Report No. 31846-BD Bangladesh Attaining the Millennium Development Goals in Bangladesh How Likely and What Will it Take to Reduce Poverty, Child Mortality and Malnutrition, Gender Disparities, and to Increase School Enrollment and Completion? June 6, 2005 Human Development Unit South Asia Region Document of the World Bank ACKNOWLEDGMENTS This report was prepared by Ani1 B, Deolalikar (Consultant, SASHD), with analytical inputs from Binayak Sen and Zulfiqar Ali (consultants), and under the overall guidance of Julian F. Schweitzer (Sector Director, SASHD), Christine Wallich (Country Director for Bangladesh, SACBG), Mansoora Rashid (Sector Manager, SASHD), and Qaiser Khan (TTL, SASHD). Assistance from Meherun Ahmed i s gratefully acknowledged. Peer reviewers were Azizur R. Khan and Arup Banerji. Comments from Mahmudul Alam, Amit Dar, Zahid Hussain, Kees Kostermans, Farial Andaleeb Mahmud, Syed Ni- zamuddin, John L. Sinclair, Jaehyang So, Venkatesh Sundararaman, Eric Swanson, Sal- man Zaidi, and Hassan Zaman on an earlier draft of the report were helpful in preparing the current revisedversion. TABLE OF CONTENTS EXECUTIVESUMMARY ............................................................................................................................ i I INTRODUCTION . ................................................................................................................................. 1 The Millennium Development Goals.......................................................................................................... 1 Data. Methodology and Caveats................................................................................................................ 1 Overview of Bangladesh's Development Record ....................................................................................... 3 I1. CONSUMPTIONPOVERTY ........................................................................................................ 5 Overall Trends............................................................................................................................................ 5 Trends in Real Agricultural Wages............................................................................................................ 6 International Comparisons .................................................................................................................... 6 Spatial Variations....................................................................................................................................... 7 Figure II.7.................................................................................................................................................. 8 Geographical Concentration of the Poor................................................................................................... 8 Growth Incidence .................................................................................................................................. 9 Profile of the Poor.................................................................................................................................... 10 TheRole of Public Interventions .............................................................................................................. Multivariate Analysis ............................................................................................................................... 13 Simulations to 2015.................................................................................................................................. 14 I11. INFANT AND UNDER-FIVEMORTALITY ............................................................................ 17 Trends....................................................................................................................................................... 17 International Comparisons....................................................................................................................... 18 Spatial Patterns ........................................................................................................................................ 19 Proximate and Socioeconomic Correlates ............................................................................................... 20 TheRole of Public Interventions .............................................................................................................. 25 Projections to 2015 .................................................................................................................................. 27 Multivariate Analysis ............................................................................................................................... 28 Simulations to 2015.................................................................................................................................. 30 I V. REDUCINGCHILD MALNUTRITION ................................................................................... 33 Trends....................................................................................................................................................... 33 Demographic Patterns ............................................................................................................................. 38 Proximate Causes..................................................................................................................................... 39 Socioeconomic and Policy Correlates...................................................................................................... 42 Projections to 2015 .................................................................................................................................. 46 Multivariate Analysis ............................................................................................................................... 47 Simulations to 2015.................................................................................................................................. 48 V . PRIMARY SCHOOLING ............................................................................................................ 51 Overall Trends.......................................................................................................................................... 51 Spatial Patterns .............................................. ..................................................................................... 53 Geographic Concentrationof Out-of-School Children ............................................................................ 54 Socioeconomic Variations........................................................................................................................ 54 Socioeconomic Variations........................................................................................................................ 55 The Role of Public Interventions .............................................................................................................. 57 Multivariate Analysis ............................................................................................................................... 59 Simulations to 2015.................................................................................................................................. 62 V I. GENDER DISPARITY INSCHOOLING .................................................................................. 64 Trends....................................................................................................................................................... 64 Gender Patterns by Age............................................................................................................................ 65 Female Secondary School Stipend Program............................................................................................ 66 Socioeconomic Variations........................................................................................................................ 67 The Role of Public Interventions .............................................................................................................. 68 Multivariate Analysis ............................................................................................................................... 69 VI1 . CONCLUSIONS ........................................................................................................................... 71 TABLES 11.1 Assumptions about various interventions to reduceconsumption poverty. 2001 to 2015..................15 111.1 Assumptions about various interventions to reduce under-five mortality. 2001 to 2015 ...................30 IV.1 Assumptions about various interventions to reduce the child under-weight rate. 2000 to 2015 ........48 V.1 Assumptions about various interventions to increaseprimary school enrollment and Completion rates. 2000 to 2015 .................................................................................................. 62 FIGURES 11.1 Headcount of poverty (%). by residence. 1991-92 to 2000 .................................................................. 5 11.2 Trends in the nominal and real agricultural daily wage rate. 1983-2002.............................................. 6 11.3 6 Headcount of poverty (%). by geographical area. 1991-92 to 2000 ..................................................... Poverty incidence (5) in the 1990s (national poverty lines). South Asia.............................................. 11.4 7 11.5 Headcount of poverty in 2000 as % of poverty in 1991.92. by geographical area ............................... 7 11.6 Relationship between rate of poverty reduction during 1991-2000 and initial poverty 8 Regional contributions to poverty in Bangladesh. 2000....................................................................... Headcount (1991-92) across geographical area............................................................................. 11.7 8 11.8 The growth incidence curve ................................................................................................................. 8 11.9 Headcount ratio of poverty (%). by occupation of household. 2000 .................................................. 10 11.10 Headcount ratio of poverty (9%). by amount of owned land. 2000...................................................... 10 11.11 Headcountratio of poverty (%). by household size and sex and age of householdhead. 2000 .........10 11.12 Headcount ratio of poverty (%). by schoolingyears of highest-educated male or female in the household. 2000..................................................................................................................... 11 11.13 Headcount ratio of poverty (%). by availability of infrastructure in village. 2000............................. 12 11.14 Headcount ratio of poverty (%). by presenceof government program in village. 2000..................... 12 11.15 Projectedpoverty headcount ratio to 2015. under various intervention scenarios (graph shows cumulative effect of each additional intervention) ........................................................... 15 111.1 Infant mortality rate. 1911-99............................................................................................................. 17 111.2 Infant mortality rate estimated from the Vital RegistrationSystem and the Bangladesh 17 111.3 Infant mortality rate. 1970.2000. selectedcountries in Asia .............................................................. DHS. 1980-2002.......................................................................................................................... 18 111.4 Infant mortality rate. by division. 1999-2000..................................................................................... 18 111.5 Infant mortality decline by residence & division. 1993-94 to 1999-2000.......................................... 19 111.6 Relationship across districts between severe child malnutrition and under-five mortality. 111.7 Measles immunization coverage (%) among children aged 1year or less. 1991-2003 ...................... 1995 & 2000................................................................................................................................ 20 21 111.8 Measles vaccination coverage among 12-23 month olds. by division and urbadrural residence. 1999............................................................................................................................ 21 111.9 Infant and under-five mortality rates. by previous birth interval. 1993-94 and 1999-2000 ................21 111.10 Infant and child mortality rates. by sex of child. 1993-94 and 1999-2000 ......................................... 22 111.11 Infant and under-five mortality rates. by mother's schooling. 1999-2000 ......................................... 23 111.12 Infant mortality decline by maternal schooling. 1993-93 to 1999-2000............................................. 23 111.13 Infant and under-five mortality rates. by wealth quintiles. 1996-97................................................... 24 111.14 Immunizationrates among children ages 12-23 months. by wealth quintiles. 1996-97..................... 24 111.15 Under-five mortality rate and district infrastructure. 2000................................................................. 25 26 111.17 Actual and projected infant mortality. 1981-2015.............................................................................. 111.16 Infant mortality rates in MCH-FP area and control areas. 1978-2000................................................ 27 111.18 Projected under-five mortality rate to 2015 (graph shows cumulative effect of each intervention) ........................................................................................................................ 30 IV.l child underweight and stunting rates. 1999-2000 (% of children inrelevant age group who are underweight or stunted) ................................................................................................. 33 IV.2 Relationship between (96) of underweightchildren aged 0-5 years and per capita GDP 34 IV.3 Underweight rates in South Asia. circa 2000 ..................................................................................... across 16 Asian countries. 1995-2000......................................................................................... 34 IV.4 Child underweight rates in Bangladesh. 1985-2000 (% of children inrelevant age group IV.5 Child underweight rates among children aged 6-71 months. by residence. 1992-2000 ..................... Who are underweight) ................................................................................................................. 35 IV.6 Percentage of children under 5 who were underweight. by division. 1996-97 and 1999-2000..........35 36 IV.7 Divisional distribution of all underweight children under 5 inBangladesh. 1996-97 and 1999-2000............................................................................................................................. 36 IV.8 Regional contribution to the total number of underweight children aged 6-71 months in 37 IV.9 Child underweight rates (%). by age. 2000 ........................................................................................ Bangladesh. 2000 ........................................................................................................................ 38 IV.10 Percent of moderately and severely underweight children. by age and sex. 2000.............................. IV.11 Severe underweight rates among children aged 6-71 months. by sex and birthorder, 2000 ..............38 39 IV 12 Underweight rates among 6-23 months old children. by feeding practice. 2000................................ . 40 IV.13 Underweight rates (%) among children aged 6-71 months. by average daily calorie consumptionper capita of household. 2000 ................................................................................ 41 IV.14 Average daily calorie intake per capita inthe households of children aged 6-23 months. 41 IV.15 Per capita daily availability of calories. South Asia. 1961-99............................................................ by per capita expenditure quintile. 2000 ..................................................................................... IV.16 Child underweight rates (5%) among children 6-23 months of age, by mother's education. 2000 ......42 42 IV.17 Child underweight rates (among children aged 6-23 months). by drinking water source IV.18 Underweight rates among children 6-71 months of age. by access to health facilities. 2000.............43 and type of toilet. 2000................................................................................................................ 43 IV.19 Child malnutrition rates among children aged 6-71 months. by village electrification status. 2000........................................................................................................... 44 IV.20 Underweight rates among children 6-71 months of age. by natural disaster in village in the past 5 years. 2000 ........................................................................................................................ 44 IV.21 Child underweight rates (6-71 months), by presenceof program in village, 2000............................. 45 IV.22 Actual and projected child underweight rates, 1985-2015 ................................................................. 47 IV.23 Projected 5% of children under 6 who are underweight to 2015, under different Interventions scenarios (graph shows cumulative effect of each additional intervention) ..........49 v.1 52 Net primary enrollment rate. by area. 2000........................................................................................ Percent of children attending school and primary school. by age. 2000............................................. v.2 53 v.3 Primary completion rate (% of 12-year olds who started and completed 5 years of primary school). by area. 2000 .................................................................................................... 53 v.4 Regional contribution to the total number of out-of-school children aged 6-10 in Bangladesh, 2000 ........................................................................................................................ 54 v.5 Net primary enrollment rate and the primary completion rate, by per capita expenditure quintile, 2000............................................................................................................................... 55 V.6 Distributionof out-of-school children aged 6-10, by per capita consumption expenditure quintile, 2000............................................................................................................................... 55 v.7 Net primary enrollment rate and the primary completion rate, by schooling years of highest educated adult female inhousehold, 2000 ...................................................................... 56 V.8 Net primary enrollment rate and the primary completion rate, by occupation of household, 2000 ...... ............................................................................................................... 56 v.9 v.10 Net primary enrollment rate and the primary completion rate, by village infrastructure, 2000..........57 Distributionof out-of-school children aged 6-10, by occupation of household, 2000 ....................... 57 v.11 Netprimary enrollmentrateandtheprimary completionrate, bytype ofgovernment assistanceprogram in village, 2000............................................................................................. 58 v.12 Projectednet enrollment rate to 2015, under various intervention scenarios (graph shows cumulative effect of each additional intervention) ...................................................................... 61 V.13 Project primary completion rate to 2015, under various intervention scenarios (graph shows) Cumulative effect of each additional intervention) ..................................................................... 63 VI.1 Ratio of females to malesinprimary school. 1991-2000................................................................... 64 VI.2 Ratio of females to malesin primary and secondary schools. by area. 2000 ..................................... 64 VI.3 65 VI.4 Percent of boys attending school inBangladesh and India. by age. 2000 .......................................... Percent of children attending school. by age and sex. 2000 ............................................................... 65 VIS Percent of girls attending school inBangladesh and India. by age. 2000........................................... 66 VI.6 Ratio of females to males in primary and secondary school. by per capita expenditure Quintile. 2000.............................................................................................................................. 66 VI.7 Ratio of females to males inprimary and secondary school. by schooling years of highest Educated adult female inhousehold. 2000.................................................................................. 67 VI.8 Ratio of females to males in primary and secondary school. by occupation of household and sex of household head. 2000................................................................................................. 67 VI.9 Ration of females to males in primary and secondary schools. by infrastructure availability in village. 2000 ......................................................................................................... 68 VI.10 Ratio of female to male students in primary and secondary schools. by pupil-teacher ratio inthe village primary school. 2000 ............................................................................................. 68 VI.11 Ratio of females to males inprimary and secondary schools. by type of government program in village. 2000 ............................................................................................................. 69 BOXES 11.1 Home-BasedNeonatalCare: Results from a Field Trial inRuralMaharashtra. India ....................... 32 MAPS 11.1 District-levelconsumption poverty headcount rates (%). 1995............................................................ 9 111.1 District-level map of under-five mortality rates, 2000 ....................................................................... 19 IV.1 District-level estimates of severe malnutrition among children (based on the Mid Upper Arm Circumference indicator) .....,......................................................37 V.l District-level estimates of the net primary enrolment rate, 2000........................................................ 54 EXECUTIVE SUMMARY 1. This report focuses on the attainment of five major human development-related MDGs in Bangladesh - consumption poverty, infant and under-five mortality, child mal- nutrition, schooling enrollment and completion, gender disparities in schooling. The se- lection of these MDGs for detailed analysis was based in large part on the availability of reliable sub-national data. The report concludes that of these MDGs, Bangladesh has al- ready attained (or nearly attained) the goal relating to elimination of gender disparity in schooling opportunities. Bangladesh i s the only country in South Asia other than Sri Lanka to have achieved parity in male and female enrollments not just at the primary level but also at the secondary level. This i s an impressive achievement for a country that i s one of the poorest countries in the world, with a per capita gross national income of only US$1,770 (inPPPterms) in2002. 2. The analysis in this report suggests that attainment of two other MDGs - in par- ticular, the reduction of consumption-poverty and under-five mortality - i s also feasible with a combination of interventions, including sector-specific interventions (such as ex- panding immunization coverage and reducing pupil-teacher ratios), economic growth, improved coverage of infrastructure, and social safety-net programs (such as the District Education Stipends Program and the Vulnerable Group Development programs). 3. However, it will be challenging for Bangladesh to attain the child malnutrition- related MDG as well as the education MDGs relating to universal net primary enrollment and primary completion. In the case of child malnutrition, the projections suggest that Bangladesh could come very close to - within 5 percentage points of - the MD goal of having no more than 34% of its children underweight. However, it will be very challeng- ing for the country to attain rates of net primary enrollment and primary completion ex- ceeding 83-86% by 2015. 4. These achievements represent extraordinary progress for a country that, until re- cently, was frequently derided as an "international basket case." Indeed, a recent article by Dreze (2004) suggests that Bangladesh i s now ahead of India on most social indica- tors. Bangladeshhas lower infant and maternal mortality rates, higher child immunization rates, better access to `improved' water sources and sanitation, and higher primary en- rolment rates than India. As noted earlier, Bangladesh has eliminated the gender gap not only in primary education but also in secondary education, while India still has a signifi- cant gender gap at both levels. Dreze admits that "Bangladesh i s no paradise of human development,. . .but social indicators are improving quite rapidly not just for a privileged elite but also for the population at large." On the other hand, Dreze contends that ... in " India, social progress i s slower and less broad-based, despite much faster economic growth. This is one indication, among many others, that India's development strategy i s fundamentally distorted and lop-sided." 5. What accounts for the extraordinary progress in improving social indicators in Bangladesh (relative to the progress made by India)? Dreze provides one possible an- i swer. According to him, Bangladesh's better performance may have to do with the fact that public expenditure on health as a proportion of GDP i s almost twice as high in Bang- ladesh (1.5%) as in India (0.9%). This was not always so. In 1990, Bangladesh spent only 0.7% of its GDP on health - less than what India spent (0.9%) (UNDP 2004). Thus, Bangladesh saw public spending on health increase very sharply during the 1990s, while India experienced stagnation inpublic spending on health (inrelation to GDP growth). 