KADARMANTO & DENZO KAMIYA

Transcription

1 PATTERN OF INCOME INEQUALITY REFLECTED IN EXPENDITURE DISTRIBUTION IN INDONESIA : AN EMPIRICAL ANALYSIS BASED ON THE MICRO-DATA SET OF SUSENAS KADARMANTO & DENZO KAMIYA Abstract. As one of the developing countries, Indonesia has been experiencing a high economic growth rate since 1970s. Economic crisis, however, has reversed the country from an economy of positive growth into one of negative growth. Some researchers have studied income distribution in Indonesia especially during rapid economic growth period in the 1970s. However, during economic crisis period, there were only few studies about income distribution in Indonesia. The purpose of this research is to carry out an analysis of income inequality approached by consumption-expenditure per capita, trying to identify how much inequality there is. Based on three-year-span micro data set drawn from Susenas, indicators of per capita expenditure level dispersion among provinces are presented. It determines the pattern of income inequality, explores inter-provincial variation in the degree of inequality, examines the sources of income inequality by means of disaggregating the inequality indices. The Gini coefficient is a tool selected for inequality measurement in this paper. In order to conceptualizing and measuring inequality, two essential graphic analytical tools such as Lorenz curve and Cumulative frequency distribution are presented here. For the sake of comparability, Atkinson and Varian of Logarithm as well as Theil indices are also chosen. Using the decomposition of Theil index as guidance, this research empirically analyzes Indonesia s inequality from a rural-urban decomposition perspective as well as a provincial perspective to assess the contribution of different population subgroups to overall inequality over the period The main findings of the study are, first, overall income inequality has been increasing both before and after economic crisis periods. The lowest level of inequality was found in 1999 whereas the most unequal income inequality was experienced in Second, urban inequality was higher not only than rural inequality but also than overall inequality. Third, the population changes or urbanization positively affected inequality to increase. Key words and phrases. Decomposition, Indonesia, Provincial, Inequality. 1. Introduction As one of the developing countries, Indonesia experienced high economic growth in the two decades of 1970s and 1980s. It was often said that Indonesia was one of the highest performing economies in Asia. In line with this, problems related to income distribution attracted much attention in this country. How is the gain from economic growth distributed among individuals? The economic crisis of the late 1997, however, turned Indonesia from an economy of positive growth into one of negative growth. Then, a second question arose: did the crisis increase the inequality of income distribution? This country still consists largely of rural area, and is in the process of urbanization, which changes sectoral income differentials. This process of urbanization may deteriorate the nation s income distribution, It is for this reason that we decompose the country into urban and rural areas to analyze changes in income distribution. The Inverted-U Hypothesis of Kuznets may be useful to understand Indonesian experiences. In addition, Indonesia is a big country, consisting of many provinces that are so diverse in resource endowments and economic opportunities. Hence, a third question: to what extent differences among provinces explain the whole picture of inequality in Indonesia? The purpose of this paper is to give a broad picture of distribution of income and level of inequality in Indonesia before and after the economic crisis of 1997, and propose some hypotheses explaining the statistical findings. More specifically, we analyze the inequality of Indonesia using techniques of modern statistical analysis, in order to assess contributions of different population subgroups to the overall inequality, to determine the pattern of income inequality at the national level and provincial level, and to show the Graduate School of Economics, Tokyo International University, Matoba-Kita , Kawagoe, Saitama , Japan. Address: kadarmanto bps@yahoo.com; kadarman@mailhost.bps.go.id Professor of Economics, Graduate School of Economics, Tokyo International University, Matoba-Kita , Kawagoe, Saitama , Japan. Address: dikamiya@tiu.ac.jp 1

2 2 KADARMANTO - DENZO KAMIYA pattern of changes in inequality both for provinces and for the nation as a whole. By doing so, we propose a tentative conclusion that trend of income distribution in Indonesia is explained by inequaliy in urban area and population movement from rural to urban area. Most of the previous works 1 on income distribution relied on the published results of the Susenas that usually show the mean expenditure per income bracket only. It is believed that inequality measures, calculated on such data, turn out to be lower than those calculated on the related raw data. Part of this research, therefore, is devoted to revision of figures indicating inequality, using the raw micro data. Moreover, many of the previous works cover, at most, three years. The present research of ours extend the period of observation to twelve years, utilizing three year span data, from 1990 to In this way, we have a clearer view on the state of income distribution in Indonesia, particularly, changes in inequality of income distribution. Setting period of observations from 1990 to 2002, we come across one problem that deserves particular attention. That is, East Timor no longer belongs to this country in It is for this reason that we recalculated inequality indices for the whole period excluding East Timor. This exclusion accounts partly for the differences between earlier figures and ours. The rest of this paper is divided into four sections. Section 2 explains the nature of the data used in this analysis. After presenting formulae defining alternative measures of inequality and their decomposition in Section 3, we report, in Section 4, on the statistical findings in detail, with some hypotheses explaining these results. Finally in Section 5, we summaries the results on the basis of the main findings revealed by dividing Indonesia into urban and rural areas. 