WHY Do DIFFERENCES IN PROVINCIAL

Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized _F PERSIST 'a rf _i- j WHY Do DIFFERENCES IN PROVINCIAL JULY 1997 INCOMES IN INDONESIA? By Jorge Garcia-Garcia, Senior Economist, and Lana Soelistianingsih, Consultant INDONESIA DISCUSSION PAPER SERIES * ^ w AT SAADPAII 8 EIO ONRNUMBER EAST AsIA AND PACIFIC REGION COUNTRY DEPARTMENT [11

SUMMARY WHY Do DIFFERENCES IN PROVINCIAL INCOMES PERSIST IN INDONESIA? By Jorge Garcia-Garcia, Senior Economist, and Lana Soelistiaringsih, Consultant Indonesia's economic and social policies, rather than good luck, contributed to the increase in provincial incomes over the past 20 years and the decrease during that time period in provincial income inequities. Prudent economic management enabled the country and its provinces to grow. Resources from the oil sector allowed the government to improve the nation's physical and social infrastructure and to reduce the differences in provincial incomes and consumption. Yet, disparities in personal and regional incomes persist. Why do differences in provincial incomes persist? We try to answer that question in this paper by studying what makes provinces grow, what makes provinces have different incomes per capita, and what caused the differences in provincial incomes over the last 20 years. DISCUSSION PAPERS PRESENT RESULTS OF COUNTRY ANALYSES UNDERTAKEN BY THE DEPARTMENT AS PART OF ITS NORMAL WORK PROGRAM. To PRESENT THESE RESULTS WITH THE LEAST POSSIBLE DELAY, THE TYPESCRIPT OF THIS PAPER HAS NOT BEEN PREPARED IN ACCORDANCE WITH THE PROCEDURES APPROPRIATE FOR FORMAL PRINTED TEXTS, AND THE WORLD BANK ACCEPTS NO RESPONSIBILITY FOR ERRORS. SOME SOURCES CITED IN THIS PAPER MAY BE INFORMAL DOCUMENTS THAT ARE NOT READILY AVAILABLE. THE WORLD BANK DOES NOT GUARANTEE THE ACCURACY OF THE DATA INCLUDED IN THIS PUBLICATION AND ACCEPTS NO RESPONSIBILITY FOR ANY CONSEQUENCE OF THEIR USE.

Why Do Differences in Provincial Incomes Persist in Indonesia? By Jorge Garcia-Garcia and Lana Soelistianingsih P rovincial incomes increased and provincial income inequality decreased in Indonesia during the last 20 years. Economic and social policies, rather than good luck, contributed to these outcomes. Prudent economic management enabled the country and its provinces to grow. Resources from the oil sector allowed the government to improve the nation's physical and social infrastructure and to reduce the differences Table 1. Distribution of Population and GDP in provincial incomes and consumption. by Main Groups of Islands: 1975, 1993 Yet, disparities in personal and regional (in percentages) incomes persist. Why do differences in provincial incomes persist? We try to Main Groups of Islands Population Gross Domestic Product answer that question in this paper by 1975 1993 19751993 studying what makes provinces grow, what Sumutra 18 21 32 23 makes provinces have different incomes per Java 63 59 50 59 capita, and what caused the differences in Bali 2 2 1 Nusa Tenggara and East Timor 4 2 2 provincial incomes over the last 20 years. Kalimantan 4 5 7 9 2. In the next sections we discuss these issues Sulawesi 7 7 5 4 and present some facts about the structure and Maluku and Irian Jaya 1 2 3 2 growth of the provincial economies. In section I we Indonesia 100 100 100 100 discuss briefly the economic structure, distribution Source: Derivedfron BPS, Population Census and Regional Accounts of population and their evolution by groups of islands between 1975 and 1993. In section II we present the evidence about the growth and distribution of regional income and consumption and examine how government consumption affected the distribution of regional per capita GDP and total per capita consumption. In section III we show that regional incomes tend to converge, and discuss the variable which account for convergence and the faster grown of regions. In section IV we present some conclusions and recommendations. I. ECONOMIC STRUCTURE AND DISTIUBUTION OF POPULATION BY GROUPS OF ISLANDS: 1975, 1993 3. Indonesia experienced sustained and rapid economic growth since the late 1960s. Prudent macroeconomic management enabled the country to grow, and it allowed the government to expand and to improve the nation's physical and social infrastructure. Yet, disparities in personal and regional incomes persist. R E G I 0 N A L D E V E I P M E N T I N I N D O N E S I A Page 1

