Republic of Madagascar Primary Education in Times of Crisis September 2013 AFTEE AFRICA World Bank Document Report No: ACS6584 . Standard Disclaimer: . This volume is a product of the staff of the International Bank for Reconstruction and Development/ The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. . Copyright Statement: . The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Table of Contents Acknowledgments ..............................................................................................................................................i Acronyms and Abbreviations ......................................................................................................................... ii Executive Summary ......................................................................................................................................... iii Introduction ....................................................................................................................................................... 1 1. THE MALAGASY CRISIS OF 2009 ...................................................................................................... 4 POLITICAL AND INSTITUTIONAL CONTEXT ........................................................................................................ 4 ECONOMIC CONTEXT ......................................................................................................................................... 4 DEMOGRAPHIC AND HEALTH ............................................................................................................................. 5 2. THE IMPACT OF THE CRISIS ON KEY EDUCATION INDICATORS ........................................... 6 EDUCATION SECTOR GOVERNANCE ................................................................................................................... 6 ACCESS .............................................................................................................................................................. 7 DROPOUT......................................................................................................................................................... 10 OUT-OF-SCHOOL.............................................................................................................................................. 12 3. THE VARIOUS CHANNELS THROUGH WHICH THE CRISIS IMPACTED ON EDUCATION OUTCOMES ...................................................................................................................................................... 14 CONCEPTUAL FRAMEWORK............................................................................................................................ 14 CHANNEL 1: DIRECT AND INDIRECT EDUCATION COSTS ................................................................................. 18 CHANNEL 2: ACCESS TO LOANS AND HOUSEHOLD INCOME ............................................................................. 22 CHANNEL 3: RETURN ON INVESTMENT IN EDUCATION ................................................................................... 28 CHANNEL 4: OPPORTUNITY COST OF EDUCATION ........................................................................................... 32 ECONOMETRIC ANALYSIS OF THE IMPACT OF THE CRISIS ................................................................................ 33 Conclusion ........................................................................................................................................................ 38 ANNEXES ......................................................................................................................................................... 41 Annex A: Tables and Figures ................................................................................................................................... 41 Annex B: South Survey in the Districts of Amboasary and Betioky ....................................................... 57 Annex C: Adjustments to Household Survey Data, 2005 and 2010 ........................................................ 59 Annex D: Econometric Analysis Methodology ................................................................................................. 60 Annex E: Schooling Age .............................................................................................................................................. 70 Bibliography ..................................................................................................................................................... 71 List of Figures Figure 1.1: Budget Trends, 2000-11 ............................................................................................................................................... 5 Figure 1.2: Trends in Health Service Usage Statistics, 2000-10 ......................................................................................... 6 Figure 2.1: Primary Enrollment, 1999/2000 – 2010/11 (ESY Data) ............................................................................... 7 Figure 2.2: Primary Enrollment, Recorded and Projected, 2008/09-2011/12 (ESY Data) ................................... 8 Figure 2.3: Primary Gross Enrollment Rates, by Grade, 2001/02-2011/12 (ESY Data) ......................................... 8 Figure 2.4: Enrollment Rates for Children Aged 6-10 Years, by Gender and Area of Residence, 2005 and 2010 (EPM Data) ..................................................................................................................................................................................... 9 Figure 2.5: Enrollment Rates in the South, by Age, 2009 and 2012 (SS Data - Sample Selection) .................. 10 Figure 2.6: Dropout Rate between Grades 1 and 4, 2001/02-2010/11 (ESY Data) ............................................... 10 Figure 2.7: Dropout Rate in the South, by Age, 2009 and 2012 (SS Data - Sample Selection) ........................... 11 Figure 2.8: Trends in Primary Dropout Rates between 2006-08 and 2008-10, by Region (ESY Data) ......... 11 Figure 2.9: Average 2006-08 Dropout Rates and Gap between 2006-08 and 2009-11, by Region (ESY Data)........................................................................................................................................................................................................... 12 Figure 2.11: Children Having Never Attended School, by Age, 2005 and 2010 (EPM Data) .............................. 