2. Data Using household as the unit of analysis, this research is done on the basis of the micro-data set drawn from the household survey undertaken by BPS-Statistics Indonesia. The series of national socioeconomic surveys, known by the Indonesian acronym Susenas were the most frequently used source for nationwide studies of income distribution relying on household consumption expenditure data since this survey is the only source of information on household expenditures covering the whole of Indonesia. Consumption expenditure data is used as a proxy of measuring society s income distribution, even though it is acknowledged that such a way in fact has a serious demerit.the use of consumption expenditure data can underestimate income. The reason is simple, total of a person consumption expenditure must not be equal to total of a person income. For example, households higher income does not always mean that their consumption expenditure is also high. To the extent that richer households saved more in addition to spending more, it may have distorted income distribution. However, because income data collection in Indonesia is still relatively difficult, so the use of household consumption expenditure data is still regarded as an alternative with the reason that as we know income consists of two kind of income; namely, (a) income in cash and (b) income in kind, this survey in collecting the consumption expenditure data also takes the original source of expenditure into consideration, such as (a) purchase or (b) self-produce especially in food-consumption items. The BPS-Statistics Indonesia has been carrying out this survey since the first time in 1963 which covered only some parts of Indonesian territory. Before 1980, however, Susenas was undertaken irregularly; for example, the second, third, forth until seventh Susenas were in 1964/65, 1967, 1969/70, 1976, 1978, and 1979 successively. But since 1980, BPS has been managing Susenas regularly every year, except in 1983 and Since 1981, the questionnaire of the survey has been divided into three different modules 2 which reappear every three years along with the core questionnaire. Hence, every year Susenas adopts core module combinations which is one of these combinations: core + module 1, core + module 2, or core + module 3. The 1963 Susenas covered only Java island composed of five provinces at that time and the number of household sample was around 16,000. Nevertheless, the number of sample size gradually increased. The purpose of increasing sample size is that the sample might be able to represent all parts of the country. With a number of households in the sample (49,000 in 1990; 65,600 in 1993; 65,664 in 1996; 49,248 in 1999; and 62,720 in 2002), the survey is capable of obtaining both national and provincial 1 For example, Akita (1998); Akita and Lukman (1995,1999); Booth (1992); Booth and Sundrum (1981); Ginneken (1976); Hasibuan (1978); Hughes and Islam (1981); Islam and Khan (1986); King and Weldon (1977); Sugito and Ezaki (1989); Sundrum (1979), and Uppal and Handoko (1986). 2 In order to collect more detailed household information, There are three Susenas modules such as household consumption expenditures as the first module; welfare, socio-culture, criminality, and tourism as the second module; and health, nutrition,education, and living condition as the third module.

3 PATTERN OF INCOME INEQUALITY IN INDONESIA, level estimates. The sample is drawn separately for urban and rural areas; therefore, independent estimates for those areas are also obtainable. The Susenas adopts multistage sampling design in selecting household sample. Provinces, regency/municipality, and urban/rural can be regarded as strata representing geographic location. Therefore, the sampling design is applied for urban/rural in each province. The first stage is usually conducted by BPS selecting primary sampling unit from sampling frame. Because of heterogeneous and often too large to handle efficiency, BPS has broken down every village into a smaller area unit called enumeration area during the 1990s and census block during the 2000s as a sampling frame. An enumeration area (EA) which should be identifiable by its natural or man-made boundaries and would presumably unchange in a ten-year period consisting of about households or unoccupied physical building or the combination of the two is generally assigned to one enumerator. The Susenas sampling frame has always been prepared as part of activities of the most recent population census, e.g., the sampling frame for the 1990 s Susenas originated from the 1990 Population Census and for the 2000 s Susenas came from the 2000 Population Census. First of all, in selecting sample, the list of enumeration area is classified into strata. Then several