Washington, DC 20006, USA c The World Bank, East Asia and Pacific Region, Jakarta Stock

Transcription

1 This article was downloaded by: [ ] On: 02 November 2013, At: 09:14 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Journal of Development Effectiveness Publication details, including instructions for authors and subscription information: Positive returns: cost-benefit analysis of a stunting intervention in Indonesia Lubina F. Qureshy a, Harold Alderman b, Claudia Rokx c, Rebekah Pinto d, Matthew Wai-Poi e & Ajay Tandon f a Human Development Sector, The World Bank, East Asia and Pacific Region, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia b International Food Policy Research Institute, 2033 K Street, NW, Washington, DC 20006, USA c The World Bank, East Asia and Pacific Region, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia d Health & Early Childhood Education, Human Development Sector, The World Bank Office Indonesia, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia e Poverty Group, The World Bank, East Asia and Pacific Region, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia f East Asia and Pacific Human Development (EASHD), Room MC 9-237, Mail Stop MC 9-919, The World Bank, 1818 H Street, NW, Washington, DC 20433, USA Published online: 31 Oct To cite this article: Lubina F. Qureshy, Harold Alderman, Claudia Rokx, Rebekah Pinto, Matthew Wai-Poi & Ajay Tandon, Journal of Development Effectiveness (2013): Positive returns: costbenefit analysis of a stunting intervention in Indonesia, Journal of Development Effectiveness, DOI: / To link to this article: PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content ) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions

2 and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at

3 Journal of Development Effectiveness, Positive returns: cost-benefit analysis of a stunting intervention in Indonesia Lubina F. Qureshy a *, Harold Alderman b, Claudia Rokx c, Rebekah Pinto d, Matthew Wai-Poi e and Ajay Tandon f a Human Development Sector, The World Bank, East Asia and Pacific Region, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia; b International Food Policy Research Institute, 2033 K Street, NW, Washington, DC 20006, USA; c The World Bank, East Asia and Pacific Region, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia; d Health & Early Childhood Education, Human Development Sector, The World Bank Office Indonesia, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia; e Poverty Group, The World Bank, East Asia and Pacific Region, Jakarta Stock Exchange Building, Tower 2, 12th Floor, Jl. Jenderal Sudirman Kav , Jakarta 12190, Indonesia; f East Asia and Pacific Human Development (EASHD), Room MC 9-237, Mail Stop MC 9-919, The World Bank, 1818 H Street, NW, Washington, DC 20433, USA Would investing to reduce stunting reap economic benefits that outweigh costs? We investigate this question by conducting a cost-benefit analysis for a large-scale integrated nutrition programme to reduce stunting in Indonesia, using actual rather than stylised data on costs. The gains are assumed to accrue from productivity enhancement from reduced malnutrition, productive earnings from deaths averted and household savings from diarrhoea costs avoided. The programme extends to six provinces over 5 years covering seven cohorts. Using a discount rate of 5 per cent, the benefit cost ratio is The study finds positive net benefits through the productivity impact of lower malnutrition even with sensitivity analysis that excludes the benefits of mortality reduction. Keywords: Indonesia; cost-benefit analysis; stunting; low birth weight; malnutrition 1. Introduction Would investing to reduce stunting reap economic benefits that outweigh costs? Economists have addressed this question in order to place investments in nutrition on the same table as competing claims for national resources to spur economic growth (Behrman, Alderman, and Hoddinott 2004). Such calculations, however, face a number of obstacles, including the time frame between the investments and the full range of economic gains, as well as the challenge of assigning a monetary value to the stream of benefits. Even when benefits from an intervention are observed after the children have reached adult status (Hoddinott et al. 2008), there are still questions as to the relevance of the cost estimates, given the changing economic environment since the initial programme, as well as the challenge of estimating costs at scale. This paper offers a real world example by undertaking a cost-benefit analysis of a nutrition programme to address stunting in Indonesia, which is in the process of being scaled up from an earlier *Corresponding author. Lubina1@gmail.com Present affiliation for Harold Alderman is International Food Policy Research Institute, 2033 K Street, NW, Washington, DC 20006, USA Taylor & Francis