6. While Dreze does not note differences between the two countries in terms of their public spending on education, it i s instructive to look at public educational expenditures inBangladesh and India as well. In 1999-2001, India's public spending on education was 4.1% of its GDP - considerably greater than public spending on education in Bangladesh, which was only 2.3% of GDP (UNDP 2004). However, as in the case of health, public expenditure on education in Bangladesh increased from 1.5% of GDP in 1990 to 2.3% of GDP in 1999-2001- an increase of more than 50%. In contrast, public spending on edu- cation as a share of GDP increased by merely 5% over the same period in India - from 3.9% to 4.1% of GDP. Additionally, there i s an important difference betweenBangladesh and India in the composition of public spending on education. While Bangladesh spends 45.1% of its total public expenditure on education at the pre-primary and primary level, the relevant figure for India i s 38.4%. At the other extreme, India spends 20.3% of its to- tal public spending on education at the tertiary level, in contrast to Bangladesh's 11.1% (UNDP 2004). Thus, the rapid growth of public spending on education and health in Bangladesh, combined with its better balance of educational spending across the primary and tertiary sectors (relative to India), are likely to be important factors in explaining the significant progress the country has made inits social indicators during the 1990s. 7. Another factor that i s likely to be important in explaining Bangladesh's relative success in attaining positive social outcomes i s the work of its NGOs. Bangladesh may well be the world's leader in using NGOs as vehicles of development. NGOs are in- volved in virtually every activity in the country - relief and rehabilitation, poverty alle- viation, health, education, social protection, and environmental protection, to name a few. A villager in Bangladesh can send his or her child to an NGO school, have family plan- ning and basic health services delivered by an NGO health worker, obtain micro-credit financing from a choice of several NGO banks, sell milk and other dairy products to an NGO dairy cooperative, and make a telephone call on an NGO telephone! Secondary education in Bangladesh i s almost entirely provided by the non-government sector - viz., the NGOs, for-profit schools, and religious schools (madrasas). Likewise, many of the family planning programs of the 1970s and 1980s, which set the stage for the subsequent decline in child mortality, were primarily delivered through NGOs. And several studies suggest that micro-credit programs, which were pioneered by one of the best-known NGOs inthe world, the Grameen Bank, have had a significant effect on reducing poverty, especially among females. 8. NGOs in Bangladesh differ from NGOs in other developing countries in an im- portant way: ". ..Several o f these organizations have become very large, very profes- sional, and they have become a model for others. Bangladesh, one of the poorest coun- 11 tries in the world and the last place you would have expected this to happen, has really become a leader in showing what the voluntary sector can do', (Smillie 1998). 9. Yet another factor in explaining Bangladesh's success, especially its ability to eliminate gender disparity in enrollment even at the secondary level, i s the use of targeted public interventions, such as the Female Secondary School Stipend Program (FSSS). The FSSS program i s essentially a Conditional Cash Transfer (CCT) or a demand-side inter- vention for rural girls (the majority of whom are poor) to attend secondary school. By all indications, the FSSS program has been hugely successful in increasing female secondary school enrollments, especially since secondary schooling in Bangladesh i s not free and parents are often unwilling to invest inthe secondary schooling of their daughters. 10. However, Bangladesh's progress on the MD indicators during the 1990s does not mean that there are no problems going forward. Indeed, there are several areas of concern highlighted in this report. First,there are very large regional disparities in virtually all of the MD indicators in Bangladesh. Districts such as Noakhali, Pathuakali, Chittagong, Ra- jshahi, and Sylhet have generally not performed well on several of the MD indicators. Even if Bangladesh as a whole attains some of the MDGs, there will be several areas of the country that will remain distantly behind. The analysis in this report suggests that many o f the MD indicators are geographically concentrated in a few regions. This inturn means that targeting interventions, central government resources, and economic growth opportunities to the lagging divisions and districts will speed up attainment of the MDGs. 11. Second, the problem of governance - in particular, poor service delivery - i s widespread in the social sectors in Bangladesh. Doctors, health workers and teachers are typically absent from their assigned posts at government health centers and schools. Membership of school management committees i s highly politicized, and teacher re- cruitment i s often subject to personal influence. Procurement of textbooks and essential drugs i s rife with corruption. The quality of health and education services offered at most government health facilities i s generally very poor. Yet the evidence presented in this re- port indicates the tremendous importance of service delivery in influencing MD out- comes. Infant and under-five mortality rates have fallen most in areas where effective family planning and MCWFP programs are delivered to rural women with low schooling; female school enrollments have increased thanks to a well-designed and well-delivered secondary stipend program that reaches its intended beneficiaries; and public transfer programs that deliver food supplies to the vulnerable in rural areas, such as Food-for- Work, Vulnerable Group Feeding and Vulnerable Group Development, are associated with large reductions in child malnutrition among the poorest children. This suggests that better governance, and improved delivery of social services in particular, would be very important to attaining the MDGs. 12. Better delivery of public services - whether in health, schooling, nutrition, or in- frastructure - i s a complex and difficult task that entails creation of the right institutions and incentives, including devolving responsibility for service delivery to local govern- ments and communities, contracting out certain types of service delivery to the non- government sector, empowering consumers to demand better services from government ... 111 health facilities, introducing competition among public providers, and ensuring the moti- vation of front-line workers (World Bank 2003). 13. There are some other findings in this report that are useful to reiterate. The report notes there i s evidence of significant synergies among the different MDGs. For instance, a reduction in the proportion of underweight children i s strongly associated with a reduc- tion of child mortality. Although maternal mortality i s an MD indicator that has not been analyzed in this report, it i s clear that interventions that reduce maternal mortality, such as tetanus immunization, expansion of antenatal care coverage, and an increase in the ra- tio of professionally-attended deliveries, will also bring about large reductions in infant (especially neonatal) mortality. Likewise, reducing child malnutrition i s likely to result in both schooling quantity and quality, as better nourished children are more likely to attend school and perform better in school. Thus there are synergies amongst the MDGs that will help in their attainment, which implies that proceeding with simultaneous action on all these measures will have the greatest impact on attainment of the MDGs. 14. At the same, it needs to be realized that the different MDGs are not necessarily internally consistent. For instance, simultaneous attainment of the poverty and child mal- nutrition MDGs by Bangladesh would result in 30% of the population being poor but 34% of the children being underweight. The contrast i s even greater when the results of the simulations undertaken in this report are considered. We find that, under plausible scenarios, Bangladesh could bring down its poverty headcount rate to 16% by 2015, but it would still have many as 39% of its children underweight. Thus, a large number of chil- dren who are classified as non-poor would in fact be underweight. This inconsistency in- dicates a problem in the manner in which poverty andor underweight thresholds are es- tablished. 15. The simulations carried out in this report also suggest that economic growth that brings about an improvement in household living standards is strongly associated with virtually every MD indicator. For example, real per capita GDP growth of 4% per annum inBangladesh could alone bringdown the under-five mortality rateby about 8 deaths per 1,000 live births and the incidence of poverty by 21 percentage points between now and 2015. In addition, this growth could bring about an increase in the net primary enrollment rate of 5 percentage points by 2015. In other words, rapid economic growth could make significant contributions to an improvement in all the MD indicators between now and 2015. 16. Finally, the importance of systematically monitoring MD outcomes at disaggre- gated levels and evaluating the impact of public programs cannot be overemphasized. There i s a paucity of reliable, time-series data on most MD indicators at the district and upazila (sub-district) levels. The lack of such data makes it virtually impossible to moni- tor progress toward attainment of the MDGs at lower levels of administration. In addi- tion, with the exception of a few food assistance and micro-credit programs, most public interventions in Bangladesh have not been subjected to rigorous, independent evaluation. Inorder to choose the right set of interventions with which to attain the MDGs, it is criti- iv cal to know which programs have been successful inimproving MD indicators and which have not. V I.INTRODUCTION The MillenniumDevelopment Goals 1.1 Since the launch of the Millennium Development Goals (MDGs) at the Millen- nium Summit held in New York in September 2000, the MDGs have become the most widely-accepted yardstick of development efforts by governments, donors and NGOs. The MDGs are a set of numerical and time-bound targets related to key achievements in human development. They include halving income-poverty and hunger; achieving uni- versal primary education and gender equality; reducing infant and child mortality by two- thirds and maternal mortality by three-quarters; reversing the spread of HIV/AIDS; and halving the proportion of people without access to safe water. These targets are to be achieved by 2015, from their level in 1990 (UnitedNations 2000). 1.2 Almost all the countries in the world, including Bangladesh, have committed themselves to attaining the targets embodied in the Millennium Declaration by 2015. Un- fortunately, there i s little understanding of whether Bangladesh will be able to attain all of the MDGs, and whether there are some MDGs that Bangladesh will be able to attain. There i s even less understanding of what it will take -by way of economic growth, infra- structural investments, and social-sector interventions - to attain the different MDGs. Further, this report argues the importance of disaggregating the MDGs for Bangladesh, given the very large geographical and socioeconomic variations in millennium develop- ment (MD) indicators across the country. 1.3 This report focuses on the attainment of five major human development-related MDGs by sub-national units in Bangladesh - consumption poverty, under-five mortality, child malnutrition, schooling enrollment and completion, and gender disparities in schooling. The selection of these MDGs for detailed analysis was based in large part on the availability of reliable sub-national data. For example, reliable data on disease preva- lence at the district or divisional level are simply not available, and this hampers useful sub-national analysis of the communicable disease-related MDG.The same i s true of an- other important MD indicator - maternalmortality. Data, Methodology and Caveats 1.4 Data. Virtually all of the analysis in this report i s based on three sets of national household surveys. First, data from three rounds of the nationally-representative Bangla- desh Demographic and Health Survey (BDHS), which were collected in 1993-94, 1996- 97, and 1999-2000, are used to analyze the levels and correlates of infant and under-five mortality and malnutrition (NIPORT 2001). Second, unit record data from the 2000 Child Nutrition Survey (CNS) conducted by the Bangladesh Bureau o f Statistics (BBS) are also used to analyze the levels and correlates of child malnutrition. Third and finally, unit re- cord data from the 2000 Household Expenditure and Income Survey (HIES), also con- ducted by the BBS, are used to analyze the levels, patterns, and correlates of consumption poverty, schooling enrollment and completion, and gender disparity in schooling. 1.5 Methodology and Assumptions. The methodological approach adopted in this report i s roughly as follows. We apply econometric estimation techniques to district or household data in order to analyze the socioeconomic and policy correlates of the se- lected MD indicators. These estimates are then used to simulate the likely trajectory of the MD indicators under alternative scenarios of change between now and 2015. 1.6 For projecting the time-path of the different MD outcomes to 2015, we consider a few common scenarios. One of these i s that mean consumption expenditure per capita in Bangladesh will grow annually at about 2.7% between now and 2015. During the 1990s, per capita consumption expenditure in Bangladesh grew by 2.2% per annum, while per capita GDP increased by 3.3% over. Assuming that the same relationship holds in the fu- ture, the 2.7% consumption growth we have assumed would be consistent with an annual GDP per capita growth rate of 4%; in other words, we assume that economic growth will be somewhat (but not substantially) higher in the future than it was during the 1990s. The other assumption that i s common to virtually all the simulations i s an increase in adult male and female schooling. Here our assumption i s that both male and female adult schooling will increase by 0.3 years annually between now and 2015. Such growth would result in mean schooling reaching a level of 7 years for females and 9 years for males by 2015 (from their 2000 values of 4.5 and 2.5 years, respectively). Admittedly, these as- sumptions are arbitrary, but, given the enormous.investments being made in education in Bangladesh during the 1990s, the assumptions arexrealistic and likely to materialize over the coming decade. More importantly, as i s noted throughout the report, none of the as- sumptions made are sacrosanct; they are only meant to illustrate the range of MD out- comes under a set of possible scenarios. The projections could be undertaken for any combination of changes inthe policy or environmental variables. 1.7 Additional sector-specific assumptions are made for projecting the individual MD outcomes.' For instance, in the poverty simulations, we assume that consumption ine- quality will continue to rise and per capita land availability will continue to decline at roughly the rates at which these variables have changed during the decade of the 1990s. In the nutrition simulations, it i s assumed that sanitation coverage and coverage of the Food-for-Work program will continue to expand and that more effective flood-control measures will reduce slightly the vulnerability o f Bangladeshis to floods. Prior to each simulation, the full set of assumptions made for the simulation i s detailed inthe text. 1.8 Caveats. By its very nature, any empirical analysis i s predicated on assumptions about data quality and measurement, inferences of causality between variables, and po- tential biases of statistical and econometric estimates. The analysis presented in this re- port i s not immune to these same concerns. It i s therefore important to note at the outset that while the results and simulations presented in this report may give an impression of The choice of `policy' variables to use in the simulations i s largely dictated by the econometric model used. In other words, only policy or environmental variables that have statistically significant associations with the MD outcome variable in the econometric analysis are usedin the projections. 2 precision, they are not that.2They should be treated as indicative of possible broad trends, and could usefully be complemented with other analyses using different methodological approaches. As long as the results are used with this understanding, they can be helpful in `rough-order' planning for MDGattainment. 1.9 Finally, it i s important to note an important limitation of the simulations per- formed in this report. The simulations are based on statistical analysis o f household sur- vey data. B y its very nature, such analysis tends to over-emphasize readily-measurable variables, such as household income or consumption, adult schooling levels, and access to infrastructure, and under-emphasize qualitative variables, such as the quality of institu- tions, governance, and empowerment. Obviously, this does not imply that the latter vari- ables are irrelevant to the MD indicators; indeed, institutional reform and good govern- ance are critical to the attainment of the MDGs. It i s therefore important to view the mes- sages of this report as complementing those from the numerous qualitative (and detailed) studies o f health, nutrition, schooling and poverty that have been conducted inthe past. Overview of Bangladesh's Development Record 1.10 Bangladesh has achieved considerable success in the areas of lowering population growth, fostering women's empowerment, reducing aid dependence, achieving success in human development, maintaining a decent level of macroeconomic stability with pro- nounced outward orientation, overcoming the shadow of famine, attaining effective disas- ter management capacity, and promoting NGOs as an alternative delivery mechanism. Not many countries at Bangladesh's level of income can list so many o f these achieve- ments. Especially inlight of Bangladesh's dismal record at development during the 1970s and 1980s, this i s remarkable progress indeed! 1.11 It i s true that the level of social indicators in the country is still modest compared to the historical standards set by the East Asian economies or, for that matter, the stan- dard set by neighboring Sri Lanka and the Indian state of Kerala. However, given Bang- ladesh's adverse initial conditions, high population density, vulnerability to disaster, and low stock of natural resources, its achievements are impressive. Since Bangladesh repre- sents livelihoods at the margin, its success at improving the quality of human existence under the most extreme of conditions holds a very important lesson for all developing countries. 1.12 Bangladesh has witnessed the sharpest decline in infant mortality of any develop- ing country (Stern 2002). It i s widely seen as a successful case in the area of population control in the face o f low income and low literacy through favorable public policy toward family planning, women's empowerment, and community involvement (Sen 1999, BIDS 2001, Dev et al. 2002). The virtual elimination of gender disparity in the enrollment rate up to the junior secondary level is another major accomplishment. Although the country i s yet to attain a sex ratio consistent with the expected biological advantage associated 2 In addition to lack of precision, the estimates presented in this report, like other econometric estimates, may be subject to systematic biases arising from measurement errors in the independent variables and from the omission o f important variables and unobserved heterogeneity from the analysis. 3 with higher female survival, it has already achieved gender parity inlife expectancy. As a result of all these achievements, Bangladesh, for the first time in its independent history, has graduated from being the "test case of development" (Faaland and Parkinson 1975) to being classified as a "medium human development" country (UNDP 2003). 11. CONSUMPTIONPOVERTY Overall Trends 2.1 At the time of its independence, the incidence of poverty in Bangladesh was very high, with nearly three-quarters of the country's population in 1973-74 being poor (Hos- sain and Sen 1992). The country has made considerable progress in reducing poverty since then, especially during the nineties. 2.2 Figure 11.1 shows Headcountof poverty (S),by residence,1991-92to 2000 the headcount ratio of 61 poverty, using data from 65 1 60; the Household Income and Expenditure Survey (HIES) and a new pov- 5 0 - erty line established by 45 the BBS and the World Bank (World Bank 40.' 2003). The figure sug- gests that, nationally, poverty fell by 9 per- 30 centage points during the 9 years between 1991-92 and 2000 - an annual National Urban Rural rate of decline of one Figure11.1 percentage point.3 From all accounts, the decline in poverty was more rapid in the 1990s than during earlier decades, possibly because of the more rapid pace of economic growth during this period. Annual per capita consumption expenditure, which grew at an annual rate of 0.6% between 1983-84 and 1991-92, grew more than four times as fast (2.7% an- nually) between 1991-92 and 2000. National income data show the annual growth in per capita GDP accelerating from about 1.5% inthe 1980s to nearly 3% during the 1990s. 2.3 Figure 11.1 also suggests that the incidence of poverty declined slightly more in the rural than in the urban areas of the country. In 1991-92, poverty in the urban areas was 73% of that inthe rural areas; by 2000, the ratio had declined further to 70%. 2.4 The data in Figure 11.1 also suggest that the rate of decline in poverty was more rapid in the early part of the decade than in the latter part. However, other data do not corroborate this differential performance across the early and the late 1990s (World Bank 2003). Inwhat follows, we largely use the poverty rates for 1991-92 and 2000. Note that the reduction in poverty duringthe 1990s i s smaller if household income (instead of consump- tion) i s used to measurepoverty. 5 Trends in Real Agricul- tural Wages Trends in the nominaland real agricultural daily wage rate, 1983-2002 70 I 2.5 Since the poorest I o f the poor typically tend 60 - +Nominal wage rate to be agricultural labor- ers, it may be worthwhile 50 1 looking at the trends in I agricultural wages. The trends in the agricultural 40 - daily wage rate are shown in Figure 11.2. The 30 - real agricultural daily wage rate for male labor- 20 - ers increased from about l0 20 taka in 1983-84 to 24 taka in 1991 - an annual increase of 2.3%. But it 1983-84 1988-89 1991 1995-96 2000 2001 2002 largely stagnated during Figure 11.2 the-1990~~increasing by only 1.3% annually between 1991 and 2000.4 International Poverty incidence(%) inthe 1990s(national poverty lines),SouthAsia Comparisons 2.6 How does 55 - the trend in pov- erty incidence in Bangladesh during 506 the 1990s compare 45 -I +-Bangladesh India t-Pakistan with the trends ob- served in other countries of the region over the same period? Fig- ure 11.3 shows the poverty headcount ratios (based on 25 national poverty lines) in Bangla- desh, India and FigureII.3 Pakistan at differ- ent points during the 1990s. While there are serious methodological problems in compar- ing poverty incidence across countries, a broad conclusion about the pace of poverty re- 4 This i s so if one uses the rural CPI deflator. However, if the price of coarse rice is used as a deflator, the increase in real wages i s greater in the 1990s. 6 duction can be made - viz., Bangladesh's performance on poverty reduction during the 1990shas bested that of Pakistan, where poverty was largely stagnant duringthe last dec- ade, but it has fallen short of poverty reduction in India, where the poverty headcount ra- tio fell from 36% in 1993-94 to 26% in 1999-2000- an annual decline of 1.7 percentage pointsa5 Headcount of poverty (%), by geographical area, 1991-92 to 2000 SpatialVariations 0 1991-92 m2000 2.7 There are large variations in poverty in- cidence across geo- graphical areas. Figure 11.4 shows that the pov- erty headcount ranges from 27% in Other Urban Dhaka to 65% in Rural Rajshahi and Pabna. The rural areas of Faridpur, Tangail and Jamalpur, as well as the rural parts of Figure 11.4 Bogra, Rangpur and Di- najpur, are among the poorest regions of the country. Headcountof poverty in2000 as % of poverty in 1991-92,by geographicalarea 2.8 Figures 11.4 and 11.5 104 104 also suggest wide regional variation in the pace of pov- erty reduction during the 1990s. For example, poverty fell by 50% between 1991- 92 and 2000 in Other Urban Dhaka and by 37% in rural Barisal and Pathuakali. In contrast, poverty actually ,;." +- ,$+Q $' e' ,e<.,@*9,9"'d rose (albeit by a modest 4%) oQE g#+ 4.c" ,9" &'* \& 9.'"$' in rural Sylhet, Comilla, 8'Q B gp &Y,J4% Noakhali and Chittagong. ,&+ d Even Metropolitan Chit- Figure 11.5 tagong saw virtually no change in poverty during this period. 5 There has been some controversy about the Indian poverty figures. Using an alternative methodology, Deaton and Dreze (2002) report the decline to be from 29% to 22% (Deaton and Dreze 2002), which works out to an annual reduction of about 1.2 percentage points a little more than that recorded for Bangladesh. - 7 2.9 In general, the pace of poverty reduc- Relationshipbetween rate of poverty reduction during 1991-2000 and initial poverty headcount (1991-92) across geographical areas tion during the 1990s was somewhat more 35 10 45 50 55 60 65 70 75 5 a - rapid in areas that had 0 w l 0 higher initial levels of poverty in 1991 (Figure 11.6), which implies modest regional con- vergence in headcount rates. goverty 2.10 Map 11.1 i s a map of Bangladesh 0 showing district-level variations in consump- Poverty headcount(Yo) in 1991-92 tion poverty in 1995.7 Figure11.6 As is obvious from the map, poverty i s pervasive in the country, with a large number of districts throughout the country having highrates of poverty. Geographical Con- centration of the Regional contributions to poverty in Bangladesh, 2000 Poor 2.11 Since some o f the most populous regions in Bangla- desh also have the highest rates of pov- erty, the poor are heavily concentrated geographically. Fig- ure 11.7 shows that three regions - com- prising the rural ar- eas of Bogra, Rang- Figure11.7 pur, Dinajpur, Faridpur, Tangail, Jamalpur, Sylet and Comilla - account for 42% of all the poor in the country. Six regions, out of a total of 14, account for nearly three-quarters o f all the poor. The geographical concentration of the poor has important implications for the targeting of poverty-alleviation interventions. Note that this does not necessarily imply reduction in overall consumption inequality over time. Indeed, the Gini coefficient increased from 0.26 in 1991-92 to 0.31 in 2000. More recent district estimates of consumption poverty are not available. The poverty rates shown in the map are based on the 'old' national poverty line, and therefore are not comparable to those discussed earlier inthis chapter. 8 Growth Incidence 2.12 The growth inci- dence curve for Bangla- 4 - desh, which shows the relative gains in per cap- 3 5 - , ita consumption during B5BE 'k 3 3 1991-92 to 2000 for dif- 5L ^ 1 ferent deciles of the .-c 2 5 population, i s reproduced 5 2 l\...."Iu"L".ll.E" ................................................. ................. from World Bank (2003) " and shown in Figure 11.8. -52 a 15 ean of gowth rates The entire curve lies 2 I above zero, indicating 0 10 20 30 40 50 60 70 80 90 100 that all segments of the % of population ranked by per capita expenditure population experienced growth in living stan- Figure 11.8 dards over this period. However, the growth rates experienced by households varied con- siderably over the expenditure distribution, with the lowest and the highest consumption groups benefiting significantly more than the middle consumption groups. t Map 11.1: District-level consumption poverty headcountrates (%), 1995 9 Profile of the Poor Headcount ratio of poverty (%), by occupation of household, 2000 80 - 77 2.13 Occupation. Who 70 are the poor in Bangladesh? One distinguishing feature 6o - of the poor i s their occupa- tion. The group having the 5 0 - highest probability o f being 4 0 ) poor i s agricultural (casual) laborers, with more than 30 three-quarters of agricultural 2o laborers in the country being Agricultural Small farmers Large farmers Salaried Self-employed Other I1hmex (< 1 ai.re\ i=,1acre) poor (Figure 11.9). The PmnlnVCCU P"tl`P"l-P"P,IT. group with the next highesl incidence o f poverty i s small Figure 11.9 farmers (those operating less than one acre of land), fol- Headcount ratio of poverty (%), by amount of owned land, 2000 lowed by self-employed en- 60 i 57 trepreneurs. Large farmers have the lowest incidence of poverty. 