How Robust Is Indonesia’s Poverty Profile? Adjusting for Differences in Needs

Abstract

Poverty profiles showing how the magnitude of poverty differs across subgroups of a population are important tools in designing effective social protection programs. Using data from the March 2013 round of the National Socio-economic Survey (Susenas) and the fourth round of the Indonesia Family Life Survey (2007–8), I explore the sensitivity of Indonesia’s poverty profile to different assumptions about the relative costs of individuals, taking into account differences in age, gender, body weight, and physical activity levels. I adopt parameter estimates for my simulation exercises from various Indonesia-specific publications, as well as from a joint intergovernmental consultation on nutrition. I compare my estimates with the per capita scale used by Statistics Indonesia (BPS), the central statistics agency. My findings suggest that the age–poverty relationship in Indonesia is sensitive to assumptions about the relative costs of individuals, with all alternative scales showing substantially lower poverty incidence among young children than by BPS’s estimate. Overall, however, I find that Indonesia’s poverty profile is relatively robust.

Profil kemiskinan yang menunjukkan besarnya perbedaan kemiskinan antar subkelompok populasi adalah alat kebijakan yang penting dalam merancang program jaminan sosial yang efektif. Menggunakan data dari putaran Survei Sosial Ekonomi Nasional (Susenas) Maret 2013 dan putaran ke-4 dari Indonesia Family Life Survey, IFLS (2007-8), penulis menelaah sensitivitas profil kemiskinan Indonesia terhadap asumsi-asumsi berbeda mengenai biaya relatif dari individu—dengan memasukkan perbedaan-perbedaan umur, jenis kelamin, berat badan, dan tingkat aktivitas fisik. Untuk simulasi, penulis mengadopsi parameter-parameter dari berbagai publikasi tentang Indonesia, serta dari forum konsultasi antarpemerintah mengenai nutrisi. Penulis juga melakukan perbandingan antara estimasi yang didapat dengan skala per kapita yang digunakan Badan Pusat Statistik (BPS). Penulis menyimpulkan bahwa hubungan antara umur dan kemiskinan di Indonesia bersifat sensitif terhadap asumsi mengenai biaya relatif dari setiap individu, di mana semua ukuran alternatif menunjukkan bahwa insiden kemiskinan pada anak-anak usia muda secara substansial lebih rendah dibandingkan dengan estimasi BPS. Namun demikian, secara keseluruhan penulis menemukan bahwa profil kemiskinan Indonesia yang ada saat ini cukup akurat.

Keywords:

• poverty
• measurement
• robustness
• child costs
• adult equivalence

JEL classification:

• I31
• I32

Notes

1 See my earlier article in BIES (Priebe 2014) for a detailed account of official poverty measurement in Indonesia.

2 Adult equivalence scales and economies-of-scale effects are likely to be interlinked much more strongly in developed countries than in developing countries. In developed countries, larger families are more likely to have relatively more children, while, in developing countries, prime-age adults and their elderly parents are more likely to live in the same household. Furthermore, richer households, conditional on having the same demographic composition as poorer households, are more likely to benefit from economies of scale in consumption, since they are more likely to consume expenditure items that can be shared with other household members. De Ree, Alessie, and Pradhan (2013), using Susenas data from 2003 and 2004, confirmed this assumption empirically for Indonesia.

3 As mentioned above, studies that try to estimate the costs of children usually focus on specific family compositions. Both Deaton and Muellbauer (1986) and Olken (2005) provide estimates for the costs of children for families consisting of two adults and one child or of two adults and two children. For this analysis, I use, from both studies, estimates for families consisting of two adults and one child. In both studies, the costs of a single child are estimated to be higher for families comprising two adults and one child than for those comprising two adults and two children. Estimates based on the former composition could therefore be considered the upper bound of child costs under the respective method. The adult equivalence scales obtained from Deaton and Muellbauer and from Olken are shown in table 2.

