E(e_{k} ) = 0\quad k = U,R$$

where Xk is a vector of explanatory variables, \({\varvec{\beta}}_{k}\) includes the parameters associated with \({\varvec{X}}_{k}\), and \(e_{k}\) is the error term, which contains unobserved factors. Moreover, we let \(\hat{\user2{\beta }}_{k}\) be a vector of the least-squares estimates for \({\varvec{\beta}}_{k}\) (\(k = U,R\)), obtained separately from the urban and rural samples, and \(\overline{{\varvec{X}}}_{k}\) be the estimate for \(E({\varvec{X}}_{k} {)}\). Then, the estimated urban–rural difference in mean per capita expenditure is expressed as a twofold decomposition:

$$\widehat{D} = \overline{Y}_{U} - \overline{Y}_{R} = (\overline{{\varvec{X}}}_{U} - \overline{{\varvec{X}}}_{R} {)}^{\prime } \hat{\user2{\beta }}^{*} + \left( {{\overline{{\varvec{X}}}_{U}}^{\prime } (\hat{\user2{\beta }}_{U} - \hat{\user2{\beta }}^{*} ) + {\overline{{\varvec{X}}}_{R}}^{\prime } (\hat{\user2{\beta }}^{*} - \hat{\user2{\beta }}_{R} )} \right)$$

where \(\hat{\user2{\beta }}^{*}\) is a vector of the least-squares estimates for the slope parameters and the intercept, obtained from the pooled sample of urban and rural households (Neumark, 1988). The first term in the above equation is the part of the urban–rural difference in mean per capita expenditure that is explained by urban–rural differences in the explanatory variables (endowments or quantity effect), and the second term is the unexplained part. Based on the above equation, we decomposed the differences in mean per capita household expenditure between urban and rural areas into several components, including educational differences.

9.3 Urban–Rural Disparity in Indonesia, the Philippines, and India: Overview

Tables 9.2, 9.4, and 9.6 present figures on mean monthly per capita household expenditure and the shares of population and expenditure in Indonesia, the Philippines, and India, respectively. The data are presented in terms of urban and rural areas and educational attainment and at two-time points. Tables 9.3, 9.5, and 9.7 report the results of urban–rural decomposition of per capita expenditure inequality by using the Theil T, in each of the three countries and at two-time points.

Table 9.2 Mean per capita household expenditure and shares of population and expenditure by groups: Indonesia
Full size table
Table 9.3 Inequality decomposition by urban and rural areas: Indonesia
Full size table
Table 9.4 Mean per capita household expenditure and shares of population and expenditure by groups: The Philippines
Full size table
Table 9.5 Inequality decomposition by urban and rural areas: The Philippines
Full size table
Table 9.6 Mean per capita household expenditure and shares of population and expenditure by groups: India
Full size table
Table 9.7 Inequality decomposition by urban and rural areas: India
Full size table

Table 9.2 indicates that, in Indonesia, the mean expenditure per capita in urban areas is 1.7 times higher than in rural areas. While the majority of the country’s population resides in rural areas, urban areas occupy a larger share of expenditure, around 60%. In both urban and rural areas, the mean per capita household expenditure increases as the level of education of the household head rises. However, there is a striking contrast between urban and rural areas in terms of educational attainment levels. The proportion of households of which household heads have received higher education is considerably larger in urban areas than in rural areas. More than 40% of rural households are headed by individuals with no schooling or incomplete primary school education.

Table 9.3 presents the results of the urban–rural decomposition of per capita expenditure inequality in Indonesia. These figures indicate an upward trend in inequality in Indonesia as a whole over the period. Inequality within urban and rural areas is more significant than inequality between urban and rural areas, in terms of both value and contribution, as measured by the Theil T. Urban areas have significantly higher within-group inequality than rural areas.

Clearly, between-area inequality is unremarkable, in comparison with within-area inequality. However, an alternative Theil index decomposition proposed by Elbers et al. (Between-area [B′] in Table 9.3) indicates that observed inequality between the two areas accounts for more than 26% of the maximum attainable between-area inequality, given the current distribution of per capita household expenditures, the relative sizes of urban and rural areas, and their ranking in terms of mean per capita expenditure.Footnote 7 This implies that the contribution of urban–rural inequality may be larger than we thought it was when using the conventional Theil decomposition approach.