13 Figure 2.12: Share of Children in the South Having Never Attended School, by Age, 2009 and 2012 (SS Data - Sample Selection) ................................................................................................................................................................... 13 Figure 3.1: Total Education Expenditure, 2000-11 ............................................................................................................... 18 Figure 3.2: Total, Recurrent and Capital Education Expenditure, 2000-11 ............................................................... 19 Figure 3.3: Household Education per Pupil Spending, 2005 and 2010 (EPM Data) .............................................. 20 Figure 3.4: Average Household Education Spending per Pupil in the South, 2009 and 2012 ........................... 21 Figure 3.5: Share of Education in Household Spending, by Wealth Quintile, 2005 and 2010 (EPM Data) .. 22 Figure 3.6: GDP per Capita, 2002-12 ........................................................................................................................................... 23 Figure 3.7: Poverty Rate, 2005 and 2010 (EPM Data) ......................................................................................................... 23 Figure 3.8: Consumer Price Index Trends, 2003-12 ............................................................................................................. 24 Figure 3.9: Distribution of Households in the South, by Heads of Cattle Owned, 2009 and 2012 (Tracer Study) ........................................................................................................................................................................................................ 24 Figure 3.10: Distribution of Households in the South, by Area of Land Owned or Cultivated, 2009 and 2012 (Tracer Study) ............................................................................................................................................................................ 24 Figure 3.11: Share of Households in the South Having Received a Loan, Suffered an Agricultural Shock or Received a Transfer, 2009 and 2011/12 (Sample Selection) ........................................................................................... 25 Figure 3.12: Enrollment, Dropout and Repetition Rates for Children Aged 6-10 Years, by Wealth Quintile, 2005 and 2010 (EPM Data) ............................................................................................................................................................. 26 Figure 3.13: Dropout Rates in the South, by Habitat and Durable Goods Index Quintiles, 2011/12 (Tracer Study) ........................................................................................................................................................................................................ 27 Figure 3.14: Main Causes of Dropout and Non-Enrollment (Children Aged 6-17 Years), 2005 and 2010 (EPM Data) .............................................................................................................................................................................................. 27 Figure 3.17: Evolution of PASEC Results, 1998, 2005 and 2012..................................................................................... 30 Figure 3.18: Channel 3 Related Reasons Indicated by Households for Out-of-School (Children Aged 6-17 Years), 2005 and 2010 (EPM Data) .............................................................................................................................................. 31 Figure 3.19: Reasons Related to the Value of Education Mentioned by Households in the South for Out-of- School, 2010-12 Average (Tracer Study) .................................................................................................................................. 31 Figure 3.20: Work-Related Reasons for Out-of-School (Children Aged 6-17 Years) Provided by Households, 2005 and 2010 (EPM Data)................................................................................................................................... 32 List of Tables Table 2.1: Enrollment Status of Children, by Age Group, 2005 and 2010 (EPM Data) ............................................ 9 Table 3.1: Main Education Spending categories, 2008-10 ................................................................................................. 19 Table 3.2: Average Household Education Spending, by Type of School and Type of Expense, 2005 and 2010 (EPM Data) .................................................................................................................................................................................. 20 Table 3.3: Share of Households in the South Having Paid Enrollment Fees, PTA Contributions and Monthly School Fees, 2009 and 2012 (Sample Selection) .................................................................................................................... 21 Table 3.4: Share of Households (with a Child Aged 6-10 Years) Having Received a Transfer, 2005 and 2010 (EPM Data) .................................................................................................................................................................................. 25 Table 3.5: Material and Human Resources Available to Primary, 2001/02-2010/11 (ESY Data) .................. 28 Figure 3.16: Public primary Teachers, by status 2006/07-2010/11 (ESY Data) .................................................... 29 Table 3.6: Teacher Characteristics in the South, 2009 and 2012 (Sample Selection) ........................................... 29 Table 3.7: Results in Math and French in the South, by Grade, 2009 and 2012 (Sample Selection) .............. 30 Table 4.1: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years, 2005 and 2010 (EPM Data)........................................................................................................................................................................................................... 34 Table 4.3: Multinomial Logit Regressions on the Activities of Children Aged 6-14 Years, 2010 (EPM Data) ...................................................................................................................................................................................................................... 36 Table 4.4: Binomial Logit Regressions on Dropout among Children Aged 6-14 Years in the South, 2009 and 2012 (Tracer Study) ................................................................................................................................................................... 