enumeration areas are chosen from each stratum. Based on the result of household listing in the field, BPS-Regional Offices then draw household samples in order to be interviewed using Core and/or Module questionnaire of Susenas. Susenas questionnaire consumption module questionnaire covers two parts of expenditure items; namely, food and non-food group. Food category is comprised from about 200 items whereas a total number of about 100 item is included in the non-food group. They are 205, 203, 216, 214, and 216 for food items and 97, 103, 103, 105, and 105 for non-food items in 1990, 1993, 1996, 1999, and 2002 questionnaire, respectively. In the case of food consumption, Susenas provides information on the original source of consumption. That is, it records whether a consumed foodstuff is initially market purchased, own-produced or received as a gift. It was collected information on quantity and money value for items on food category whereas on non-food category items, information gathered was related only to money value. 3. Methodology 3.1. Inequality Measures. Income 3 inequality measures for individuals are examined by adjusting the total consumption-expenditure of a household with the number of its member (size). Hence, the living standards among the households are the same. If the total consumption expenditure is not adjusted by the household size, then if two households have the same level of consumption-expenditure but one household has a size twice as large as the other, individuals in the household with a smaller size have access to a higher level of spending. In order to gauge the level of income inequality and its changes over time, we need to have an appropriate yardstick. By considering desirable properties that index of inequality should have, such as mean independence (income-zero-homogeneity), population size independence, and the Pigou-Dalton principle of transfers (Bourguignon 1979; Shorrocks 1980), we use five inequality measures the Gini coefficient, Atkinson index, the variance of log income, and two measures from Generalized Entropy family i.e. Theil-T and Theil-L. The Gini coefficient satisfies all properties above. This index is sensitive to changes in the middleincome range which is calculated using the following formula 4 : G = 2 nµ cov(j, y j) (3.1.1) where n, µ, j, y j are, respectively, total number of households, mean income, rank of income, and income of the j th household. Atkinson index, which is one of the normative 5 measures of inequality, takes the form as the following: 1 I A = 1 1 n ( ) 1 ε 1 ε yj (3.1.2) n y j=1 3 In this paper, income is approached by consumption-expenditure. Therefore, the terms of income/expenditure will be used interchangeably throughout the rest of the discussion. 4 There are numerous ways of calculating Gini coefficient which can be seen in Anand (1983). 5 Normative measures are based on welfare economic theory and related to loss of social welfare resulting from the unequal distribution of income so that a higher degree of inequality corresponds to a lower level of social welfare for a given total income.

4 4 KADARMANTO - DENZO KAMIYA where ε is an unknown parameter measuring inequality aversion, that is, the relative weights to be attached to persons at different levels of income. The variance of log income (V ), which will also be used, is defined as follows: V = 1 n n [ln y j ln g(y)] 2 (3.1.3) j=1 where g(y) is the geometric mean of y. Using this geometric mean, this measure places greater emphasis on the lower incomes of the distribution. According to Anand (1983), the formulas for two Generalized Entropy measures, Theil-T and Theil- L, are defined, respectively, as: T = 1 n y ( j n m log yj ) (3.1.4) m L = 1 n j=1 n ( ) m log j=1 y j (3.1.5) where m is arithmetic mean income of the population. E(1) is another name of Theil-T whereas Theil-L is also known as E(0) or Mean Log Deviation Decomposition. Of the indices we choose, only the entropy measures and V are decomposable without bias (Coulter, 1989). Relating to decomposability property, we can classify the inequality measures into three types as follows: Strictly Additively Decomposable, Weakly Additively Decomposable, and General Decomposable. Theil-L and V belong to the first whereas Theil-T belongs to the second. Atkinson index, on the other hand, is classified into the third class. Decomposition refers to determination of the contribution of each of several constituent parts to the total. Constituent parts might be subsets of components or geographic regions. Bhattacharya (1992) points out that there are two types of decompositions: (a). Decomposition by Population Subgroups which considers a partition of the population into disjoint subgroups of households or persons and (b). Decomposition by Factor Components which considers total income of each individual or household as the sum of amounts earned from different sources. This paper deals only with the former. For decomposing an inequality index by population subgroups, in this paper, the country may be grouped by: (i) provinces, (ii) regions, i.e., Java and the Outer Islands, (iii) rural and urban areas ignoring provincial boundaries, (iv) rural and urban areas of each province. Suppose the country may be split up into G group, i = 1, 2,..., G; hence, the overall inequality is additively decomposed into within-group and between-group inequalities. In the simple notation, additively decomposability of generalized entropy measures can be expressed as follows: E θ (y 1,..., y G ) = G w i E θ (y i ) + E θ (y B ) (3.2.1) In the case of (i) above, group i will be a province. The terms within-group and between-group are then equivalent to intra-provincial and inter-provincial, respectively. For cases ranging from (ii) to (iv), the interpretation varies accordingly. Supposing that the population is grouped into mutually exclusive and collectively exhaustive socioeconomic groups, the decomposition for Theil-T and Theil-L into within-group and between-group components are as follows: where: T = G ( ni n m ) i T i + m G G ( ni ) G L = L i + n }{{} within-group ( ni n ( ni m ) i log m ) ( m log ( mi ) m ) n m i } {{ } between-group (3.2.2) (3.2.3)