Table 2. Distribution of Main Islands and Indonesia's GDP by Sectors: 1975, 1993 (in pertentages) Regions 1975 1993 Agriculture Oil, Gas, Industry Services Total Agriculture Oil, Gas, Industry Services Total Mining Mining Sumatra 22 52 4 23 100 20 29 14 38 100 Java 33 3 13 51 100 15 4 24 57 100 Bali 48 1 3 49 100 22 1 7 69 100 Nusa Tenggara and East Timor 65 1 2 32 100 39 2 4 55 100 Kalimantan 30 34 3 33 100 18 31 13 37 100 Sulawesi 51 1 3 44 100 35 3 10 52 100 Maluku and Irian Jaya 34 42 1 23 100 22 36 8 35 100 Indonesia 31 22 8 39 100 18 12 19 50 100 Source: BPS, Provincial Income in Indonesia, 1975-79 for 1975, and Regional Income of Provinces in Indonesia By Industrial Origin 1988-1993 for 1993 Note: Oil, gas and mining includes manufacturing of oil and gas 4. Java dominates Indonesia's economic activity followed by Sumatra. The two islands produce about 80 percent of Indonesia's GDP and hold about 80 percent of the country's population (See Table 1). Their relative importance as economic and population centers changed between 1975 and 1993. Java increased its share in Indonesia's GDP from 50 percent to 59 percent, while Sumatra reduced its share from 32 percent to 23 percent. Java held about 63 percent of Indonesia's population in 1975, but 59 percent in 1993. Sumatra, on the other hand, held about 23 percent of Indonesia's population in 1993 while it held 18 percent of it in 1975. 5. Services and manufacturing increased their share in Indonesia's GDP from 39 and 8 percent in 1975 to 50 and 19 percent in 1993 (See Table 2.) The gains occurred at the expense of the primary sector (agriculture and mining), whose contribution to GDP fell from 53 percent in 1975 to 30 percent in 1993. The gains in services and manufacturing occurred in all the islands, but only in Java the primary sector generates less than 20 percent of the island's GDP. Bali follows Java. The primary sector generates less than 25 percent of Bali's GDP. In the other islands, the primary sector generates a larger share of GDP, from 38 percent in Sulawesi to 49 percent in Sumatra and Kalimantan and 58 percent in Maluku and Irian Jaya. 6. Three metropolitan-industrial areas in Java (Greater Jakarta, Bandung and Greater Surabaya) concentrate most of Indonesia's modern industries and infrastructure. The manufacturing sector of West Java, Jakarta and East Java produced about 60 percent of the manufacturing GDP of Indonesia, excluding oil and gas.' Hill (1990) calls this a pronounced 'bipolar' industrial development around its main eastern and western population centers. 7. Large changes also occurred in the composition and origin of exports in Indonesia. Oil and gas from Aceh, Riau, and East Kalimantan generated more than half of total Indonesia's exports in the 1980s, while Java exported mainly manufactured goods. In the 1990s, the combined exports of timber products from Kalimantan and of manufactured goods from Java exceeded exports of oil and gas. Exports of wood products and manufactured goods (Sectors 2, 5-9 of SITC classification) accounted for about 60 percent of Indonesia's exports in 1993. Exports of oil and gas that year reached US$11 billion, while exports of wood products and manufactured goods reached about US$23 billion. That remarkable growth occurred mainly after 1983. R E G I O N A L D E V E L O P M E N T I N IN D O N E S I A Page 2

8. Java's economic and population power continues to cast a shadow over Indonesia. Its metropolitan areas pull people from outside Java and from rural areas in Java. The challenge for the government continues: how can a financially constrained central government induce a more balanced regional development? The financial constraints oblige the central government to select better its fewer interventions and to rely more on indirect approaches such as deregulating the trade regime, eliminating barriers to entry and competition, and creating incentives for accumulating human capital. 9. This paper reviews the experience of regional growth in Indonesia and identifies possible causes why differences in regional incomes exist and persist. The paper shows that lower-income provinces tend to catch up with better-off provinces, and that they tend to do so faster when certain conditions exist. For example, more and better health and education help to reduce the differences faster because they reduce the rate of population growth and increase the quality and quantity of the country's human capital. 10. For over 20 years the Biro Pusat Statistik (BPS) has monitored personal incomes and expenditures through expenditure surveys (SUSENAS), and it has followed provincial economic activity through industrial, agricultural and labor market surveys. We use the information from BPS to look at the evolution of the levels and the distribution of provincial per capita GDP and per capita consumption. 2 II. FACTS ABOUT PROVINCIAL GROWTH AND THE DISTRIBUTION OF WELFARE Figure 1. Annual growth of real per capita consumption: 1983-93 (percent) 6.0 Indonesia 4.0-3.5 X E a - - o -o o o E 2 I 0 N A L D E V E L 0 P M E N T I N N D 0 N E S I A Page 3

8.0 Figure 2. Annual growth of per capita GDP by province: 1983-93 (percent) ' ~~~~~Indonesia 6.0 - Avgl Sg g g.... 6.0 2.0 0.0 X_ GROWTH AND WELFARIE 11. Economic welfare (per capita consumption) increased. The average Indonesian becarne better-off between 1983 and 1993 because his private per capita consumption increased 40 percent during the period.3 These gains appear more remarkable since 1986, after the government deregulated the financial and foreign trade sectors, as private per capita consumption increased 30 percent. Welfare increased in all provinces, but in some it grew faster than in others (Figure 1.) East Java, West Kalimantan, Lampung and South Sulawesi show the highest rates of growth in provincial welfare, while Bengkulu, Riau and South Kalimantan show the lowest growth rates during these 10 years. 12. All regions grew between 1983 and 1993. Between 1983 and 1993 total and non-oil real per capita GDP in Indonesia grew at 4.8 percent and 5.5 percent per year respectively (obtained as the average of annual growth rates of total regional GDP). Total GDP per capita grew in all regions (See Figur-e 2.) Bali grew the fastest, at 7.5 percent, followed by Lampung-7.3 percent-, Central Java-6.1 percent-, Jakarta and South Sulawesi-6.0 percent- and North Sumatra-5.9 percent-. Riau, East Kalirnantan, South Sumatra and Irian Jaya grew at the lowest rates-minus 1.1 percent, 0.4 percent, 1.7 percent and 1.9 percent-. Consumption also grew in all provinces, at an average of 3.6 percent per year. The fastest annual growth occurred in Irian Jaya -7.8 R E G I O N A L D E V E L O P M E N T I N N 0 0 N E S I A Page 4