37 Acknowledgments This Report is the result of a collaborative work between the Republic of Madagascar and the Education Unit, Africa Region of the World Bank. The Ministry of Education actively participated in the design of the study and supported its development during the data collection and analysis, as well as the dissemination of results. The Ministry’s team was led by Pascal Rabetahiana, Permanent Secretary and included Joel Sabas Randrianalizandry (Head of the Department of Planning), Jacqueline Ralisiarisoa, Arsene Ravelo, Olivier Razafindranovona and Lina Rajonhson. In addition, the Regional Directorates for Education (DREN) in Anosy and Atsimo Andrefana, as well as the districts (CISCOs) of Amboasary-Sud and Betioky provided support in the field. The World Bank team was led by Fadila Caillaud (Senior Economist, Africa Region) and included Nelly Rakoto-Tiana, Marie-Hélène Cloutier, Rohen d’Aiglepierre, Harisoa Rasolonjatovo, Feda Kebede, Patrick Rakotomahefa, Ramahatra Rakotomalala, Erick Rabemananoro, Norosoa Andrianaivo, Lalaina Rasoloharison and Barnaby Rooke. Ritva Reinikka, Sector Director, Human Development, Africa Region provided detailed guidance from the initial stages of the Report, and supported the team throughout its development. Haleh Bridi, Country Director for Madagascar, provided overall guidance and useful insights into the process. Valuable comments were also received from Shanta Devarajan, Sajitha Bashir and Deon Filmer. The team is particularly grateful for the comments received from peer reviewers: Margo Hoftijzer, Alain d’Hoore, Keiko Inoue, Stefano Paternostro, Patrick Premand and Patrick Ramanantoanina. Finally, the team would like to thank all the parents, teachers, children and community members who participated in the data collection; the members of civil society organizations who facilitated the development of the analysis and provided extremely useful feedback on the design and the results; the members of the panels and participants in the various dissemination events; and development partners, especially the World Food Programme and UNICEF, who supported some of the data collection. The production of this document has been partly funded by the Rapid Social Response (RSR) Program, generously supported by the Russian Federation, Norway, the United Kingdom and Sweden. i| Acronyms and Abbreviations BEPC End of lower secondary exam (Brevet d’études du premier cycle du secondaire) CEPE End of primary exam (Certificat d’études primaires élémentaires ) CISCO Education district (Circonscription scolaire) DREN Regional education directorate (Direction régionale de l’éducation nationale) EPM Household survey (Enquête périodique auprès des ménages) ESY Education statistical yearbooks Fokontany Administrative area (village, sector or part of town) FRAM Fikambaba’ny Ray Aman-dreninb’ny Mpianatra (Parent-teacher association) GDP Gross domestic product GER Gross enrollment rate INSTAT National Statistical Institute (Institut National des Statistiques de Madagascar ) MEN Ministry of National Education MFB Ministry of Finance and Budget PASEC International learning outcomes assessment programme ( Programme d'analyse des systèmes éducatifs de la CONFEMEN) PIE Interim education plan (Plan Intérimaire de l’Éducation) PPS Public primary school (including government-aided) PTA Parent-teacher association PTR Pupil-teacher ratio SS South survey carried out in the districts of Amboasary and Betioky ZAP Administrative areas that comprise about ten schools on average ii | Executive Summary The Malagasy population has undergone a severe political and economic crisis since 2009, whose impact on children’s enrollment is little documented. 1 In a context of limited data, including the absence of a recent population census, documenting changes in education outcomes is difficult, and in particular, it is impossible to monitor precisely the extent to which the number of out-of-school children may have changed. However, various factors indicate a significant impact on education outcomes. In particular, the stagnation of primary school enrollment in a context of sustained demographic growth points to a rapid deterioration in the access to basic education services. Primary School Enrollment since 2007, actual and projections using past enrolment trends 5.0 Millions 4.95 4.8 4.78 4.6 4.4 4.37 4.2 4.0 3.8 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2002-08 trend 2004-08 trend Actual enrollment Source: Authors’ calculations based on education statistical yearbook data, 2006-11. The present report aims to fill this information gap by analyzing education trends in Madagascar between 2009 and 2012. Based on multiple data sources, the study combines a descriptive analysis of the changes related to education access, supply and demand with an econometric analysis that aims to identify the channels through which the political and economic crisis may have influenced the education sector.2 The analysis highlights a rapid deterioration in education indicators that coincides with the current Malagasy political crisis: › Primary enrollment is stagnant at 4.3 million pupils since 2009, despite sustained estimated demographic growth. The deficit in terms of the number of out-of-school children resulting from the crisis is estimated at between 400,000 and 600,000. › According to the household surveys, the enrollment rate for children aged 6 to 10 years has witnessed a serious drop over the 2005-10 period, from 80 percent to 75 percent. The decline was more pronounced for boys than for girls, (6 percentage points for boys decline vs 4 percentage points for girls), especially in rural areas, where the enrollment 1 The political crisis rapidly translated into a long economic crisis, in particular leading to: (i) a contraction of GDP of 4.1% in 2009; (ii) a drop in total public expenditure of 12% over 2008-11; (iii) a 50% reduction in capital expenditure; and (iv) an increase in unemployment and underemployment rates. 