In the Philippines, as shown in Table 9.4, urban areas have higher mean household expenditure per capita than rural areas, in both 1997 and 2006. These numbers suggest a reduction in urban–rural inequality in mean expenditure per capita during the ten-year period; however, the mean expenditure per capita in urban areas is more than twice as large as in rural areas in 2006. While more than half of the country’s population resides in rural areas, urban areas account for around two-thirds of consumption expenditure. Mean expenditure per capita increases monotonically with the education attainment of the household heads in both urban and rural areas. However, the educational attainment level of household heads is very different between urban and rural areas. The proportion of households of which the household heads have received higher education is significantly larger in urban areas than in rural areas.

In Table 9.5, it is shown that, over the study period, inequality in terms of mean expenditure per capita in the Philippines has improved, as a whole; however, it still remained at a high level. Inequality within urban and rural areas is more significant than inequality between urban and rural areas, judging from both value and contribution measured by the Theil T. Looking within urban and rural areas, urban inequality is considerably higher than rural inequality.

Urban–rural inequality accounts for slightly <20% of overall inequality. When using the alternative Theil index decomposition method, inequality between urban and rural areas can explain around 40% of overall inequality. This suggests that urban–rural inequality in the Philippines is severe.

Table 9.6 illustrates the mean per capita household expenditures and the shares of population and expenditure in India in 1999/2000 and 2011/12. Urban areas have higher mean expenditure per capita than rural areas. The shares of population and expenditure are larger in rural areas than in urban areas. Although the center of the country—in terms of the size and share of population—is still in rural areas, the shares of population and consumption expenditure have been shifting from rural to urban areas over the period. This implies the advancement of dynamic urbanization and the expansion of inequality between urban and rural areas in India.

The mean per capita household expenditure increases monotonically with the education of the household heads in both urban and rural areas. However, the mean per capita household expenditure is higher in urban areas than in rural areas, even at the same educational attainment level; this tendency can be clearly seen in groups of higher-educated people. Furthermore, the proportion of households in which household heads have received higher education is considerably larger in urban areas than in rural areas.

Table 9.7 presents the results of the urban–rural decomposition of per capita expenditure inequality in India in 1999/2000 and 2011/12. These figures show an upward trend of inequality in India as a whole during the period. In India, similar to the other two countries, inequality within urban and rural areas is more salient than inequality between urban and rural areas, in terms of both value and contribution, as measured by the Theil T. Moreover, urban areas have higher within-group inequality than rural areas.

During the observation period, the value of between-area inequality rose to 0.035, and its contribution to total inequality increased to 14%, as measured by the Theil T. In addition, an alternative decomposition measure indicates that inequality between urban and rural areas accounts for nearly a quarter of the maximum attainable between-area inequality in 2011/12. India should therefore address inequality between urban and rural areas urgently.

As described earlier, urban areas have higher mean household expenditures per capita than rural areas in all three countries. While the centers of the three countries in terms of population are still on rural areas, population and consumption expenditure have been flowing into urban areas.

Common in these three countries, in both urban and rural areas, mean per capita household expenditure increased as the level of education attained by household heads rose. However, mean per capita household expenditure was shown to be higher in urban areas than in rural areas, even in the same educational attainment level; this tendency becomes clear in the groups of higher-educated people. Additionally, the proportion of households in which household heads had received higher education is larger in urban areas than in rural areas.

As measured by the Theil T, in Indonesia and India, the overall inequality in terms of per capita household expenditure increased between two-time points. In the Philippines, inequality in mean expenditure per capita was slightly reduced, although it still remained at a high level. When using the conventional Theil index decomposition method, the share of inequality between urban and rural areas was shown to be relatively lower than that of inequality within urban and rural areas. However, when using the alternative Theil index decomposition, the share of inequality between urban and rural areas increased substantially in Indonesia, the Philippines, and India.

This suggests that the gaps between urban and rural areas are not necessarily small enough for their impact to be ignored. Our study focuses on and examines the role of educational differences in urban–rural inequality in these three countries.

9.4 Accounting for Urban–Rural Disparity in Indonesia, the Philippines, and India: The Blinder–Oaxaca Decomposition Method

The preceding section provides an overview of the urban–rural inequality in consumption expenditure in each of the three countries. Previous studies on inequality in Asian economies point out that household income or expenditure disparities are generated by unequal access to education.Footnote 8 The current section thus analyzes the degree of impact of various household characteristics, including differences in educational attainment on urban–rural inequality in per capita consumption expenditure, using the Blinder–Oaxaca decomposition method.

Specifically, we decomposed differences in mean per capita household expenditure between urban and rural areas into the following common components of household features, as the determinants of the urban–rural inequality in each of the three countries:

  1. (i)

    household size;

  2. (ii)

    gender of household head (female = 0; male = 1);

  3. (iii)

    age of household head;

  4. (iv)

    squared age of household head;

  5. (v)

    years of education of household head; and

  6. (vi)

    job sector of household head (agriculture/mining = 0; non-agriculture/mining = 1).