percent- followed by East Java-5 percent per year-, followed by South Sulawesi, Lampung and West Kalimantan-4.9, 4.7 and 4.6 percent per year-. Consumption grew the lowest in Benkgulu-0.4 percent per year. 13. Provinces can grow fast, both in Western and in Eastern Indonesia. The Government has expressed its concern about the welfare of people in the provinces of Eastern Indonesia (Kalimantan, Nusa Tenggara-East and West-, Sulawesi, Maluku and Irian Jaya), but the experience shows that they can grow as fast or faster than Western Indonesia. Although provinces in Eastern Indonesia have low incomes and high poverty incidence, some of them performed better than the average since 1983. In fact, the evidence on provincial growth (Figure 2) for the period between 1983 and 1993 shows that some Eastern Indonesia provinces grew as fast or faster than provinces of Western Indonesia. 14. Provinces that started with the highest per capita GDP in 1983 finished with the highest per capita GDP in 1993. Most provinces kept their ranking in the income ladder during the period. The richest and poorest provinces in 1983 continued being the richest and poorest in 1993 (See Figure 3). Riau, East Kalimantan, Aceh, Jakarta, South Sumatra and Irian Jaya had the highest per capita GDP in 1983 and in 1993. East Timor, East and West Nusa Tenggara, and Lampung had the lowest per capita GDP in 1983 and in 1993. The high-income provinces, except Jakarta, have a good endowment of mineral resources (oil, gas and other minerals) which accounts for their high per capita income. The low-income provinces have a poor endowment of land, water and mineral resources which accounts for their low starting GDP. Figure 3. Provincial per capita GDP in 1983 and 1993: Few changes in relative income (Note: Provinces graphed from lowest to highest GDP per copita in 1983) 15.0 ** Increase Between 1993 and 1983 S GDP per calita, 1983 5D 12.5 *~10.0. 7.5 o a,0 = ** Jg 5 == 03 0E X a, J,C= = f 9c- Soiirce: Biro Piisat 2tisti R E G I O N A L E V E L 0 P M E N T I N 0N 0 N E S I A Page 5

15. The starting conditions of a province determine to a large extent its final status, but rankings change over time. That happened in the group of middle-income provinces, where some grew faster than others and moved ahead in the ranking of per capita income. Bali grew the fastest (at 7.5 percent) and moved from the eleventh to the seventh highest provincial per capita GDP, and Yogyakarta moved ahead of South East Sulawesi and Bengkulu. 16. Income grew in densely populated areas. High population densities have not prevented fast growth rates, and low population densities have not ensured high growth rates. Over the last 35 years, Indonesia grew faster as her population density increased. Per capita income decreased during the 1960s when population density reached about 60 people per square kilometer. Per capita income grew at about 4.5 percent per year during the 1980s and 1990s while population density reached about 90 people per square kilometer. Among the provinces, Bali grew the fastest while its population density reached 500 people per square km. In East Kalimantan and Irian Jaya, on the other hand, per capita income grew at the lowest rates while their population densities reached about 9 and 4 people per square km. Although low population densities allow farmers to cultivate more land (extensively), they also hinder specialization because they reduce the size of the market and increase the costs of providing good transport and cornmunications infrastructure. 17. Natural resources do not ensure continuous prosperity. The experience of regions with large amounts of oil, gas, minerals and forests (Aceh, Riau, South Sumatra, East and Central Kalimantan and Irian Jaya) shows that large pockets of natural resources help generate high GDP levels but do not ensure fast growth. During 1983-93, these regions had six of the seven lowest provincial growth rates. B. THE LEVEL AND DISTRIBUTION OF WELFARE 18. In this section we examine whether: (a) welfare increased in all provinces; (b) large differences in provincial welfare exist; and (c) economic growth and govemment consumption reduced the dispersion of welfare among provinces. To answer these questions we examine first how private real per capita consumption evolved between 1983 and 1993, and then how each province's relative consumption changed during the period. We measure welfare by the level of consumption per capita. Real per capita consumption measures human welfare better than real per capita income, as income may not show the true level of expenditure in a province; for example, the oil rich provinces consume much less than their GDP per capita because the central government takes a large part of the oil revenue from these provinces and distributes it to the non-oil provinces. Last, we look at how government consumption changed the level and the distribution of welfare among provinces. 19. Levels of provincial zoelfare. Economic growth brought higher provincial welfare and changed the distribution of provincial income and consumption during 1983-93. To examine what happened to provincial welfare during this period we look at what happened to provincial real per capita consumption between 1983 and 1993. Figure 4 shows real per capita consumption by province in 1983 and 1993. The black thick line shows the level of welfare (total per capita consumption) in 1993, and the thin line with white dots shows the level of welfare in 1983. The black thick line falls above the thin-with-white-dots line, thus showing that all provinces increased their welfare between 1983 and 1993. But did the distribution of welfare improve during these 10 years? 20. The distribution of welfare and its evolution. To look at how the distribution of welfare evolved during these 10 years we classify provinces as low-, middle-, and high-income. We define low-income provinces as those where per capita total consumption (private consumption plus government consumption) falls below mean provincial consumption minus one standard deviation. We define high-income provinces as those where per capita total consumption exceeds mean provincial consumption plus one standard deviation. In middle-income provinces per capita total consumption falls within one standard deviation of mean consumption. Figure 4 shows these deviations around mean consumption for 1983 and 1993. The area between the R E G I 0 N A L D E V E I O P M E N T IN N N D O N E S I A Page 6