2 The analyses presented in this study are based on the following sources: (i) Monetary and budgetary data; (ii) School statistics produced by the Ministry of National Education (MEN); (iii) the last two household surveys (EPM), of 2005 and 2010; (iv) a tracer survey of the pupils, households and schools of the districts of Amboasary and Betioky in the South of Madagascar, carried out in 2012 on the basis of an original pre-crisis survey of 2009; and (v) workshops with key education sector players. iii | rate for boys decreased from 78 percent to 72 percent. In the South, decomposing enrollment rate by age shows that children aged 6-8 and 12-15 are those who have been heat the hardest, with enrollment rates declining by more than 10 percentage points.. › The dropout rate between Grade 1 and Grade 4 increased by 5.5 percentage points between 2008 and 2011. The number of pupils dropping out before Grade 5 has thus increased from 469,000 in 2008 to 724,000 in 2011, by about 255,000. In addition, the regional gaps in primary dropout rates have substantially widened. Indeed, dropouts have increased more rapidly in the Southern (including in Androy, Anosy and Atsimo- Andrefana) and in the Eastern parts of the country (including in Boeny and Melaky). › The primary survival rate has dropped significantly over the period. Of 100 pupils that access primary, less than half currently reach Grade 5, against 63 in 2007/08. › The number of civil servant teachers has dropped steeply over recent years, reaching 33 percent in 2010/11. The share of trained teachers has equally dropped, with a supposed negative impact on learning outcomes. › All learning outcomes (CEPE, PASEC and French and math in the South) have deteriorated, pursuing the trend started 10 years ago. Key Dimensions of the Deterioration of Education since the Crisis Enrollment Dropout Primary enrollment has been stagnating since The number of children dropping out before 2009. Grade 5 has risen from 469,000 in 2008 to 724,000 in 2011, about 255,000 more. The deficit in terms of the number of out-of- school children stemming from the crisis is Of 100 pupils who access primary, less than half estimated at between 400,000 and 600,000. currently reach Grade 5, against 63 in 2007/08. . Changes noted since 2009 Teachers Learning Outcomes The share of civil servant teachers has dropped over the decade, reaching 33 percent All learning outcomes (CEPE, PASEC and French in 2010/11 and math in the South) have deteriorated, pursuing the trend started 10 years ago. The share of trained teachers has also dropped, with a supposed negative impact on learning outcomes. The analysis identifies four groups of factors that may have affected schooling patterns:  Direct and Indirect Education Costs › Public education expenditure has been reduced by close to Ar 176 billion over 2008-11 (about 25 percent), equivalent to a drop in public (recurrent) unit costs of about 15 percent. iv | › Household education spending has increased considerably between 2005 and 2010, both in the public and private sectors (+36 percent on average). The increase is higher in rural areas (+45 percent, against +20 percent in urban areas). › This has led to a substantial increase in education’s share of household spending, from 2.1 percent in 2005 to 3.1 percent in 2010. › In the South districts, a greater number of households with a child enrolled in Grade 2 had to pay enrollment fees (+7 percentage points), parent-teacher association contributions, (+4 percentage points) and monthly school fees (+1 percentage point). › In the South, the high level of direct and indirect costs are often voiced by parents and school directors as the reason for dropout (mentioned by 15 percent of the households of children followed by the South surveys over 2009-12).  Household Income, including loans and transfers › GDP per capita dropped by 4.6 percent between 2008 and 2012. The share of the Malagasy population considered to be living under the poverty line increased from 67.7 percent in 2005 to 76.5 percent in 2010, by nine percentage points on average. Recent estimates suggest that this percentage may have reached 92 percent, indicating that only very few Malagasy leave with more than $2 a day. › Access to loans is relatively weak according to the household surveys, although increasing. In 2010, barely seven percent of households had received a repayable transfer. › Nevertheless, data also show that households had received repayable transfers six times more often in 2010 than in 2005. › Between 2009 and 2012 in the South, the share of households declaring to have received a loan was on the decline, from 37 percent in 2009 to 26 percent in 2011.  Return on Investment in Education3 › Since the crisis, pupil-teacher and pupil-class ratios have improved. › Teachers’ training, the availability of textbooks and instructional time have all worsened on the other hand. › Learning outcomes have continued their downward trend  Opportunity Cost of Education › Overall the salary gap between uneducated and educated workers has been reduced. The average salary for uneducated adults dropped from Ar 48,411 to Ar 29,243 between 2005 and 2010 whereas that of adults having at least completed primary has dropped from Ar 107,137 to Ar 76,691, which potentially indicates a reduction of the opportunity cost for education as a result of the crisis. 3 For the purpose of the analysis, the return on investment in education is the perception, by parents mainly, of the capacity of education to contribute to greater future income. This perception is influenced by the perceived quality of education, to which factors such as the number and training of teachers may contribute. v| › On the other hand, the necessity for children to work as a result of declining households’ income has increased. Indeed, although children have fewer economic activities in 2010, they are more often involved in household duties, even when they are enrolled in schools. Potential Factors Affecting Education in Times of Crisis Direct and Indirect Costs Household Income  Increase in household education spending  Drop in GDP per capita of 4.6% since 2008. (+36% on average; +45% in rural areas).  Increase in the number of repayable transfers  Increase in the share of education in total (times 6) over 2005-10. household spending (+1pp).  Drop in household loans in the South  Increase in the number of households with a (-11pp). pupil in Grade 2 who paid enrollment fees (+7pp) or PTA contributions (+4pp). Return on Investment in Education Opportunity Costs  Since the crisis, pupil-teacher and pupil-class  On the one hand, the contraction of ratios have improved. household income results in an increase in e  The level of teacher training, the availability in the opportunity cost of schooling due to of textbooks and instructional time have the greater need for the potential income that rather worsened. children can generate.  