5 PATTERN OF INCOME INEQUALITY IN INDONESIA, n i = number of households belonging to the i th group m i = arithmetic mean income of the i th group T i = the Theil-T index in equations (3.1.4) corresponding to the i th household group = the Theil-L index in equations (3.1.5) corresponding to the i th household group. L i Let f i = n i n be the population share of the ith group and λ i = m i m be arithmetic income share of the i th group. Then equations (3.2.2) and (3.2.3) can be rewritten, respectively, as: G G T = (f i λ i )T i + f i λ i log λ i (3.2.4) and L = G G f i L i f i log λ i. (3.2.5) The first term on the right side of equation (3.2.4) and (3.2.5), which is a simple weighted sum of the subgroup inequality values, refers to the within-group component whereas the second term can be regarded as the between-group component, explaining that the inequality is solely caused by differences in the subgroup mean income. Theil-L uses population share while Theil-T uses income share as a weight. In order to examine the extent to which the changes in different factors contributed to changes in aggregate inequality, we consider only Theil-L index. Following Mookherjee and Shorrocks (1982), applying the difference operator to both sides of equations (3.2.5) gives: G G G G L = f i L i + L i f i log λ i f i f i log λ i (3.2.6) where represents the change in the variable from period t to period t+1. Equation (3.2.6) decomposes the change in inequality into four terms. The first term refers to Pure Inequality Effect ; the sum of the second and the third term gives Demographic Effect whereas the last term represents Income Effect. For the computational convenience, Mookherjee and Shorrocks (1982) and Tsakloglou (1993) have proposed an approximation given as the following formula: G G G ( ) G ( L = f i L i + L i f i + λi log λ i fi + f i λi 1 ) log m i (3.2.7) }{{} Pure Effect (PE) } {{ } Demographic Effect (DE) where the over-bar represents a simple average. } {{ } Income Effect (IE) 4. Pattern of Income Inequality in Indonesia This section contains results that are needed in analyzing income inequality in Indonesia. The results presented here are mainly divided into three subsections; namely, Overall Inequality, Urban- Rural Perspective, and Provincial Perspective Overall Inequality. Table 1 provides a summary of inequality indicators for the period The table contains information on different inequality measures. In addition to the Gini index, two members of Generalized Entrophy Class, Variance of Logarithm and Atkinson Index with three parameters of aversion (0.5, 2, and 4) are also presented. What we can infer from Table 1 is that in 1999 all indices show that the degree of income inequality has the lowest value which reveals that 1999 became the year in which distribution of total household expenditure per capita was most equal, and all indices also show that the total household expenditure per capita was most unevenly distributed in Our estimates for the Gini coefficient in Indonesia as a whole are slightly larger than the previous estimates. For instance, the Gini coefficient calculated by BPS were 0.321, 0.335, 0.355, 0.308, and in 1990, 1993, 1996, 1999, and 2002, respectively. For this analysis we will concern with the figures excluding East Timor (E). Table 1 confirms that the evolution of inequality over the period is marked by an increase from 1990 to 1996, a decline from 1996 to 1999, and a subsequent deterioration over the remaining period. The Gini coefficient increased from in 1990 to in 1993 and further increased to in Gini Coefficient and all other inequality measures show a

6 6 KADARMANTO - DENZO KAMIYA Table 1. Income Inequality Indicators Gini Index Theil-T Theil-L VarLog Atkinson - ɛ = ɛ = 2 Year (1) (2) (3) (4) (5) (6) E I E I E I E I E I E I E ɛ = 4 I Source: Own calculation from the Susenas. decreasing trend in This reduction in income inequality appeared during a period when financial crisis swept into Indonesia in late The financial crisis has made the Indonesian situation get worse everyday. The effect of the monetary crisis was severe for Indonesia....The monetary crisis affects Indonesia worse than it affects any other countries, because there are also racial problems in Indonesia... (Son,n.d.,p.3). In addition, it caused also political crisis. The crisis has turned Indonesia from an economy of positive growth into one of negative growth. GDP growth fell from 8 per cent in 1996 to minus 13 per cent in Many workers were laid off from their jobs during this period, especially non-agriculture workers. The manufacturing sector fluctuated very violently with the negative growth resulting in a severe contraction of GDP. The fluctuation of GDP growth is exhibited in Figure GDP Agriculture Manufacturing Percentage Year Figure 1. Economic Growth at the Constant Prices, Source: BPS, Statistics during 50 Years Indonesian Independence & Statistical Yearbook of Indonesia various years.