Figure 4. Real per capita consumption by province 1983 and 1993 900.................. 1993 800 700 600 400 300 100 E 0, = E e o = B 0 ' v ] 0 -E 0,, o dotted straight lines shows the values of consumption that fall within one standard deviation of mean consumption in 1983; the limit values for consumption above and below one standard deviation from the mean are Rp. 188,000 and Rps. 354,000. The area between the continuous straight lines shows the values of consumption that fall within one standard deviation of mean consumption in 1993; the limit values for consumption above and below one standard deviation from the mean are Rps. 248,000 and Rps. 490,000. 21. Most provinces in Indonesia belong to the middle-income category and the number of provinces within that group increased from 20 to 23 during the period. Five results from Figure 4 stand out. First, Lampung, a low-income province, moved to the middle-income category. Second, South Sumatra, Central and East Kalimantan moved from the high- to the middle-income category. Third, East Thnor, and West and East Nusa Tenggara increased their consumption and improved their welfare, but they continue as the poorest regions. Fourth, only Jakarta remained in the high-income group, moving further away from the other provinces. Real private consumption in Jakarta increased more than in any other province, from Rps. 495,000 rupiahs in 1983 to Rps. 658,000 in 1993. East Java followed; real consumption there increased from Rps. 209,000 to Rps. 355,000. Fifth, the distribution of consumption within the middle-income group became less equal; its relative variability (standard deviation/mean consumption) rose from 14 to 17 percent. That dispersion is small, however. 22. Did provincial consumption become more evenly distributed during 1983-93? In other words, did total and private per capita consumption become more evenly distributed among provinces? How much did economic growth and government actions contribute to these results? To answer these questions we look at how dispersed provincial per capita consumption became during the period. We look at the evolution of dispersion for all the provinces (27) and for all the provinces but Jakarta (26 provinces). The standard devia- R E G I O N A L DE V E L O P M E N T IN N DO N E S I A Page 7

Figure 5. Annual dispersion of private consumption: Figure 6. Annual dispersion of total consumption: 1983-93 1983-93 All Provinces~~~~~~~~~~~~~~~~~~~~~~~~llPovne 8.13 D.12 0.12 0.11 All Provinces except Jadaria 0.11 0.10 0.10-0.09 -.... _ --- 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1983 1984 1985 1986 19817 1988 1589 1990 1991 1992 1993 Figure 7. Comparison of dispersion of total and private Figure 8. Comparison of dispersion of total and consumption in all provinces except Jakarta: 1983-93 private consumption in all provinces: 1983-93 e.1..............................................................0.12 r...... Private Consumption 0.13 Private (OsusmptiDn. f 0~~~~~~~~~~~~~~~~~~~~~~~~~11-X 812~~ ~ > /'~ Total Consumnptio... n Total Comsumptiom 11 0.11..... 0.99 ~_....... 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 Note: Vertical axis measures the standard deviation of the logarithm of per capita income. tion of the natural logarithm of per capita consumption measures its dispersion. Figures 5, 6, 7 and 8 present the evolution of dispersion of total and private consumption. Figure 5 present the annual dispersion of private consumption for 27 and 26 provinces. It shows that inequality increased since 1986, and that including Jakarta increases provincial inequality (the line that shows the dispersion of consumption for 27 provinces falls above the line that shows the dispersion for 26 provinces.) Figure 6 presents the dispersion of total consumption. It shows that inequality decreased until 1986, it then increased until 1988 and it stabilized after that. It also shows that including Jakarta increases inequality (again, the line that shows the dispersion of consumption for 27 provinces falls above the line that measures that dispersion for 26 provinces.) 23. Governmenzt consumption and the distribution of zvelfare. Last we examine how government consumption affects the distribution of welfare among provinces. To do that we compare how the dispersion of total and private per capita consumption evolved for 26 and 27 provinces. Figures 7 and 8 show that evolution. From there we can conclude the following. First, government consumption reduced the inequality in provincial welfare. We can see that because the line that displays the dispersion of total consumption falls below the line R E G I 0 N A L D E V E L O P M E N T I N IN D O N E S I A Page 8

that displays the dispersion of private consumption. Second, the dispersion of total and private consumption in 27 provinces was lower in the 1990s than in the early 1980s. Third, government consumption offset the growing dispersion of private consumption after 1988, thus helping to keep provincial consumption less dispersed in the 1990s than in the 1980s. 4 24. Personzal Income Distribtutionz Withint Regionts. Economic growth has not led to widespread personal income inequalities in Indonesia, a country which enjoys one of the most equitable distributions of income in East Asia. 5 The distribution of personal incomes zvithin provinces also seems to be quite similar across provinces, as data on the distribution of consumption expenditures from household surveys suggest (Figure 9). Provincial Gini coefficients, which measure the distribution of personal consumption in each region, have similar values across provinces, and their values decreased for most provinces between 1983 and 1993. That fall shows that the personal distribution of income within each province improved. III. DiFFERENCES IN PROVINCIAL INCOMES FELL 25. Poverty and differences in provincial incomes still exist, but the evidence shows that in Indonesia poor people can come out of poverty and poorer provinces can catch up with the richer ones. In this section we deal only with how poorer provinces can catch Up with the richer ones. Below we: (a) look at whether low-income provinces tend to catch-up with middle- and high-income provinces and (b) identify what made provinces grow and poorer provinces catch-up with richer ones. The forces that influence growth could then be influenced to help the poorer provinces grow faster. 0.40 Figure 9. Provincial Gini coeffitients: 1983-94 0.25 0.15 ez~~ E i, a, E, a' c E a,.e, a,,, a, a,> a ~~-~ a _, a3 * F E E,E E a." E G i 0 K A L D E V E L O P M E N T I N I N D O N E S I A Page 9