Note that variable (v), the number of years of education, is calculated according to the following, in each of the three countries.

In Indonesia, household heads’ years of education are calculated as (1) no schooling (0 years); (2) incomplete primary school (3 years); (3) general and Islamic primary schools (6 years); (4) general and Islamic junior high schools (9 years); (5) general, Islamic, and vocational senior high schools (12 years); (6) diploma I and II (13 years); (7) diploma III (15 years); (8) diploma IV (bachelor’s degree) (16 years); and (9) master’s or doctoral degree (18 years).

In case of the Philippines, the length of education is calculated as (1) no schooling (0 years); (2) incomplete elementary education (3 years); (3) elementary education (6 years); (4) incomplete secondary education (8 years); (5) secondary education (10 years); (6) incomplete tertiary education (12 years); and (7) tertiary education including postgraduate education (14 years).

As for India, based on Cain et al. (2008), the number of years of education is calculated in the following way: (1) illiterate (0 years); (2) literate through non-formal schooling (i.e., NFEC [Non-formal Education Courses], ALC [Adult Literacy Centers], EGS [Education Guarantee Scheme], TLC [Total Literacy Campaign], and [other]) (1 year); (3) literate, but incomplete primary education (3 years); (4) primary education (5 years); (5) middle schools/lower secondary education (8 years); (6) secondary education (10 years); (7) higher secondary education (12 years); (8) diploma/certificate courses (12 years); (9) undergraduate education (15 years); and (10) postgraduate education (17 years). For details on the education system in India, refer to National Sample Survey Office (2015).

Tables 9.8, 9.9, and 9.10 indicate the results of the Blinder–Oaxaca decomposition of urban–rural differences in mean per capita expenditure in Indonesia, the Philippines, and India, respectively, at two-time points.

Table 9.8 Blinder–Oaxaca decomposition of urban–rural differences in Mean per capita household consumption expenditure: Indonesiaa
Full size table
Table 9.9 Blinder–oaxaca decomposition of urban–rural differences in Mean per capita household consumption expenditure: The Philippinesa
Full size table
Table 9.10 Blinder–Oaxaca decomposition of urban–rural differences in Mean per capita household consumption expenditure: Indiaa
Full size table

Table 9.8 shows that, in Indonesia, the mean of the natural log of per capita expenditure in 2008 is 12.973 for urban households and 12.482 for rural households, yielding an urban–rural expenditure gap of 0.492. The same figures for 2010 are almost at the same level. The Blinder–Oaxaca decomposition method can divide this expenditure gap into two parts. The first part—that is, the explained part (endowments or quantity effect)—reflects the increase in mean per capita expenditure if rural households had the same endowments as urban households, assuming that urban and rural households have the same coefficients, obtained from the pooled sample of urban and rural households. The second part is a residual or unexplained part that captures all potential effects of differences in unobserved variables. In the table, the increases of 0.226 in 2008 and 0.239 in 2010 indicate that differences in endowments (household size, gender, age, education, and job sector) as a whole account for 46% and 48%, respectively, of the urban–rural expenditure gap.Footnote 9

In the explained part, the components related to household size, gender, and age have only a marginal effect. The most significant component is education, followed by the job sector component. Educational attainment differences measured by the length of education of the household head are the largest contributor to differences in mean per capita expenditure between urban and rural areas. This education component accounts for approximately 36% of the urban–rural expenditure gap. Differences in the job sector also explain 13–14% of the gap. Non-agricultural jobs or off-farm opportunities in rural areas would have an effect on reducing the gap between urban and rural areas. This result suggests that urban–rural disparity is largely associated with educational attainments and job sectors.

Table 9.9 exhibits the results of the decomposition of urban–rural differences in mean per capita household expenditure in the Philippines. The mean of the natural log of per capita expenditure in 1997 is 9.896 for urban households and 9.278 for rural households, yielding an urban–rural expenditure gap of 0.617. These levels in 2006 are not much different from those in 1997. The increases of 0.341 in 1997 and 0.322 in 2006 show that differences in endowments (household size, gender, age, education, and job sector) as a whole account for 55% and 48%, respectively, of the urban–rural expenditure gap.

In the Philippines, similar to Indonesia, the components associated with household size, gender, and age do not play a prominent role in the explained part. Most noticeable is the education attainment component, followed by the job sector component. Educational differences have the largest influence on differences in mean per capita household expenditure between urban and rural areas. This education attainment component accounts for approximately 30–33% of the urban–rural expenditure gap. Differences in the job sector also explain around 16–21% of the gap. This result implies that educational attainments and job sectors have a large impact on urban–rural disparity in the Philippines.