7 PATTERN OF INCOME INEQUALITY IN INDONESIA, Following the crisis, in May 1998, there was a students movement which caused the downfall of President Soeharto; Vice President Habibie was pointed as the third president of the Republic of Indonesia, which ruled only for one year, and after that there were many riots. In the May 1998 riots more than 500 people died and thousands of shops, homes, and offices were destroyed, where many of them belonged to ethnic Chinese. The Chinese minority has therefore suffered a severe traumatic shock since many have lost their property. Much of the violence has been directed at the Chinese minority, which is resented partly because of its relative wealth. There is at least an image among native Indonesian that Chinese are rich and control the economy. Some Indonesians often blame the Chinese for the monetary crisis; even though, the Chinese has nothing to do with the cause of the crisis. Some left the country temporarily or permanently. Since then, ethnic-chinese businessmen were reported to be further sending their wealth abroad.... The flight of wealthy, educated Chinese following last May s riots came as a shattering blow in a nation that has since become the frailest of Asia s once prosperous and now shattered economies. An estimated 70,000 of them left the country, taking with them tens of billions of dollars... (Washington Times, November ). Sim, Susan from Singapore Straits Times reported that accompanying an exodus of ethnic Chinese entrepreneurs about US$80 billion (S$136 billion) had already left the country and 25,000 ethnic Chinese businessmen were likely to take their families out of Indonesia costing the country another US$500 million in lost expenditure. Reuters reported that Indonesia suffered an estimated US$20 billion in capital outflows before, during and after the May riots and saw little chance of the money returning soon. It made a serious problem of the Indonesian economy. According to Hill (2000), the number of Chinese community was about 3 per cent of the population but they controlled perhaps up to 40 per cent of the economy. Moreover, Bardsley (1999) states that 3.5% of the Chinese in Indonesia is in control of 70% of the economy. Hill shows a table of the major business conglomerates in Indonesia on page 113. The table exhibits that only four out of the top 40 conglomerates in 1993 were owned by pribumi 6 and two of the three largest were owned by the former President Soeharto s sons. Based on another table listing the largest 40 private business groups in Indonesia, Booth (1992) estimates that total annual sales of them in 1988 were estimated to be Rp. 37 trillion in which the top 10 accounted for almost 57 per cent; all were controlled by Indonesians of Chinese extraction. From 40 conglomerates, only 12 were controlled by pribumi interest which all together accounted for only 17.9 per cent of total sales. It shows the degree to which the Indonesian economy was dominated by a relatively small number of business groups, most of which were owned by Indonesians of Chinese descent. Hence, all these factors economic, political, and social situation all together might be able to be taken into considerations as factors reducing income inequality in Indonesia during According to Pigou-Dalton principle of transfer, any income transfer from a richer to a poorer that does not reverse their relative ranks in income reduces the value of index. This statement implies that the income share of the rich falls. It is the case in Indonesia during the crisis, if we look at the above situation even though not transfer but it still reduces the income share of the rich in which that of the poor is unchanged or increases. Accordingly, this can explain the decreasing in inequality level during the period When the economic situation was rather stable from prolonged economic crisis shown by a positive economic growth since 2000, income inequality began to rise again. The Gini coefficient increased by 7.53 per cent from to during the period Among the inequality indices presented in Table 1, Atkinson index can help us determine which expenditure per capita of Indonesia among the years is better in terms of social welfare. Assuming low sensitivity against inequality (ɛ=0.5) in the society, income distribution in 1996 is most unequal compare with other else. The increase in the index from in 1990 to in 1996 indicates the increase in income inequality. If the income in the society were distributed absolutely equally, it would be possible to reach the same social welfare level with just 90.91% of the total income in 1990, whereas only 88.89% of the aggregate of income, if equally distributed, would result in the same social welfare level in In other words, the loss of welfare due to inequality in 1990 was 9.09% and went up to 11.11%. In conformity with this, the loss of welfare due to inequality in other years we analyze; namely, 1993, 1999, and 2002 were 9.99%, 8.43%, and 9.83%, respectively. In order to conceptualizing and measuring inequality visually so as to get a complete picture of income distribution in Indonesia, two essential graphic analytical tools are employed here. These two graphical presentation devices include Lorenz Curve and Cumulative Frequency Distribution Function 6 pribumi refers to indigenous people, as distinct from non-pribumi, mainly ethnic Chinese.