A. POORER PROVINCES CAN CATCI-H-UP WITH RICHER PROVINCES IF... 26. The economic literature uses an extension of the standard model of economic growth to examine if lower-income countries catch up with higher-income countries. (For a detailed analysis of growth models and survey of the literature see Barro and Sala-i-Martin, 1995.) In that model the rate of growth depends on the initial level of income. A negative association between incomes and rates of growth means that incomes tend to converge absolutely; that convergence is known as absolute convergence. If that negative association does not exist, absolute convergence does not exist. However, if after taking into account the influence of other variables, the negative association exists, then per capita incomes tend to converge conditionally; this convergence is known as conditional convergence. 27. Since the late 1960s the Government adopted policies which helped provinces grow fast. The Government avoided large fiscal deficits and an overvalued rupiah. When necessary, it devalued the rupiah and reduced the fiscal and current account deficits quickly. The relative size of government fell as the private sector grew faster than the government sector. Since the 1980s the Government has deregulated the economy, thereby reducing distortions in relative prices and improving resource allocation. Although provinces faced the same general economic policies, the special conditions and factor endowments of each province, at the end, made some provinces grow faster than others. We review these conditions below, before going into the regression analysis. 28. Trade interventions fall. Import and export restrictions favored Java and hurt other provinces. Import restrictions protected the commodities that Java produced (like rice, sugar and manufactured goods), but outside-java provinces lost because they had to pay higher prices for these commodities. Export restrictions fell mainly on agricultural and forestry products that Sumatra and the Eastern Islands produced. Since the mid-1980s, when falling oil prices forced Indonesia to reduce tariff barriers to diversify her exports, exports of services like tourism and of labor intensive goods boomed. Bali and Java grew rapidly, but other provinces have not benefited as much from those changes because the Government continued to restrict the trade of primary commodities (like rattan, cloves, logs) which they produce. 29. Human skills improve. After 1970 individuals and the Government invested large amounts of resources in education, and Indonesia experienced one of the fastest rates of educational transition in the world. In 1984 Indonesia achieved near universal net primary school enrollment (93 percent), a feat achieved by few uppermiddle-income countries. The illiteracy rate (the percentage of people older than 10 years which can read and write) fell from 40 to 16 percent between 1971 and 1990. A more educated labor force accounts for the reduction in relative earnings differential and income inequality. The large increase in the supply of better-educated workers reduced relative wage differentials between skilled and non-skilled workers and helped to reduce income inequality. 6 30. Schooling improved the ability to learn and raised productivity in the market and in the household in rural and urban areas. The introduction of new technologies raises the returns to schooling if the new technologies increase the need for learning and reduce the scope for misuse of inputs. For example, the "green revolution" in agriculture led to an increased premium for acquisition of information. The new high-yielding seed varieties that drove the growth of the green revolution responded more to the use of inputs as water and fertilizer. Literate farmers adopted the technology earlier than illiterate farmers because they could acquire information faster. Evenson (1992) showed that technical assistance to Indonesian farmers that produced rice, maize, soybean, mungbean, cassava and peanuts yielded rates of returns between 50 and 90 percent. Literate farmers adopted the new seed varieties faster than the illiterate ones, thereby boosting the returns to investment in research and extension 7. 31. Fertility and mortalityfall. Education increased the rate of growth as it reduced fertility and the rate of population growth. More educated women marry later, postpone their first pregnancy and use more contra- R E 6 I O N A L D E V E L. OP M E N T I N I N D O N E S IA Page 10

ceptives than less educated ones. Women in Jakarta, better educated than in other parts of Indonesia, had an average of 2 children in 1990-95, while in West Kalimantan and East Nusa Tenggara they had 3.6 children. More children died and were born in the regions with the highest illiteracy rates. In 1990, infant mortality (children that died before one year of age) in Jakarta reached 40 per 1000 infants but 145 per 1000 in West Nusa Tenggara. In 1990, 92 per cent of women could read and write in Jakarta but only 62 percent could do so in West Nusa Tenggara. (CBS, Welfare Indicators, 1994). 32. Overall hiealtl conditions improve. Indonesians improved their health significantly. Between 1980 and 1990 males' life expectancy at birth increased from 51 to 58 years and females' from 53 to 61 years; child and infant mortality decreased from 162 and 109 deaths per 1000 to 103 and 71 deaths per 1000. Certainly, health conditions improved, but people can still improve their health substantially. Few Indonesians have access to safe drinking water. In 1990, 34 percent of Indonesians had access to safe drinking water, up from 23 percent in 1980, but the rates of urban coverage stayed at 35 percent, and only 33 percent of rural people had access to safe drinking water. Rural areas still have the largest group of people with poor health conditions in Indonesia; a broad based expansion of rural health services would improve their welfare and productivity. 8 B. RELATIVE PROVINCIAL INCOME INEQUALITY DECREASES IF... 33. The economic literature says that incomes tend to converge when low-income regions catch up with high income regions. Convergence, thus, occurs when the lower-income provinces grow faster than the higher-income provinces. The faster the low-income provinces grow relative to the rich ones, the sooner the poor will catch up with the rich; that is, the faster the convergence. In a graph which shows rates of growth and initial levels of income, the graph will show convergence if the regression line that links the points in the 0.1 Figure 10. Annual growth rate in 1975-93 and per capita GDP in 1975 0.06-0.04-0.02-0- -0.02-0.04 11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5 Natural logagthm of real per (apita GDP in 1975 R E G I 0 N A L DE V E L 0 P M E N T I N IN D 0 N E S I A Page 11