According to Table 9.10, the mean of the natural log of per capita expenditure in India in 1999/2000 is 7.535 for urban households and 7.203 for rural households; these yield an urban–rural expenditure gap of 0.332. Likewise, those in 2011/12 are 7.748 for urban households and 7.508 for rural households, yielding an urban–rural expenditure gap of 0.240. The increases of 0.262 in 1999/2000 and 0.164 in 2011/12 demonstrate that differences in endowments (household size, gender, age, education, and job sector) as a whole account for around 79% and 68%, respectively, of the urban–rural expenditure gap.

In India, while the gender and age components do not play a major role in the explained part, the education component is the most influential. Similar to the other two countries, educational differences are the primary factor of differences in mean per capita expenditure between urban and rural areas. This education component accounts for approximately 54–59% of the urban–rural expenditure gap. Next to the education component, the household size and job sector component have a large effect on urban–rural inequality in India. However, the contribution of the job sector to the urban–rural expenditure gap decreases markedly in 2011/12. This change could also be explained by a recent increase in non-agricultural jobs and off-farm business opportunities in rural areas.

The results of the analyses of the three countries using the Blinder–Oaxaca decomposition method suggest that the household components of educational attainments and job sectors make a large contribution to differences in per capita consumption expenditure between urban and rural areas.

9.5 Conclusions

This study selects Indonesia, the Philippines, and India among Asian developing countries and, based on household survey data, examines the determinants of urban–rural disparities in per capita consumption expenditure in these three countries, with a focus on education, using the Blinder–Oaxaca decomposition method. The results of the analysis and the implications drawn from them are summarized as follows.

Indonesia, the Philippines, and India have achieved steady economic development, with an average annual growth rate in real GDP per capita of 3.8%, 2.8%, and 5.1%, respectively, between 2000 and 2010. Among Asian countries, Indonesia, the Philippines, and India are large countries, with populations in 2010 of approximately 240 million, 94 million, and 1.2 billion, respectively.

In all three countries, urban areas have higher mean per capita consumption expenditure than rural areas. Moreover, the shares of population and consumption expenditure tend to shift from rural to urban areas over the period. Common in all three countries, the mean per capita household consumption expenditure increased monotonically with the education attainment of the household heads in both urban and rural areas. However, the proportion of households of which household heads had received higher education is considerably larger in urban areas than in rural areas.

In both Indonesia and India, inequality in per capita consumption expenditure, as measured by the Theil index, tended to expand during the observation period. In the Philippines, inequality in per capita expenditure improved over the period, although the level of inequality still remains high. The share of inequality between urban and rural areas is relatively lower than that of inequality within urban and rural areas, due to the use of the conventional Theil decomposition method. However, the gaps between urban and rural areas are not small enough for their impact to be ignored. Furthermore, when using Elbers’ alternative decomposition approach as a supplementary tool for the conventional Theil decomposition method, the share of inequality between urban and rural areas increased substantially, in all three countries.

This study therefore attempts to decompose the differences in mean per capita consumption expenditure between urban and rural areas into several household features, including education, using the Blinder–Oaxaca decomposition method. As a result, in Indonesia, the Philippines, and India, differences in educational endowments appear to have been a key determinant of urban–rural disparity, accounting for approximately 30–60% of the urban–rural expenditure gap.

In addition, differences in job sectors (agricultural sector vs. non-agricultural sector) also contribute to the expenditure gap, albeit to a lesser extent. As indicated in Tables 9.2, 9.4, and 9.6, in these three countries, while the proportion of households of which the household heads have completed tertiary education is about 10–20% in urban areas, it is merely around 2–5% in rural areas. In rural areas, the share of the population engaged in agriculture—an industry with low productivity and high risks—is large. It can be assumed that the differences between urban and rural areas in terms of access to education and employment opportunities in industries with high productivity and value added would become a major factor, causing the urban–rural disparity in household consumption expenditure.

As countermeasures, it seems that the expansion of education in quantity, improvement in agricultural productivity, and the creation of employment opportunities in non-agricultural sector could contribute to the reduction of urban–rural disparity in per capita consumption expenditure in Indonesia, the Philippines, and India. An important issue would be whether these three countries, in which more than 40% of household heads in rural areas have not received primary education and agriculture sector with low productivity is a key industry, can expand the provision of education services to people in rural areas, create educational opportunities for them through social policy tools—such as a conditional cash transfer program—and enhance their income-earning capacity in either the agricultural or non-agricultural sectors.