8 8 KADARMANTO - DENZO KAMIYA (CDF) which are plotted in Figure 2. Due to the relatively slightly different Gini coefficient, the Lorenz curve are depicted in the several figures so that we will be able to see them more clearly while in presenting CDF, the expenditure level is truncated at the level of 1,000,000 rupiahs. (a). Lorenz Curve (b). Lorenz Curve (c). Lorenz Curve 1993 and 2002 (d). Truncated Cumulative Distribution Function: Figure 2. Lorenz Curve and Cumulative Distribution Function Source: the 1990, 1993, 1996, 1999, 2002 Susenas. If one Lorenz curve lies everywhere above another it is said to Lorenz dominance over the other curve and all inequality measures will show inequality to be lower for the higher curve. If distribution function A lies nowhere above and somewhere below that of B, then A displays first-order-stochastic dominance over B. First order stochastic dominance of distribution A over B implies that any social welfare function that is increasing in income, will record higher levels of welfare in distribution A than in distribution B as stated by Saposnik (1981;1983.) Table 2 gives the Lorenz/Stochastic dominant results of the period where LD (FSD) in cell (i,j) indicates that the distribution for year i Lorenz dominates (first order stochastically dominates) that for year j. As far as our data on consumption expenditure can be a good proxy which is capable of representing an actual figure of income, the results imply that Indonesia s economic development during succeeded in raising social welfare even though raising inequality in some periods. The increase in social welfare from the beginning to the end of the period is remarkable. Incomes rose so much that, despite growing disparities, social welfare increased unambiguously between 1990 and Inequality indices are convenient to show the degree of inequality but do not reveal what kind of changes happened to the income share of each income group. The changes are more clearly brought out by an examination of the amount of income accruing to each group of the population. Table 3 shows the income share by household group. The results suggest that there was an increase in the share of total household income accruing to the top 10 per cent during

9 PATTERN OF INCOME INEQUALITY IN INDONESIA, Table 2. Stochastic Dominance Analysis Results: Year Year " LD LD - LD 1993 FSD " LD FSD FSD " FSD / LD FSD / LD FSD / LD " LD 2002 FSD FSD / LD FSD / LD FSD " Source: Figure 2. Table 3. Distribution of Income by Household Group Decile Group Percentage Share of Income (1) (2) (3) (4) (5) (6) I II III IV V VI VII VIII IX X Ratio Top 20% to Bottom 20% Source: Own calculation from the Susenas. The ratio of highest Quintile 7 to lowest Quintile in Table 3 can also represent the income gap. It can be inferred that during the 1990s, the percentage share of income of lowest 20 percent increased while the percentage share of highest 20 per cent decreased. On the other hand, during the late 1990s till the early 2000s, the percentage share of lowest 20 per cent declined. It is pertinent to note the increase for the highest 20 per cent during the same era. Consequently, the income gap that declined during the showed an increasing trend hereafter. The Gini coefficient also performs the same trend-line. At the national level, Fifty per cent of the population receives per cent of the total national income in This share gradually falls in 1993 and 1996, viz and per cent respectively. It rises of percentage points in 1999 becoming and decreases again to per cent in 2002, however. In the years 1990 and 1999, the share of the top 10 per cent of the population is lower than that of the bottom 50 per cent in terms of income approximated by household expenditure per capita. The opposite situation happens in the other years. The most equally distributed income can be seen in the case of 1999 in which the lowest 50 percent has the highest income share of per cent among other years. It shows that the poor were not probably much influenced by economic crisis, and the impact of the crisis, on the other hand, was more severe for the rich. The movement of the share of the top 10 per cent and lowest 50 per cent has the opposite direction in which if the former increases then the latter decreases; in contrast, if the former decreases then the latter increases. That of the top 10 per cent has consistency with the change in income inequality. If there is an increase in the share of the top 10 per cent, then there will also be an increase in the Gini index. Conversely, if the share of the top 10 per cent falls, the index will also decrease. It means that 7 Quintiles constitute one method of categorising a population by income (or other variables).the lowest quintile represents the 20% of the population whose income is lowest.by the same token, the highest quintile represents the 20% of the population whose income is highest. Therefore, the highest quintile consists of decile IX and decile X while the lowest quintile is composed of decile I and decile II in Table 3.

10 10 KADARMANTO - DENZO KAMIYA the trend of the overall inequality in Indonesia during the period is mainly determined by the change in the income share of the top 10 per cent. Looking at the distribution of income by decile groups in Table 3, we can see that the income share rises relatively slowly as we move from the lowest group upwards, but there will be a big gap between the ninth decile to the top 10%. It is probably a feature of the income distribution pattern in the developing countries as the case of Indonesia. Figure 3 demonstrates this situation more clearly. In this figure, the income share of each decile is expressed as a percentage of the top decile s share of total income over the whole periods. In 1990 the income share of the lowest decile, which was 3.72%, amounted to 13.52% share owned by the top decile. By 1993 the income share of the lowest decile had declined marginally to 3.58% of the total income. But the share of the top decile had increased to 28.65%. An increase in income at the top decile means that people in other deciles deteriorated relative to the top, and generally the deterioration was greater for those in the middle income shown by the relative distance of the two curves from the two different years widening in the middle in Figure 3. It appears that the changes have been much greater over the period than over the period. There was, however, substantially amelioration in income share of the poor relative to the rich demonstrated by the movement of 1999 curve which is much higher than that of 1996 in which all deciles but the ninth as well as the top decile made improvement of their income share to the national income. It is an even higher position than the position in the initial year of 1990 but thereafter the relative share worsened again in Percentage (%) Decile (Population Group) Figure 3. Share of Total Income Received by Each Group as a Percentage of the Top Decile Share Source: Table Urban-Rural Perspective. The present subsection consists of two parts as the following: Urban/Rural Area Income Inequality, and Decomposition of Overall Inequality by Area Urban/Rural Area Income Inequality. Inequality in income distribution among society group takes place not only for the country as a whole but also spatially or between areas urban and rural. Income was distributed more evenly in rural than in urban area; in other words, income distribution in rural area is better than in urban area. It is shown in Table 4 that all rural Gini coefficients are always lower than that of urban. The table also tells us that average urban household expenditure per capita is much higher than that of rural. The Gini coefficient for Indonesian urban dwellers increased slightly from in 1990 to in 1993 and dramatically rose to in 1996, while for the rural population the increase was also experienced during this period from in 1990 to in 1996 during the pre-crisis period. Although income was evenly distributed in rural area, urban area appeared to be benefiting relatively more from economic growth. The ratio of average urban household expenditures to average rural household expenditures indicates the disparity showing the income gap between rural and urban areas.