4 2 L.F. Qureshy et al. programme. Thus, we offer ex ante estimates suitable for planning using actual data from the earlier programme rather than stylised data on costs gleaned from the literature. As this is to guide a future investment, we cannot avoid making assumptions regarding eventual outcomes, but we address this inherent challenge by beginning with conservative assumptions where they are needed and by offering sensitivity analysis on the implications of various assumptions. While we focus on a specific project, the approach as well as the magnitude of the results is generalisable Stunting and its consequences Stunting or low height-for-age indicates a cumulative nutritional deprivation that sets in early in life in utero until the age of two. Early life consequences of stunting for those who survive include higher susceptibility to infections and reduced cognitive ability, which can carry on to school-going age. Long-term effects include the possibility of reduced earnings and an increased risk of contracting chronic diseases such as diabetes and hypertension (Almond, Chay, and Lee 2005; Alderman and Behrman 2006; Alderman, Hoddinott, and Kinsey 2006; Victora et al. 2008). Stunting interventions have a critical window of opportunity: from the time a child is conceived until the age of two. Thus, interventions focusing on improving both maternal and child nutrition through improved dietary intakes and enhanced childcare, including growth monitoring and promotion, are the most effective. In a study covering 36 countries, stunting prevention interventions were found to bring about as much as a 36 per cent reduction in stunting and a 25 per cent fall in mortality from 0 to 36 months (Bhutta et al. 2008) Previous cost-benefit analyses Alderman and Behrman (2006) demonstrate the economic benefits of reducing low birth weight for one infant in a developing country using stylised cost data. They assume that the long-term and short-term benefits stem from a rise in productivity due to reduced stunting and increased ability, reduced costs of chronic diseases, intergenerational benefits, and aversion of costs resulting from infant deaths neonatal care as well as from infant or child illness. The present discounted value of all these benefits is estimated at US$ 510, at a discount rate of 5 per cent, and approximately US$ 257 at a 10 per cent discount rate, indicating that any intervention costing less than this would make it economically worthwhile to invest. Their study does not include any costs necessary to achieve these benefits. Behrman, Alderman, and Hoddinott (2004), however, apply this methodology to determine priorities for investment in nutrition. Other estimates of benefit cost ratios for specific nutrition interventions, such as breastfeeding promotion and micronutrient initiatives, include Horton, Alderman, and Rivera (2009), who estimate these ratios for deworming to be 6:1 and salt iodisation to be as high as 30:1. Similarly, iron fortification of staples has benefits eight times the cost (Horton and Ross 2003). Other studies have arrived at the cost effectiveness of the interventions per disability-adjusted life year (DALY) saved or in terms of cost per death averted. Returns to vitamin A supplementation, for instance, range from US$ 3 to US$ 16 per DALY saved (Ching et al. 2000; Fiedler 2000). These and similar studies focus on components of interventions; an exception being Hoddinott, Rosegrant, and Torero (forthcoming), who include an estimate of a package of services in their study. An

5 Journal of Development Effectiveness 3 integrated delivery, also addressed in this paper, is a more realistic measure of public health interventions in most countries. 2. The Indonesia programme The primary health care system of Indonesia provides basic prenatal and post-natal services. Since 1980, these are delivered at the village level through the Posyandu, a health post serviced by community health workers or cadres. In utero or prenatal interventions under the Posyandu include foetal and maternal growth monitoring, micronutrient supplements (particularly iron-folate) and immunisations. Post-natal interventions until age two include weighting, immunisations, basic breast-feeding counselling, micronutrient supplementation (vitamin A, zinc and iron) and using oral rehydration salts (ORS) for treating diarrhoea. While micronutrients and immunisations were already available free of cost at the primary health centres, the Posyandu was a change from a situation where the villagers would otherwise have to travel to the closest health centre to access these services, thus increasing compliance by making the services easily available. Wai-Poi (2011) attributes an incremental 40 per cent reduction in all-cause under-five mortality in Indonesia to the Posyandu programme relative to secular trends. The analysis compares cohorts born before and after the Posyandu intervention using a difference-indifference approach based on village rollout and data from the longitudinal Indonesia Family Life Survey. However, the effectiveness of the Posyandu since then is questionable with reports of understaffing, poor training, low motivation, high cadre dropout rates, low utilisation rates and resource constraints (Strauss et al. 2004; Khomsan et al. 2007). There is evidence that an incentivised block grant scheme, the National Community Empowerment Programme, Healthy and Smart Generation (PNPM Generasi), first launched in 2007, made possible a much needed revitalisation of the Posyandu interventions. Generasi targets the Millennium Development Goals of maternal and child health and universal primary education. The villagers, with assistance from trained facilitators and service providers, use a participatory planning process and block grant funds to identify problems in improving performance on the different indicators and designing appropriate village-level activities to address impediments to accessing services. Each village is allocated a fund and progress on a predetermined set of health and education indicators, as compared to other villages in a sub-district, determines 20 per cent of the following years funding. Generasi also allows for flexibility in the use of funds. For example, if public provision is considered inadequate, the community has the flexibility to contract private providers to improve service quality or performance. Village communities are assisted by facilitators at the sub-district, district and provincial level in achieving these objectives. An impact evaluation of the Generasi for the period , revealed increase in the number of mothers and children participating in village health post activities to receive the targeted prenatal, neonatal and child health services and, specifically, a decline in stunting in treatment areas, as compared to control areas, large improvements in the frequency of weight checks for young children and an increase in the number of iron sachets pregnant mothers received through antenatal care visits (Olken, Onishi, and Wong 2012). Given the success of the Generasi in increasing the utilisation of the existing Posyandu services and influencing health outcomes, it is being used as a platform to deliver enhanced Posyandu activities, aimed at reducing severe malnutrition and child stunting at the community level. The enhanced service package consists of shifting the focus on stunting from weight gain alone and from monitoring to growth promotion. It,