2.14 Land ownership. An- other common trait of the poor in Bangladesh i s their lack of ownership o f land. This i s not surprising as the Bangladeshi 1 5 ~ 36-50 >50 ~ Male Female c=35 the poor throughout the work years ~ yeas ~ years i s their large family size. 11: Household sue Sex of householdhead Age of householdhead Bangladesh, as well, there i s a very strong association be- Figure11.11 10 tween poverty incidence and household size (Figure 11.11). For example, in 2000, the in- cidence of poverty among households with three or fewer members was only 35%, as against 54% among households with more than five members. 2.16 Inaddition to householdsize, the age of the household head i s an important corre- late of poverty, reflecting life-cycle patterns of poverty. Households with younger heads are much more likely to be poor than households with older heads (Figure 11.11). How- ever, the data do not indicate any association between poverty and the sex of the house- hold head (Figure 11.11).* 2.17 Schooling. There i s Headcount ratio of poverty (%), by schooling years of highest-educated a great deal of evidence male or female in the household, 2000 from around the world that 66 human capital - especially 60. in the form of schooling - provides an important means 1 of escaping poverty. School- 40 ing increases an individual's 30 4 skills and productivity, al- 2 0 J lowing him or her to in- crease earnings and lower 1 the likelihood of poverty. Figure 11*l2 that shows ISchooling yeaa of highest-educatedfemale in houaeholdl Schooling yeas of highest-educdieumdle,ti ~ schooling of females i s even more strongly associated with reduced poverty for the Figure 11*12 household than the school- ing of male members. For instance, the incidence of poverty among households in which the highest-educated adult female has 8-9 years of schooling i s 22%; the comparable fig- ure for households where the highest-educated adult male has 8-9 years of schooling i s significantly higher (35%). Likewise, the reduction in poverty from having 10 or more years of schooling i s much larger when the schooling accrues to females within the ho~sehold.~ The Role of Public Interventions 2.18 Infrastructure. An inverse association between infrastructure in a community and the incidence of poverty in that community does not imply any causal relationship between the two variables. The association could reflect that better-off communities are able to invest in more and better-quality infrastructure, or it could reflect that infrastruc- However, female headship i s associated with higher levels of extreme poverty. 9Of course, the association between poverty and schooling may not reflect causality, since there i s no con- trol for parental background. If better-off parents provide their children with more schooling and also be- queath them more land, an inverse association between the poverty status and schooling of the children ' would not necessarily reflect the poverty-reducing effects of schooling. The available data do not permit extensive exploration of such possibilities. 11 ture increases the productivity and earnings of individuals and thereby reduces the likeli- hood of their being poor. 2.19 It is nevertheless useful to examine the association between poverty and infra- structure using the HIES data for 2000. Figure 11.13 suggests strong positive associations between poverty and three types of infrastructure - paved roads, bus transport, and elec- tricity. For instance, the incidence of poverty in districts having more than 15% of the roads are paved i s 46%, as compared to an incidence of 56% if districts where 10% or less of the roads are paved. Likewise, availability o f a Headcount ratio of poverty (%), bus station in a village is by availability of infrastructure in village, 2000 associated with a reduction 60 - 59 of 8 percentage points in the incidence of poverty in that village. Finally, an electrified village i s ob- served to have a 15% re- duced incidence of poverty. 2.20 Government pro- grams. The Government of 1 Bangladesh has had several in-kind food assistance programs going back to the Figure11.13 1970s, many of which bene- fit the poor. These include, among others, Food-for-Work (FFW), Test Relief, Food-for- Education, Gratuitous Relief (GR), Vulnerable Group Development (VGD), and the Vul- nerable Group Feeding (VGF) program. 2.21 Since the community Headcount ratio of poverty (%), by presence of government programin village, 2000 module of the HIES 2000 i collected information on 55 whether selected govern- ment programs were operat- 53 ing in the village during the 51 past year, it i s possible to 51 50 test the hypothesis that pub- 49 lic transfer programs are as- sociated with reduced pov- 47 erty. The same caveat dis- cussed earlier in the context 45 1 of infrastructure applies here - viz., an observed associa- tion between public transfer programs and poverty does Figure 11.14 not reflect causality. Indeed, 12 given that these programs are typically well-targeted to the poor in Bangladesh (World Bank 2003), one would expect them to be associated with higher rates of poverty. 2.22 The evidence from HIES 2000 certainly seems to bear out this result (Figure 11.14). In most cases, the presence of a government transfer program in a village i s asso- ciated with a higher incidence of poverty in that village. In most cases, the difference in poverty incidence between villages having the program and not having the program i s modest - about 2-5 percentage points. 2.23 Role of NGOs. While NGOs have been important actors in development - and particularly in poverty alleviation - in most developing countries, nowhere has their im- portance been as great as in Bangladesh. NGOs, such as the Bangladesh Rural Advance- ment Committee (BRAC), PROSHIKA, and the Grameen Bank, are well known throughout the world for their pioneering work in micro-credit, skill development, and employment generation. It i s estimated that nearly 80% of the villages in Bangladesh are now covered by some NGO program or project. 2.24 One of the most rigorous attempts to estimate the impact of group-based micro credit programs - a very common NGO intervention in the area of poverty alleviation - found significant effects on many household outcomes (Pitt and Khandker 1998).loIn particular, the study observed a significant positive effect of participation in the credit programs on household consumption per capita (and thereby on the risk of poverty). The study also found that participation in the micro-credit programs by females had a much stronger effect on household living standards (as well as on other outcomes, such as chil- dren's schooling) than participationby males. Multivariate Analysis 2.25 To examine the likelihood of Bangladesh attaining the poverty MDG, we have estimated a multivariate model of poverty incidence, usinghousehold data from the HIES 2000.l The multivariate model has the advantage of controlling for several variables that 1 may be simultaneously associated with poverty. The estimation results are reported in Annex Table 1, while only the broad findings of the empirical analysis are discussed here. 2.26 The multivariate model confirms many of the bivariate relationships discussed earlier. Even after controlling for other variables, agricultural labor households are ob- served to have the highest probability of being poor. The amount of land owned per cap- ita i s strongly inversely associated with poverty, with a one percent increase in land own- ership being associated with a reduction in the probability of being poor by 0.2%. Like- wise, larger households and households headed by younger heads are significantly more likely to be poor than smaller households and households headed by older heads, respec- '"The study used a quasi-experimentaldesign and allowed for the fact that certain types of individuals might self-select themselves into the creditprograms. 11Since the dependent variable in the model i s a dichotomous variable (Le., whether or not a householdi s poor), the model has beenestimated by the maximum-likelihood probit method. 13 tively. The schooling of adult males and females has a strong inverse association with the probability of being poor, with the schooling of females having a slightly (but not signifi- cantly) larger association. 2.27 All three infrastructure variables - the extent of pavedroads, electricity coverage, and availability of bus transport - have significant associations with poverty. However, neither of the two government program interventions included - the VGD and the Food- for-Work program- has any significant correlation with poverty. 2.28 The variable that has the strongest association with poverty i s the log of mean dis- trict consumption expenditure per capita, which i s used as a proxy for average living standards in a district. The implied `growth elasticity' of poverty - i.e., the percent reduc- tion inthe poverty headcount ratio associated with a one percent increase in mean district consumption expenditure per capita - i s slightly more than -1 (viz., -1.14), implying that economic growth has a one-for-one association with poverty reduction. l2 2.29 In addition to mean district consumption expenditure per capita, we include an explanatory variable in the probit equation reflecting income inequality (proxied by the Gini coefficient of per capita consumption expenditure). The empirical results suggest that, controlling for mean consumption, poverty i s associated positively and strongly with consumption inequality. The elasticity of poverty with respect to the Gini coefficient i s estimated to be 0.7, implying that a one percent increase in the Gini coefficient of con- sumption inequality would be associated with a 0.7% increase in poverty. These results suggest that worsening consumption (or income) inequality can substantially offset - even reverse - the beneficial effect of economic growth on poverty reduction. 2.30 Another variable that i s included in the probability-of-poverty model i s the aver- age per capita availability of land in a district. The latter variable i s included to test whether aggregate natural resource constraints have an additional association with pov- erty over and above household resource (land) constraints. However, the results do not indicate any such significant association. Simulationsto 2015 2.31 Based on the multivariate probit model estimated above, we have undertaken simulations of the poverty headcount ratio in Bangladesh to 2015 under certain assump- tions. The nature and magnitude of the interventions are detailed in Table 11.1.The scope and magnitude of the assumed interventions are only meant to illustrate the likely reduc- tion in poverty under one possible scenario. It i s obviously not possible to predict whether the assumed interventions will indeed take place, and, even if they do, whether they will proceed as the pace assumed in Table 11.1. Usingregional data over three time periods (1991-92, 1995-96 and 2000), the World Bank (2003) esti- mated the growth elasticity of poverty to be much larger (viz., 2.12). The difference in the two estimates could be accounted for by (i)the use of aggregated regional data in the other study versus our use of house- hold data, and (ii)our control for other socioeconomic variables (such as schooling, land ownership, etc.) versus the lack of control for these other factors in the other study. 14 Table 11.1: Assumptions about various interventions to reduce consumption poverty, 2001to 2015 Starting Assumed Ending value in changeper value in Intervention 2000 year 2015 Adult male schooling (years) 4.5 0.3 9.0 Adult female schooling (years) 2.5 0.3 7.0 Mean of district monthly consumption ex- penditureper capita (Taka) 900 2.7% 1,342 Gini index of consumption inequality 30.0 0.5 37.5 Per capitaavailability of land(acres) 0.16 -2% 0.12 Availability of bus station (%) 21 1%points 36 Electricity coverage(9%) 64 1%points 79 2.32 The assumption that mean district monthly consumption expenditure per capita will grow annually at about 2.7% to 2015 needs some elaboration. During the 1990s, per capita consumption expenditure grew by 2.2%% per annum, while per capita GDP in- creased by 3.3% over. Assuming that the same relationship holds in the future, the 2.7% consumption growth we have assumed would be consistent with an annual GDP per cap- ita growth rate of 4%; in other words, we assume that economic growth will be somewhat (butnot substantially) higher inthe future than it was during the 1990s. 2.33 In addition, we assume that consumption inequality will increase and per capita land availability will decline at the same rate at which these variables have changed dur- ing the decade of the 1990s. As noted earlier, none of these assumptions are sacrosanct; they are only meant to be illustrative. The projections could be undertaken for any com- bination of changes inthe policy or environmental variables. 2.34 Figure 11.15 Projected poverty headcount raeio to 2015, under various intervention scenarios (graph shows cumulative effect of each additional intervention) shows the projected 65 1I changes in the inci- 60 - I dence o f consump- 55 50 tion poverty in 45 Bangladesh when 40 all seven policy and 35 environmental vari- 30 - 30 ables proceed as Intervention 25 -Falling per capita land availability shown inTable 11.1. 20 Increasing consumption inequality The declining I:" -Expansion of female schooling 15 "Expansion of male adult schooling t l5 availability of land Improcing bus transport -Expanding elecrncity coverage per capita i s associ- ---Real per capita economic growth ated, understanda- bly so, with rising poverty (from about Figure 11.15 50% to 52%). Increasing inequality i s further associated with another 8 percentage point 15 increase in poverty (to 60%). However, all the other interventions contribute to reduc- tions in poverty. Both the expansion of male and female schooling are associated with large declines (10-12 percentage points each) in poverty incidence, but the contribution of transport and electricity access are very small (less than one percentage point each over the entire period). Finally, annual per capita GDP growth of 4% i s associated with the largest decline in poverty (of about 21 percentage points). Together, the seven policy and environmental variables are associated with a reduction of about 33.5 percentage points in the incidence of poverty - bringing the poverty headcount rate well below the MDG level (16% versus 30%). Indeed, the projections suggest that real per capita eco- nomic growth of 4% annually, without any increase in consumption inequality, would by itself allow Bangladesh to just meet its MD target. However, if consumption inequality increases at the same rate as it has during the 1990s, this would not be possible. 2.35 What these simulations underscore i s that attainment of the poverty MDG cer- tainly appears plausible inBangladesh, but only if the country maintains strong economic growth and continued expansion of male and female schooling, and prevents income and consumption inequality from rising, inthe years ahead. 16 111. INFANTAND UNDER-FIVEMORTALITY 3.1 The mortality of children i s often seen as the criteria o f "success and failure of nations" (Sen 1998). It i s an important indicator of well-being in its own right, as recog- nized by its inclusion among the MDGs. The mortality of children not only represents an enormous waste of human resources, but also a major cause of suffering in the popula- tion. The millennium development goal for Bangladesh i s to reduce the under-five mor- tality rate from about 150 in 1990 to 50 by 2015. Trends 3.2 The historical trends in infant mortality, culled Infant mortality rate, 1911-99 from various sources and 200 - surveys, are shown in Fig- ure 175 - I I I . l . 1 3The IMR ap- pears to have dropped 150 - sharply in the early 1900s, but barely dropped from 125 - 168 infant deaths per 1,000 100 - live births to 161 deaths during the two decades be- 75 - tween 1951 and 1971. In the immediate aftermath of Bangladesh's independ- Source: See Annex Table I ence, the IMR actually increased to 173. But since Figure 111.1 then the IMR has fallen secularly and rapidly, Infant mortality rate estimatedfrom the Vital RegistrationSystemand the BangladeshDHS, 1980-2002 reaching a level of 125 by 1984-85, 80 in 1994-95, and 66 currently. It i s only after 1989 does one see a definitive and a faster trend o f decline. 3.3 Two data sets pro- vide much of the recent in- formation on infant mortal- ity. One is the vital registra- tion survey (VRS) data of the Bangladesh Bureau of BDHS Figure 111.2 13Note that the time increments in Figure 111.1are not equal. 17 Statistics (BBS) and the other i s the Bangladesh Demographic and Health Survey (BDHS) data. Both suggest dramatic improvements in infant mortality in the 1990s. The VRS data of the BBS represents the longest series on IMR based on a single source. The VRS data show virtually no improvement in infant mortality during 1980-88 (and an in- crease in 1980-82 (Figure 111.2). In 1988, the IMR still stood at 116, but by 1995 it had dropped to 75. The rate fell even faster during the late 1990s and early 2000s - to 57 by 1998, 53 by 2000, and 51 by 2002. While the very low infant mortality rate of 51 esti- mated by the VRS for 2002 i s probably the result of a Infant mortality rate, 1970-2000,selected countries in Asia -2 6% -3 3% death registration system 145 1 138 01970-75 01980-85 130 -3.9% 109 mortality in recent years I05 - Annuul riife of ciionge in IMR 1970-2000 suggested by the BBS data 8 5 - i s confirmed by the BDHS -3 I% data, which also show the 65 -4 3% 60 -5 6% n IMR halving from its levels 1, in the last decade (Figure 45 111.2). ,,: ~ l4 5 International Compari- sons Figure 111.3 3.4 How does Bangladesh's performance at infant mortality reduction compare to that of other countries in the region? Over the period 1970-2000, infant mortality has fallen by anywhere from 2.6 to 5.6% annually inthe countries shown inFigure 111.3, with South Korea and Sri Lanka being the stellar performers. Infant mortality rate, by division, 1999-2000 Bangladesh has, however, done very well, managing Sylhet 127 to reduce its infant mortal- ity rate at a rate compara- Dhaka ble to that of Thailand and Rajshahi much faster than that of India. Indeed, what i s sur- Bzuisal prising i s that the level of infant mortality i s now Chlttagong 69 lower in Bangladesh than Khulna in India - a country that 7 has two times the per ca - 130 ita GDP of Bangladesh.19 Figure 111.4 14 Note that the BDHS data, which are used in the rest of this chapter, refer to infant mortality during the five years precedingthe survey. l5Dreze (2004) notes this as well: "... Whether one looks at infant mortality, or vaccination rates, or school participation, or child nutrition, or fertility rates [in Bangladesh], the message i s similar: living conditions are rapidly improving, not just for a privileged elite but also for the population at large. InIndia, social pro- 18 Spatial Patterns Infant mortality decline by residence & division, 1993-94to 1999-2000 3.5 Divisional varia- 12(, tions. The BDHS data indicate wide variations in infant mortality across divisions, with the divi- 80 sion of Sylhet having an infant mortality rate that i s nearly two times as high as that in Khulna 40 (Figure 111.4). Indeed, 20 Sylhet has an infant mor- tality rate that i s more " than 50% higher than the Dhaka Chittagong Bansal Rajshahi Khulna Urban Rural division having the sec- ond-highest rate of infant Figure 111.5 mortality (Dhaka). There i s greater similarity among the other five divisions in terms of their infant mortality rates. 3.6 The decline in infant A mortality during the 1990s has also varied significantly across divisions (Figure 111.5). Since Sylhet was carved out of Chittagong in 1998, changes in infant mor- tality can only be shown for the five divisions that ex- isted in the early 1990s (when Sylhet formed part of Chittagong). Figure 111.5 indicates that the division having the highest level of infant mortality in 1993-94 - Dhaka - experienced the level slowest rate of infant mortality decline (14%) over M a p 111.1:District- map of under-five mortality the following six years. In rates, 2000 contrast, Khulna, which enjoyed the lowest level of infant mortalit in 1993-94, experienced a rate of infant mortality de- cline that was two times as much as that experienced by Dhaka. Thus, regional variations ininfant mortality appear to havebecome morepronounced. gress i s slower and less broad-based, despite much faster economic growth. This i s one indication, among many others, that India's development strategy i s fundamentally distorted and lop-sided." 19 3.7 However, Figure 111.5 shows that there has been a remarkable convergence in in- fant mortality rates across rural and urban areas, thanks to much more rapid decline in infant mortality in the rural areas. While the infant mortality rate in rural areas was 27% higher than inurban areas in 1993-94, it was only 8% higher in 1999-2000. 3.8 District variations. A district-level map of Bangladesh by the under-five mortal- ity rate i s shown in Map 111.1below. There are large variations inthe under-five mortality rate, with a number of districts having an under-five mortality exceeding 97 deaths per 1,000 live births. These districts have no particular location; they are scattered in the North, Northeast, and South of the country. Some of them are even contiguous to low- mortality districts. Proximate and Socioeconomic Correlates 3.9 Child Malnutrition. After the first month of life, child malnutrition becomes an important contributing factor to infant and child mortality. Malnu- Relationshipacross districtsbetweensevere child malnutritionand under-five trition in Bangladesh often sets in mortality, 1995 & ZOO0 early, often owing to improper 250. 0 feeding practices, such as early termination of exclusive breast- feeding and introduction of (in- adequate) supplementary feeding. In addition, even during the ex- clusive breastfeeding period, in- fants may be malnourished owing to insufficient quantities of breast milk - in turn the result of poor nutrition and heavy workload of 3 6 9 12 15 18 21 24 Severe child malnutrition (mid-upper arm circumference) poor women. Malnourished in, fants are more prone to diarrheal respiratory and other infections, Figure 111.6 which, when untreated, can lead to infant death. 3.10 It i s often difficult to establish the correlation between child malnutrition and mortality in the absence of longitudinal data, since anthropometric data from most cross- %sectional surveys are available only for children living at the time of the survey. But dis- trict-level data on under-five mortality rates and severe malnutrition rates (based on the mid-upper arm circumference measure) for 1995 and 2000 show a positive, albeit not perfect, association between the two rates (Figure 111.6). Of course, some districts, such as Cox's Bazar, have unusually high under-5 mortality rates relative to their child malnu- trition rates, while other districts, such as Chandpur, have unusually low under-5 mortal- ity relative to their child malnutrition rates. 20 3.11 Immunization. In addition to Measles immunization coverage( W ) amongchildren aged 1year or less, 1991-2003 69 ity to be considered as a separate MDG 5 5 - 54 (with the goal being universal immuni- 5 0 - zation of one-year olds against mea- 45 - evaluation surveys suggests that cover- 2003 age of measles vaccination has increaseu Figure 111.7 very slowly since the mid-1990s (Fig- ure 111.7). hIeaslesvaccinationcoverage among 12-23 month olds, by divisionand urhadrural residence,1999 3.12 The DHS 1999-2000 data also Urban 81 indicate large variations in measles Rural 69 coverage. In Khulna, 81% of children aged 12-24 months are vaccinated Khulna 81 against measles (Figure 111.8). How- , Chitlagong 77 ever, the coverage rate i s as low as 58% Rajshh inSylhet. Likewise, there i s a large dis- 70 parity in measles coverage between the Dhaka 66 rural and the urban areas. Sylhet - 3.13 Previous birth intervals. The timing of successive births is an impor tant risk factor in both infant and under- Figure 111.8 five mortality. As Figure 111.9 shows, the Infant and under-five mortality rates,by previousbirth interval, risk of death for an infant or a child de- 1993.94 and 1999-2000 clines dramatically when he/she i s born Previous birth interval 224 four or more years after the previous birth (as compared to being born less than two 200 - Infmmonaliry years after the previous birth). Interest- 163 ingly, however, high-risk births - those spaced within two years of the previous birth- experienced a larger proportionate decline in mortality risk than low-risk births between 1993-94 and 1999-2000 suggesting that overall infant and under five mortality decline in Bangladesh oc- curred independently of longer birth Figure 111.9 spacing by mothers. 21 3.14 Gender. Muchhas been written about sex differentials ininfant andchild mortal- ity in South Asia. South Asia i s said to be one of the few regions in the world where fe- male infants and children have a higher risk of mortality rate than males. The Bangladesh DHS data do not show this, at least for infant mortality; indeed, infant mortality rates for observed to be greater for males than for females (Figure 111.10). However, the survival advantage enjoyed by female infants appears to have narrowed over time; while female infant mortality was 87% of male infant mortality in 1993-94, it had risen to 94% by 1999-2000. Infant and child mortality rates, by sex of child, 3.15 Figure 111.10 also sug- 1993-94 and 1999-2000 gests that the survival advantage enjoyed by females i s lost and even reversed as they grow be- yond infancy. The child mortal- ity rate, which measures the probability of death between the ages of one and five, i s actually significantly greater (by about a third) for females than for males. The gender disparity in child mortality rates increased slightly between 1993-94 and 1993-94 1999-2000 c/i decline 1993-94 1999-2000 R decline 1999-2000 - from 32% to 36%. Parental neglect toward girls - Figure 111.10 symptomatic of the generally low social status of women - i s an important cause of the gender disparity in child mor- tality. Girls are less likely to receive adequate food allocations and medical treatment for their illnesses than boys. 3.16 A study examining data on nearly 12,000 birthsduring 1973-74 from the Demo- graphic Surveillance System data inthe Matlab region of Bangladeshfound that the mor- tality risk for females exceeded that for males around the age of 8 months - typically the age when an infant cannot survive on breast-feeding alone and needs nutritional supple- mentation (Koenig and D'Souza 1986). 