Learning outcomes pursue their downward  On the other hand, the impact of the trend. economic crisis on salaries may have reduced the opportunity cost. Among these factors, financial considerations are those that have the greatest impact on primary access and retention: › In 2010, 26 percent of households mention financial problems as the main reason for children’s dropout (against 19 percent in 2005) and non-enrollment (against 24 percent in 2005). › In the South districts, dropout is more often explained by children’s work and the need for family labor (in 2010, 16 percent of pupils had dropped out of school for work, against 26 percent in 2011 and 19 percent in 2012). › The importance of the financial factor is also perceived through the greater deterioration of schooling patterns among the poorer population: for children aged 6 to 10 years, enrollment has dropped by 11 percentage points among the poorest households (Quintile 1), while remaining stable for the wealthiest (Quintile 5). › Household poverty has a positive and significant impact on being out of school between 2005 and 2010. › The existence of parent-teacher association (PTA) contributions in 2009 tends to increase the probability of dropout between 2009 and 2012, according to the South survey. › Dropout does not appear to be significantly related to teacher characteristics, but is significantly increased in schools where the number of community teachers (FRAM teachers) has increased between 2009 and 2012. › Overall, the effects of the political and economic crisis appear to have negatively impacted enrollment decisions through education costs and household income, whereas vi | the effects on the perceived return on investment in education are difficult to assess with the available data. In addition, the impossibility of precisely estimating opportunity costs may explain the difficulty in assessing the impact of this factor on recent enrollment trends. Determinants of Household Education Behavior - Education Statistics, Household Surveys, South Survey and Regressions - Impact on Macro/ESY EPM South education outcomes 2005 and 2009 and 2005-10 and 2000 - 2011 Variables 2010 2012 2009-12 Channel 1: Direct and Indirect Education Costs Public Spending per pupil Decreased (–) Education Price Index Increased (+) Share of education in household spending Increased (+) Per pupil household education spending Increased (+) Increased (+) Share of households paying fees and contributions Increased (+) Negative * Channel 2: Access to Loans and Household Income GDP per Capita Decreased (–) Consumer Price Index – Staple goods Increased (+) Access to loans Increased (+) Decreased (–) ns Poverty (1) Increased (+) Increased (+) Negative * Agricultural shocks Decreased (–) Non-repayable transfers Positive * Channel 3: Return on Investment in Education Pupil-teacher and pupil-class ratios Decreased (–) Decreased (–) Share of trained teachers Decreased (–) Textbook availability Decreased (–) Number of hours of instruction Decreased (–) Share of FRAM teachers Increased (+) Increased (+) ns Learning outcomes Decreased (–) Decreased (–) Channel 4: Opportunity Cost of Education Wage differential Decreased (–) Combining activities (work, chores) with school Increased (+) Increased (+) Source: Authors. Note: (1) Poverty levels, the gap in wealth quintiles and the mention of financial problems as the cause of out-of-school for the household surveys. vii | Introduction Since 2009 the Malagasy population has undergone a profound political crisis whose impact on children’s education is still little documented. The political crisis of March 2009 was followed by the suspension of most external funding and a drastic drop in domestic financing, especially for the social sectors. The economic crisis, unfolding in parallel over the past four years, has entailed a rapid deterioration of household living conditions, with potential consequences on children’s schooling. Many anecdotes were collected in preparation for this study that confirm the fast change in the sector, and the stagnation of enrollment also testifies to the halt in progress in primary access. Nevertheless, because of the insufficiency of recent coverage data limited systematic and detailed evaluation has been undertaken until now. The objective of this study is thus to evaluate the effects of the current crisis on household education choices, to orient education policy decisions over the short and medium term. The study combines recent available qualitative and quantitative data, to propose a set of assumptions on the effects of the crisis and its impact channels on education. Short of being able to carry out a full impact analysis of the crisis, due to its national coverage, the following data and sources are analyzed in detail to better document its effects: (i) Monetary and budget data; (ii) Education statistics from the Ministry of Education (MEN); (iii) the last two household surveys (EPM), of 2005 and 2010; (iv) a follow-up survey of the pupils, households and schools of the districts of Amboasary and Betioky in the south of Madagascar, carried out in 2012 on the basis of an original pre-crisis survey, of 2009; and (iv) workshops with the main sector players. The conceptual framework approaches the determinants of household education choices and behavior from the perspective of education supply and demand. The crossed impact of supply and demand-side factors enables the identification of four channels of impact of the crisis on household education choices: direct and indirect education costs, household income and access to loans, the perceived return on investment in education and the opportunity costs of schooling. The analysis first reviews the evolution of education behavior through the main indicators, including access, dropout and equity. The evolution of the channels and the impact of the crisis on each are then reviewed in greater detail. Finally, the effects of a certain number of pupil and household characteristics on school access are determined through econometric analyses that enable the evaluation of the relevance of each channel. Available Data To describe the potential effects and impact channels of the crisis on education, several information sources are used in complementary fashion. As each source carries a series of advantages and shortcomings to respond to the questions raised, the idea is to compare the results based on each to reinforce the degree of reliability and coverage. 