Figure 11. Cross-sectional standard deviation for the log of per capita provincial incomes for all provinces except East limor: 1975-1993 0.35-0.30 0.25 -_ p co M- _ e4 c 10s C 0 _ C' CV r. r. I_ C c o Co C= co co C= co co O O7 0s. 0' 0' 0' 0' 0' 0' 0' 0' 0' 0 0' 0' '0 '0 '0 graph slopes downwards; that is, if initial high per capita incomes are associated with lower growth rates over the observation period (See Figure 10). 34. To determine if provincial income inequality decreased, we look at it in two ways. First, we study if the dispersion of provincial per capita income decreases over time. Barro and Sala-i-Martin (1995, p. 383) call this reduction in dispersion 6 convergence. To determine if a convergence exists we compute the dispersion of provincial per capita income. We measure that dispersion as the standard deviation of the logarithm of provincial per capita income. Second, we study if poor regions tend to grow faster than rich regions, so that per capita income in poor regions tends to catch up with per capita income in rich regions. Barro and Sala-i- Martin (1995, p. 383) call this catching up /3 convergence. The P convergence (poor regions growing faster than richer regions) tends to produce a convergence (reduced dispersion of per capita regional income.). a convergence 35. Figure 11 shows the cross-sectional standard deviation for the log of per capita provincial GDP for 26 provinces between 1975 and 1993 (all but East Timor.) The dispersion of per capita incomes declined form 0.39 in 1975 to 0.30 in 1982, but then rises to 0.33 in 1983 9. Such increases may reflect one of two things. First it may reflect the adverse shocks to the oil sector which prevented the government from using oil revenues to finance a larger amount of infrastructure and education in the poorer provinces. After that shock the government introduced adjustment measures to stabilize the economy and reduce the current account deficit (Woo, Glassburner and Nasution, 1994). After that, the government also obtained other, more stable, sources of revenues which allowed it to continue expanding infrastructure and education throughout the archipelago. After 1984 the dispersion fell, and in 1993 it reached 0.28. Second, it may reflect changes in national accounts data arising from changes in the base and in the calculation of the regional accounts of Indonesia. 36. Barro and Sala-i-Martin (1995, Chapter 11) find a similar evolution for the dispersion of personal income across 47 or 48 US states for the period 1880-1990 and for 47 Japanese prefectures for the period 1930-1990. For the US the levels of dispersion exceeded 0.5 in the 1880s, but reached about 0.2 in the late 1980s. For R E G I O N A L D E V E L O P M E N T I N N D O N E S I A Page 12

Japanese prefectures they found dispersions of 0.5 and more in the 1930 and 1940s. After that the dispersion of income among Japanese prefectures declined sharply, and reached about 0.15 in the 1980s. Barro and Sala-i- Martin also studied cconvergence for Italy, Spain, Germany, France, and the United Kingdom for the 1950-1990 period. They find that Italy has the highest dispersion of regional incomes and the United Kingdom the lowvest dispersion. Italy started with a high level of dispersion, over 0.4, and ended with a level of 0.25 in 1990. The United Kingdom had a dispersion of just over 0.15 in 1950; that dispersion fell to about 0.05 in the mid-1970s and increased to around 0.1 in 1990. /3 convergence 37. To test for,b conivergenzce we look first if absolute convergence (poorer provinces grow faster than richer provinces) exists and then if conditional convergence (poorer provinces grow faster than richer ones if other variables besides initial income levels are included) occurs. To test for absolute convergence we use estimating equation 1 ln(yt/y)/(ty) a + bln(yo) (equation 1) where In stands for natural logarithm, and y, and y 0 stand for final and initial level of per capita income (years t and 0). To test for conditional convergence we use estimating equation 2 ln(y,/yo) /(t)=a + bln(yo) + j'yjz (equation 2) where Z. stands for other selected variables (e.g., education, fertility, health) which also influence the rate of growth. Many variables could obviously be used to explain the differences. For the regions of Indonesia we have selected the variables that seem to account better for the differences in provincial income growth rates. (Appendix 1 summarizes the results of Barro and Sala-i-Martin on the variables that account for the differences in the growth of countries.) 38. To test for convergence we run regressions for equations 1 and 2. The regressions cover the periods 1975-1993, 1980-93 and 1983-93. Per capita GDP of 1975, 1980 and 1983 constitute the initial income levels for the regressions of 1975-93, 1980-1993 and Table 3. Absolute Convergence: Regression 19830-1993. We test for convergence for 26 provinces for 26 Provinces (all but East Timor, for which information exists only Independent Variables Dependent Vanable (Annual growth rate of GDP) since 1985.) 1975-93 1980-93 1983-93 Results 39. Poorer Provinces Grozw Faster thant Richier (1) (2) (3) Provinces (Absolute Convergenice). We test for absolute convergence using equation 1 above. Table 3 displavs Constant parameter 0.3406 0.2829 0.2639 the results of the regressions. Equations 1, 2 and 3 (5.82) (5.64) (5.64) show the correlation between the rate of growth in Per Capita GDP in 1975-0.0234 1975-93,1980-93 and 1983-93 and the logarithm of (-5.00) per capita income in 1975, 1980 and 1983 respectively. Per Capita GDP in 1980 - -0.0187 '0 The rate of growth for each period is measured as Per GDP (-4.99) the logarithm of the ratio (real per capita income in Capita in 1983 - - -0.0170 1993/real per capita income in 1975, 1980 or 1983) R-squared (-4.64) divided by the number of years (18, 13 or 10). The 0.54 0.48 0.49 estimated coefficient for 1975-93 (2.4 percent) shows Notes: Figures in parentheses t-statistics. Regwssion faster convergence than the estimated coefficients for excludes East Tlimor R E G I O N A L D E V E L O P M E N T I N I N D O N E S IA Page 13