3.17 In South Asia, gender interacts with birth order in exerting a powerful influence on the survival probabilities of infants and children. Another study from Matlab for the period 1981-82 observed that girls with surviving older siblings faced higher risks of death than first-born girls having no older siblings.16The probability of dying was higher for girls having older brothers and lower for boys with older sisters. Boys with older male siblings also faced a somewhat higher risk of death than single male children (although not as high a risk as girls with older male siblings). These data clearly suggest that excess female mortality i s attributable, in large part, to conscious and selective parental neglect o f girls. l6Muhuri PK, Preston SH. 1991. "Effects of family composition on mortality differentials by sex among children in Matlab, Bangladesh." PopulationDevelopment Review 17(3): 415-34. 22 Infant and under-five mortality rates, by mother's schooling, 1999-2000 3.18 A longitudinal study using the Matlab data over n 140 130,1 Under-5 mortdlity fate the period 1970-95, how- Infant monality rate 104.8 ever, found that child mor- tality among females has fallen faster than among males.17Of course, the Mat- lab region has benefited from a major MCH-FP pro- ject intervention over the years, and the MCH-FP pro- None Primary Primary Secondary None Primary Primary Secondary gram may have contributed incomplete complete or higher incomplete complete or higher to the reduction in excess Mother'sschooling Mother's schooling female deaths. Figure111.11 3.19 Female schooling. As i s widely observed in many countries (including Bangladesh), mother's schooling i s strongly associated with infant mortality (Figure 111.11). What i s interesting i s that while both infants and un- der-five children Infantmortalitydeclineby maternalschooling, 1993-94to 1999-2000 benefit (in terms of reduced risk of mor- 1201 113 1996-97 0 1999-2000 1% 1993-99 decline, tality) from having their mothers even 100 slightly schooled - 93 (e.g., primary incom- 82 plete), the propor- 80- tionate benefit to un- 65 der-five children i s greater than that to infants. This likely occurs because a lar- 40- ger proportion of older children's 20 1 deaths are prevent- able due to interven- tions such as nutrition 0 -r and prompt medical None Primary incomplete Primarycomplete Secondaryt attention, which tend Mother's schoolinglevel to be strongly associ- ated with mother's Figure111.12 schooling. l7Ashish K. Dutta and Radheshyam Biaragi. "Improvement in Female Survival: A Silent Revolution in Bangladesh." ICDDR, 1999,mimeo. 23 3.20 What i s also interesting to observe i s that women with no schooling experienced almost as much relative decline in the mortality of their infant children between 1993-94 and 1999-2000 in Bangladesh as did women with primary schooling (Figure 111.12). Women with secondary or more schooling showed the smallest decline in infant mortality over the 6-year period. This suggests that a large portion of the infant mortality decline in Bangladesh during the 1990s appears to have occurred independently of the expansion in female schooling. 3.21 Household wealth. Data from the 1996-97 DHS have Infant and under-five mortality rates, by wealth quintiles, 1996.97 been compiled by Wagstaff et 140 Wrjderfne Monniin al. (1999) to examine the asso- ciation between mortality rates lZo and household wealth status. As Figure 111.13 shows, the asso- 90 ciation i s sharply negative, with :: bottom quintiles having an un- 60 der-five mortality rate that i s 50 nearly two times that of the top Baitom Second Third Fourth Top Bottom Second Third Fourth Top Afpregated wealth quinttles quintile. between ~h~disparity Source Coinpilcd by Wagstaffcf nl (1999) based an DHS 1996-97 data the infant mortality rates of the Figure 111.13 bottom and top quintiles i s somewhat lower, but still very large; for instance, the bottom ciaintile experiences an infant mortality rate that i s 70% greater than that experienced by the top quintile. The disparities in mortality across asset groups reflect differences in pa- rental education, levels of child nutrition, and access to health and medical services across the poor and the non-poor. 3.22 The lower access of the poor to public health measures, such as child immuniza- tion, i s evident from Fig- ure 'I1*which 14, reports Immunization rates among children aged 12-23 months, by wealth 100 quintiles, 1996-97 the child immunization % im&otized rigniriar menriey ratesdata the DHS 1996- from 8 0 ) 74 97 across wealth 3 7 0 - 62 quintiles. While measles '7c eceiviiig no , 6o - if? iiiriizatiori of any ope immunization coverage i s as high as 83% among the 30 children (aged 12-23 2O 18 months) belonging to the IO toD wealth auintile. the coverage rate i s as low as Bottom Second Third Fourth Top Bottom Second Third Fourth Top Aggregated wealth quintiles 62% among the children ~ource:Compiled by Wagstaff ern[. (1999) basedon DHS 1996-97 daw o f the bottom wealth quin- Figure111.14 tile. The percentage of children aged 12-23 months who receive no immunization of any type i s about 5% in the top two wealth quintiles; among the bottom quintile, the corre- sponding figure i s 18%. 24 The Role of Public Inter- Under-fivemortalitj rate and district infrastructure, 2000 ventions 130 1 127 126 120 I18 3.23 Rural infrastruc- ture. Rural infrastructure I10 107 106 108 can have powerful influ- 100 ences on infant and child 100 mortality outcomes. For in- 91 91 90 stance, access to paved roads allows easier transport 80 to health centers and referral 70 C/c of householdsin distnct having Q of roads paved in distiict 9i of householdsin district 3.24 Data at the district level for the periods 1995 and 2000 show significant associa- tions between under-5 mortality and access to infrastructure (Figure 111.15). The under-5 mortality rate i s nearly 40% higher in districts having fewer than 25% of households with access to a sanitary toilet than in districts where more than one-half of households have sanitation access. Likewise, the under-5 mortality rate i s 38% higher in districts where fewer than 7.5% of roads are paved than in districts where more than 15% of the roads are paved. Among the infrastructural variables, access to electricity appears to have the smallest impact on under-5 mortality, with the under-5 mortality being only 15% lower in districts with more than 40% electricity coverage as compared to districts with less than 20% electricity coverage. 3.25 Family planning and MCH programs. The importance o f family planning in- terventions inbringingabout infant and child mortality decline cannot be discounted. It i s well-known that fertility decline and mortality decline often go hand in hand with each other. Bangladesh has had one of the most successful family planning programs in the developing world. The program has achieved extraordinary results by building an exten- sive network of health and family welfare clinics throughout the country, training thou- sands o f female workers to take family planning advice directly to women, and using mass media campaigns to create awareness about family planning in the population. The program has enjoyed strong political commitment from the government, grassroots-level partnership with NGOs, and generous and coordinated assistance from donors. Indeed, Bangladesh's experience has shown that it i s possible to bring about fertility and mortal- 25 ity decline in poor countries even in the absence of strong economic growth and improv- ing socioeconomic conditions. 3.26 Some of the best evidence of the role of effective family planning and M C H in- terventions on infant mortality decline in the developing world comes from the Matlab area of Bangladesh, where the Maternal Child Health and Family Planning (MCH-FP) Project has been operating since 1977. This project has provided more accessible and bet- ter-quality family planning services to a "treatment" area in comparison to those offered in nearby "control" areas. The more accessible and better-quality family planning ser- vices have included more frequent visits from female welfare assistants who provide counseling and deliver contraceptives, as well as closer access to a network of family planning sub-centers operated by the InternationalCenter for Diarrheal Diseases Re Figure 111.16: Infant mortality rates inMCH-FParea and control areas, 1978-2000 Source: Center for Health and PopulationResearch, ICDDR,Bangladesh search (ICDDR, B). Figure111.16 suggests that the MCH-FP project has contributed to a decline of 10-30% ininfant mortality since its inception. 3.27 The Role of Service Delivery. Despite its highly successful family planning pro- gram, the quality of service delivery in the overall public health sector in the country re- mains poor. There i s widespread absenteeism of doctors and paramedics at government health centers and sub-centers; most government health facilities are in disrepair; and the availability of drugs and medical supplies at public health facilities i s very limited. For example, when unannounced visits to health clinics were made with the intention of dis- covering what fraction of medical professionals were present at their assigned post, a na- tionwide survey found that the average number of vacancies over all types of providers in 26 rural health centers was 26% (Chaudhury and Hammer 2003). Not surprisingly, vacancy rates or unfilled posts were generally higher in the poorer parts of the country. Absentee rates for doctors were observed to be as high as 40% at the larger health clinics, while a shocking 75% of the doctors at smaller, single-doctor sub-health centers were not present at their posts during the unannounced visits. The poor quality of health services reflects many factors at work, including lack of accountability among public health providers and disenchantment with working conditions among health workers. Chaudhury and Hammer (2003) found that proximity of the medical provider's residence to the health facility, ac- cess to a road, and rural electrification were among some of the variables strongly associ- ated with the absenteeism of doctors from a clinic. 3.28 The NGO sector in Bangladesh has played an important role in the delivery of quality health services. Among some of the important NGOs in the country's health sec- tor are the Voluntary Health Services Society, Bangladesh Rural Advancement Commit- tee (BRAC), Bangladesh Association for Voluntary Sterilization (BAVS), Bangladesh Women's Health Coalition (BWHC), Family Planning Association of Bangladesh, Gonoshasthaya Kendra, Proshika, and Thengamara Mohila Sabuj Sangha (TMSS). These and other NGOs have worked in the areas of water and sanitation, M C H and family plan- ning, child survival, and AIDS/STD prevention, among other things. Projectionsto 2015 3.29 The Bangladesh q DHS data suggest that the Actual and projected infant mortality, 1981-2015 decline in infant mortality T 120 in Bangladesh between 0 Actual IMR -Projected IMR at 4 7% decline + 110 1979-83 and 1995-99 has -Projected IMR at 3 6% decline 100 averaged an impressive 0 3.6% annually. The decline e .I0: 80 e during the 1990s has been 70-I 470 even more rapid - about 4.7% annually. Figure 111.17 suggests that if the rate of infant mortality de- cline experienced between 1979-83 and 1995-99 con- 20 120 tinues into the future, infant mortality rate in Bangladesh Figure 111.17 could be expected to reach a level of 34 in 2015 -just slightly above the MDGlevel of 31. If the future rate of decline remains at the (higher 4.7%) rate experienced inthe 1990s, the infant mortality rate could decline to 29 by 2015. Thus, Bangladesh could expect to attain the infant mortality MDG -orcomeveryclosetoattainingit-ifitsimplycontinuesthetrendithasseeninthere- cent past. 27 3.30 In fact, however, this i s unlikely to be the case. The decline in infant mortality experienced by Bangladesh during the past 10-15 years i s unprecedented - both in rela- tion to the country's own earlier experience as well as in relation to the experience of other developing countries. The latter suggests that declines from very high initial levels of infant mortality are driven largely by reductions in the number of post-neonatal deaths (Le., deaths occurring between the age of one month and twelve months). These deaths are more easily averted by the typical (and relatively inexpensive) child survival interven- tions, such as child immunizations and oral rehydration therapy. However, as the overall level of infant mortality comes down, further reductions in overall infant mortality can only be obtained via reductions in neonatal mortality. Averting neonatal deaths typically requires more expensive interventions, such as professionally-attended deliveries or de- liveries in institutions as well as post-delivery and hospital-based emergency care. Thus, sustained infant mortality reduction becomes increasingly more difficult and expensive. 3.3 1 Bangladesh's extraordinary success in bringing down the infant mortality rate has meant that neonatal mortality currently accounts for about two-thirds of infant deaths and more than half of the deaths among children under 5 years of age. In fact, the ratio of neonatal mortality to under-five mortality has increased by about 40% over the last dec- ade (Status of Peq5ormance Indicators 2002). Therefore, future interventions to reduce infant or under-five mortality will need to focus on averting neonatal deaths. Although neo-natal mortality reduction typically requires hospital-based care, it i s possible to pro- vide a relatively inexpensive package of home-based neonatal services, as shown by a hignly-successful field trial in India's Maharashtra state in 1995-98 (see Box II.1 for a detailed description of the intervention). Multivariate Analysis 3.32 In order to undertake further simulations about the likelihood of Bangladesh meeting the under-five mortality MDG, we have estimated a multivariate model of under- five mortality using unit record data from the Demographic and Health Survey 1999." The multivariate model has the advantage of controlling for several variables that may be simultaneously associated with under-five mortality. The estimation results are reported in Annex Table 2, while only the broad findings of the empirical analysis are discussed here. 3.33 After controlling for the other factors associated with under-five mortality, urban areas are actually observed to have significantly higher under-five mortality than rural areas. This suggests that the urban areas enjoy lower rates of infant and under-five mor- tality than rural areas because of their higher living standards and adult schooling and generally better health services. Once these variables are controlled for, urban residence i s actually correlated with higher under-five mortality rates. l8 Since the dependent variable i s dichotomous (viz., whether or not a child dies within 60 months of its birth), the model has been estimated by the maximum-likelihood probit method. As noted earlier, the refer- ence period for calculating under-five mortality i s the five years preceding the survey. 28 3.34 The results also confirm that while the risk of mortality i s not significantly differ- ent across girls and boys, higher birthorder girls have a significantly greater likelihood of dyingthan higher birth order boys. These results are consistent with many earlier studies that indicate a peculiar form of intrahousehold gender discrimination in South Asia against higher birthorder daughters. 3.35 The results also highlight the extreme vulnerability of multiple (twin) births. Con- trolling for other factors, such as parental schooling and household living standards, a multiple birthi s nearly 20 times more likely to end in death than a single birth. 3.36 As in other studies from around the world, maternal schooling - but not father's schooling - i s observed to be significantly and inversely associated with under-five mor- tality, with each additional year of schooling (of the mother) reducingthe under-five mor- tality rate by about 4 deaths per 1,000 live births. Even after controlling for mother's schooling, the mother's age at the time of a child's birth has .a strong inverse association with the risk of that child dying within 5 years of its birth. A delay of each year inbearing a child reduces the under-five mortality rate by about 4 deaths per 1,000 live births. 3.37 The standard o f living of a household, as proxied by predicted log of monthly consumption expenditure per capita,l9 has a strong and significant (inverse) association ' with under-five mortality, with the elasticity of the under-five mortality rate with respect to household consumption expenditure per capita being estimated at 0.25. 3.38 Surprisingly, neither the availability of piped drinking water nor access to toilet facilities i s observed to have any significant association with under-five mortality, after controlling for household living standards and parental schooling. While a number of studies in South Asia have failed to find a significant association between the availability of piped drinking water inside the household and infandchild mortality, the lack of sig- nificance of the sanitation access variable i s surprising. Electricity coverage in the district also does not have a significant association with under-five mortality. 3.39 Bangladesh has made tremendous progress in expanding child immunization cov- erage over the last two decades. The WHO Vaccine Preventable Diseases Monitoring System indicates that Bangladesh went from virtually no measles vaccination coverage in 1980 to 72% coverage by 1998. The empirical results suggest that district-level immuni- zation coverage of measles has a strong (inverse) association with under-five mortality, l9Inthe probit model, we have included an explanatory variable - predicted log of household consumption expenditure per capita - to proxy household living standards. The DHS i s a rich data set, but it has the limi- tation that it does not contain information on income or expenditure, both of which are widely used as measures of household welfare. Usingdata on land ownership, ownership of consumer durables (radio, TV, bicycle, refrigerator, motorcycle, watch or clock, and sewing machine), and the type of materials used for the roof and wall of the household's dwelling (which are available in both the DHS 1999 and the HIES 2000), we predicted log monthly consumption expenditure per capita for each of the DHS households on the basis of an econometric relationship between actual log monthly consumption expenditure and land assets, consumer durables and housing quality variables that was estimated with unit record data from the HIES 2000 data. The distribution of predicted log monthly consumption expenditure per capita in the DHS sample was observed to be very similar to that in the HIES sample. 29 with each percentage point increase in measles vaccine coverage being associated with a reduction o f 0.4 child deaths per 1,000 live births. These estimates imply that universal measles vaccine coverage would be associated with a reduction in under-five mortality of about 16 deaths per 1,000 live births. Simulationsto 2015 3.40 Based on the multivariate probit model estimated above, we have undertaken simulations of the under-five mortality rate in Bangladesh to 2015 under certain assump- tions. The nature and magnitude of the interventions are detailed inTable 111.1.The scope and magnitude of the assumed interventions are only meant to illustrate the likely reduc- tion in under-five mortality under one possible scenario. It i s obviously not possible to predict whether the assumed interventions.wil1 indeed take place, and, even if they do, whether they will proceed as the pace assumed inTable 111.1. Table 111.1: Assumptions about various interventions to reduce under-five mortality, 2001to 2015 Starting Assumed Ending value in change per value in Intervention 2000 vear 2015 Adult female schooling (years) 2.5 0.3 7.0 Mother's age at child's birth (years) 23.8 0.2 26.8 Mean of district monthly consumption ex- penditure per capita (Taka) 900 2.7% 1,342 Measles vaccination coverage in district (%) 63 2.5 100.0 3.41 Based on the esti- mates presented in Annex Projected under-five mortality rate to 2015 Table 2, we have projected (graph shows cumulative effect of each intervention) the likely decline in the un- -90, der-five mortality rate un- der the assumption that the policy variables change over time as shown in Ta- ble 111.1 Figure 111.18 shows the projected trajec- Interrention tory o f the under-five mor- tality under this scenario. De1a)ing child birth (increasingmother'sage at child s birth) The under-five mortality i s -Real pes capitaeconomic growth ---Expanding measlesimmunization coverage MDG observed to decline suk 30 stantially - by more tha IS 50% - -over this period. The largest decline (of 18 Figure 111.18 deaths per 1,000 live births) comes about from the expansion of female schooling, followed by expanded mea- sles vaccination coverage (15 deaths per 1,000 live births). Delayed child bearing, which reflects both a delayed age at which the first child i s borne as well as better spacing 30 among subsequent children, i s also associated with a large reduction (of about 11 deaths per 1,000 live births) in the under-five mortality rate. The smallest association i s ob- served with living standards improvement. The results suggest that real annual GDP per capita growth of 4% (or annual growth of household consumption expenditure per capita of 2.7%) would be associated with a reduction in under-five mortality of 8 deaths per 1,000 live births. Together, the four interventions are associated with a reduction of 52 deaths per 1,000 live births in the under-five mortality rate - bringing that rate below the MDGlevel (46 deaths per 1,000 livebirths). 3.42 Thus, the simulation confirms the results of the simple trend analysis conducted earlier. It should be possible for Bangladesh to attain the child mortality-related MDG, but only with a package of interventions that includes strong economic growth, expansion of female schooling, family planning programs that motivate women to delay child bear- ing, and expanded child immunization coverage. 31 Resultsfroma FieldTrialinRuralMaharashtra,India specialized hospital care i s either inaccessible to the rural populationor prohibitively expensive. area were receiving home-based care. The team had selected 47 control villages in an adjacent area of the same district, collected baseline data in both treatment and control villages. An evaluation done after the t sive health education and home-based neonatal care. Source: Banget al. (1999). 32 IV. REDUCINGCHILD MALNUTRITION 4.1 Reducing child malnutrition i s one of the surest ways of reducing income-poverty. A high degree of child malnutrition i s one of the most important factors constraining the future productivity of a country. Child malnutrition leads to poor schooling and cognitive outcomes, which shapes occupational choice, which in turn has implications for future productivity as well as intergenerational mobility. Child malnutrition also has a direct adverse impact on labor productivity in adulthood. In addition to pursingbetter child nu- trition for its impact on future labor productivity and income potential, improved child nutrition i s also an important human development goal in and of itself, since malnutrition significantly reduces the quality of life of children. In addition, of course, child malnutri- tion i s an important contributing factor to the highrates of infant mortality in developing countries; by some estimates, as many as half of all the infant deaths inpoor countries are directly or indirectly related to child malnutrition. Trends 4.2 Levels. Child mal- nutrition rates in Bangla- Child underweightand stunting rates, 1999-2000 (% of childrenin relevantage group who are underweightor stunted) desh are very high- among I A 0 51 48 the highest in the world. The two most recent child nutrition surveys - the 19 Child Nutrition Survey 20 18 2000 and the Demographic 10 2 and Health Survey 1999- 2000 - indicate that nearly one-half of children below the age o f 5 or 6 years are Severe Moderate Severe moderately underweight or I1 ~ BBS Child Nutiition Survev 2M" ianer 6-71 Demomaohic and Health Survev 1999-2000 (ages stunted (Figure IV.1).20,21 About 10-18% of children are severely underweight or Figure IV.l stunted in the sense of being more than three standard deviations below the relevant NCHS standards. This suggests that children inBangladesh suffer from short-term, acute food deficits (as reflected inlow weight-for-age) as well as from longer-term, chronic under-nutrition (as manifested in highrates of stunting). 20As in the literature, a child i s considered underweight when his or her weight-for-age i s more than two standard deviations below the NCHS reference weight. A child i s stunted when his or her height-for-age i s more than two standard deviations below the NCHS reference. Severe underweight and stunting occur when the relevant nutrition indicator i s more than three standard deviations below the NCHS reference. 21Note that not only were the CNS and DHS surveys conducted during slightly different periods, but the age groups o f children covered by the two surveys were also somewhat different (see Figure1V.l). 33 children aged 0-5 years and - e e fourth in terms o f the pro- 4.4 But Bangladesh i s Underweight rates in South Asia, circa 2000 also among the poorest countries in the world, and 60 - 58 Source: Various DHSrepom child malnutrition rates are 55 - typically strongly correlated 50 - with household living stan- dards. I s Bangladesh's child 45 - malnutrition level in line with what would be ex- pected of a country at its 30 level of per capita income? Figure IV.2, which plots the 25 relationship between child underweight levels and per capita GDP for 16 countries ir. Figure 1v.3 Asia,23 suggests that the per- centage of underweight chil- dren in Bangladesh i s approximately 16 percentage points higher than would be expected at its level of per capita GDP, given the observed relationship between child underweight rates and per capita GDP across the 16 Asian countries. In other words, based on its per 22Note that the UNDP data are 1995-2000averages, so the reported figure for BangladeshinFigure IV.2 i s different from that reportedinFigureIV.1. 23These include Myanmar, Nepal, Bhutan, Cambodia, Lao PDR, Bangladesh, Mongolia, Pakistan, Viet- nam, India, Indonesia, Sri Lanka, China, Philippines, Thailand and Malaysia (in order of increasing per capita GDP). 34 capita GDP, Bangladesh would be expected to have a child underweight rate of 40% (similar to that of Lao PDR) - not the 56% it had in 1995-2000. 4.5 How does Bangladesh compare to other countries in South Asia? Bangladesh's overall child underweight rate i s comparable to that of India, although it i s significantly larger than the underweight rates,observed inPakistan and Sri Lanka (Figure IV.3). 4.6 Trends. Bangladesh has made impressive gains in reducing its child underweight rates during the last 15 years. The decline in under- Child underweight rates in Bangladesh, 1985-2000 weight rates has been espe- (% of childrenin relevantage group who are underweight) 75i cially steep since the early 1990s (Figure IV.6). For instance, between 1992 and 2000, underweight rates dropped from 68% to 51%, implying an annual decline o f 3.7%. The decline i s con- firmed by the Demographic and Health Surveys (DHS) o f 1996-97 and 1999-2000 (Figure IV.4). -2000 ths) 4.7 Figure IV.5 below Figure IV.4 shows that the decline in child malnutrition rates dur- Child underweight rates amongchildren aged6-71months,by residence, ing the 1990s occurred in 1992-2000 both the rural and urban ar- eas o f the country. Indeed, * O1 7n the rate o f decline in both indicators of malnutrition 57 was approximately similar in the urban and rural areas of the country (26% versus 24%). 