1| Sources of Data Available for the Study Source Period Coverage Advantages Shortcomings Macroeconomic 2000-11 MFB, INSTAT National Year-on-year change National averages statistics (annual) Education MEN (Original National 2000-11 School-level perspective statistical survey (All PPS and Year-on-year change (annual) only yearbooks questionnaires) their pupils) National and Household-level 2005 Includes poverty, Household regional perspective only. Dates are INSTAT and consumption and income surveys (EPM) (Sample- not ideal to assess the 2010 levels based) impact of the crisis. Includes various supply Selection bias (only Amboasary and demand-side factors; households with children South survey in 2009 and Betioky a tracer study of a group in Grade 2 included; Amboasary and World Bank and districts of children, households schools in unsafe areas and Betioky 2012 (Sample- and schools; and data on children who were out-of- based) learning outcomes and school in 2009 are dropout excluded). Macroeconomic statistics These data enable the analysis of the context of the ongoing crisis in Madagascar as well as the supply-side factor of education financing. Data are drawn from the economic and financial reports of the Ministry of Economy and Industry (yearly national averages are available until 2011) as well as estimations carried out by the technical team of the MEN for the Interim Education Plan, 2013-15. Education Statistical Yearbooks The statistical yearbooks of the Ministry of Education (MEN) enable the description of the school- level situation. The data are taken directly from the primary survey questionnaires, sent to each school in October every year and returned completed by headmasters in December. Data therefore provide a view of the official administrative situation of schools at the beginning of the school year. Data are compiled and checked by each district-level authority (CISCO) and sent to the MEN for consolidation. These data present the advantage of being exhaustive and available until school year 2010/11. Household Surveys (EPM) The most recent household surveys were carried out by the National Statistical Institute (INSTAT) in 2005 and 2010. These surveys are based on broad samples of households that are representative at the national and regional levels. The main objective of these surveys is to collect information on the characteristics and living conditions of households, but they also provide detailed information on enrollment. The data collection exercises were carried out in October and November 2005, and between June and October 2010. For 2005 the database includes 11,780 households and just over 54,000 individuals, spread over 561 local authorities of the 22 regions of Madagascar. The 2010 sample includes 12,460 households and 59,300 individuals, spread over 623 local authorities. The 2005 survey data are therefore used to illustrate the pre-crisis situation, and those of 2010 to illustrate the post-crisis situation. It is important however to note that this study is based on the strong assumption that education behavior was grossly stable over 2005-08, and that most of the 2| changes observed in the inter-census period having occurred as a result of the crisis, between 2009 and 2010. It is also worthy of note that the 2010 survey was only carried out a year and a half after the beginning of the crisis: on the one hand therefore, a certain number of other factors are likely to explain the differences between 2005 and 2010, and on the other, the 2010 data may only reflect part of the impact of the crisis, ongoing today. The South Survey in the Districts of Amboasary and Betioky Finally, a survey was conducted in 2009, just before the events leading to the crisis, in the districts of Amboasary and Betioky. The survey, focusing on education, nutrition and household wealth, constituted the base-line study for a school feeding programme. A follow-up survey was therefore organized in 2012, targeting a selection of schools and households surveyed in 2009, linking the evolution of household living conditions with education indicators in the context of the crisis (See Annex B). More precisely, the objective of this survey is to deepen the level of understanding of the non-enrollment and dropout phenomena in the South districts covered, and to explore changes in terms of access, equity, quality and the costs of primary education over the 2009-12 period. The 2009 sample included 299 schools offering primary Grade 2, randomly chosen. In 2012, the target sample included 155 schools. Two approaches were adopted: (i) a targeted follow-up of children surveyed in 2009 (panel data), and (ii) a new selection of children according to the same methodology used in 2009 (cross-section data). The tracer study enabled to follow-up on close to 1,700 pupils, whereas the database of the new sample selection includes about 150 schools, 3 000 pupils and 700 households. 3| 1. THE MALAGASY CRISIS OF 2009 Political and Institutional Context Over the last decade, Madagascar has undergone two profound political crises, affecting the quality of institutions. The arrival to power of Marc Ravalomanana in 2002 produced a first important crisis, although of short duration. His eviction by Andry Rajoelina in March 2009 created a second. This latter crisis, still ongoing at the time of publication of this study in June 2013, has had multiple political and institutional consequences: (i) The government has undergone numerous changes since the crisis and its efforts are mainly focused on the preparation of elections and the search of a consensus between different political factions. This political instability has translated into the suspension of numerous projects and programmes funded by international aid and into the absence of a clear policy orientation for the key sectors; (ii) The level of governance also appears to be in recession since the crisis. According to the Ibrahim Index, Madagascar was ranked 35th out of 52 countries in 2011, a stark retreat since 2008 (16th out of 44 countries). 4 Likewise, Madagascar’s place in Transparency International’s ranking of countries according to the perception of corruption has slipped, from 85th in 2008 to 100th in 2011; and (iii) Insecurity is rising steeply. Economic Context The political crisis has had a profound impact on the national economy, affecting gross domestic product.5 Over 2000-11, annual average GDP growth has been of 2.9 percent in constant terms, with high volatility over the crisis years (See Annex Table A1). Although growth resumed very quickly after the 2002 crisis (GDP increased by 9.8 percent in 2003 and by 6.1 percent on average per year over 2004-08), it has been much slower in the context of the current crisis (growth is estimated at 0.5 percent for 2010 and 1.9 percent for 2011). Whereas national income and public expenditure increased considerably over the 2003-08 period, the drop in national income since 2009 has mainly affected capital expenditure. Between 2008 and 2011, although recurrent expenditure continued to rise, capital expenditure was cut by more than half. Total public expenditure was thus reduced by 12 percent over the period (See Figure 1.1 and Annex Table A2). Grants, both for recurrent and capital expenses, have also been drastically reduced since 2008. 