11 PATTERN OF INCOME INEQUALITY IN INDONESIA, Table 4. Mean Household Expenditure Per Capita and Gini Coefficient by Area Area Mean Household Expenditure Per Capita Gini Coefficient (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Overall 30, , , , , Urban 44, , , , , Rural 24, , , , ,784, Ratio ///// Source: Own calculation from the Susenas. ///// ///// ///// ///// The ratio has been increasing from about 1.81 in 1990 to 1.93 in There was no further rise in urban-rural disparity until the year During the period covering economic crisis, the Gini coefficient for both Indonesian urban and rural population dropped drastically until below its initial level in in the case of urban and in the case of rural. They, however, rose again in 2002 to and in urban area and rural area respectively. The decrease and increase in Gini coefficient from 1996 to 1999 and to 2002 are accompanied by the decline and the rise in the ratio. It declined in 1999 and rose again to 1.79 in the year It indicates that in 2002 urban population is 79 per cent better off than that of rural. We can observe also that all year Gini coefficients of urban are higher not only than that of rural but also than that of the country-wide. The smaller is the gap, the more equal is income distribution in the country as a whole. Generally speaking, given the distribution within rural and urban economies, the inequality of overall distribution of income varies directly with the inequality between average urban and average rural income. A rise in rural income, as a proportion of national income, shown by the smaller ratio in Table 4 has a strongly equalizing effect on the distribution of income for Indonesia as a whole reducing the level of the inequality index. The movement of urban Gini index is similar to that of the overall Gini. The increase in income inequality in urban area played an important role in increasing overall income inequality. Changes of the top 10 per cent of urban income share affected the top 10 per cent of the whole Indonesia directly because urban population predominates the top 10 percent of the country as a whole. This relationship also happens in the regard to the income share of the lowest income between the whole Indonesia and the rural area. Table 5 provides the population share of the richest decile and the poorest decile as well as the poorest 50%. As given by Table 5, Seventy-three point nine per cent of the richest 10% lived in urban area accounting for per cent of total urban population in 1990 and became per cent composing of per cent of urban population in the last period. On the other hand, more than 85 per cent of the poorest 10% or 74 per cent of the poorest 50% lived in the rural area. It even reached to per cent in the case of the lowest 10% and 85.13% in the case of the lowest 50% in The fact that many urban people lost their jobs during the economic crisis has made them become unemployed increasing the number of the poor who lived in urban area. It is shown by a very sharp increase in the percentage of the poor residing in urban area from only 8.13 per cent in 1996 to per cent in Therefore, we might be able to say that the decreasing (increasing) income share of the rich of the country as a whole is mostly attributable to the decreasing (increasing) income share of the rich dwelling in urban area over the full period. Table 5. Percentage of Population Contribution Percentage Share of Population Group Urban Rural (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Lowest 10% Lowest 50% Top 10% (24.31) (24.07) (22.09) (20.43) (19.13) (3.75) (2.98) (3.16) (3.23) (2.69) Source: Own calculation from the Susenas.