4.8 H o w does the de- cline in child underweight rates in Bangladeshcompare to those observed in other countries of the region? Figure IV.5 Data from India (two rounds o f the National Family Health Surveys) indicate a decline of about 1.9% per year be- tween 1992-93 and 1998-99 (from a rate of 52.7% to 47%). Thus, Bangladesh's rate of decline o f 3.6% per year in child underweight rates i s significantly greater than India's rate o f decline. However, the underweight rate in Sri Lanka (based on DHS data), fell 35 from 38% in 1993 to 29% in 2000 - an annual decline of 3.9%. Vietnam, where data on underweight rates are available for roughly comparable periods, the child underweight rate fell from 49% in 1992-93 to 36% in 1998-99 - an annual rate of decline of 5.3% (World Bank 1999)! Thus, while Bangladesh performed better than India, its perform- ance i s roughly on par with that of Sri Lanka but pales in comparison to that of Vietnam. Spatial Patterns Percentage of children under 5 who Mere underweight, by division, 1996-97 and 1999-2000 4.9 DivisionaI Varia- ElSevere 1996-97 Sebere 1999-2000 tions. The prevalence of OModerate, 1996-97 ElModerate, 1999-2000 64 child underweight rates by division, as well as changes in the prevalence between 1996-97 and 1999-2000, are shown in Figure IV.6.24 Sylhet, which had the highest prevalence of un- derweight children in 1996-97 (at 64%), saw the slowest relative decline Barisal Chittagong Dhaka Khulna Rajshahi Sylhet (11%) in underweight rates between 1996-97 and Figure IV.6 19913-2000. But Chittagong, which also had very high underweight rates in 1996-97 (60%), saw the sharpest relative decline over the three years (24%). Thus, there appears to be no patternto the decline in child malnutrition across divisions. 4.10 The same data are shown in the form of each division's contribution to the total number of (moderately and severely) underweight children in the country in Figure IV.7. These charts suggest show the high degree of concentration in child malnutrition in Divisionaldistributionof all underweightchildrenunder 5 in Bangladesh, Divisionaldistribution of all underweightchildren under 5 in Bangladesh, 1996-97 1999-2000 7 3 6 6 OBarisal 8 7 6 6 0Barisal 6Chitlagong BCh~lia~ang ODhaka 0Dhaka OKhulnd 4 OKhulna MKajshahi 6Kqshahi 105 OSjlhet 23 I Sylhet 29.3 Figure IV.7 Bangladesh. In 1999-2000, for instance, two divisions - Chittagong and Dhaka - ac- 24These data are from the Demographic and Health Surveys of 1996-97 and 1999-2000. 36 counted for more than one- Regionalcontribution to the total number of Underweight children aged half of all underweight 6-71 monthsin Baneladesh,2000 children, while three re- 100 gions - the above two plus Rajshahi - accounted for three-quarters of all under- 60 50 weight children. Between 40 + 0Cumulative share(70) HShare(Q) 1996-97 and 1999-2000, the largest relative decline in malnutrition occurred in 4 Chittagong, which saw its share of national under- weight children fall from 31% to 21%. In contrast, Rajshahi and Khulna ac- counted for an even larger Figure IV.8 share of underweight children in the country in 1999-2000 as compared to 1996-97. 4.11 Because the more populous regions of Bangla- desh also have very high child underweight rates, the number of underweight children i s highly concen- trated in a few regions in the country. Figure IV.8 shows the extent of this concentra- tion. Three regions (out of 14) - comprising the rural districts of Sylhet, Comilla, Faridpur, Tangail, Jamalpur, Noakhali and Chittagong - account for nearly one-half o f all underweight children in the country. Six regions (out of a total of 14) account for three-quarters of all un- derweight children. The re- gional concentration of mal- nutrition means that geo- malnutrition among children (based on the graphical targeting of nutri- MidUpper Arm Circumference indicator) tional interventions can be 37 highly effective in achieving the largest absolute reduction in child malnutrition in the country. 4.12 Inter-district variations. Data on malnutrition below the level of the division are available but only for the mid-upper arm circumference (MUAC) indicator. The MUAC i s an alternative measure of severe child malnutri- tion - but one that has been shown to be highly correlated with the child underweight measure. A district-level map of Chi1 inderweight rates (701, y age, 2000 Bangladesh (Map IV.1) 60 - shows large variations in the child malnutrition rate. 50 - The rate varies from a low 40 - of about 2% - inTangail - +Moderate -4-SCb.erC to a high of 14% (in 30 1 Bhola). About 42% of the 14 districts have a severe 10 11 4 child malnutrition rate of 6 5% or more. 0 6-11 12-23 24-35 36-47 48-59 60-71 Age (months) Demographic Patterns Figure IV.9 4.13 Percent of moderately and severely underweight children, Age Patterns. by age and sex, 2000 Malnutrition for a large 6 0 - proportion (about a third) 56 56 of children begins in the Males Females 5 0 - first year of life (Figure 4 0 - IV.9). Reasons for this I may be low-birth 3 0 ; weights,25 sustained and nurtured by inadequate breast-feeding and com- 10 plementary feeding prac- tices. But the risk o f mal- 6-11 12-23 24-35 nutrition increases sharply Age (months) in the second year of life (beginning at age 12 Figure IV.10 months), when most children stop breastfeeding and begin relying almost exclusively on solid foods. The insufficiency and inadequacy of weaning diets in Bangladesh increases the risk of malnutrition among infants. Not surprisingly, the probability of dying also in- 25Other data sources suggest that nearly 30% of infants born in Bangladesh weigh less than 2,500 grams at birth (UNDP 2001). 38 creases at this age. Figure IV.9 suggests that child malnutrition rates begin declining modestly after age 2. 4.14 Gender Disparities. Gender differences in malnutrition are most pronounced at young ages. Girls aged 6-11 months are significantly more likely than similarly-aged boys to be underweight (Figure IV.10).26Between ages one and three years, there i s vir- tually no gender difference in underweight rates. 4.15 The apparent lack of Severe underweight rates amongchildren aged 6-71 months, gender disparity in malnutri- by sex and birth order, 2000 tion at older ages may be explained by the higher rate of mortality of female rela- Mdle f3Female 16 - tive to male children aged 1- 5 years.27The surviving co- 14 - 13 hort of female children aged 12-35 months is likely to show (deceptively) lower levels of malnutrition than similarly-aged boys, since the most-severely under- weight girls in this age group are the ones mosl likely to die (and drop out of Figure IV.ll the sample). 4.16 There i s another dimension in which a child's sex matters to child malnutrition, and that i s its interactive association with birth order. As seen in Figure IV.ll, first-born females are significantly (75%) more likely to be severely underweight relative to first- born males, reflecting the strong cultural preference inBangladesh for having a first-born son. The gender disparity in underweight rates progressively diminishes with birth order, and, among children of birth order six or greater, there i s no gender difference in rates of severe malnutrition (with, of course, both sexes suffering very high rates of severe mal- nutrition). Proximate Causes 4.17 Infant Feeding Practices. An important proximate cause of child nutritional status i s nutrient intake, which in turn depends on the nature and duration of feeding (in- cluding breastfeeding) practices. Feeding practices are especially critical during the first few days and months of an infant's life, since growth i s faster and protection against ill- nesses and infections i s most needed during this crucial period. The fact that a large pro- 26Interestingly, however, the gender difference in severe malnutrition for this age group (6-11 months) i s reversed. Boys are nearly two times more likely than girls to be severely underweight. 27Mortality rates are actually higher for male than for female infants in Bangladesh, but the mortality gap reverses after age 1. 39 portion of Bangladeshi mothers wait for several hours or even more than a day after a baby i s born to initiate breastfeeding i s thus detrimental to the nutritional well-being of the child. The delay in breastfeeding may be related to an incorrect perception that the first breast milk (colostrum) i s an inferior food. In fact, colostrum i s rich in antibodies and highly beneficial to the new-born infant. 4.18 The CNS 2000 indi- Underweightratesamong 6-23 months old children,by feedingpractice, cates that only about 18% 2000 of children receive 55 7 53 mother's milk or colostrum 50 - as their first food (upon birth). Nearly two-thirds of 45 - children are fed sugar, misri water or honey as their first 40 food, while another 8% re- ceive cow/goat milk or in- 35 fant formula. About 3% of 30 infants are fed mustard oil Colostrum, breastmilk Other milk,sugar, honey, 4 months or younger 5 months or older as their first food. Prema- mustardoil ture introduction of foods First food at biith was: Age at which supplementary feeding was other than breast milk greatly increases the risk of Figure IV.12 infection in the small infant, and this sets in motion the process of malnutrition. 4.19 Another common feeding practice inBangladesh that has adverse implications for child malnutrition i s the early termination of exclusive breast-feeding and introduction of supplementary feeding. One reason why mothers give up exclusive breastfeeding early i s their perception that they are producing insufficient quantities of milk, inpart because of their poijr nutrition and heavy workload. Premature introduction o f supplemental foods puts the infant at greater risk of malnutrition, since weaning diets are often inadequate in Bangladesh, as in other developing countries. Supplementary feeding begins with a thin gruel of rice, often heavily diluted with water and with some vegetables or legumes added as a relish depending on season and availability, but generally in very small quanti- ties. The consequent low energy density o f this weaning food leads to a reduced intake of calories and protein, and i s an important cause of growth faltering during the weaning period, from six months to two years of age. The CNS 2000 suggests that supplementary feeding i s introduced within the first four months of birthfor 36% of children and within the first 6 months for 91% of children, which i s not in line with the recommendations of WHO and UNICEF that exclusive breastfeeding continue for the first six months of a child's life. 4.20 The CNS provides very clear evidence that the provision of foods other than co- lostrum or breast milk as a first food to the new-born infant, as well as the early introduc- tion o f supplementary feeding, significantly increase the risk of subsequent malnutrition in the child. The incidence of underweight rates significantly greater among children 40 whose first food immediately after birthwas cow's milk, sugar, honey or mustard oil than among children whose first food was colostrum or breast milk (Figure IV.12). Likewise, underweight rates are significantly lower among children aged 6-23 months who were started on supplementary feeding (in addition to breast-feeding) at ages 5 months or older relative to those whose supplementary feeding within four months of birth (Figure IV.12). 4.21 Family Food In- take. As children get older, Underweightrates (%) among childrenaged 6-71months,by average daily calorie consumption per capitaof household, 2000 which could be as early as one year of age in Bangla- Moderately underweight desh, they begin eating the 60 same foods as adults in the 50 Severely underweight household. Thus, the quan- 40 tity, quality and distribution of household food intakes 30 are likely to have an impor- 20 tant bearing on the nutri- I0 tional status of toddlers and young children. Figure 0 <1500 1501- 2001- 22350 <1500 1501- 2001- >2350 IV.13 clearly shows that 2000 2350 2000 2350 child malnutrition rates in- crease as household calorie FigureIV.13 consumption per capita falls. The inverse association of calorie intake and malnutri- Average daily calorieintakeper capita in the householdsof children tion i s much stronger for aged 6-23 months, by per capita expenditurequintile, 2000 severe than for non-severe 2,400 1 2,350 (moderate and mild) malnu- trition rates. For instance, children aged 6-71 months in households with an aver- age daily calorie intake per person of less than 1,500 kcal. are 41% more likely to be stunted and 69% more likely to be severely stunted than children in households with an average daily calorie intake of more than 2,35C kcal. FigureIV.14 4.22 Since calories are a normal good, average calorie intake per capita i s likely to be related to household living standards. Figure IV.14 shows a strong relationship between the two variables, with individuals in the top consumption quintile consuming nearly a third more calories than those in the bottom quintile. Average daily calorie intakes per 41 person among the bottom two quintiles appear inadequate relative to recommended al- lowances. 4.23 In this connection, it is important to note that Bangladesh has had among the slowest growth of calorie availability in South Asia (Figure IV.15). While daily per cap- ita availability of calories in Pakistan grew from 1,782 in 1961-62 to 2,459 calories in 1998-99 (representing an annual rate of increase of 0.9%), the corresponding growth in Bangladesh over the same period was only about 0.08% per year. India saw its daily per capita availability of calories go from 2,081 calories to 2,408 calories (an annual increase of about 0.4%). The slow growth of calorie availabil- Per capita daily availability of calories, South Asia, 1961-99 ity, which i s an important (although not the only) fac- 0 1998-99 v? h tor in the continuing high 2s00 rates of child malnutrition in the country, reflects the 2,250 lack of strong, sustained agricultural productivity 2,000 growth in the last 2-3 dec- ades. 1,750 Socioeconomic and Policy 1,500 Bangladesh India Pakistan Correlates Source:FA0 food balancesheers: USDA Economic ResemchService. Food 5 curit). Arsessmenr, March 4.24 Mother's Educa- Figure Iv.15 tion. There i s a large litera- ture documenting the many benefits of maternal education for child outcomes - infant and child mortality, child nutrition, or child schooling, The CNS data show a sharp asso- ciation between the prevalence of child malnutrition, especially severe malnutrition, and mother's education (Figure IV.16). For instance, 57% of children (aged 6-23 months) of mothers with no schooling are moderately underweight, as compared to only 27% of children with mothers who have had 8 or more years of Child underweight rates (%) among children 6-23 months of age, schooling. Severe under- by mother`s education, 2000 weight rates are even more Moderate underweight sharply associated with mother's schooling. Severe undeiweight 4.25 Water and Sanita- tion Facilities. Contamina- tion caused by unsafe drinking water and lack of sanitation are important causes of diarrheal and other infections in develop- ing countries. These infec- Figure IV.16 42 tions, when they affect a child repeatedly, can cause him or her to become malnourished. The CNS data indicate a strong association between rates of child malnutrition and household sources of drinking water (Figure IV.17). In general, tap water i s observed to be the 'safest' water source (in terms of being associated with the lowest rates o f child malnutrition), followed by water from wells and water obtained from ponds and rivers. Indeed, children who obtain their drinking water primarily from ponds or rivers are nearly 80% more likely to be underweight than children who obtain their drinking water from taps. 4.26 For the same rea- Child underweight rates (amongchildren aged 6-23 months), by drinking sons discussed above, the water source and type of toilet, 2000 type of toilet in a home can ,,,. 69 also have a bearing on child malnutrition rates. The 6o - CNS data suggest that flush or sanitary toilets offer the 5o best protection against child 40 - 3* malnutrition, followed by 1 pit latrines (Figure IV.17). 30 As would be expected, the 20 use of open space as a toilet Well or tube Pond or river 1 1 i s associated with the high- sanitary toilet temporary est rates of child malnutri tion. Children who use open spaces for their sani- FigureIV.17 tation needs are two times as likely to be underweight as those who use flush toilets. 4.27 Access to Health Facilities. The availability of and access to health facilities i s likely to reduce child malnutrition by increasing the utilization of health services, which are an important input into child nutrition and child health, However, if health facilities are located (by governments 1 or NGOs) in those villages Underweightrates amongchildren 6-71 months of age, by access to having the worst health and health facilities, 2000 nutritional conditions, a 62 (perverse) .positive associa- tion between child malnutri- i 63 60 62 58 tion and availability o f 56 health facilities would be 54 observed. 52 50 \ hotillage Locatecvillage 5 kms in 1more 4.28 Data from the CNS or than5 ~ closer kms suggest that proximity to a away ,away away j away thana (district) health center Thana health NGOclinic Private health Thana health NGO clinic Private health ~ rOntel. ,-li"i^ rPnt.3. does lower child malnutri- clinic tion (Figure IV.18). Interest- eholds ingly, however, the largest FigureIV.18 43 declines in child malnutrition from having a thana health center within a distance of 5 kilometers are observed for the poorest consumption quintile. For example, a thana health center in close proximity to a village i s associated with a reduction of the overall incidence of underweight children inthat village from 53.5% to 51.9%. But the decline i s far greater - from 63.1% to 57.9% - among children belonging to the bottom consump- tion quintile. 4.29 The results with respect to the availability of an NGO health clinic are very simi- lar. Availability an NGO health clinic in a village i s associated with a sharp reduction in child underweight rates, but only among the bottom consumption quintile. 4.30 However, the results with respect to the proximity of private health clinics are dif- ferent; close proximity to a private clinic i s associated with a negligible decline in child underweight rates among the poorest quintile, suggesting perhaps that the poor do not - make as much use of pri- vate health clinics as of Child malnutrition rates among children aged 6-71 months, by village electrification status, 2000 government and NGO health clinics. 651 63 1 4.31 Village Infra- structure. Infrastructure, 60 - such as roads and electric- , ity, can also have indirect associations with child malnutrition by improving access to health facilities and by improving condi- I Village not electrified 1 Village electrified Village not electiified 1 Village electrified tions of food storage and I Entire sample Bottom consumption quintile i preparation. The CNS data do not, however, FigureIV.19 show a strong association between village electrification and prevalence of under- nutrition (Figure IV.19). 4.32 Natural Disasters. Underweight rates among children 6-71months of age, by natural Households in poor coun- disaster in village in the past 5 years, 2000 tries, such as Bangladesh, 65 1 62 62 are often inadequately 60 - protected against weather- 57 and environment-induced shocks, such as floods, droughts and epidemics, because of the absence of well-functioning credit ~o 1 Yes and insurance markets. ~ Flood in viliage Cyclone I" Epidemic in These types of shocks can i n past 5 years? "lllafe I"pas15 rlllage I"pas15 years7 years? have adverse impacts on All households FigureIV.20 44 household consumption and thereby on child nutritional outcomes. The CNS data suggest that weather shocks - especially floods and cyclones - can have lasting impacts on child malnutrition (Figure IV.20).28For instance, rates of underweight children in villages that experienced a flood in the five years preceding the CNS 2000 are 7 percentage points greater than in villages that experienced no flood, with virtually no difference observed between the overall sample and the poorest consumption quintile. The occurrence of a cyclone has an association of similar magnitude, while an epidemic in the village in the preceding five years i s even more strongly associated with higher child underweight rates. Child underweight rates (6-71months), by presence of programin village, 2000 75 i 68 55 4 54 54 50 z " / c * Food-for- Vulnerable Vulnerable Grameen Work Group Group Bank, BRAC Work Feeding Development or Proshika .^__ ! " - L :--....:-A:l- i Figure IV.21 4.33 Public Nutrition Programs. Depending upon their geographical reach and pro- grammatic effectiveness, government in-kind (food) transfer programs are likely to influ- ence child nutritional levels. The Government of Bangladesh has had several nutritional intervention programs going back to the 1970s. These include, among others, Food-for- Work (FFW), Test Relief, Food-for-Education, Gratuitous Relief (GR), and Vulnerable Group Development (VGD). In addition, in response to the devastating floods of 1998, the Government started the Vulnerable Group Feeding (VGF) program, which provides some four million vulnerable households in the country with 16 kg of wheat and rice per household per month. 28Although seasonal flooding i s nothing new to Bangladesh (owing to the fact that its three major rivers drain a vast basin twelve times their own area), the country suffered its worst floods in living memory in 1998. The 1998 floods affected some 30 million people and caused over 1,000 deaths. In addition to se- verely damaging public and private infrastructure and assets, they resultedin extensive loss of rice crop. 45 4.34 Since the community module of the HIES 2000 collected information on whether selected govemment programs were operating in the village during the past year, it i s possible to test the hypothesis that public transfer programs are associated with reduced prevalence of child malnutrition. However, an observed association between nutritional interventions and nutritional outcomes could be spurious, especially if government nutri- tional interventions are targeted to villages with severe child malnutrition rates. 4.35 The CNS 2000 data shown in Figure IV.21 indicate that, while public food trans- fer programs, such as Food-for-Work, VGF and VGD, are weakly associated with overall prevalence rates of child malnutrition, they appear to have large (inverse) associations with child malnutrition rates among the poorest quintile of children. For instance, the Food-for-Work program i s associated with a reduction of 9 percentage points in under- weight rates among the bottom quintile of children aged 6-71 months old. Likewise, un- derweight rates among children in the bottom quintile are 7 percentage points lower in villages having a VGF program than in those not having a VGF program. 4.36 During the 1990s, there has been a gradual shift in emphasis within the targeted food assistance programs operated by the Government of Bangladesh from purely relief to a more explicit development orientation. Given the finding that relief programs, such as the FFW and VGF, have a strong association with child nutritional status among the poor, and given the propensity for natural disasters in Bangladesh, it would be clearly important for the government to retain some food relief programs, such as the VGF and GR, that could be rapidly scaled up intimes of need to provide relief ard short-term risk- coping. 4.37 The CNS data show very large associations betweenthe presence of an NGO pro- gram in a village and child malnutrition rates among the poorest quintile of children in that village. For instance, the percentage of underweight children i s 16 percentage points lower in villages having a Grameen Bank, BRAC or Proshika program than in villages not having one. BRAC and the Grameen Bank are the two largest NGOs in Banglade~h.~~ Both operate micro-credit schemes that attract some of the poorest households as their members, In addition to its micro-credit activities, BRAC operates community-based ag- ricultural, health, education, and water and sanitation projects. Thus, the NGO programs can be viewed as comprehensive income-generating and community development inter- ventions. Projections to 2015 4.38 The Bangladesh CNS data suggest that child underweight rates in Bangladesh have declined at an annual rate of 2.4% between 1985 and 2000. The decline during the 1990s has been even more rapid - about 3.5% annually. Figure IV.22 suggests that ifthe rate o f decline of child malnutrition experienced between 1985 and 2000 continues into the future, the child underweight rate in Bangladesh could be expected to reach a level of 36 in 2015 -just slightly above the MDGlevel of 34. If the future rate of decline remains at the (higher 4.5%) rate experienced in the 1990s, the child underweight rate could de- 29Indeed, BRAC i s often described as the largest NGO in the world. 46 cline to 30 by 2015. Thus, Bangladesh could expect to attain the child underweight MDG -orcomeveryclosetoattainingit-ifitsimplycontinues thetrendithasseeninthere- cent past. -75 derlying factors that deter- 70 -Predicted at 2 4% annual decline -- ' O @ * . mine child malnutrition. A 65 -Pledicted at 3 5% annualdecline -. 65 projection based on the un- 6o I -~60 derlying factors i s at- 5 5 , -- 55 4.41 The multivariate model confirms many of the bivariate relationships discussed earlier. After controlling for other variables, neither age nor gender i s a significant corre- late of malnutrition. However, birth order is, with higher birth-order children being sig- nificantly more likely to be underweight than lower birth-order children. 4.42 Maternal schooling has a strong association with underweight rates, with each ad- ditional year of schooling of the mother being associated with a decline of about 2 per- centage points in the child underweight rate. The log of monthly consumption expendi- ture per capita (proxying for a household's living standards) also has a strong association, with a one percent increase in per capita consumption expenditure being associated with a 0.2% decline. in underweight rates. However, consumption inequality, as measured by the Gini index of per capita consumption expenditure, has no significant association with child malnutrition. 4.43 Infrastructure generally has strong inverse associations with child malnutrition. Children in households having a flush toilet are, on average, 15% less likely to be under- weight than children in households not having access to a flush toilet. As was observed in _____ 30Since the dependent variable in the model i s a dichotomous variable (i.e., whether or not a child i s un- derweight), the model has been estimated by the maximum-likelihood probit method. 47 the bivariate results discussed earlier, village electrification has no significant association with underweight rates. However, proximity to a bus station appears to have a significant association, indicating the importance of transport and road access to the probability o f a child being underweight. 4.44 The results also indicate that natural disasters - in particular, floods - have a sig- nificant inverse association with child nutritional status. Children residing in villages that experienced a flood in the 5 years preceding the CNS 2000 were 7% more likely to be underweight than children in villages that did not experience a flood. 