4 The Ibrahim index of African governance is an attempt to statistically monitor governance levels throughout Africa. Financed and managed by the Mo Ibrahim Foundation, the index uses a number of indicators to compile a general ranking of countries, with the aim of helping civil societies improve the accountability of their governments (See http://www.moibrahimfoundation.org/). 5 It is worthy of mention that Madagascar’s economic trends over the past three years are the combined result of the national crisis and the international financial and economic one. 4| Figure 1.1: Budget Trends, 2000-11 Thousands of Ariary (Constant 2011 Prices) 3,000 Total Public Expenditure (Constant 2011 Prices) 2,500 Thousands ofAriary 2,000 National Income 1,500 Recurrent Expenditure 1,000 Capital Expenditure 500 0 Source: Economic and financial reports of the MEN and PIE, 2013. Note: * Provisional data. ** Estimations. The crisis has also been marked by the rapid reduction in employment, due in part to the bankruptcy of countless family production units and businesses. Greater underemployment and a drop in the employment rate have also been noted in various key sectors of the economy (McRam, 2010). Hundreds of thousands of people working in duty-free businesses, for instance, have lost their jobs because of bankruptcy. In Antananarivo, an employment survey carried out in 2010 shows a degradation in labor market characteristics between 2006 and 2010, with greater unemployment and underemployment, a boom of the informal subsistence sector and an increase in disparities (INSTAT, 2010). Madagascar was already ranked among the poorest of the world’s countries, but the crisis has made the situation worse. According to the human development index compiled by United Nations for 2011, Madagascar was ranked 151st out of 187 countries. Preliminary estimations indicate that the share of the population living under the poverty line has increased by 10 percentage points over 2008-12, the greatest part of the change having occurred over 2011-12 with the worsening of the crisis (World Bank, 2012). Furthermore, a great part of international aid for Madagascar was suspended due to international sanctions and numerous projects and programmes financed by development partners had to be suddenly interrupted. Demographic and Health The demographic transition is ongoing in Madagascar, with high fertility, greater life- expectancy and reduced infant mortality. The Malagasy population is estimated at 20.7 million for 2011 according to the latest household survey projections for 2010 (and 23.4 million according to the 2005 projections).6 The annual growth rate of 3.0 percent (3.7 percent according to the 2005 survey) is higher than the Sub-Saharan African average. The share of children has remained stable over the 1993-2011 period (20 percent of the population is aged under 5 years, 16 percent is aged 6 to 10 years and 13 percent is aged 11 to 14 years). Finally, the population is increasingly urban (currently 30.2 percent of the population lives in urban areas), although the majority continues to live in rural areas. 6 The last population census dates back to 1993. 5| Since 2009, signs of a decline in the health and nutrition status of children are increasingly apparent. With the withdrawal of donor support, the government has carried out draconian cuts in social sector budgets. Public health expenditure was 30 percent lower in 2010 than in 2009, whereas the 2012 budget is barely 50 percent of that of 2011. These cuts have entailed significant delays in the payment of health sector workers and hundreds of community health centers have had to close. Health indicators have rapidly declined. According to health sector statistical yearbooks, the vaccination coverage rate, the rate of use of medical appointments in basic health centers and the rate of use of maternity wards have all shown downward trends over recent years (See Figure 1.2 below). Furthermore, in terms of nutrition, UNICEF estimations indicate that in 2008, up to 42 percent of children aged under five years suffered from severe malnutrition in the arid regions of the south and the east, west and north coasts.7 Figure 1.2: Trends in Health Service Usage Statistics, 2000-10 Percent 100 80 Percent 60 40 20 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 DTC3 Health Center Hospital Health Center Political Coverage Maternity Maternity Appointment Crises Rate * Usage Usage ** Usage Source: Jacquet et al., 2012. Note: * Infants aged 0-11 months; ** Includes district, university and regional hospitals. 2. THE IMPACT OF THE CRISIS ON KEY EDUCATION INDICATORS Education Sector Governance With no clear strategy and repeated staff replacements at the management level, education sector governance has deteriorated. The 2005-08 period was marked by the preparation of sector strategies including important reforms aiming to improve the functionning and performance of the education system. These reforms included among others the extension of the primary cycle from five to seven years as well as the revision of curricula and the training of teachers. Implemented as pilots in many education districts (CISCOs), the reforms were not scaled up to the national level and in some cases were even interrupted because of the crisis. The lack of continuity in the management of the sector, both at the central, regional (DREN), district (CISCO) and local (ZAP) levels has translated into a lack of strategic vision. The elaboration and adoption of an Interim Education Plan (PIE) for the 2013-15 period should allow to temporarily mitigate this situation by guiding the short- term development of the sector (PIE, 2013). 7 “The Anosy and Amoron’ Imania regions have high shares, meaning severe, of nutritional deficit, close to 40.8 percent and 41.7 percent of children under 5 years in 2008. In regions such as Androy, Atsinanana, Upper Matsiatra, Bongolova, Itasy and Vakinankaratra, between 30 percent and 34 percent of children are affected by severe nutritional deficits” (UNICEF, 2011). 6| Numerous strikes and delays in the payment of teachers obstruct the smooth running of the sector. At the end of the 2011/12 school year, some regions had accumulated more than four months of primary and secondary teacher strikes. This has impeded the completion of the curriculum in the course of the year and entailed the deferral of national examinations (CEPE, BEPC and baccalaureate) to the summer of 2012. 