12 12 KADARMANTO - DENZO KAMIYA The increased inequality in urban area from in 1990 to in 1993 and further increased to in 1996 is due to the decreasing number of population below poverty line 8 tremendously from 9.4 million (16.8%) in 1990 to 7.2 million people or 9.7% of total population in that year the corresponding figure for 1993 is 8.7 million people or 13.4%. During this period the number of population below poverty line in the rural area also decreased even though it was not substantially changed as presented in Table 6. The declining number of urban population below poverty line can be explained by the fact that the number of population engaging in the construction, trade, manufacturing and the financial sectors always increases as far as the period is concerned. Even the manufacturing sector employment and trade sector employment as well continue growing until the end of the year of this study (Table 7). Manufacturing sector has become a magnet for workers seeking jobs at higher wages attracting many workforce to be engaged in. Wages outside the manufacturing sector were typically lower, with the exception of wages in some services such as finance and banking. As average incomes grow, the manufacturing workforce also grow as shares of the total workforce which is graphically shown by Gillis et.al. (1996) on page 486. These four sectors are a typical of urban formal sector. The government and large-scale enterprises such as banks, insurance companies, factories, and trading houses are its examples (Gillis et.al., 1996). The main attractions of employers of these sectors not only manufacturing sector are that they pay the highest wages and offer the steadiest employment. In line with the growing of urban dwellers engaged in these sectors, urban income inequality of the richest or the top 10 percent also rises which is presented in Table 8 that reveals the Gini coefficient of urban poorest and richest. This table shows that during the period the inequality of the richest group shows the increasing trend whereas that of the poorest gives the mixing pattern. Hence, the rise in urban inequality is also found to be due to increasing inequality in the top group of urban society. Table 6. Poverty Line, Percentage, and Number of Population Below the Poverty Line, Year Poverty Line (Rp.) Percentage of Population Below the Poverty Line Urban Rural Urban Rural Total Number of Population Below the Poverty Line (Million) Urban Rural Total Abs. Growth Abs. Growth Abs. Growth (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) ,614 13, ,905 18, ,246 27, ,032 31, ,959 72, ,409 74, ,499 96, Based on the 1998 standard which is adjusted to account for the shift in consumption pattern of the respective year. Source: BPS-Statistics Indonesia: 2003 Statistical Yearbook of Indonesia. As the nation s inequality, urban inequality also decreased at the time of economic crisis. As has already been explained in the previous section that the decreased nation inequality is partly explained by the behavior of Chinese ethnic sending their wealth abroad which make the income share of the rich fall. Most of the rich Chinese ethnic reside in urban area. Therefore, it can also explain why urban inequality also decreased during the period When economic crisis hit Indonesia, although the effect of the crisis were initially observed in both rural and urban area, the effect on urban area were more severe. Regarding poverty, as presented in Table 6, poverty increased in both urban and rural areas. This increase, however, was more marked in urban area. While the number of population below the poverty line in urban area increased by 83.33% (from 9.6 million in 1996 to 8 The poverty line referred to the daily minimum requirement of 2,100 kcal per capita plus the non-food minimum requirement, such as for living, clothing, schooling, transportation, household necessities, and other basic individual needs. The value of expenditure needed for fulfilling the basic minimum requirement including food and non food is called poverty line. Since December 1998 a new standard has been adopted. As the previous (1996) standard, it is also dynamic, as it adjusts to the shift in consumption pattern and to differences in consumption pattern across regions. The differences between the 1998 and the 1996 standard, however, are primarily due to the extension of the commodity coverage to be accounted in estimating the basic needs.

13 PATTERN OF INCOME INEQUALITY IN INDONESIA, Table 7. Distribution of Employment by Sector, Sector Abs. % Abs. % Abs. % Abs. % Abs. % (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) a Total Note: % is share to total employment. is growth rate from previous period. 1. Agriculture, Forestry, Hunting, Fisheries; 2. Mining and Quarrying; 3. Manufacturing Industry; 4. Electricity, Gas and Water; 5. Construction; 6. Wholesale Trade, Retail Trade, Restaurant and Hotel; 7. Transportation, Storage, Communication; 8. Financing, Insurance, Real Estate and Business Services; 9. Public Services; 0. Others. a including sector 4. Source: Statistical Yearbook of Indonesia various years. Table 8. Gini Coefficient of Urban Poorest and Richest, Year Group (1) (2) (3) (4) (5) (6) Lowest 10% Lowest 50% Top 10% Source: Susenas. Table 9. Labour Productivity (major sectors, Rp. 000, 1993 prices) Year Sector (1) (2) (3) All sectors 5,083 3,966 Agriculture 1,854 1,635 Manufacturing 9,885 9,544 Construction 8,373 5,996 Trade & tourism 4,336 3,543 Transport & communication 7,798 6,753 Finance & accomodation 58,977 45,924 Govt & private services 2,994 2,894 Source: BPS, Sakernas(National Labor Force Survey) 97 & million people in 1998) or by 62.5% in 1999 (from 9.6 million to 15.6 million), it only increased by 28.11% in rural area. It is because the crisis had a profound impact on employment mainly in urban area. According to Dhanani (2001), In 1998, some 2.5 million workers, or 3 per cent of the total work force, were displaced. Job losses came principally from the manufacturing and the construction sectors (1.1 and 0.7 million). The mining, trade, and service sectors each lost about 0.2 million jobs, while the utilities and finance sector shed around 0.05 million jobs each. The crisis has forced many urban workforce to come back to villages or engaged in the agriculture sector where the agriculture jobs can be found easily resulting in an increase in labour force of agriculture sector. Even though working force who are engaged in the agriculture sector increases, the labour productivity in this sector fell but not by as much as that in most other sectors (Table 9). As mentioned earlier, those sectors which has lost jobs are mainly concentrated in urban area; thus, the result is the dramatic increase in urban