4.45 Among the various government nutritional programs, the Food-for-Work program appears to have a significant inverse association with child underweight rates. Controlling for other variables, children in villages having a Food-for-Work program are 5% less likely to be underweight than children invillages not having such a program." 4.46 Finally, the results indicate that, even after controlling for household living stan- dards, the scarcity of land in a community (as proxied by mean per capita land ownership in a child's district of residence) is significantly and inversely associated with child un- derweight rates. A one percent increase in per capita land availability in district i s associ- ated with a reduction of about 0.1% in child underweight rates. Simulationsto 2015 4.47 Based on the multivariate probit model estimated above, we have undertaken simulations of the child underweight rate in Bangladesh from 2001 to 2015 under certain assumptions. The nature and magnitude of the interventions are detailed inTable IV.1. Table IV.l: Assumptions about various interventions to reduce the child un- derweight rate. 2000 to 2015 Starting Assumed Ecding value in change per value in Intervention 2000 year 2015 Adult female schooling (years) 2.5 0.3 7.0 Food-for-work program coverage (9%) 68 1%points 83 Monthly consumption expenditure per capita (Taka) 900 2.7% 1,342 Flushtoilet coverage (%) 9 0.5% point 17 Distance to nearestbus station (kms.) 5.4 0.15 3.15 Mean of district land availability per capita (acres) 0.16 -2% 0.12 Probability of village experiencing a flood in 5-year period 0.80 -0.1 -0.65 31Of course, this result might reflect that the Food-for-Work program i s located in better-off communities that happen to have lower child malnutrition rates. However, such a possibility appears unlikely, given the design of the program. 48 As noted previously, the scope and magnitude of the assumed interventions are only meant to illustrate the likely reduction in child malnutrition under one possible scenario. It is obviously not possible to predict whether the assumed interventions will indeed take place, and, even ifthey do, whether they will proceed as the pace assumed inTable IV.1. 4.48 As in the previous chapter, we assume that mean district monthly consumption expenditure per capita will grow annually at about 2.7% to 2015, which would be consis- tent with an annual per capita GDP growth rate of 4%, given the historical relationship between GDP and consumption growth in Bangladesh over the 1990s. In addition, we assume that per capita land availability will continue to decrease at the rate at which it has declined during the 1990s. Finally, we assume that flood prevention and management measures, such as construction of storage reservoirs in the upper catchments of the Ganges, the Brahmaputra and the Megna Rivers (in cooperation with other countries sharing the basins of these rivers), drainage improvements, and use of better embankment materials, will reduce the likelihood of floods. As noted earlier, none of these assump- tions are sacrosanct; they are only meant to be illustrative. The projections could be un-. dertaken for any combination of changes inthe policy or environmental variables. 4.49 Figure IV.23 Projected % of children under 6 who are underweight to 2015, shows the projected path under different interventionscenarios of the child underweight (graph shows cumulative effect of each additional intervention) rate in Bangladesh to 2015 with all of the seven 50 policy and environmental changes shown in Table 45 IV.l occurring. As would be expected, the declining 1 40 availability of land per Expanding coverage of Food-for-Workprogram MDG 35 I Real Der cams economic srowth ---Bette; acceis to bus transport I, capita i s associated with . ._ --Impmving . . .~ ~ ~ sknitation acrew , ,,,,I , , , .- - _.__ ,- , '-Reducing vulnerability to flood; ~ rising underweight rates (from about 51% to 53.5%). However, all the 30 4 1 3 0 'j w m r E r! other interventions con- N 3 w tribute to reductions in Figure 1v.23 poverty. The largest de- cline in child underweight rates (about 8 percentage points) comes about with the expan- sion o f female schooling, with economic growth also contributing to an appreciable de- cline (of 3% percentage points). The other interventions - flood control and management measures, improved bus transport, sanitation access, and improved coverage of the Food- for-Work program - are all associated with smaller declines (of about one percentage point each) in child underweight rates to 2015. Together, the seven interventions are as- sociated with a reduction of about 12 percentage points in the child underweight rate - bringingthe child underweight rate down from 51% to 39% - about 5 percentage points above the MDG level (34%). 4.50 These results suggest that even though attainment of the child nutrition MDG will be challenging in Bangladesh, it should be possible to bringchild underweight rates down 49 sharply (and relatively close to the MD target) with a package of interventions that in- cludes economic growth, flood control and management, expansion of female schooling, improved physical infrastructure (transport and sanitation access), and greater coverage by food assistance programs, such as Food-for-Work. 50 V.PRIMARY SCHOOLING 5.1 Universal primary enrollment i s one of the main education-related MDGs. The primary enrollment ratio i s loo%, and that all the pupils entering grade 1are retainedun- millennium development goal i s to ensure that, by 2015, all children are in school, the net til grade 5 (typically the last year of primary school). 5.2 The numerous benefits of schooling are well-known and have been widely dis- cussed in the literature on economic development. Schooling i s one o f the most powerful instruments for reducing poverty, unemployment and inequality; improving health and nutrition; and promoting sustained, human development-led growth. It i s also self- perpetuating across generations, with educated parents much more likely to provide schooling to their children. Both the pecuniary and non-pecuniary returns from schooling have been well documented inthe literature for several countries, including Bangladesh. OverallTrends 5.3 Enrollment Rates. Bangladesh has achieved rapid progress in schooling during the last two decades. The gross primary enrollment rate, which was only 61% in 1980, increased to 72% by 1990 and to 96% by 2000. 5.4 However, as in other developing countries, gross enrollment rates tend to be greater than net primary enrollment rates because of the late entry of children (Le., be- yond age 6) into primary school and the resulting enrollment of overage children &e., those above age 10) at the primary level. In the case of Bangladesh, the net primary en- rollment rate from administrative data i s estimated to be about 86% in 2000. 5.5 Household surveys present yet another set of estimates. For example, data from the HIES 2000 indicate a net primary enrollment rate of only 65.4% in 2000.32There are many reasons for the discrepancy between household survey-based and school adminis- trative records-based enrollment rates. First, household surveys typically obtain informa- tion on whether a child i s attending school at the time of the survey, while administrative data refer to students enrolled in the registers of the school at the beginning of the school year. The latter may be greater than the former if students enroll in school at the start of the school year but then do not attend it during the remainder of the year. Second, gross enrollment rates from administrative records are very sensitive to incorrect estimates of the population of school-aged children. Third and finally, there are incentives for school administrators and district officials to overstate the number of enrolled students, since many types of government education expenditure allocations to districts and schools are often based on the number of enrolled students. 32Data from UNICEF's Multiple Indicator Cluster Survey (MICS) indicate a much higher net primary en- rolment rate - 79.8% for 2000. 51 5.6 Figure V.l shows the percentage of children of different ages that were attending school, and those that were attending primary school in particular, in 2000, based on the HIES.33 School enrollment rates are observed to increase from 29% at age 5 to 55% at age 6 and then peak at 86% at age 9. Thereafter, enrollment falls gradually until age 14, and sharply beyond that age. For the age group 6-10 Percent of children attending school and primary school, by age, 2000 years, the school enroll- m -- ment rate i s 75.2% - sig- nificantly larger than the 80 65.4% net primary enroll- 70 ment rate. The difference 60 arises largely because a 50 number of children aged 6- 40 10 years attend pre-primary --Q attending any school 30 school. The HIES data in- -7c attendingprimaiy school 20: / dicate that one-third of en- rolled students aged 6 and 16% of students aged 7 do - .^ .. .. .. .- ., .- . ~ , " " ,~ not, in fact, attend primary > school. This suggests an age FigureV.1 at entry into primary school of closer to 7 or 8 years instead of the 6 that i s officially expected. On the other hand, nearly half of all enrolled students aged 12 and 30% of students aged 13 report attending primary school. This overage enrollment results in high rates of gross (relative to net) primary enrollment. 5.7 Primary Completion.Universal primary enrollment i s only one of the education- related millennium development goals. Another goal i s retention of students - viz., to en- sure that the entire cohort of children who begins grade 1remains in school untilgrade 5. School completion i s an indicator - albeit imperfect - of the quality of schooling. It i s possible that in the rush to expand access to schooling, policy makers might compromise the quality of schooling. The compromise in quality would likely show up in lower rates o f student retention and primary school completion. 5.8 Calculating the true primary completion rate requires longitudinal data on chil- dren, but in the absence of such data, one can use household survey data on children's ever-schooled, currently-in-school, and current grade status. The HIES 2000 data reports whether a child ever went to school, whether he/she was currently attending school at the time o f the survey, the grade currently attending, and the grade last completed. 5.9 The above information can be used to calculate the primary completion rate for children aged 12 years. Obviously, 12-year olds who never attended school are excluded from the calculation of the primary completion rate. A child i s considered to have com- 33 Inwhat follows, we use the term 'enrollment' for rates estimated from both administrative and survey data, since the term 'attendance rate' refers to the percentage of school days that a student attended school. Data on such rates are rarely available from most multi-purposehousehold surveys. 52 pleted primary school if he/she reported having completed class 5 at the time of the sur- vey and if he/she was not reported as never having attended school. In 2000, the primary completion rate thus calculated was 66.3%. A similar calculation for India for approxi- mately the same year (1999-2000) yields a primary completion rate of 61.4% (World Bank 2004). 34,35 Spatial Patterns Net primary enrollment rate, by area, 2000 5.10 There are large 75 1 71 74 variations in the net primary enrollment rate across regions (Figure V.2). The rural areas of Faridpur, Tangail and Jamalpur, for instance, have net primary en- rollment rates of only 48%. On the other hand, the rural areas of Khulna, Jessore and Kushtia have net pri- mary enrollment rates of 74%. Figure V.2 Primary completion rate ('3 of 12-year olds who started and completed 5.11 Likewise, there are 5 years of primary school), by area, 2000 84 large variations in the pri- 80 mary completion rate 15 across areas (Figure V.3). 70 At 84%, Other Urban 65 Dhaka (i.e., urban areas 60 outside the Standard Met- 55 50 ropolitan Area of the city) 45 has the highest primary +% 9 @% B -8 @' +' qp+'.""v*$< $' ** -&* $9 completion rate, followed $9Gr'.&* ,>e,p"$' 6 +,.?" .**e c"dGr' .&& a?8 %@,~~,G2+o~o,@& +? ,." +s .$'Go by rural Dhaka, urban ,e' 5.""d'e 9; .I'' +* +?. , Khulna, and rural Sylhet and Comilla. The rural areas o f Noakhali and FigureV.3 Chittagong rank at the bot- tom, with primary completion rates of 52%. '' 34Increasing the potential pool of children from age 12 to those aged 12-13 years does not make an overly large difference to the estimated primary completion rate. In 2000, the estimated primary completion rate oes up from 66.3% to 71.6% when the age group 12-13 years i s considered. Another, more widely-used method of calculating the primary completion rate i s to compare the size of the first grade cohort in a given year with that o f the sixth grade cohort five years later. 53 Inter-district variations in the net primary enrollment rates are even larger (Map V.l).36The low enrollment districts are found in the eastem, western and south- ern parts of the country. Geographic Concentra- tion of Out-of-School Children 5.12 It i s useful to know the absolute population of out-of-school children, since this group would need to be targeted by schooling interventions. 5.13 The HIES 2000 in- dicates that nearly 4.7 mil- lion children aged 6-10 years (out of a total popula- tion of 18.8 million in that age group) do not attend school in Bangladesh. Fig- ure V.4, which breaks Map V.l: District-level estimates of the \ 1 down this number by area, net primary enrolment rate, 2000 suggests a high degree of regional concentration of Regional contribution to the total number of out-of-school chilriren aged this variable. The rural ar- I 6-10 in Bangladesh, ZOdO I eas of Faridpur, Tangail and Jamalpur account for 80 nearly one-fifth of all out- 70 of-school children aged 6- - c 8 50 13Cumulative sharein total number of out-of-school children aged 6-10 10 in the country. Three -- 0 40- -J Share in total number of out-of-school children aged regions (out of a total of 6-10 14) together account for ei nearly one-half - and six , , regions account for three- quarters - of all out-of- school children nationally. Figure V.4 36The district-level estimates are from administrative sources. Hence, they are not strictly comparable to the area estimates in FigureV.2, which are calculated from the HIES 2000. 54 Socioeconomic Varia- tions Net primary enrollment rate and the primary completion rate, by per capita expenditure quintile, 2000 household income con- strains primary schooling Distribution of out-of-schoolchildren aged 6-10, by per capita opportunities for the poor, consumptionexpenditure quintile, 2000 probably because the poor 2.0- I - 1 T 45 face a high opportunity 180 40 cost of sending young -E 35 children to 30 b 252 5.15 Figure V.6 shows L that o f the 4.67 million 20e out-of-school children 0.6 15y; aged 6-10, nearly two- 0.4 10 thirds (about 3 million) 0.2 5 belong to the bottom two 0.0 0 consumption quintiles. Thus, the problem of chil- dren not attending school Figure V.6 i s largely a problem of the poor. 5.16 Parental schooling. Schooling tends to transmit itself across generations, with schooled parents being much more likely to send their children to school and to keep them in school for the duration of the primary course, Figure V.7 suggests that this i s true for Bangladesh as well, but what i s interesting i s that there i s a threshold effect inthe association between adult female schooling on the one hand and net primary school en- 37 As in other developing countries, the direct costs of attending primary school are typically quite low in Bangladesh. 55 rollment and primary com- pletion rates and adult fe- Net primary enrollment rate and the primary completion rate, by male schooling on the schoolingyearsof highest-educatedadult female in household, 2000 other.38 The data suggest % of children uged 6-10 years riffending R of children q e d 12years who hove that the major difference in primor?. school coni~iieted5 jenr,s ofprinioi?. school 79 primary school enrollment 77 and completion rates occurs 75 between households in 70 which the highest-educated 65 adult female has no school- 60 ing at all and those in 55 which the highest-educated female has some schooling. 50 The number of years of ~Schooling years of highest-educatedadult female I Schooling yexs of highest-educatedadult female ~ schooling of the adult fe- in hourehnlrl in hniiwhnld male, conditional on her having some schooling, does Figure V.7 not appear to make much of a difference to the net primary attendance and completion rates of children inher household. 5.17 Occupation. There Net primary enrollment rate and the primary completion rate, are also sharp differences in by occupation of household,2000 primary schoo! enrollment Rofchildren uged 6-10years atfending %of children rig& 12 iars who huve : and completion rates across 85 1 yrimnr) school conzplered 5 Fears of primor). schooi occupational groups (Fig- 73 73 ure V.8). Some of these 4 mirror the differences ob- 70 67 served across economic groups. For instance, the children of agricultural and casual laborers, who consti- tute the poorest occupa- tional group in Bangladesh, have very low rates of pri- mary school enrollment and completion among. The Figure V.8 children o f small farmers (Le., those operating less than an acre of land) and self-employed entrepreneurs are next, while the children of large farmers, salaried employees, and other occupational groups have roughly comparable (and higher) rates of primary school enrollment and comple- tion. 38 The HIES data do not permit identification of the mother of each child in the sample (unless the sample i s restricted to biological children of the household head). We have therefore used the schooling years of the highest-educated adult female in a household as a proxy for maternal schooling. 56 5.18 The facts that casual laborers constitute a large Distribution of out-of-school children aged 6-10, by occupation of household, 2000 share of Bangladesh's 2 5 T population (about 28%) and I 2 0 9 T T t45 that their children have the 2 0-- -40 O N u m b e i (million) lowest rates of primary -tShae (70) -- 35- school enrollment result in 2 1.5-- -~30 3E a very high concentration 2 -- 25 L,0 -- of out-of-school children in 1.0 -- 20E m the country in this occupa- -- 15 0.5-- -- 10 tional group. Figure V.9 -- 5 indicates that fully 45% of T O all out-of-school children Agricultural Small farmers Large farmers Salaried Self-employed Other belong to the agricultural laboress (< 1acre) (=> 1acre) employees entrepreneurs .. . .. labor profession. Another 17% belong to self- Figure V.9 employed households, which implies that nearly two-thirds of all out-of-school children in the country are either children of agricultural laborers and self-employed entrepreneurs. These findings have obvious implications for efforts at targeting school interventions to selected groups of children. The Role of Public Interventions 5.19 Infrastructure. The HIES data suggest that ac- me, by village infrastructure, 2000 cess to infrastructure i s as- 52 of children aged 6-IOyears arrending Q of children aged 12 years who have 7.5 priman sciiool sociated with compiered 5 years of yriman sciiool generally higher primary school en- 71 12 rollment and completion 70 67 67 rates. Villages having elec- tricity coverage and a bus 65 station tend to have higher 60 net primary enrollment r- ates, but tap water and 55 sealed toilet coverage do not NO i Yes KO Yes NO 1Yes 1 NO 1 I l l 1 I Yes appear to have any signifi- Tap water Sealed Electricity Bus station ,Tap watei t..:,... cant association with ne1 primary enrollment. On the other hand, the availability Figure V.10 of tap water, sealed toilets, and electricity - but not ac- cess to a bus station - are strongly positively associated with higher primary completion rates (Figure V.10). 5.20 Government programs. The Government of Bangladesh has, for a number of years, operated several income assistance programs, one of which directly links its bene- fits to the school attendance of primary school-aged Net primar) enrollment rate and the primar) completion rate, by type of government assistance program in killage, 2000 children (viz., the Food-for- % oJchr1dreri aged 6 10 i e o u nrreriding %of children aged I 2 \ U I ? S1010 itale Y'irr1Lir~%~ c d u d p M ,e'',> u, ~ _I / i o o l Education program).39 An I 70 70 important question i s the extent to which these pro- 6.5 grams influence primary school attendance and 60 completion rates of chil- 5 5 , , , , , , , , , , dren. Figure V.ll suggests 1 NO 1Yes NO 1yes i ~ oYes , NO 1, 1 1 i 1 ' 1 ~Yes NO Yes NO Yes NO Yes NO Yes that some of these pro- Food-for- Food-for- Vulnerable Vulnerable Food-for- Food-Cor- ,Vulnerable Vulnerable work GroupFeeding ~ work ~education Group Group program 1 1 ~ ~education Group grams, especially the Food- program evelapinen program program program /levelopmen/ Feeding Programonerational in village? for-Education, Vulnerable Group Feeding and Vulner- FigureV.ll able Group Development programs are associated with higher net primary enrollment rates. But there does not ap- pear to be any association of these programs with primary completion rates. Indeed, rather surprisingly and counter-intuitively, one o f these programs - Food-for-Education - i s associated with sharply lower school completion rates. 5.21 Governance. Poor governance i s pervasive in the educational sector of Bangla- desh. Membership of school management committees i s rife with politics, and teacher recruitment i s often subject to personal influence. As in the public health sector, teacher absenteeism i s rampant, with teachers placing much greater emphasis on private tutoring than on teaching at schools. There have been numerous textbook production and pro- curement scandals over the years, with books that are supposed to be distributed for free showing up for sale in markets (World Bank 2003). Corruption in procurement has also resulted inpoor quality of school construction. These types of governance problems con- tribute to the poor quality of education in Bangladesh, and undermine the tremendous gains made inexpanding access. 5.22 Reducing teacher absenteeism and making schools accountable to students and the community i s no simple task. As the World Development Report 2004 points out, it requires broad-ranging institutional reform, incorporating, among other things, empow- erment of citizens and communities who can hold the state accountable for performance, devolution of administrative and financial powers to communities, greater autonomy to schools, involvement of parents in school management, and ensuring the motivation of front-line workers. 39 One study found that the Food-for-Education program, which was launched in 1993 to increase primary school enrollment and completion among children from very poor and landless families, did raise enroll- ment and completion rates, but that it suffered from high levels of leakage (costing 1.6 taka to transfer one taka in benefits) and was poorly targeted (with 50% of the beneficiaries coming from households above the food poverty line) (World Bank 1998).The program was discontinued in June 2002, but has been replaced by another program called the Primary Stipend Education Program. 58 5.23 Role of NGOs. As in other sectors, NGOs have played an important role in pro- moting basic education in Bangladesh. In 1990, mainstream NGOs working in the educa- tion sector formed a coalition called the "Campaign for Popular Education or CAMPE." The goal of CAMPE was to eradicate illiteracy via mass formal and informal literacy programs. At the present time, more than 400 NGOs are engaged in non-formal education programs in the country. Some of these NGOs have established innovative projects to promote basic education and literacy. For instance, the BRAC Non-Formal Primary Edu- cation Program caters to older children who never attended formal school and takes them from grade 1to 3. This program i s by far the largest single non-government primary edu- cation program in the country, with more than 30,000 schools and about a million pupils. More than 90% of the children who start in BRAC schools graduate, and a large propor- tion of the program graduates are admitted into grade 4 or higher of the government school system (Sharafuddin 1998). Multivariate Analysis 5.24 To examine the likelihood of Bangladesh attaining the child education-related MDGs, we have estimated multivariate models of net primary school enrollment and completion, using unit record data from the HIES 2000.40The multivariate models have the advantage of controlling for several variables that may be simultaneously associated with primary school enrollment and completion. The estimation results are reported in Annex Tables 4 and 5, while only the broad findings of the empirical analysis are dis- cussed here. 5.25 Net primary school enrollment. The multivariate model confirms many of the bivariate relationships discussed earlier. Older children are observed to have a higher probability of primary school enrollment (as compared to children aged 6 years), with primary school enrollment peaking at age 9. Interestingly, after controlling for age and other factors, girls are significantly more likely than boys to be enrolled in primary school, although the difference (about 3.6 percentage points) i s modest. 5.26 Household living standards, as measured by the log of monthly consumption ex- penditure per capita, have a strong positive association with primary school enrollment, with a one-percent increase in per capita consumption expenditure being associated with a 0.23% increase in the net primary school enrollment rate. 5.27 Interestingly, however, while the likelihood of primary school enrollment i s sig- nificantly and positively associated with adult female schooling in the household, it has a stronger association with adult male schooling. Each additional year of schooling o f the highest-educated adult male in the household i s associated with a 1.4 percentage point increase in the net primary enrollment rate, but the corresponding increase associated with the schooling of the highest-educated adult female in the household i s only half as 4" Since the dependent variable in the model i s a dichotomous variable (i.e., whether or not a child of a given age i s attending primary school or has completed primary school), the models have been estimated by the maximum-likelihoodprobit method. 59 much (0.7 percentage points). 41 This result is counter-intuitive and flies in the face of evidence from around the world indicating stronger associations between mother's and children's schooling than between father's and children's schooling. The result might re- flect the fact that the highest-educated adult female in the household may not necessarily be a child's mother, nor might the highest-educated adult male be a child's father. At any rate, since it i s merely net primary school enrollment that i s being analyzed here, there i s nothing in the results to suggest that adult female schooling in a household i s less impor- tant than adult male schooling to regular school attendance and improved learning out- comes among children. 