6 4 L.F. Qureshy et al. thus, includes height measurement, intensified personal counselling on nutrition, hygiene, active feeding and prevention and control of diarrhoea, including counselling on the use of zinc as adjunct therapy in its treatment, accompanied with improved training and financial incentives for cadres, provision of multi-micronutrient sachets and a sanitation triggering campaign, following the Community-Led Total Sanitation (CLTS) approach (Kar with Chambers 2004). The package is also complemented by a national campaign to raise awareness on these issues to further increase utilisation rates and compliance. Thus, delivery of these enhanced services, coupled with Generasi motivational services, is expected to positively address stunting. We cost the existing Posyandu services and the additional component of enhancing the effectiveness of the interventions through better service provision by giving incentives and providing rigorous training. We do not cost procurement, such as of vitamins, vaccinations or zinc, with the exception of two items not currently available at the Posyandu, height measuring equipment and micronutrient sachets, thus restricting the costing only to the new enhanced services. Since the intervention will use the social infrastructure of Generasi to deliver the services, we also include the block grants as a part of the cost of delivering the enhanced package (see Table 1 and Section 3.3 for cost details). This paper adds to the literature by estimating the benefit cost ratio of an intervention that provides a range of services delivered as a package through the existing primary health system and the block grant scheme. It, therefore, conforms more to the full-scale programmes as administered in most environments and less to the vertical interventions, following a systems rather than a components approach (Horton, Alderman, and Rivera 2009; Horton et al. 2010). Indeed, given the general agreed best practices in nutrition (such as the Lancet series, 2008), research that seeks to isolate the impact of a single component by excluding some of these in control areas may be challenged on ethical grounds. Collectively, the interventions are expected to provide the following benefits: (1) reduced mortality; (2) reduced incidence of low birth weight (LBW) and stunting; (3) reduced disease incidence (mainly diarrhoea); and (4) reduced incidence of chronic diseases. We monetise benefits using earnings gained through lives saved, incremental earnings gained from productivity increase attributed to a reduction in LBW and stunting, and savings accrued due to reduction in diarrhoea morbidity as well as chronic diseases. While we draw on various sources of literature to compute the benefits, we consider the two main benefits of reduced mortality and stunting based on Wai-Poi s analysis of Posyandu expansion and the impact evaluation of the PNPM Generasi, respectively. 3. Method and data We first detail the programme coverage and costs, and then discuss the benefits. Under each benefit, we explain our assumptions and how we monetise the benefit. We use a discount rate of 5 per cent for computing the core net present value of the programme, although a range of options is employed in the relevant literature. The Centers for Disease Control and Prevention (CDC 2013) use a 3 per cent rate of discount for public health interventions. In calculating the DALYs, the Global Burden of Disease study uses a 3 per cent time discount rate (Murray and Acharya 1997). Horton, Alderman, and Rivera (2009) apply both 3 per cent and 6 per cent rates of discount, referring to the former as a social and the latter as a close to commercial discount rate. On the other hand, Alderman and Behrman (2006) use a baseline rate of 5 per cent when calculating the present value of

7 Journal of Development Effectiveness 5 Table 1. Costs of the stunting intervention. Cost Method of costing Total costs ( 000 US$) Generasi costs Block grants No. of villages * grant based on Government guidelines * no. of years Cost of facilitation, socialisation, monitoring Lump sum per village * no. of villages * no. of years 3474 Additional/enhanced programme costs Cost of height measuring equipment No. of villages * cost of equipment 437 Cost of micronutrient sachets, first 2 years No. of children * cost of sachet * Incentives to service providers at village level (cadres, nurses, midwives) to advocate importance of good nutrition and work on improving compliance Development of materials for technical training and advocacy training (includes cost of consultants to aid in this) Training a of cadres, nurses and midwives in measuring and monitoring growth; giving advice on growth promotion; administering vitamins, etc.; nutrition and hygiene counselling; advocacy and compliance training to improve participation and achieving targets Training of doctors, nutritionists and sanitarians in training cadres, nurses and midwives (training of trainers) Awareness training/capacity building for district and province level health officials 81,648 No. of providers * incentive 5443 Lump sum based on government regulations * no. of years 950 [Transport cost + (Days of training * accommodation expense * per diem) + training material] * number of participants [Transport cost + (Days of training * accommodation expense * per diem) + training material] * number of participants [Transport cost + (Days of training * accommodation expense * per diem) + training material] * number of participants National Awareness Campaign Cost of campaign * no. of provinces 3600 Sanitation triggering Cost of campaign * no. of villages 5443 Sub-total 107,374 Project management cost 7% of above total , Notes: a Training costs computed are higher for first-time training in the first year and are lower over the following years. We report the final costs. Further details on costing are available upon request from the authors.