1980-93 and 1983-93 (1.9 and 1.7 percent). The estimated coefficients imply that if conditions continued as during the regression period (1975-93, 1980-93, 1983-93) it would take 29, 39 and 41 years to reduce provincial income differences by half. The regressions show that the initial level of income accounts for about half or less than half of the differences in the rates of regional growth. Other variables account for the other half of the variability in the rates of regional growth, and the following section looks at how much these other variables contribute to reduce that difference. 40. Evidencefor India, Japan, United States and Europe on Absolute Convergence. Cashin and Sahay (1996) studied if regional incomes in India converged during the period 1961-1991. They found that regional incomes among the 20 states tended to converge at 1.5 percent per year. That means that it takes about 45 years to close 50 percent of the gap between any state's initial level of per capita income and the 20 states common long-term level of per capita income. Barro and Sala-i-Martin (1995, Chapter 11) present evidence for,b convergence for states of the United States, prefectures in Japan and regions in Europe. For the United States they find that differences of personal incomes across states declined at about 1.7 percent per year for the 110 years period from 1880. For Japan they found that the per capita income of 47 prefectures during the period 1930-90 tended to converge at 2.79 percent (they estimated a coefficient of convergence of 2.79 percent.) For 90 regions of Europe they found that per capita GDP converged at 1.9 percent for the period 1950-90. The values of the convergence coefficients for per capita GDP that we have estimated for regions of Indonesia (2.4 percent for 1975-93 and 1.7 percent for 1983-93) fall between the values of the coefficients estimated for the United States (1.7 percent) and Japan (2.8). 41. Education and Other Factors Help (Conditional Convergence). This section shows that regions can reduce their differences in per capita income in less than the 29-41 years mentioned above. This happens when we take into account other variables in the regressions, as indicated in equation 2 above. In the neo-classical growth model income per capita depends on the rate of investment in physical capital, the rate of population growth and the level of human capital. In this model high population growth rates lower income per capita because physical and human capital spread more thinly over the population (Mankiw, Rome, Weil 1992.) The external trade environment can also affect the growth of regions. Countries more open to external trade tend to grow faster than countries less open to trade, but a country's trade regime tends to influence regional economies differently because regions have different factor endowments. Sometimes regions have their own trade and regulatory policies which hurt the region's economic performance. 11 Regions or countries affected more frequently by terms of trade shocks face more uncertainty in their external environment, so they may have lower rates of growth. Institutions can also affect the rate of growth (Olson 1996). Institutions which facilitate production and distribution and law enforcement make regions grow faster. 42. We have included some of these variables in the regressions that estimate the conditional convergence of provincial incomes. We have left other variables out of the regressions because we lack data for them. Initial income, population growth, human capital, external shocks and natural resource endowment enter into the estimated growth equation. Fertility (the number of children a woman in child bearing age has) represents population growth; regions with higher fertility rates are expected to have lower rates of growth. Fertility is measured by the average number of children born per woman in 1980 and 1993. 43. Regions which have more human capital per capita should grow faster than regions which have less of it. We use educational attainment to represent human capital, and educational attainment is measured by the share of population over 10 years of age with junior or senior high school education. The higher the share of people with secondary education in a region, the higher the region's human capital. We calculate educational attainment for 1977-93 and for 1982-93. The value for 1977-93 corresponds to the average of the observations for 1977, 1978, 1982, 1984 and 1986-1993, and the value for 1982-93 corresponds to the average of the observations for 1982, 1984 and 1986-1993. Regions which have higher rates of education should build more human capital and grow faster than regions which don't. The number of students per teacher could also R E G I O N A L E V EL 0 P M E N T I N N ONE SI A Page 14