5.28 Infrastructure generally has mixed associations with primary school enrollment rates. Paved roads in a district are strongly associated with primary school enrollment, with a one point increase in the percentage of roads in a district that are paved being as- sociated with a 0.4 percentage point increase in the net primary school enrollment rate. Similarly, distance to a bus station i s associated negatively with primary school enroll- ment, reflecting the importance of transport and road access to schooling decisions. The results suggest that every one-kilometer reduction in the distance of a village from the nearest bus station i s associated with an increase in the net primary enrollment rate of about 0.6 percentage points. However, electricity coverage does not appear to be a sig- nificant correlate of primary school enrollment. 5.29 Among the various government programs, the Food-for-Education program ap- pears to have a very significant and large association with primary school enrdlment. Controlling for other variables, the net primary enrollment rate in villages having a Food- for-Education program i s 8.5 percentage points higher than the corresponding rate in vil- lages not having a Food-for-Education program. The Vulnerable Group Development (VGD) program is also observed to have a very strong positive association with net pri- mary school enrollment; VGD villages on average have a net primary enrollment rate that i s 6 percentage points greater than non-VGD villages.42 5.30 An interesting question i s the extent to which access to schools and the quality of schools in a community are associated with net primary school enrollment rates. The dis- tance (as measured in minutes of walking) to the nearest primary school in a village - an indicator of school access - has no significant association with primary school enroll- ment. The lack of significance of access i s surprising, but perhaps reflects the fact that 85% of villages in Bangladesh have a primary school located in the village. On the other hand, lowering the pupil-teacher ratio at the primary level in a district - an indicator of increased schooling quality - i s significantly associated with higher rates of school en- rollment. The absolute size of the association i s small, however; the results suggest that, controlling for other factors, a one-percent reduction inthe pupil-teacher ratio in a district i s associated with an increase of only 0.1% in the net primary enrollment rate. These re- sults thus indicate that primary school enrollment in Bangladesh i s currently not con- 41 The results do not, however, show a statistically significant (at the 5% level or below) difference between the point estimates of adult male and adult female schooling. 42 Of course, these results might reflect that the Food-for-Education and VGD programs are (unintention- ally) targeted to better-off communities that happen to have higher net primary enrollment rates. 60 strained by the availability of primary but that enrollment would likely benefit (modestly) from school quality improvements in the form of a reduction of the pupil teacher ratio. 5.31 The results also suggest that the number of female teachers inthe village school i s not significantly associated with net primary school enrollment rates. This result i s sur- prising, as there is a great deal of anecdotal evidence from Bangladesh and other coun- tries indicating that parents are less reluctant to send their children, especially daughters, to school when the school teacher i s female. However, it may be the case that having fe- male teachers in school i s especially important to improving regular school attendance and learning among children, especially girls - effects that this analysis i s unable to cap- ture. 5.32 Primary completion. The results of the multivariate analysis of primary comple- tion are disappointing, since few explanatory variables are significantly associated with primary completion. The only variables that are significant are log of per capita consump- tion expenditure, adult male schooling, and the presence of the Food-for-Education pro- gram in the village. Of these, the last variable has a perverse (negative) association, sug- gesting that the Food-for-Education program i s associated with lower rates of primary completion. Such a result makes little sense, especially given the earlier finding that the Food-for-Education program i s strongly associated with higher rates o f primary school enrollment. Likewise, the significance of adult male schooling, but lack o f significance of adult female schooling, i s troubling, given the large body of evidence indicating stronger associations of mother's (relative to father's) school- Projectednet enrollmentrate to 2015, under variousinterventionscenarios ing with children's primary - 100 school enrollment and com- Intervention I ~ pletion. Interestingly, this 95 -Reducingpupil-teacher ratio result i s consistent with the -"'Improving bus transport 9o . pavedroad earlier finding that net pri- -Real per capitaeconormc growth -Expanding co\erage of VGD mary enrollment rates have 8 5 - Expandingco\erage of PrimaryEducationStipendsProgram a stronger association with ---Expansion of maleschooling 8o -Expansion of female schooling adult male schooling in a household than with adult 75 - female schooling. However, it i s not clear how much cre- 70 -70 dence one can place in these 65 unusual and counter- intuitive findings. Ln 460 5.33 The empirical results Figure v.12 suggest that a one percent increase in consumption expenditure per capita i s associated with a 0.25% increase in primary completion rates. The observed association of primary completion with adult male schooling i s also strong, with a one-year increase in the schooling of the highest- 43Naturally, given the linear prediction,this result would hold only up to some limit. 61 educated male in the household being associated with an increase of 2.5 percentage points in the primary completion rate. None of the other variables, including adult female schooling, i s significantly associated with primary completion. Simulationsto 2015 5.34 Based on the multivariate probit models estimated above, we have undertaken simulations of the primary school enrollment and completion rates in Bangladesh from 2001 to 2015 under certain assumptions. The nature and magnitude of the interventions are detailed in Table V.1. As noted previously, the scope and magnitude of the assumed interventions are only meant to illustrate the likely reduction in child malnutrition under one possible scenario. It i s obviously not possible to predict whether the assumed inter- ventions will indeed take place, and, even if they do, whether they will proceed as the pace assumed in Table V.1. 5.35 Figure V.12 shows the projected changes in the primary school enrollment rate in Bangladesh when the eight interventions shown inTable V.1are pursued simultaneously. It is obvious that, while each of the interventions contributes to the increase in primary school enrollment, the ones that are associated with the largest increases in primary school are expansion of male and female schooling, increases in household consumption expenditure per capita, and paving of rural roads. Together, the eight interventions con- sidered are associated with an increase of about 21 percentage points in the net primary enrollment rate - brhging that rate to 86% or well below the 100%MDG rats. Table V.l: Assumptions about various interventions to increase primary school enrollment and completion rates, 2000 to 2015 - Starting Assumed Ending value in change per value in Intervention 2000 year 2015 Adult male schooling 4.5 0.3 9.0 Adult female schooling (years) 2.5 0.3 7.0 Primary Education Stipends Program (suc- 2% points to cessor to Food-for-Education program) cov- a maximum erage (%) 22 of 40 40 VGD program coverage (%) 56 1% points 71 Monthly consumption expenditure per capita (Taka) 900 2.7% 1,342 % of roads in district that are paved 13 0.5% point 20 Distance to nearest bus station (kms.) 5.4 0.15 3.2 Pupil teacher ratio invillage primary school 78 -1% point 63 5.36 Figure V.13 shows the projected changes in the primary completion rate when the only three interventions that are significantly associated with primary completion are pur- sued simultaneously. Given the (perverse) inverse association between primary comple- tion and the Food-for-Education (or its successor, the Primary Education Stipends Pro- gram), an expansion in the coverage o f that program i s projected to reduce primary com- pletion rates by about 2 percentage points. Expansion of adult male schooling i s associ- 62 ated with an increase of 11 percentage points in the primary completion rate, while an- nual per capita GDP growth of 4% i s associated with an increase of 9 percentage points. Thus, the primary completion rate i s projected to increase from its base level of 66% to 80% by 2015. 5.37 What these simulations sug- I Projectedprimary completion rate to 2015, undervarious intervention gest i s that there i s a great deal of scenarios(graph shows cumulativeeffectof eachadditionalintervention) scope for raising both the primary 100- 100 Interventio V school enrollment and the primary 95 : -Real per capita economic growth 95 completion rate in Bangladesh over -Expansion of male schooling -Increased coverageof Primary Education Stipends Program 90 the next 12 years with a package of 85 i MDG 85 interventions that include economic growth, expansion of adult male and female schooling, improved physical infrastructure (mainly roads and transport), and greater coverage by food assistance pro- grams, such as the Primary Educa- tion Stipends Program. However, the achievements are still likely tc fall short of the levels called for b] the education MDGs. Figure V.13 63 VI. GENDER DISPARITY INSCHOOLING 6.1 One of the Millennium Development Goals i s to reduce gender disparities in schooling, such that the ratio of girls to boys enrolled at all schooling levels - primary and secondary - i s 100%. This report focuses on the gender disparity situation inBangla- desh and explores how far Bangladesh i s from attaining that MDG. Trends 6.2 Levels and Trends. Ratio of females to malesin primary school, 1991-2000 School-based ad- ministrative data suggest 96 1 that Bangladesh has made impressive gains in reduc- 9 2 ~ ing gender disparities in primary and secondary 90 - schooling. Data from the 8 8 - Directorate of Primary Education show that the ratio of females to males in 84 primary schools has stead- 82-1 ily increased from about I 83% in 1991 to 96% in 8o 1300 2000 (Figure VI.1). At the secondary level, thanks Figure VI-1 largely to the Bangladesh Female Secondary Stipend program, there are already more girls enrolled than boys. In 2000, Ministry of Education statistics indicate that, of the 7.65 million children enrolled injunior secondary and secondary schools, 4 million were females, which would imply a ratio o f females to males in secondary schools of 112%. 6.3 Regional Varia- Ratio of females to malesin primary andsecondary schools,by area, 120- 2000 117 tions. There are large spa- 1 1 5 - tial variations in the extent 110- of gender disparity in 105- schooling (Figure VI.2). - loo Data from the HIES 2000 95 indicate that, while the ratio 90 o f females to males in pri- 85 8O mary and secondary Q$-" ,+? #' 6' ..",..." schools in the entire coun- ++- &\*' ,."'+5* ,'*e+' +.Q s-c \to q.?+' @& &+**+WA' $P vpQ 69' &?' ** I.+ 3 ,Q-" 0 try i s 97%, it varies from a +0' B','&46. &@eo &."'3+@ 3 i r .a*. 9- ,." .d*9.5'8' 8 low o f 86% in SMA Dhaka to a high o f 117% in Other Urban Dhaka. Six regions, Figure 64 out of a total of 14, have a female majority in primary and secondary schools. These in- clude the rural areas of Noakhali, Chittagong, Rajshahi, Pabna, Barisal and Pathuakali and the urban areas of Khulna, Rajshahi and Dhaka. Gender Patternsby Age Percent of children attending school, by age and sex, 2000 6.4 Figure VI.3 shows the pattern of male and fe- loo 90 male school attendance by 1I age. Until age 9, approxi- 70 80 I mately the same proportion 60 of males and females attend 50 school. However, beyond 40 -Males age 9, the percent of fe- 30 -Females -Female& males attending schools i s 20 - consistently higher than the 10 percent of males attending 0 school, and this trend con- 5 6 7 8 9 10 11 12 13 14 15 16 17 18 A ^^ ^^_^\ , tinues untilage 18. Figure VI.3 6.5 These results are nothing short of astonish- ing, since they are so dif- ferent from the pattern found in the other countries of South Asia as well as in other countries at Bangla- desh's level of per capita GDP. Figures VI.4 and i VIS contrast the pattern of 40 sex-specific schooling at- + 20 tendance in Bangladesh -India -Bangladesh 10 - I O against that found in India 0 (using data from the 55th L O 6 7 8 9 10 11 12 13 14 15 16 17 18 round of the National Sam- ple Survey conducted in 1999-2000). Between ages Figure VI.4 10 and 18, age-specific school attendance rates for boys are higher in India than in Bangladesh (Figure VI.4). However, the pattern i s completely reversed for girls (Figure VIS). At virtually every age, Bangladeshi girls have higher rates of school attendance than Indian girls. 65 FemaleSecondary School StipendProgram Percentof girls attendingschool in BangladeshandIndia, by age, 2000 loo 1 T loo 90 6.6 What i s responsi- 8o ble for these unusual re- sults in Bangladesh? The 7o uncommonly large en- rollment of girls in secon- dary school i s largely the result of a government ini- tiative - the Female Sec- 2 0 - ondary School Stipend lo , (FSSS) program - launched in 1994. Under 6 7 8 9 10 11 12 13 14 15 16 17 18 the FSSS, the government provides a cash incentive 01 FigureVI.5 stipend to households to cover a large portion of direct school expenses incurred by girls in grades 6-10.44The sti- pend i s paid directly to an account specially set up for each girl in a nearby commercial bank. The recipient girls are expected to pay miscellaneous school fees (but not tuition fees) out of their stipend. The FSSS program also provides tuition assistance, though this part of the financial assistance i s paid to the school where the girl i s enrolled, rather than to the girl directly. The coverage of other costs rises with grade because extra inceiitives I , are needed in the upper grades to reduce high dropout rates. The program simultaneously has attempted to raise the number of teachers - especially female teachers - in secondary school; provide occupa- tional skills training to girls Ratioof femalesto malesin primaryand secondary school,by per capita who are about to graduate; expenditurequintile,2000 make schools more attrac- tive to provide a healthier and safer setting for girls; and strengthened govern- ment institutions for secon- 95 - dary education. 6.7 The program ap- pears to have been huge1 ~ successful in its twin ob- jectives of increasing the Figure Bo[rom aecono iniia rouirn ' OP number o f girl students Per capita expendituie quintile entering secondary school as well as keeping them in school until graduation. Clearly, with this program, Bangladesh has become a pioneer in South Asia in increasing female secondary enrollments and innarrowing gender disparities at the secondary level. 44 Unlike primary schools, which are free, secondary schools require payment of tuition fees in Bangladesh. In addition, households have to incur all other costs, such as transportation, books, uniforms, school sup- plies, and examination fees. 66 Socioeconomic Variations 6.8 Living standards. There i s no clear association between gender disparity in schooling and household living standards (Figure VI.6). The ratio of females to males in primary and secondary i s highest for the second consumption quintile, followed by the third quintile. The other three quintiles have approximately similar ratios of females to males inprimary and secondary school. 6.9 Mother's school- ing. There i s likewise little Ratio of femalesto males in primary and secondary school, by schoolingyears of highest-educatedadult female in household, 2000 relationship between gen- 105 105 der disparity in schooling and adult female schooling in a household (Figure ion - V1.7). Households where the highest-educated female 95 - 94 has 8-9 years of schooling 92 are observed to have the smallest ratio of females to males in primary and sec- ondary school. .. I 1 n-4 I 5-7 I R-9 i n + ! 6.10 Occupation and Figure sex of household head. There are some unusual patterns of gender disparity Ratio of femalesto males in primary and secondaryschool, by occupationof household and sex of householdhead,2000 across occupational groups (Figure VI.8). The gender 106 disparity in schooling i s greatest among large farm- ers @e., those operating one or more acre of land) and smallest among small farmers (those operating less than one acre). Agri- cultural laborer households also have relatively low Occupation of household Sex of household gender disparity. 6.11 Interestingly, gender Figure VI.8 disparity varies significantly by the sex of the householdhead (Figure VI.8). Inhouseholds where the head i s male, the ratio of females to males in primary and secondary i s 96%, but in households where the head i s female, the corresponding ratio i s as high as 108. This suggests that female- 67 headed households are more likely than male-headed households to encourage girls to attend and stay in school. The Role of Public Interventions 6.12 Infrastructure. The HIES 2000 data suggest that access to infrastructure i s gen- erally associated with higher ratios of females to males in primary and secondary school (Figure VI.9). Inparticular, access to tap water, electricity, and a bus station are all asso- ciated with significantly lower gender disparity in schooling. The availability of tap water probably relieves girls in the household from time-consuming water-collection chores, which in turn makes time for them to attend school. Likewise, a bus station in the village provides easier (and safer) access to secondary schools that are typically located outside most villages. Since parents are more likely to not send girls to far-away schools owing to safety concerns, the avail- ability of bus transport Ratio of femalesto males in primary and secondary schools, by infrastructure availability in village, ZOO0 helps girls proportionately more than it does boys. 110 109 - 104 6.13 School quality. 105 - Does school quality in the 100 99 99 form of lower pupil-teacher - ratios influence gender dis- 96 95 94 parity? The HIES 3000 data 92 certainly suggest so (Figure 90 VI.10).The ratio of femdles to males in primary and 85 secondary schools i s sig- nificantly greater invillages where the primary school has a pupil-teacher ratio of Figure VI.9 less than 50 than in villages Ratioof femaleto malestudents inprimary andsecondaryschools,by where the ratio exceeds 50 pupil-teacherratiointhe village primaryschool, 2000 (107% versus 94%). This suggests that parents are `lo1I less likely to send their 107 daughters (relative to their sons) to `overcrowded' schools that have a large 100 number of students relative to teachers. The reasons for this might be concern for 95 94 their daughters' security or the perception that their daughters might not benefit \JU I >=JU much in such environ- Pupiliteacher ratio in village primary school ments. 68 6.14 Government programs. The two government programs that have the largest as- sociation with reduced gender disparity in schooling are the Food-for-Work program and the Vulnerable Group Development program (Figure VI.11). Since neither of these pro- grams explicitly target girls' schooling, it i s not clear why these programs are associated with significantly higher ratios of females to males in primary and secondary school. Again, it i s probably the case that availability of these programs releases girls from hav- ing to perform time-consuming chores in the household, and this is what results in their higher rates of school attendance. MultivariateAnalysis Ratio of femalesto males in primary and secondary schools, by type of governmentprogram in village, ZOO0 6.15 Bangladesh has al- IO5 ready achieved the MDG 102 related to gender disparity 100 99 in schooling opportunities, 1 97 97 as the ratio of females to 95 males in primary and sec- 93 ondary schools was 97% in 90 89 2000. There is, therefore, no point in undertaking simulations of this ratio 85 through 2015. Neverthe- Fwd-for-work program Fmd-for-education Vulnerable Group Vulnerable Group Feeding ~ 11 less, as it might still be use- program De\,elopment program program ful to know which variables influence male and female Figure VI.11 school attendance differen- tially, we have estimated a multivariate model of school enrollment at the primary and secondary levels separately for boys and girls aged 6-18 years, using unit record data from the HIES 2000.45The multivariate model has the advantage of controlling for sev- eral variables that may be simultaneously associated with school enrollment of 6-18 year olds. The estimation results are reported inAnnex Tables 6, while only the broad findings of the empirical analysis are discussed here. 6.16 The multivariate model confirms that even after controlling for other variables, girls have a significantly higher probability of attending school, and a lower likelihood of dropping out, than boys past the age of 10 (and untilthe age of 17). 6.17 Household living standards, as measured by the log of monthly consumption ex- penditure per capita, have a stronger positive association with the school enrollment of boys than of girls, although the difference i s not ~ i g n i f i c a n tHowever, greater inequal- . ~ ~ 45 Since the dependent variable in the model i s a dichotomous variable (i.e., whether or not a child aged 6- 18 years i s attending primary or secondary school), the model has been estimated by the maximum- likelihood probit method. 46 Interestingly, a similar exercise for India produced the same result (World Bank 2004), suggesting that economic growth might actually widen gender disparities in schooling. 69 ity in consumption in a district i s associated with reduced enrollment of girls, but the as- sociation i s not significant for boys. 6.18 Interestingly, adult male schooling in the household has a stronger positive asso- ciation with boys' than girls' enrollment, while adult female schooling has exactly the opposite associations. This suggests that better-educated fathers (or other adult males in the household) favor boys (in terms of offering them more schooling opportunities), while better-educated mothers (or other adult females inthe household) favor girls. 6.19 Among the infrastructure variables, proximity to a bus station has a stronger asso- ciation with girls' than with boys' enrollment, implying that improved transport would be associated with a narrowing of gender disparities in schooling. 6.20 Two school-related variables also have differential associations with male and female enrollment. The availability of a secondary school in a village has a strong posi- tive association with female - but not male - school enrollment. This finding reinforces the earlier finding relating to bus transport - viz., when the distance and difficulty of reaching a school i s reduced (either by having better roads or transport or having a sec- ondary school in the village), the enrollment of girls increases much more than that of boys. 6.21 The second school-related variable relates to the quality of schools. Higher pupil- teacher ratios in the village primary school are observed to have a stronger inverse '.:sso- ciation with the enrollment of girls than with the enrollment of boys. This result could indicate that parents are less willing to send their daughters (relative to their sons) to h w - quality schools, or it could reflect that parents are concerned for their daughters' security when there are fewer teachers to supervise a large number of students. 70 VII. CONCLUSIONS 7.1 Of the five MDGs analyzed here, Bangladesh has already attained (or nearly at- tained) the goal relating to elimination of gender disparity in schooling opportunities. Bangladesh i s the only country in South Asia other than Sri Lanka to have achieved par- ity in male and female enrollments notjust at the primary level but also at the secondary level. This i s an impressive achievement for a country that i s one of the poorest countries inthe world, with a per capita gross national income of only US$1,770 (inPPPterms) in 2002. The analysis inthis report suggests that attainment of two other MDGs - inparticu- lar, the reduction of consumption-poverty and under-five mortality - i s also feasible with a combination o f interventions, including sector-specific interventions (such as expanding immunization coverage and reducing pupil-teacher ratios), economic growth, improved coverage of infrastructure, and social safety-net programs (such as the District Education Stipends Program and the Vulnerable Group Development programs). However, it will be challenging for Bangladesh to attain the child malnutrition-relatedMDG as well as the education MDGs relating to universal net primary enrollment and primary completion. In the case of child malnutrition, the projections suggest that Bangladesh could come very close to - within 5 percentage points of - the MD goal of having no more than 34% of its children underweight. However, it will be very challenging for the country to attain rates of net primary enrollment and primary completion exceeding 83-86% by 2015. 7.2 These achievements represent extraordinary progress for a country that, until re- cently, was frequently derided as an "international basket case." Indeed, a recent article by Dreze (2004) suggests that Bangladesh i s now ahead of India on most social indica- tors. Bangladesh has lower infant and maternal mortality rates, higher child immunization rates, better access to `improved' water sources and sanitation, and higher primary en- rolment rates than India. As noted earlier, Bangladesh has eliminated the gender gap not only in primary education but also in secondary education, while India still has a signifi- cant gender gap at both levels. Dreze admits that "Bangladesh i s no paradise of human development,. .. but social indicators are improving quite rapidly not just for a privileged elite but also for the population at large." On the other hand, Dreze contends that ... in " India, social progress i s slower and less broad-based, despite much faster economic growth. This i s one indication, among many others, that India's development strategy i s fundamentally distorted and lop-sided." 7.3 What accounts for the extraordinary progress in improving social indicators in Bangladesh (relative to the progress made by India)? Dreze provides one possible an- swer. According to him, Bangladesh's better performance may have to do with the fact that public expenditure on health as a proportion of GDP i s almost twice as highin Bang- ladesh (1.5%) as in India (0.9%). This was not always so. In 1990, Bangladeshspent only 0.7% of its GDP on health - less than what India spent (0.9%) (UNDP 2004). Thus, Bangladesh saw public spending on health increase very sharply during the 1990s, while India experienced stagnation inpublic spending on health (inrelation to GDP growth). 