8 6 L.F. Qureshy et al. moving an infant out of a low birth weight status and employ a sensitivity analysis using 3 per cent and 10 per cent. As one would expect, 3 per cent and lower rates would result in higher benefit cost ratios for investments in children. We, therefore, choose the relatively conservative social discount rate of 5 per cent for our analysis and present a sensitivity analysis using a higher rate of discount of 10 per cent Programme coverage Six provinces, with an estimated seven districts in each province, benefit from the programme. The provinces covered are West Java, East Java, Nusa Tenggara Timur (NTT), Nusa Tenggara Barat (NTB), Gorontalo and West Sulawesi, since they have under-five stunting rates at or above the national average of 36.7 per cent (RISKESDAS 2007). The programme covers 3528 villages in the first year and by year five the coverage will be expanded to 7056 villages. The programme will be implemented for five years, covering seven cohorts: five new (s 1 5) and two existing (s 1 and 2). s 1, 2 and 3 will be covered from gestation until the age of two. 4 will be covered during gestation until the age of one as it falls in the last two years of the programme implementation. 5 will only benefit from prenatal interventions, coinciding with the final programme year, as currently designed. These children would benefit further should the programme continue past the fifth year. However, for the purpose of this illustration, we will take the duration of the programme (and the associated costs) as if it had a finite planning horizon. In the year that the programme will be initiated, two existing cohorts s 1 and 2 will also benefit from the nutrition interventions though only in the post-natal stage. Children from 1 will be infants in the first year of the programme and will be covered for the first two years of their life and 2 children who will be 1-year old in the programme year 1 will be covered for one year until the age of 2 years Number of new children We obtain the average number of new children per province as the product of the Crude Birth Rate for Indonesia (Statistics Indonesia and Macro International 2008) and the population for four main provinces, West Java, East Java, NTT and NTB (Statistics Indonesia 2011). On an average, across all provinces, this amounts to 306,518 children across 3528 villages in the first year of intervention and 87 new children per year per village. The costs of the programme include (1) block grants that support the current work under Generasi and includes the cost of running the Posyandu and paying a stipend to the village-level community health workers (cadres) for their services; (2) training costs for service providers (cadres, nurses, midwives, doctors, nurses, nutritionists and sanitarians) and awareness training costs for government health officials (administrative); (3) equipment to measure the height of children; (4) micronutrient sachets for the first two years of the programme; (5) National Awareness Campaign; and (6) a sanitation triggering campaign Programme cost We estimate the total costs of the programme at million dollars over the 5-year period of implementation (see Table 1). A part of these costs are channelled into block

9 Journal of Development Effectiveness 7 grants for community services that are used to improve education as well as nutritional outcomes. We include the full costs of the block grants because it is not possible to earmark an exact proportion to education or health, given the flexible nature of the Generasi services. However, since the Generasi impact evaluation has shown an improvement in Posyandu utilisation, it forms an important component in improving the nutrition outcomes and reflects how in real-life settings actual costs can extend beyond the immediate project costs. We, therefore, incorporate the block grant costs, while restricting the benefits to gains from nutrition alone. This would likely result in a relatively conservative benefit cost ratio. We also estimate per child costs for the package, adding to the otherwise sparse literature on such costs for integrated multi-sectoral programme interventions. One such community-based integrated child care programme in Honduras estimates average cost per child less than two at US$ 6.82 (Fiedler 2003). That programme, however, focuses on costs of delivering health and nutrition services alone. Our programme costs, which include block grants, average at US$ 18 per child. Computing the costs without the block grant component yields a total cost of US$ 27.5 million with an average per child cost at US$ 4.7. However, as mentioned, the grant may increase the efficacy of delivery and we, thus, consider it integral to the project costs Benefits Reducing mortality. Wai-Poi (2011) estimated a 40 per cent reduction in all-cause under-five mortality due to the introduction of the Posyandu over We believe that the same reduction in all-cause mortality for the current study is plausible, given the evidence of a decline in infant mortality, the largest contributor to under-five mortality, under Generasi s first year of implementation, 1 the introduction of enhanced Posyandu services, coupled with a sanitation triggering campaign and a national community awareness campaign to increase utilisation and compliance. However, Lopez et al. (2006) attributed a 30 per cent reduction in all-cause mortality for children below 5 years over the period in developing countries. We use the latter estimate in keeping with our strategy of using lower bound numbers. Using a total all-cause mortality reduction of 30 per cent following Lopez et al. (2006), we attribute a 33 per cent reduction to nutrition interventions and 15 per cent to reductions in diarrhoea prevalence, following Black et al. s (2010) analysis of 193 countries. We do not incorporate a possible reduction due to the use of zinc as adjunct therapy, even if it is a promising policy, as the evidence is based on very few studies conducted in controlled environments, such as Baqui et al. (2002), who attribute as much as a 50 per cent mortality reduction to the use of combined zinc and ORT. Following from above, we estimate that the Indonesia programme would contribute 48 per cent (33 plus 15) of the 30 per cent reduction in all-cause mortality, amounting to an overall 14 per cent reduction in mortality by way of nutrition and hygiene and sanitation interventions. We now calculate the mortality rate on which we apply this 14 per cent reduction. The Indonesia Demographic Health Survey (IDHS) (Statistics Indonesia and Macro International 2008) reports under five mortality rate (UMR) at 44 per thousand live births, additively equal to the neonatal mortality rate (NMR) from birth to 1 month of age at 19 per thousand, post-neonatal mortality rate (PNR) from 1 month to 1 year at 15 and child mortality (CMR) from 1 year to 5 years at 10. However, 12 per cent of all the neonatal deaths from birth to 1 month of age are due to congenital abnormalities (Black et al.

