The road to pro-poor growth: the Indonesian experience in regional perspective

Indonesia's long-run ‘pro-poor growth’ record is among the best in Asia. It shows that appropriate policies can free societies from poverty's worst manifestations in a generation, a crucial message as democracy begins to influence the policy process. This paper places Indonesia's record in regional perspective, analysing determinants of income distribution in Asia and connecting this analysis to Indonesia's pro-poor growth process and the policy mechanisms that encourage pro-poor growth. Using a data set for eight Asian countries, it examines patterns of change in incomes and distribution across countries and over time. Building on Indonesian experience, the paper presents a pro-poor growth model encompassing three levels: improving the ‘capabilities’ of the poor, lowering transactions costs in the economy, especially between rural and urban areas, and increasing demand for goods and services produced by the poor. It finds that rapid pro-poor growth requires simultaneous and balanced interaction between growth and distribution processes. Notes Source: Adapted from Bourguignon (Citation2004). aThe horizontal axis shows the Gini coefficient; the vertical axis shows per capita GDP (log scale) for all countries except South Korea and India, for which the log of per capita income is shown; see note 4 for data sources, and appendix 1, table 3, for the period covered for each country. Since economic growth has generally been positive, the movement of points upward on the vertical axis also corresponds to movements over time. Each point represents a year for which data on income distribution are available. Sources: See note 4. aData are from various rounds of the Susenas, Indonesia's National Socio-Economic Survey, conducted by the Central Bureau of Statistics, which provides detailed expenditure data on more than 60,000 households every three years. Less detailed data from a ‘core’ questionnaire are available annually for about 200,000 households. Because of differences in approach, it has been difficult to match up results from the two questionnaires. Inevitably, many intellectual debts are incurred in writing a paper such as this. I cannot hope to thank all my colleagues in Indonesia and elsewhere who over the years have helped me understand the processes of economic growth, income distribution and poverty reduction. The intellectual framework used here goes back at least to Food Policy Analysis (1983), and my co-authors for that volume, Wally Falcon and Scott Pearson, have remained close colleagues and sounding boards over the years. The Food Policy Support Activity (FPSA), funded by USAID in Indonesia, has provided continuing support and colleagues for the development of my ideas. My daughter and co-author on a related manuscript, Ashley Timmer, has provided wonderful technical assistance on the data analysis for this paper. Finally, several reviewers for BIES offered helpful comments. Many thanks to all, and blame to none for continuing problems and errors. Bourguignon (Citation2004), drawing on Datt and Ravallion (Citation1992), explains the simple mathematics of this identity. ‘A change in the distribution of income can be decomposed into two effects. First, there is the effect of a proportional change in all incomes that leaves the distribution of relative income unchanged, i.e. a growth effect. Second, there is the effect of a change in the distribution of relative incomes which, by definition, is independent of the mean, i.e. a distributional effect … A change in poverty can then be shown to be a function of growth, distribution and the change in distribution’ (pp. 3–4). Three recent volumes from the World Institute for Development Economics Research of the United Nations University (UNU/WIDER) address these issues directly (CitationCornia 2004; CitationShorrocks and Van der Hoeven 2004; and CitationVan der Hoeven and Shorrocks 2003). These issues can also be approached from a political perspective. In the face of resentment over rapidly changing income distribution, political challenges can spill directly into economic policy, which affects the speed and direction of economic growth. A paper pursuing these political economy dynamics in the Asian context is under way (see Timmer and Timmer forthcoming). The eight countries selected are the only countries in Asia with consistent and reliable data on income distribution over at least two decades—each country has at least eight observations on the distribution of household income or expenditures. The basic data are from World Bank Indicators (www.worldbank.org/data/wdi2003/) and a special data set developed for the project ‘Operationalizing Pro-poor Growth’ (CitationWorld Bank 2004). Appendix 1, table 3, shows the period covered by the data for each country. The last recorded Gini value for India in figure 2 suggests a dramatic increase in inequality during the period of more rapid growth, but there is some doubt as to whether this reflects a trend. The actual regression is as follows (with t-statistics in parentheses): Ldgap20 = 0.798 + 0.426 * Pcigrow + 0.692 * Lgapstart, (0.9) (9.2) (5.4) (R squared = 0.956) where Ldgap20 = logarithm of the change in the income gap over 20 years (calculated from the rate for the entire period); Pcigrow = growth rate in per capita income over the entire time period (% per year);and Lgapstart = logarithm of the income gap at the start of the period (in constant 1993 $PPP). Most researchers have found that relationships that hold for the bottom quintile hold closely for other definitions of the poor as well; see Cord, Lopez and Page (Citation2003) for a review. Income distribution has widened within sectors as well, although roughly two-thirds of the rise in the overall Gini in China can be attributed to the growing wage gap between urban and rural workers. But it has been argued that in urban areas income distribution has also deteriorated and poverty increased, largely because of increases in food prices and the need to spend more on services previously supplied by the state (CitationMeng, Gregory and Wang 2004). A similar attempt to disaggregate changes in income distribution into urban and rural differentials is described in Eastwood and Lipton (Citation2004). Regressions similar to those presented in table 3 for Loggini as the dependent variable were run with the log of the income gap as dependent variable, as well as the relative gap. In general, the results were similar to those for Loggini; they are not presented here in the interests of space. There are two points here. The first is the economic significance of the variable, in contrast to its statistical significance as reflected by the t-statistic. To judge this, each estimated coefficient is multiplied by one standard error in the variable itself, to judge the size of the impact such a change would have. The following ‘importance’ coefficients result: 1. Logsect 0.075 2. Wagegap 0.030 3. Gdpserv 0.014 4. Saving 0.013 5. Ricertof 0.015 6. Ssrresid 0.038 7. Kcalresid 0.015 If all of these variables had a one standard deviation move in the ‘bad’ direction at the same time, the Gini coefficient would rise from 38.0 to 46.3, which would represent almost exactly a one standard deviation movement in the Gini for the whole sample. The second point is whether sectoral ‘parity’ means policies that protect domestic agricultural producers during the structural transformation. Although that is one option, followed aggressively by Japan and South Korea, the alternative is to invest in more flexible connections between the rural and urban economies, including migration and financial flows. This path is still open to Indonesia (CitationTimmer 2003). For an excellent short summary of the Soeharto years, see Hofman, Rodrick-Jones and Thee (Citation2004). For greater historical detail, see Hill (Citation1996) and the ‘Survey of Recent Developments’ in each issue of the Bulletin of Indonesian Economic Studies. A recent working paper from the SMERU Research Institute attributes two-thirds of the reduction in the overall poverty headcount index achieved at the provincial level between 1984 and 1996 to growth in agriculture. Industrial growth was only marginally significant in reducing urban poverty (CitationSumarto and Suryahadi 2004). Sharp differences between transactions costs in Holland and in colonial Netherlands East Indies (Indonesia) have been offered as a major reason for their different development paths from the 18th through the early 20th centuries (see CitationVan Zanden 2002, Citation2004; CitationVan der Eng 2002). A BIES reviewer suggested that this was a very controversial interpretation of Bulog's activities from 1970 to 1996. The analytical arguments and empirical defence are summarised in Timmer (Citation1996), with more institutional detail available in Timmer (Citation1991). See Artadi and Sala-i-Martin (Citation2003) for an explanation of the economic policies that led to Africa's decline since 1960. Every variable in their list of contributing factors has direct parallels to issues now facing Indonesia.