4 The WNPGs also took place in 1972, 1976, 1986, 2000, and 2008. The documents related to the 1972, 1976, and 1986 workshops are lost in the archives, however, and the 2000 and 2008 workshops did not address minimum calorific requirements.

5 In contrast to Deaton and Muellbauer’s and Olken’s scales, the WNPG scales exhibit differences across age and gender driven entirely by differences in minimum calorificintake requirements. This feature of the WNPG scales is in line with the way that statistical offices in other developing countries obtain their adult equivalence scales for official poverty measurement. However, there are caveats involved in letting gender- and age-specific differences be determined exclusively by differences in calorific requirements. First, there exist a lot of individual specific genetic and metabolic differences that cannot be captured by using such general scales. Second, the quality of food intake cannot be expressed just in terms of calories; it would ideally be captured by other nutritional indicators. Third, there is a certain circularity involved in letting nutritional intake requirements determine poverty status when, in some cases, it is the other way around. For instance, children from poor families are more likely to be stunted than children from better-off families. Stunted children thus need fewer calories than non-stunted children and would, ceteris paribus, have a higher change of being classified as non-poor than poor on the basis of strict nutritional minimum requirements. The WHO approach, discussed in the extension section of this article, is particularly prone to this problem.

6 What I refer to as food equivalence scales are the ratios of age- and gender-specific calorific intake values, benchmarked against the calorific intake value of my chosen reference standard (men aged 40–45). Since there is no reason why differences in calorific requirements by age and gender should be related to age- and gender-specific non-food expenditure patterns, I need to make assumptions about the gender-specific ones. There is no information available on the differences in non-food costs or needs by age and gender, so I use instead the non-food adult equivalence scales, for which BPS simply assumes that every person has the same non-food needs. The weighting factor of 75% for the food adult equivalence scale is based on the consideration that the food poverty line in 2013 consisted of a food share of approximately 70% in urban areas and 78% in rural areas, with the population-weighted average close to 75% (for the national poverty line).

7 Official poverty rates are based on poverty lines that differ by province and by rural or urban area.

8 The basis is the province-specific rural and urban poverty lines used by BPS for official poverty measurement in Indonesia.

9 For the choice of a reference household, I classify children as individuals aged 14 and under.

10 Although I do not report it here, I find that all scales that give a low relative weight (≤0.75) to young children reproduce the age–poverty pattern derived from Deaton and Muellbauer’s (1986) adjusted scale: poverty rates among the 14–18 age group increase while those among the 0–4 and 25–40 age groups decrease. Results can be obtained from the author on request.

11 Additional problems arise when modelling the influence of higher prevalence of sickness among the elderly. Scales of minimum calorific intake tabulate values for healthy elderly adults, whereas the amount of non-food expenditures needed for an elderly adult depends on the individual’s health.

12 The analysis in this article discusses changes to the national average only. At the province and district level, there could be relatively large changes to the poverty profile, depending on the specific age structure and family living arrangements in the particular province or district.

13 See Van der Eng’s (2000) article for an overview of historical data on differences in individual consumption behaviours by activity level in Indonesia.

14 Following the WHO recommendations, I assign pregnant women additional energy requirements of 85 kilocalories per day in the first trimester, 285 kilocalories per day in the second, and 475 kilocalories per day in the third. To avoid overestimating the amount needed (in rupiah) to satisfy calorific needs at the household level, I take into consideration the amount of calories needed for breastfed children but not the extra amount of calories needed for breastfeeding mothers.

15 I have to move from the individual to the household because consumption (expenditure) data are collected in surveys at the household level. Likewise, consumption (expenditure) and welfare decisions are usually made at the household level.

16 BPS uses province-specific rural and urban poverty lines to calculate official poverty rates. Since the underlying BPS food poverty lines are referenced to an intake of 2,100 kilocalories per day, the lines are divided by 2,100 to obtain the average kilocalorie price. Furthermore, the official food poverty lines from the year 2007 are used because the dataset that illustrates the WHO approach is from that year.

17 See the article by Thomas et al. (2012) for more information on the nature and quality of the IFLS.