10 8 L.F. Qureshy et al. 2010). We, therefore, revise the NMR that is potentially affected by the programme downward by this percentage. Thus, the NMR is 17 per 1000 live births and the UMR is, correspondingly, We apply to this the 14 per cent reduction in UMR due to the proposed intervention to arrive at a mortality of 35.7 per 1000 live births. 2 We split the final post-intervention UMR of 35.7 per 1000 into NMR, PNR and CMR. Using the 2007 IDHS data, we calculate the percentage contribution of NMR, PNR and CMR to the total under-five mortality. Thus, with an NMR of 17 per 1000 live births and a UMR of 41.7, the contribution of NMR to UMR is as much as 40 per cent. Likewise, we get the PNR contribution at 36 per cent and CMR at 24 per cent. We then apply these percentages to the post-intervention UMR of 35.7 per 1000 live births to get the postintervention mortality rates. Thus, for example, post-intervention NMR is the product of UMR post-intervention and the percentage contribution of this age group to UMR, resulting in 14.3 per 1000 live births. Using the post-intervention mortality rates, we arrived at the number of children remaining in the programme each year and the number of children who would otherwise have died. Among the newborns, the number to graduate to the second year of the programme will be the difference between the number of newborns and the product of the NMR and the number of newborns. Likewise, we estimate the number of children who graduate to year two of the programme using the PNR of 12.8 per For the third year of the programme, which these children enter at age one and exit at age two, we use the CMR of 8.6 per 1000 live births post-intervention. These children exit the programme at age two. In order to maintain the conservative nature of the calculations here, we assume that additional benefits of the programme do not accrue once the children have exited the programme. Thus, we use the CMR of 10 per 1000 pre-intervention to arrive at the number of children who will turn 3, 4 or 5 years. After 5 years of age, we assume a mortality rate of 5.4 per 1000, obtained as an average of child mortality and adult mortality of 2.7 per 1000, in the absence of mortality data on the 5 15 age group (Statistics Indonesia and Macro International 2007). 3 The number of children that remain alive at 15 years would then be those who will become adults. The difference between with and without intervention estimates of number of children (adults) gives us the number who would otherwise have died. Table 2 provides the estimates for one cohort as an example. The same principle has been applied to other cohorts. Table 3 summarises the number of children who live to become adults and the number of deaths averted by cohort. The number of deaths differ by cohort since, as explained, the full benefits of the programme are only realised for cohorts covered from gestation until the age of 2 years. We use the numbers from Table 3 to calculate the earnings gains from reduced LBW and stunting, as discussed further below Reducing incidence of LBW and stunting. The pre-intervention percentage of LBW in Indonesia is 8.8 (World Development Indicators 2007). Pena-Rosas et al. (2012a, 2012b), in a review of the effect of iron supplementation during pregnancy, conclude a 19 per cent reduction in LBW. We use this estimate of the impact of iron on LBW for our computations, although greater declines in LBW may result from the full package of the prenatal interventions. The resulting percentage of LBWs averted, arrived at as the difference between the post-intervention and pre-intervention percentage, is 1.7 per cent. We multiply this percentage with the number of children to arrive at the number of cases of LBW averted. We calculate this separately for the category of children who would have lived anyway and those who would otherwise have died.