71 7.4 While Dreze does not note differences between the two countries in terms of their public spending on education, it i s instructive to look at public educational expenditures inBangladeshand Indiaas well. In 1999-2001, India's public spending on education was 4.1% of its GDP - considerably greater than public spending on education inBangladesh, which was only 2.3% of GDP (UNDP 2004). However, as in the case of health, public expenditure on education in Bangladesh increased from 1.5% of GDP in 1990 to 2.3% of GDP in 1999-2001- an increase of more than 50%. Incontrast, public spending on edu- cation as a share of GDP increased by merely 5% over the same period in India - from 3.9% to 4.1% of GDP. Additionally, there i s an important difference between Bangladesh and India in the composition of public spending on education. While Bangladesh spends 45.1% of its total public expenditure on education at the pre-primary and primary level, the relevant figure for India i s 38.4%. At the other extreme, India spends 20.3% of its to- tal public spending on education at the tertiary level, in contrast to Bangladesh's 11.1% (UNDP 2004). Thus, the rapid growth of public spending on education and health in Bangladesh, combined with its better balance of educational spending across the primary and tertiary sectors (relative to India), are likely to be important factors in explaining the significant progress the country has made inits social indicators duringthe 1990s. 7.5 Another factor that i s likely to be important in explaining Bangladesh's relative success in attaining positive social outcomes i s the work of its NGOs. Bangladesh may well be the world's leader in using NGOs as vehicles of development. NGOs are in- volved in virtually every activity in the country - relief and rehabilitation, poverty alle- viation, health, educatiori, social protection, and environmental protection, to name a Cew. A villager in Bangladesh can send his or her child to an NGO school, have family plan- ning and basic health se.vices delivered by an NGO health worker, obtain micro-crrdit financing from a choice of several NGO banks, sell milk and other dairy products to an NGO dairy cooperative, and make a telephone call on an NGO telephone! Secondary education in Bangladesh i s almost entirely provided by the non-government sector - viz., the NGOs, for-profit schools, and religious schools (madrasas). Likewise, many of the family planning programs of the 1970s and 1980s, which set the stage for the subsequent decline in child mortality, were primarily delivered through NGOs. And several studies suggest that micro-credit programs, which were pioneered by one of the best-known NGOs in the world - the Grameen Bank, have had a significant effect on reducing pov- erty, especially among females. 7.6 NGOs in Bangladesh differ from NGOs in other developing countries in an im- portant way: "...Several of these organizations have become very large, very profes- sional, and they have become a model for others. Bangladesh, one of the poorest coun- tries in the world and the last place you would have expected this to happen, has really become a leader in showing what the voluntary sector can do" (Smillie 1998). 7.7 Yet another factor in explaining Bangladesh's success, especially its ability to eliminate gender disparity in enrollment even at the secondary level, i s the use of targeted public interventions, such as the Female Secondary School Stipend Program (FSSS). The FSSS program i s essentially a Conditional Cash Transfer (CCT) or a demand-side inter- vention for rural girls (the majority of whom are poor) to attend secondary school. B y all 72 indications, the FSSS program has been hugely successful in increasing female secondary school enrollments, especially since secondary schooling in Bangladesh i s not free and parents are often unwilling to invest inthe secondary schooling of their daughters. 7.8 However, Bangladesh's progress on the MD indicators during the 1990s does not mean that there are no problems going forward. Indeed, there are several areas of concern highlighted in this report. First,there are very large regional disparities in virtually all of the MD indicators in Bangladesh. Districts such as Noakhali, Pathuakali, Chittagong, Ra- jshahi, and Sylhet have generally not performed well on several of the MD indicators. Even if Bangladesh as a whole attains some of the MDGs, there will be several areas of the country that will remain distantly behind. The analysis in this report suggests that many of the MD indicators are geographically concentrated in a few regions. This in turn means that targeting interventions, central government resources, and economic growth opportunities to the lagging divisions and districts will speed up attainment of the MDGs. 7.9 Second, the problem of governance - in particular, poor service delivery - i s widespread in the social sectors in Bangladesh. Doctors, health workers and teachers are typically absent from their assigned posts at government health centers and schools. Membership of school management committees i s highly politicized, and teacher re- cruitment i s often subject to personal influence. Procurement o f textbooks and essential drugs i s rife with corruption. The quality of health and education services offered at most government health facilities i s generally very poor. Yet the evidence presented in this re- port indicates the tremendous importance of service delivery in influencing MD out- comes. Infant and under-five mortality rates have fallen most in areas where effective family planning and MCWFP programs are delivered to rural women with low schooling; female school enrollments have increased thanks to a well-designed and well-delivered secondary stipend program that reaches its intended beneficiaries; and public transfer programs that deliver food supplies to the vulnerable in rural areas, such as Food-for- Work, Vulnerable Group Feeding and Vulnerable Group Development, are associated with large reductions in child malnutrition among the poorest children. This suggests that better governance and improved delivery of social services, in particular, would be very important to attaining the MDGs. 7.10 Better delivery of public services - whether in health, schooling, nutrition, or in- frastructure - i s a complex and difficult task that entails creation of the right institutions and incentives, including devolving responsibility for service delivery to local govern- ments and communities, contracting out certain types of service delivery to the non- government sector, empowering consumers to demand better services from government health facilities, introducing competition among public providers, and ensuring the moti- vation of front-line workers (World Bank 2003). 7.11 There are some other findings in this report that are useful to reiterate. The report notes there i s evidence of significant synergies among the different MDGs. For instance, a reduction in the proportion of underweight children i s strongly associated with a reduc- tion of child mortality. Although maternal mortality i s an MD indicator that has not been analyzed in this report, it i s clear that interventions that reduce maternal mortality, such 73 as tetanus immunization, expansion of antenatal care coverage, and an increase in the ra- tio of professionally-attendeddeliveries, will also bring about large reductions in infant (especially neonatal) mortality. Likewise, reducing child malnutrition i s likely to result in both schooling quantity and quality, as better nourished children are more likely to attend school and perform better in school. Thus there are synergies amongst the MDGs that will help in their attainment, which implies that proceeding with simultaneous action on all these measures will have the greatest impact on attainment of the MDGs. 7.12 At the same, it needs to be realized that the different MDGs are not necessarily internally consistent. For instance, simultaneous attainment of the poverty and child mal- nutrition MDGs by Bangladesh would result in 30% of the population being poor but 34% of the children being underweight. The contrast i s even greater when the results of the simulations undertaken in this report are considered. We find that, under plausible scenarios, Bangladesh could bringdown its poverty headcount rate to 16% by 2015, but it would still have many as 39% of its children underweight. Thus, a large number of chil- dren who are classified as non-poor would in fact be underweight. This inconsistency in- dicates a problem in the manner in which poverty and/or underweight thresholds are es- tablished. 7.13 The simulations carried out in this report also suggest that economic growth that brings about an improvement in household living standards is strongly associated with virtually every MD indicator. For example, real per capita GDP growth of 4% per annum inBangladeshcould alcne bringdown the under-five mortality rateby about 8 death: per 1,000 live births and the incidence of poverty by 21 percentage points between now and 2015. Inaddition, this growth could bring about an increase in the net primary enrollment rate o f 5 percentage points by 2015. In other words, rapid economic growth could make significant contributions to an improvement in all the MD indicators between now and 2015. 7.14 Finally, the importance of systematically monitoring MD outcomes at disaggre- gated levels and evaluating the impact of public programs cannot be overemphasized. There i s a paucity of reliable, time-series data on most MD indicators at the district and upazila (sub-district) levels. The lack of such data makes it virtually impossible to moni- tor progress toward attainment of the MDGs at lower levels of administration. In addi- tion, with the exception of a few food assistance and micro-credit programs, most public interventions in Bangladesh have not been subjected to rigorous, independent evaluation. Inorder to choose the right set of interventions with which to attain the MDGs, it is criti- cal to know which programs havebeen successful in improving MD indicators and which have not. 74 ANNEX TABLES Annex Table 1:Maximum likelihoodprobit estimatesof the probability of a household being poor, 2000 Asymp. Independent variable Parameter z-ratio Owned acres of land -0.5792 -10.96 Whether agricultural labor household?* 0.2381 11.24 Whether salaried employee household?* 0.1001 3.58 Whether small farm (<1 acre) household?* 0.0805 2.83 Whether large farm (1 or more acres) household?* 0.0647 1.84 Age of household head (years) -0.0018 -2.70 Whether household head female?* 0.0036 0.10 Schooling years of highest-educated adult male -0.0237 -9.64 Schooling years of highest-educated adult female -0.0273 -8.95 Log of mean district consumption expenditure per capita -0.5671 -13.53 Household size 0.0423 12.08 Whether village has electricity?" -0.0518 -2.86 Whether village has Food-for-Work program?* 0.0223 1.23 Whether village has Vulnerable Group Development program?" 0.0099 0.57 % of roads in district that are paved -0.0002 -0.10 Whether village has bus station?* -0.0518 -2.48 Giniindex of consumption inequality indistrict 0.0119 5.92 Per capita land availability in district 0.2264 1.47 Number of observations 4,700 Chi-squared test 1,620 Log likelihood ratio -2,440 Pseudo R-squared 0.249 Notes: Estimation employs unit record data from the 2000 HIES,merged with relevant district- and village-level data. Standard errors are corrected for heteroscedasticityusing the Huber-white method. All coefficients are expressed as marginaleffects (Le., the change in probability of being poor with a one-unit change in the right-side variable.) An "*" implies the variable is dichotomous. Figures in bold indicatestatisticallysignificanceof the marginaleffect at the 10% or lower level. 75 Annex Table2: Maximum likelihoodprobitestimatesof the probability of a child dyingbefore the age of 60 months,2000 Asynzpt. Independentvariable Parameter z-ratio Whether urban resident?* 0.0279 2.44 Birthorder 0.0009 0.34 Whether child female?" -0.0154 -1.60 Whether child female* x Birth order 0.0044 1.65 Predicted log of monthly consumption expenditure per capita -0.0224 -2.03 Gini index of inequality of predictedhousehold con- sumption expenditure per capita -0.0003 -0.51 Per capita availability of land (acres) in district 0.0201 0.84 Whether piped water available to household?* 0.0059 0.38 Whether household has no access to toilet?" -0.0020 -0.28 Mother's schooling years -0.0043 -3.55 Father's schooling years -0.0007 -0.72 Whether child was multiplebirth?* 0.4024 13.96 Mother's age at child's birth -0.0038 -4.96 Whether household head i s female?* 0.0062 0.51 % of children in district who have been vaccinated for measles -0.0004 -2.92 % of pregnant women in district who had profes- sional prenatal care pregnancy -0.0001 -0.46 % of households indistrict with electricity connec- tion -0.0001 -0.75 Numberof observations 10,761 Log likelihoodratio -3,280 Chi-squared test 308.91 Pseudo R-squared 0.045 Log likelihoodratio 0.045 Notes: Estimation emdovs unit record data from the 1999 DHS. Standard errors are I , corrected for heteroscedasticity using the Huber-white method. All coefficients are expressed as marginal effects (i.e,, the change in probability of a child dying with a one-unit change in the right-side variable.) An "*"implies the variable is dichotomous. Figures in bold indicate statistically significance of the marginal effect at the 10% or 76 Annex Table 3: Maximumlikelihoodprobit estimates of the probability of a childbeing underweight, 2000 Asympt. Independent variable Parameter z-ratio Age (months) 0.0037 1.51 Age squared -0.0001 -1.63 Log per capita consumption expenditure -0.0933 -4.09 Mother's schooling years -0.0187 -5.27 Whether child female?" -0.0348 -0.96 Birthorder -0.0129 -1.94 Whether child female?" x Birthorder 0.0067 0.70 Whether village has Food-for-Work program?" -0.0480 -2.20 Whether village i s electrified?* 0.0132 0.60 Distance (kms.) from nearestbus stop 0.0027 1.87 Whether village experienced flood in last 5 years?* 0.0707 2.65 Whether household has flush toilet?" -0.1477 -3.12 Gini index of inequality of household consumption expenditure per capita -0.0030 -0.97 Per capita availability of land (acres) in district -0.6356 -2.29 Number of observations 2,625 Chi-squared test 123.42 Pseudo R-squared 0.034 Log likelihoodratio -1,753 Notes: Estimation employs unit record data from the 2000 CNS, merged with house- hold data from the 2000 HIES and with relevant village-level data. Standard errors are corrected for heteroscedasticity using the Huber-white method. All coefficients are expressed as marginal effects (Le., the change in probability o f a child being under- weight with a one-unit change in the right-side variable.) An "*" implies the variable i s dichotomous. Figures in bold indicate statistically significance o f the marginal effect at the 10%or lower level. 77 Annex Table 4: Maximum likelihoodprobit estimates of the probability of a child aged 6-10 years attendingprimary school, 2000 Asymp. Independent variable Parameter z-ratio Whether child aged ... 7 years?" 0.2341 9.70 8 years?" 0.3383 14.76 9 years?" 0.3654 15.84 10 years?" 0.3225 14.11 Whether child female?" 0.0358 1.98 Log monthly household consumption expenditure per capita 0.1484 6.27 Whether household head female?" -0.0196 -0.5 1 Schooling years of highest-educated adult male 0.0136 4.7 1 Schooling years of highest-educated adult female 0.0070 1.73 % of roads in district that are paved 0.0038 1.89 Whether village has electricity?" 0.0114 0.55 Whether village has Food-for-Education program?" 0.0853 4.04 Whether village has Food-for-Work program?" -0.0024 -0.11 Whether village has Vulnerable Group Developmentprogram?" 0.0564 3.00 Distance (kms.) from nearest bus stop -0.0062 -3.69 Time (hours) to reach nearest primary school in village 0.0721 1.12 Pupil teacher ratio in village primary school -0.0007 -2.72 Number of female teachers in village primary school -0.0028 -0.42 Gini index of inequality of household consumptionexpenditure per capita -0.0026 -1.27 Per capita availability of land (acres) in district -0.1783 -1.11 Numberof observations 3083 Chi-squared test 586.93 Log likelihoodratio -1706 Pseudo R-sauared 0.1467 Notes: Estimation employs unit record data from the 2000 HIES,merged with relevant district- and village-level data. Standard errors are corrected for heteroscedasticity using the Huber-white method. All coefficients are expressed as marginal effects (Le., the change in probability of a child aged 6-10years attending primary school with a one-unit change in the right-side variable.) An "*" implies the variable i s dichotomous. Figures in bold indicate statistically significance of the mar- ginal effect at the 10%or lower level. 78 Annex Table 5: Maximumlikelihoodprobitestimatesof the probabilityof a child aged 12 years having completedprimary school (Class5>,2000 Asymp. Independent variable Parameter z-ratio Whether child female?* 0.0067 0.14 Log monthly household consumption expenditure per capita 0.1676 2.50 Gini index of inequality of household consumption expenditure per capita 0.0012 0.20 Per capita availability of land (acres) in district 0.2935 0.62 Whether household head female?* 0.0151 0.14 Schooling years of highest-educated adult male 0.0253 3.85 Schooling years of highest-educated adult female -0.0017 -0.20 % of roads in district that are paved 0.0024 0.50 Whether village has electricity?* 0.0756 1.29 Whether village has Food-for-Education program?* -0.1454 -2.45 Whether village has Food-for-Work program?* -0.0173 -0.3 1 Whether village has Vulnerable Group Development program?* -0.0739 -1.44 Distance (kms.) from nearest bus stop 0.0011 0.22 Time (hours) to reach nearestprimary school in village -0.0520 -0.3 1 Pupil teacher ratio in village primary school 0.0001 0.14 % of female teachers in village primary school -0.0111 -0.65 Number of observations 410.0000 Chi-squared test -237.8900 Log likelihoodratio 50.5800 Pseudo R-squared 0.0961 Notes: Estimation employs unit record data from the 2000 HIES, merged with relevant district- and village-level data. Siandard errors are corrected for heteroscedaszcity using the Huber-white method. All coefficients are expressed as marginal effects (i.e., the change in probability of a child aged 12 years having completed primary (class 5) with a one-unit change in the right-side variable.) An implies the variable i s dichotomous. Figures in bold indicate statistically significance of the marginal effect at the 10% or lower level. 79 Annex Table 6: Maximum likelihood probit estimatesof the probability of a female or male child aged 6-18 years attending school, 2000 Females Males Asymp. Asymp. Independent variable Parameter Z-ratio Parameter z-ratio Whether aged 7 years?* 0.1988 6.3 1 0.1572 4.16 Whether aged 8 years?" 0.2726 9.22 0.2700 7.3 1 Whether aged 9 years?" 0.2740 8.60 0.2785 7.10 Whether aged 10 years?* 0.2701 9.23 0.2206 6.20 Whether aged 11 years?* 0.2695 8.13 0.1957 4.52 Whether aged 12 years?* 0.2021 6.43 0.0714 1.85 Whether aged 13 years?" 0.1706 4.54 0.0258 0.56 Whether aged 14years?* 0.1056 2.80 -0.0559 -1.24 Whether aged 15 years?* -0.0005 -0.01 -0.2529 -5.62 Whether aged 16 years?" -0.0751 -1.48 -0.2085 -4.68 Whether aged 17 years?* -0.0178 -0.28 -0.3775 -6.95 Whether aged 18 years?" -0.5020 -10.57 -0.3698 -8.27 Log monthly consumption expenditure per capita 0.1680 7.23 0.1943 8.30 Gini index of inequality of household consumptionexpendi- ture per capita -0.0036 -1.69 0.0033 1.62 Per capita availability of land (acres) in district 0.0336 0.20 0.1415 0.86 Whether household head femalc?" 0.0051 0.14 0.0502 1.30 Schooling years of highest-educated adult male in household 0.0137 4.89 0.0200 7.67 Schooling years of highest-educared adult female in household 0.0193 4.97 0.0136 3.81 % of roads paved in district 0.0027 1.36 0.0026 1.35 Whether village electrified?* 0.0074 0.36 -0.0036 -0.17 Whether Food-for-Education program operates in village?" 0.0632 2.99 0.0725 3.45 Whether Food-for-Work program operates in village?* -0.0056 -0.27 0.0159 0.79 Whether Vulnerable Group Development program operates in village?* 0.0154 0.81 0.0079 0.42 Distance to nearest bus station (kms) -0.0066 -3.72 -0.0043 -2.42 Pupil teacher ratio invillage primary school -0.0010 -3.77 -0.0005 -1.89 Percentage of female teachers in village primary school -0.0050 -0.75 -0.0056 -0.82 Whether secondary school available in village?* 0.0499 2.52 0.0154 0.78 Number of observations 3,038 3,415 Chi-squared test 923.54 832.15 Pseudo R-squared 0.2411 0.1814 Log likelihood ratio -1,454 -1,877 Notes: Estimation employs unit record data from the 2000 HIES, merged with relevant district- and village-level data. Stan- dard errors are corrected for heteroscedasticity using the Huber-white-method. All coefficients are expressed as marginal effects (Le., the change inprobability of achild aged 6-18 years attending school with a one-unit change in the right-side variable.) An "*" implies the variable is dichotomous. Figures inbold indicate statistically significance of the marginal ef- fect at the 10%or lower level. 80 SourceNear Infant mortality rate SourceNear IMR Davisa Kabirg 1911 205 1974 153 1921 198 BRSFM~ 1931 179 1975 153 Censusb BFS (1975) 153 1951 168 CPS' PGE' 1979 145 1962-65 144 1983 143 1971 161 1973 173 Source: Compiledby Begum (2001) 81 REFERENCES Ahmed, R., S. Haggblade, and T.E. Chowdhury (2000). Out of the Shadow of Famine. Evolving Food Markets and Food Policy in Bangladesh, Johns Hopkins University Press, Baltimore and London. Anand, S. and A. Sen (1997). "Concept of Human Poverty." Background paper prepared for the 1997 Human Development Report, New York. Anand, S. and M. Ravallion (1993). "Human Development in Poor Countries: On the Role of Private Incomes and Public Services", Journal of Economic Perspectives, Vol. 7, NO.1,pp. 133-150. BIDS (2001). Fighting Human Poverty: Bangladesh Human Development Report 2000, BangladeshInstitute of Development Studies- PlanningCommission, Dhaka. Bauman, Z. (2004). WastedLives: Modernity and its Outcasts, Polity Press, London. Beck, U. (1992). TheRisk Society: Towards a New Modernity, Sage: London. Chowdhury, Md. and Fazlul Karim (1994). "Participation of Girls in Education: Bangla- desh Experience," paper presented at seminar on Girls' Education in Asia, Guilin, China, mimeo. Daniels, N.(1985). Just Health Care. Cambridge University Press, Cambridge. Dev, S.M., K.S. James, and B. Sen (2002). "Causes of Fertility Decline in India and Bangladesh: The Role of Community", Economic and Political Weekly, Vol. 37, No. 43, October 26. Dreze, Jean (2004), "Bangladesh shows the way," opinion page of The Hindu, 17 Sep- tember. Easterly, W. (2001). The Elusive Questfor Growth: Economists'Adventures and Misad- ventures in the Tropics,MIT Press, Cambridge, Mass. Faaland, J. and J.R Parkinson (1975). Bangladesh: The Test Case of Development, Uni- versity Press Limited, Dhaka. Government of Bangladesh (GOB)(2003). "A National Strategy for Economic Growth, Poverty Reduction, and Social Development," Economic Relations Division (ERD), Ministry of Finance, Govt. of Bangladesh, May. 82 Hulme, D., K. Moore, and A. Shepherd (2001). "Chronic Poverty: Meanings and Ana- lytical Frameworks," CPRC Working Paper 2, Institute for Development Policy and Management, University of Manchester. Koenig M.A. and D'Souza S (1986). "Sex differences in childhood mortality in rural Bangladesh." Social Science and Medicine 22 (1): 15-22. Kabeer, N. (2002). ThePower to Choose,Verso. Khan, A.A. (1996). Discovery of Bangladesh: Explorations into the Dynamics of a Hid- den Nation, University Press Ltd, Dhaka. Khandker, Shahidur (1996). "Education Achievements and School Efficiency in Rural Bangladesh," World Bank DiscussionPaper 319, Washington, D.C. Khuda, B, and A. Barkat (1992). "An Impact Study of the Female Education Scholarship Program in Bangladesh," study prepared for the World Bank for its Bangladesh Female Secondary School Assistance Project, University Research Corporation, Dhaka. Matin, I. (2004), "Reaching the Poorest with Microfinance," Background paper prepared for the Bangladesh Chronic Poverty Report, Dhaka, draft. Mujeri, M.K. and B. Sen (2002), "A Quiet Transition: Some Aspects of the History of Economic Growth in Bangladesh, 1970-2000," Paper prepared for the Global Growth Project. Pitt, Mark and Shahidhur Khandker, 1998, "The Impact of Group-BasedCredit Programs on Poor Households in Bangladesh: Does the Gender of Participants Matter?" Journal of Political Economy 106(5): 958-996. Rahman, H.Z., M. Hossain, and B. Sen (eds.) (1996). "1987-94: Dynamics of Rural Pov- erty in Bangladesh," Bangladesh Institute of Development Studies (BIDS), Dhaka (Mimeo). Ravallion, M. and B. Sen (1994). "Impacts on Rural Poverty of Land-contingent Targeting: Some Further Results for Bangladesh", World Development, Vol. 22, No. 6, pp. 823-838. Rawls, J. (1971). A Theory of Justice, Cambridge: Mass.: Harvard University Press. Rawls, J. (1993). Justice as Fairness, Cambridge: Mass.: Harvard University Press. Rawls, J. (1996). Political Liberalism, New York: Columbia University Press. Sen, A. (1985). "Well-Being, Agency and Freedom", Journal of Philosophy, 82, April. 83 Sen, A. (1998). "Mortality as an Indicator of Economic Success and Failure", Economic Journal, Vol. 108 (January), pp. 1-25. Sen, A. (1999). Development as Freedom, Oxford University Press, Oxford. Sen, A. (2004). "Understanding Economic Reforms," Key-Note Speech given at the GDNConference inDelhi, 27 January 2004. Sen, B. and S. Begum (1998). "Methodology for Identifying the Poorest at Local Level," WHO Technical Paper No. 27, "Macroecono-mics, Health and Development" Series, Ge- neva. Sen, B. (2000). "Beyond Nation: A Re-Considerationof Akbar Ali Khan's Discovery of Bangladesh: Explorations into the Dynamics of a Hidden Nation", Bangladesh Development Studies, March. Sen, B. (2001). "Rethinking Anti-Poverty" in Changes and Challenges. A Review of Bangladesh's Development 2000, Centre for Policy Dialogue and University Press Ltd, Dhaka, pp. 115-160. Sen, B. (2003). "Anti-Poverty: Which Way Now?" Report prepared for the Asian Devel- opment Bank, Dhaka, November. Sen, B. and D. Hulme (Eds.) (2004). "Chronic Poverty in Bangladesh: Tales of Ascent, Descent, Marginality and Persistence," BIDS, Dhaka- IDPM, University of Manchesler, May, draft. Sharafuddin, A.M. (1998). "Innovations in Primary Education in Bangladesh," in Baid- yanath Saraswati, ed., The Cultural Dimension of Education, New Delhi: Indira Gandhi National Center for the Arts. Smillie, Ian (1998). "Relief and development: the struggle for synergy," Institute for In- ternational Studies, Occasional Paper #33. Brown University. Stern, N.H.(2002). "The Investment Climate, Governance, and Inclusion inBangladesh," Public Lecture given at the Bangladesh Economic Association, Dhaka (Processed). UNDP(2003). Human DevelopmentReport 2003, Oxford University Press, New York. UNDP(2004). Human Development Report 2004. Oxford University Press, New York. Valad, Paula T (1995). "The Female Secondary School Assistance Project in Bangladesh: Enhancing Women's Opportunities for Empowerment and Development through Sti- pends for Secondary Education," paper prepared for Education Week 1995 Module on Household Financing of School and Local Control of Educational Resources, World Bank, Washington, D.C. 84 Wagstaff, A. et a1 (1999). "Bangladesh: Socio-Economic Differences in Health, Nutri- tion, and Population," World Bank, Washington, D.C. World Bank (1990). "Bangladesh: Strategies for Enhancing the Role of Women in Eco- nomic Development," World Bank Country Study, World Bank, Washington, D.C. World Bank (1993). "Bangladesh Female Secondary Assistance Project," Staff Appraisal Report 11386-BD, World Bank, Washington, DC. World Bank (2002). "Poverty in Bangladesh: Building on Progress," Poverty Reduction and Economic Management Sector Unit, South Asia Region, World Bank, Washington, D.C. (Report No. 24299-BD). World Bank (2003). "Bangladesh: Education Sector Development Support Credit Initiat- ingMemorandum," June 18, mimeo. 85