measure the coverage of the educational effort and the efficiency in the use of resources spent in education. Thus, regions that have more students per teacher would tend to grow faster. 44. Good endowment of natural resources also affects the growth of regions or countries. Regions well endowed with natural resources can grow faster if their inhabitants use wisely the income from these resources. The experience of many countries indicates, however, that a good endowment of natural resources does not ensure high levels of income or rapid growth. For example, several oil exporting countries mismanaged the windfall gains from high oil prices and their economies stagnated or went into deep recessions. 12 Indonesia, on the other hand, managed these resources well. The central government took these resources and distributed them to all the provinces of Indonesia. But did oil affect the growth Table 4. Conditional Convergence: Regression Results of regions in Indonesia? We answer this 4o 26 Povines question by including a variable which for 26 Provinces measures the share of mining resources (oil, gas and mining) in the region's GDP Explanatory Variables Dependent Variable (Annual Growth Rate of GDP) for the periods 1975-93, 1980-93 and 1983-1975-93 1980-93 1983-93 93. (1) (2) (3) 45. As noted above, terms of trade Constant parameter 0.6367 0.6062 0.3813 shocks create uncertainty and increase the (9.46) (7.86) (2.93) costs of doing business, especially for Per Capita GDP in 1975-0.0480 - - economies with little diversification in (-4.15) their trade. Therefore, we would expect Per (apita GDP in 1980 - -0.0446 - that the more volatile the terms of trade (-7.28) the lower the growth rate. The volatility Per (apita GDP in 1983 - - -0.0254 of terms of trade is measured by the (-2.66) absolute value of the percentage changes Fertility rate, 1980-93 -0.0235-0.218-0.0219 in their terms of trade; the average of the (-4.15) (-3.41) (-3.60) absolute values during 1983-93 measures Education attainment their volatility during the period. Data for Educational Attainment 1977-93 0.2024 - - earlier years are not available. (5.08) C. CONVERGENCE TEST Educational Attainment 1982-93 - 0.1980 0.1072 (5.29) (2.17) per teacher, 1980-93 0.0020 0.0017 0.0021 46. The regression results we've Students already presented in Table 3 indicate that (2.11) (2.11) (5.04) it may take between 29 and 41 years to Share of oil & gas in GDP reduce differences in provincial income Share of oil & gas in GDP 1975-93 0.0983 - by half. Thirty or forty years seem to be a (5.09) long period to wait for income differences Share of oil & gas in GDPI 980-93 - 0.0715 - to fall by half. Do regions have to wait (3.46) that long to reduce per capita income Share of oil & gas in GOP] 983-93 - - 0.0105 differences? That does not appear to be (0.36) the case in Indonesia. Other variables Absolute change in terms-of-trade, 1983-93 - - -0.0289 such as education, mortality, changes in (-3.51) terms of trade and endowment of natural R-squared 0.89 0.81 0.83 resources can influence the rate of growth, as we show in Table 4 below. Table 4 presents the results of estimating Notes: figures in parentheses are t-statistics. Regressions exclude East Timor R E G I O N A L D EYE L OP M E N T I N N DO N E S I A Page 15

conditional convergence. Equations 1-3 show that other forces affect growth. Some of these forces are outside the control of individuals and governments, and others within their control. The estimated coefficients for log (initial per capita income) and the other variables (except oil share in 1983-93) have the expected signs and are significant at levels of 98 percent or more. The estimated rates of convergence equal 4.8 percent, 4.5 and 2.5 percent for the period 1975-93, 1980-93 and 1983-93. At these rates of convergence differences in per capita income fall by half in 15, 16 and 28 years, faster than the ones calculated initially. The estimated coefficients for fertility in equations 1, 2 and 3 stays the same. 47. Oil and gas and the coverage of education services show positive correlation with the rates of growth, and the volatility of terms of trade has a negative correlation with growth. The estimated coefficient for the share of oil and gas in 1975-93 and 1980-93 indicates that a one percentage point increase in the share raises the rate of growth of provincial GDP by less than 1/10 of one percent. For the 1983-93 period the coefficient does not differ significantly from zero, so the oil and gas sectors influenced regional growth little during that period. The coverage of education services, as measured by the student/teacher ratio, also increases the rate of growth of provinces. 48. The evidence suggests that poor provinces tend to catch-up slowly with rich provinces when left on their own. Differences in provincial incomes can fall more rapidly if people have better education and population grows at lower rates. Moreover, the recent growth experience of South East Asia's poor countries shows that low-income countries can grow fast, and Indonesia exemplifies that experience well. The same forces that made these countries grow fast could also make the poor regions of Indonesia grow fast. Government policies can encourage or discourage growth, sometimes powerfully. Institutions can also slow down or accelerate growth (See Olson 1996). Provinces face for the most part the same set of policies from the central government, but provincial institutions may be different and may work differently, so that policies which work well in one province may not do so in other. The previous analysis did not take into account the de facto institutions under which provinces operate. The de facto institutional differences in the provincial and local governments no doubt account for some of the differences in provincial growth rates. 49. Evidence for A Cross Section of Countries. Barro and Sala-i-Martin (1995, Chapter 12) present evidence for growth rates and conditional convergence for a cross-section of 87 countries for 1965-75 and 97 countries for 1975-85. They include the initial level of per capita GDP in the growth equation, so that the coefficient of this variable represents the rate of convergence. They also include other variables which reduce or increase the rate of growth, and which we discuss and summarize in Appendix 1. They estimate 24 growth equations and find that the rate of convergence varies between 1.4 percent and 2.8 percent, but the most frequent values vary between 2.5 and 2.7. The estimated coefficient of convergence for the 27 and 26 regions of Indonesia vary between 4.8 percent for 1975-93 and 2.5 percent for 1983-93; these estimated values represent higher rates of convergence than the estimated values by Barro and Sala-i-Martin for the cross-section of countries. IV CONCLUSIONS AND OTHER CONSIDERATIONS 50. Persistent regional imbalances continue to mark the economic development of Indonesia. Regional income disparities tended to fall (incomes converged), but regions in the top and bottom of the distribution in 1983 finished in the top and bottom of the distribution in 1993. The central government has adopted policies and programs to reduce such disparities. Yet many regions present socioeconomic indicators which show that the efforts to reduce inequality must be continued. 51. Better education can increase the rate of growth in the poorer provinces. The poorer provinces can grow much faster because they have the largest gains to make in health and literacy In fact, the poorer provinces have the lowest literacy rates, the less skilled labor force and the worst health indicators. Literacy rates and levels of educational attainment of women fall behind those of men. Women in rural areas have R E G I 0 N A L D E V E L O P M E N T I N N D O N E S IA Page 16