11 Journal of Development Effectiveness 9 Table 2. Number of children in each programme year for 1. Programme year No. of children (1) Age (months) (2) With intervention (3) Without intervention (4) Children who would have died 1 New children per year (=pregnant women in year 0 of a cohort, year 1 of programme) 9 to 0 306, ,518 Neonatal mortality rate (per thousand) (adjusted for congenital abnormalities) a 2 Children who will live past the neonatal stage (graduate to year 2 of the programme) , , Post-neonatal mortality rate (per thousand) Children who will live graduate to 3rd year of the programme , , (enter at 1 year exit at 2 years) Child mortality rate (per thousand) Children who are 2 3 years old (4th year of the programme but , , exited already) Child mortality rate (per thousand) 10 b 10 5 Children 3 4 years (5th year of the programme but exited , , already) Child mortality rate (per thousand) Children 4 5 years , , Mortality for 5 15 years 60-adult 5 5 Number of children who become adults c 288, , Number of children who die 18,280 20, Notes: a We make this downward adjustment since congenital abnormalities cannot be reversed by nutritional interventions; b The mortality rate is assumed, conservatively, to be the same as the pre-intervention rate since these children have exited the programme; c The number of children who would otherwise have died = Column 2 Column 3 in this row or the next row.

12 10 L.F. Qureshy et al. Table 3. Number of deaths averted and children who will live to become adults by cohort Children who would have lived to become adults anyway Children who will become adults who would otherwise have died Total number of children who will become adults 286, , , , , , , , , , , , , ,473 We subtract the 8.8 per cent of LBWs from the 37 per cent of under-fives stunted in Indonesia to avoid any possible double counting, resulting in a baseline percentage of stunted children, who were, though, of normal birth weight, at 28. We then calculate the decline in stunting due to our programme based on Indonesia-specific studies. Wai-Poi (2011) finds a 25 per cent reduction in stunting because of nutrition interventions under the Posyandu. The impact evaluation report of Generasi (Olken, Onishi, and Wong 2012), on the other hand, found a decline of 21 per cent in severe stunting in the province of NTT. We use this smaller result for our baseline calculations. Assuming a 21 per cent stunting decline, the post-intervention percentage of stunted is then 22, implying that 6 per cent (28% 22%) of the children will avert stunting. Once again, we calculate the actual number of children who avert stunting, as in the LBW case, for those who would have lived anyway and those who would otherwise have died. Following Alderman and Behrman (2006), we assume that the productivity gains resulting in higher earnings for the child who has been moved from LBW to non-lbw status can be derived as the product of the impact of LBW on adult height and the incremental earnings attributed to stature. The increase in earnings as such are attributed to: (a) improved schooling outcomes, such as enrolment rates due to improved stature alone, resulting in a gain of 5.3 per cent of annual earnings, and (b) an additional 2.2 per cent, 4 which is in keeping with higher wages associated with increased stature conditional on schooling. As reported in Alderman and Behrman, wage increments such as these are commonly reported in empirical studies, although it is not always possible to distinguish the role of stature per se from associated and improved cognition nor, however, is it necessary to do so for the purpose of estimating overall benefits. From 0 to 24 months, the productivity gains from reduced stunting in terms of additional earnings are in addition to the gains from reduced LBW. Alderman, Hoddinott, and Kinsey (2006) estimate a 7 per cent increase in annual earnings from reduced stunting at this stage based on estimates from a longitudinal study in Zimbabwe as a result of increased years of schooling. To monetise the gains, we compute earnings as productivity benefits to those who would have lived anyway and those who would otherwise have died. For each of these categories we obtain the total benefit from the programme for all children who benefit.

13 Journal of Development Effectiveness 11 We assume that earnings accrue from the age of 15 to 60 years. The current median income in Indonesia is Rp.700,000 (US$ 72) per month (SAKERNAS 2007), while the average earnings in manufacturing for non-supervisory staff (Statistics Indonesia 2009) is Rp. 1,150,000 (US$ 119) per month, close to the Indonesian minimum wage. The minimum wage covers only formal workers, who currently make up around 40 per cent of the workforce. With continued economic growth, an increase in formality is likely in 15 years time, when the first children born under the programme would begin working. Even so, we use the median income of Rp.700, 000 per month accounting for annual earnings of US$ 869. We assume the work of those in informal employment as well as those involved in unpaid activities to be at least as high as the current wage offer (Schultz 1993). We assume there is no growth in real income in our baseline scenario. As an alternative scenario, we assume the actual average real income growth of 1.2 per cent per year from 1990 to 2009 (World Bank 2011) so that real earnings each year rise by this percentage. However, this period includes the Asian Financial Crisis, thus the growth used is likely lower than that of the last decade. 5 With annual earnings of US$ 869 and an increase in earnings of 7.5 per cent from a reduction in LBW from prenatal interventions, following Alderman and Behrman (2006) above, the annual incremental earnings amount to 7.5% * US$ 869 = US$ Similarly, using the 7 per cent increase in annual earnings from reduced stunting interventions targeted at 0 2 years, following Alderman, Hoddinott, and Kinsey (2006), the incremental earnings are: 7% * US$ 869 = US$ These annual increments are discounted for the present value calculations and summed over the working lifetime. s 1, 2 and 3 get the full incremental gain in earnings since they benefit from 9 months of prenatal interventions and 2 years of post-natal interventions. 4 only benefits from 9 months of prenatal and 1 year of post-natal intervention. Thus, while the incremental earnings from reduced LBW would be the same, we assume the incremental earnings for this cohort because of reduced stunting to be 50 per cent of that for s 1, 2 and 3. That is, we assume the reduction in the risk of stunting is evenly distributed between the ages of 0 24 months. Similarly, 2 also benefits from 1 year of postnatal interventions. As explained earlier on, both s 1 and 2 benefit from stunting interventions and not prenatal interventions since the children in this cohort would be born before the programme starts. 6 5 benefits from prenatal interventions but not from post-natal interventions since the children in this programme would be below 2 years when the programme will complete its course. For those who would have lived anyway, we compute the earnings benefits as the sum of the earnings from productivity increase due to reduced LBW and the productivity increase due to reduced stunting. For those who would otherwise have died, the total earnings benefits include, in addition, the annual earnings expected when these children start working. Table 4 presents the earnings for the different cohorts for these two categories of children. The total benefits of the programme from reduced stunting and LBW as well as reduced mortality (for children who would otherwise have died) is then the sum of these two streams of earnings: those that come from children who would have lived anyway and those who would otherwise have died, obtained as the product of the number of children (see Table 3) and earnings (Table 4) by cohort. The benefits are then discounted using a 5 per cent rate Reducing diarrhoea morbidity. In a WHO study (Kosek, Bern, and Guerrant 2003), analysing the global burden of diarrhoeal disease from 1992 to 2000, it was

14 12 L.F. Qureshy et al. Table 4. Summary of earnings benefits (US$). s 1, 2, Average annual earnings From reduced LBW From reduced stunting Total increase in earnings for those who would have lived anyway (Row 2 + Row 3) Total increase in earnings for those who would otherwise have died (Row 1 + Row 2 + Row 3) found that on average children in developing countries experience 3.2 episodes of diarrhoea a year. The estimates for Indonesia used by Aitken et al. (2007) are much lower at 1.3 episodes per child per year. The IDHS data reports the occurrence of diarrhoea two weeks prior to the survey. Using their data the number of diarrhoea episodes per year amount to 3.9 per child, slightly higher than the WHO average for all developing countries. In keeping with our preference for lower bound estimates within plausible ranges, our baseline estimates assume 1.3 episodes per child per year (for all age groups). A reduction in diarrhoea morbidity in the programme would potentially come from two channels: use of zinc and hygiene improvements. Baqui et al. (2002) report a 15 per cent decline in diarrhoea morbidity through the use of zinc. Esrey et al. (1991) report the average diarrhoea morbidity reduction as the median value from six studies on hygiene at 33 per cent. A number of these studies focused on handwashing alone, whereas our study involves a sanitation triggering campaign as well. A recent impact evaluation of a project undertaken in East Java concludes a 30 per cent reduction in diarrhoea prevalence in treatment groups, but suggests that this was also influenced by handwashing behaviour (Cameron, Shah, and Olivia 2013). We can, therefore, expect the programme to induce at least the same decline as Esrey et al. (1991). It is not clear whether the effects of zinc and hygiene together are independent (additive), complementary (multiplicative) or substitutable (less than additive), such that better hygiene may preclude the need to use therapeutic zinc when an episode occurs. We, therefore, exclude the benefits of zinc and assume a 33 per cent reduction overall from the programme. Starting with 1.3 episodes of diarrhoea, a 33 per cent reduction in morbidity implies a post-intervention morbidity of 0.87 episodes per child per year. We assume that benefits of the programme continue even after a child reaches the age of two, and once the child is 5- years old diarrhoea morbidity drops to zero. We assume the same number of episodes per child for all the 5 years. Savings from diarrhoea morbidity accrue from the costs saved in treating diarrhoea, which depends on: (a) cost of treating an episode of diarrhoea at home with ORS; (b) cost of outpatient visits; (c) cost of medicines; and (d) cost of hospitalisation. We obtain the percentages of children who visit the outpatient, treated at home with ORS and those treated with medicines from IDHS, The estimates for expenses on each of these categories and data on percentage of children hospitalised for diarrhoea are

15 Journal of Development Effectiveness 13 from Wilopo et al. (2009). The percentage of hospitalisations from diarrhoea in Indonesia until 5 years of age is 9.9 per cent, calculated as the ratio of total diarrhoea hospitalisations to the total diarrhoea cases, which we break up by age using available literature such as Quintanar-Solares et al. (2011). The total hospitalisations are then calculated as the episodes per child per year times the probability of hospitalisations in age group x. The total expense on hospital stay per year per child for age group x is the product of the cost of hospital stay for one episode and the number of diarrhoea hospitalisations per child per year in x age group. The cost of hospital stay incorporates both the medical and nonmedical (transport) direct costs and non-medical (wages lost for caregivers) indirect costs of hospitalisation. For the cost of physician visits, Wilopo et al. (2009) compute direct costs of medication and diagnosis, registration cost and cost of transport. We adjusted Wilopo et al. s (2009) estimates downward by excluding the cost of medicines alone since we include them in a separate category. The indirect costs are computed in terms of lost wages. The total outpatient costs as such are US$ 4.04 per visit. The IDHS reports that 52.3 per cent of 4 5 year olds consulted a health facility. This probability times the number of episodes per year gives the total number of physician visits per child per year. The cost of home treatment is the product of the probability of home treatment by age group times the number of episodes per child in this age group times the cost of home treatment. These, in turn, are the product of the cost of an Oralit sachet of US$ 0.05 (Ministry of Health 2008), the number of sachets used per day, assumed two (Rahman and Bari 1990) and the number of days an episode lasts, assumed three, less than Wilopo et al. s (2009) 5 days. The costs of treatment with pills are arrived at using the same method. We assume that the diarrhoea savings are the same across all the cohorts regardless of years of hygiene education received. Table 5 presents the costs by type of treatment and age group. As in the earnings case, we obtain the difference in the total number of diarrhoea cases pre-intervention by each age group and the number post-intervention to obtain diarrhoea cases averted. The total diarrhoea costs averted are then a product of averted cases and diarrhoea costs by age group. We assume diarrhoea savings as such to occur 1 year after the programme starts. This assumption is realistic, given that we are focusing on improving hygiene and other behaviour as a means for reducing diarrhoea morbidity and mortality. The diarrhoea savings accrue until the child is 5-years old since we assume a Table 5. Costs of diarrhoea by treatment type and age group. Mode of treatment Age group Hospitalisation Physician/health facility visit ORS (home treatment) Pills/ medicines Total expense 0 6 months months months months months months Total expense Note: The above calculations are based on an assumption of 1.3 episodes per child per year with duration of 3 days for one diarrhoea episode.

16 14 L.F. Qureshy et al. drop to zero morbidity after 5 years of age to be conservative. These are discounted, as in the earnings case, at 5 per cent Reducing chronic diseases. We compute benefits from reduced chronic diseases using the same methodology and assumptions as Alderman and Behrman (2006). Thus, chronic diseases set in at 60 years and cost 10 years of annual earnings in a developing country, assumed US$ 500 per year, resulting in a total loss of US$ The probability of reduction in chronic diseases if a baby moves to a non-lbw category is Thus, the cost of chronic diseases in year 60 for each LBW prevented is US$ 5000 times or US$ 435. This is then discounted to get the present value Results Table 6 presents the conservative baseline results for the entire programme covering all cohorts. It includes costs of a block grant programme that supports both health and education services estimated at a total of US$ million. As indicated earlier, without the block grants the costs are US$ 27.5 million. We also use a low range of assumptions for the benefits from the programme, such as applying a 30 per cent reduction in all-cause mortality, although Wai-Poi (2011) estimates a 40 per cent reduction, with a final reduction of 14 per cent to account for mortality benefits attributed to our programme alone. The overall BC ratio is 2.08, with an economic rate of return (ERR) 8 of 9.4 per cent (see Table 7, last row), implying that any discount rate below this rate would result in a benefit cost ratio greater than one. The largest contribution to the benefits comes from income gains from those who would otherwise have died at 41 per cent of the total (last column, Table 6). Savings from reduced diarrhoea are slightly greater than other productivity gains, both being more than a quarter of the total. In contrast, reduced chronic illness contributes little to the overall gains. Even with country specific data on large-scale interventions, any calculations of benefit streams requires a range of assumptions. While this study has opted for lower ranges of most assumptions in developing the baseline, we show the sensitivity of the results to alternative assumptions. Table 7 compares the baseline to a number of other scenarios and also presents an optimistic scenario. Part A of the table lists the assumptions under each scenario (column). The emboldened numbers indicate a change in an assumption relative to the baseline in the first column. The second part of the table under B presents the net present value or the difference between the discounted benefit and cost stream summed over all cohorts and the benefit cost ratios under each scenario for the entire 5-year programme. The last row under C uses the same assumptions as reported from rows numbered 2 to 14 under A but calculates the ERR. The first scenario in Table 7 applies a 10 per cent discount rate. The BC ratio falls to just under one. In the second scenario, which modifies the assumption on diarrhoea episodes, applying 3.2 episodes of diarrhoea per child per year following Kosek, Bern, and Guerrant (2003), the BC ratio rises to 4.1. Next, we assume a modest real income growth of 1.2 per cent relative to the baseline of zero growth in the economy and find a slight rise in the BC ratio to 2.3, relative to 2.1 in the baseline. This result indicates that economic growth does not make nutrition interventions less important to a development strategy. In fact, growth in other sectors contributes to higher returns from investments in health. Scenario 3 excludes the costs of the block grant to the communities, thus concentrating only on the costs of the actual nutrition programme, which are estimated at US$ 27.5 million. The BC ratio is as high as 8.4. If the costs of the inclusive block grant were