8 Likewise, frequent delays in the payment of teachers, in particular community teachers (or FRAM teachers, named after the PTAs that recruit and pay them) have important fall-backs on their motivation, presence in class and ultimately on the quality of teaching. Access Enrollment Malagasy primary enrollment numbers have been stagnating since 2009, despite the growth of the population. Indeed, the trend of regular growth in primary enrollment noted over the 1999/2000 to 2008/09 period was interrupted in 2009/10, coinciding with the beginning of the crisis. During the second year of the crisis (2010/11) numbers even dropped (See Figure 2.1 below). Overall, the number of primary school-aged children enrolled since the beginning of the crisis has remained stable at 4.3 million. The inversion of the growth in enrollment has affected the private sector relatively more than public schools (See Annex Tables A3 and A4). The share of the private sector has dropped considerably at the primary level, from 21.4 percent in 2001/02 to 17.8 percent in 2010/11. Figure 2.1: Primary Enrollment, 1999/2000 – 2010/11 (ESY Data) Thousands of Pupils 5,000 Total 4,000 Enrollment Thousands of Pupils 3,000 Public Sector Enrollment 2,000 New Grade 1 1,000 Entrants Private Sector 0 Enrollment Source: Education statistical yearbooks, 2000-11 (See Annex Table A3). The crisis has interrupted a decade of progress towards universal primary education, and has impeded the enrollment of about half a million children. The stagnation of primary enrollment is all the more striking when compared with enrollment trends over the period immediately preceding the crisis. Thus, before the crisis the annual growth rate of primary enrollment is estimated at 7.8 percent on average over 10 years and 6.2 percent on average over 3 years. On this basis, and the assumption that enrollment would have continued to grow at these rates without the crisis, between 400,000 and 600,000 additional children could have been enrolled over these four years (See Figure 2.2). 8 The CEPE (Certificat d’études primaires élémentaires, or primary education certificate) is sat by pupils in primary Grade 5 and the BEPC (Brevet d’études du premier cycle du secondaire, or lower secondary completion examination) at the end of lower secondary. 7| Figure 2.2: Primary Enrollment, Recorded and Projected, 2008/09-2011/12 (ESY Data) Thousands of Pupils 5,000 4,947 Thousands of Pupils 4,800 4,783 4,600 4,400 4,366 4,200 4,000 3,800 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2002-08 trend 2004-08 trend Actual enrollment Source: Authors’ calculations based on education statistical yearbooks, 2007-11. Coverage Primary enrollment rates have been dropping steeply since 2009, for all grades. According to the education statistical yearbooks, the gross enrollment rate (GER) for primary dropped from 157 percent in 2008/09 to 147 percent in 2010/11, equivalent to a 9.6 percentage point drop (See Figure 2.3). The rate has decreased especially for Grades 2 and 5, losing 16 and 14 percentage points over the period, respectively. According to the household surveys, the enrollment rate of children aged 6 to 14 years has also receded strongly over the 2005-10 period, from 77 percent to 74 percent. This drop is most apparent for primary school-aged children (aged 6 to 10 years), for whom the rate was reduced by 4.6 percentage points, against 1.2 percentage points for children aged 11 to 14 years (See Table 2.1). Thus in 2010 it is estimated that close to 800,000 Malagasy children aged 6 to 10 years were out-of-school, or about 25 percent, an increase of 214,000 children since 2005. Figure 2.3: Primary Gross Enrollment Rates, by Grade, 2001/02-2011/12 (ESY Data) Percent 300 250 Grade 1 200 Grade 2 Percent 150 Grade 3 100 Grade 4 50 Grade 5 0 Total 2001/02 2002/03 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 Source: Education statistical yearbooks, 2000-12 and EPM, 2005 (See Annex Table A5). Note: Based on a population growth assumption of 3%. 8| Table 2.1: Enrollment Status of Children, by Age Group, 2005 and 2010 (EPM Data) Number and Percent 2005-10 Gap 2005 2010 (Number – (% Points) Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out Enrolled out Enrolled out Enrolled 6-10 yrs 2,268,834 80,347 502,655 2,388,293 91,774 705,251 119,459 11,427 202,596 % 79.6 2.8 17.6 75.0 2.9 22.1 -4.58 0.06 4.52 11-14 yrs 1,471,436 298,651 224,493 1,549,049 323,452 261,517 77,613 24,801 37,024 % 73.8 15.0 11.3 72.6 15.2 12.3 -1.18 0.18 1.00 6-14 yrs 3,740,270 378,998 727,148 3,937,342 415,226 966,768 197,072 36,228 239,620 % 77.2 7.8 15.0 74.0 7.8 18.2 -3.16 -0.01 3.17 Source: Household surveys, 2005 and 2010 (See also Annex Table A6 that provides the status by single age). The drop in enrollment rates is unequal, especially affecting rural zones and the southern regions. On the basis of the household surveys, the enrollment rates for children aged 6 to 10 years have dropped for all socioeconomic categories, the greatest reductions and levels of exclusion being observed in rural areas, especially for boys (See Figure 2.4 below). In rural areas, the share of children having never attended school has increased over the past five years, whereas the dropout rate has varied little (See Annex Table A8). Regional differences in terms of enrollment rates are equally important. The regions where rates have receded the most are those of Anosy, Atsimo- Andrefana, Boeny and Menabe (in the south and west), whereas other regions such as those of Itasy, Diana and Vatovavy-Fitovinany (in the north, centre and east) have improved their rates (See Annex Tables A9 and A10). Figure 2.4: Enrollment Rates for Children Aged 6-10 Years, by Gender and Area of Residence, 2005 and 2010 (EPM Data) Percent 90 87.0 86.0 2005 2010 85.1 85.0 85 83.7 82.3 Percent 79.6 80.1 79.1 78.4 80 78.0 77.7 76.2 75.0 74.4 75 73.1 73.7 71.7 70 National Urban Rural Boys Girls Urban Urban Rural Rural Boys Girls Boys Girls Source: Household surveys, 2005 and 2010 (See Annex Table A8). These trends are confirmed by the South survey which shows that between 2009 and 2012 enrollment rates dropped for all ages. For children aged 9 years, for whom enrollment levels are the highest, the rate has dropped from 94 percent in 2009 to 89 percent in 2012; for children aged 14 years, usually completing the junior secondary cycle, the rates have dropped from 72 percent to 60 percent over the period. 9| Figure 2.5: Enrollment Rates in the South, by Age, 2009 and 2012 (SS Data - Sample Selection) Percent 2009 2012 92.6 94.1 92.7 90.5 90.0 100 89.2 80.9 84.6 Percent 80 71.7 69.7 89.3 86.0 82.3 84.7 79.6 79.0 60 48.3 61.1 73.3 45.5 60.0 59.8 40 5 6 7 8 9 10 11 12 13 14 15 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012 (See Annex Table A7). Note: Only includes households with a child enrolled in Grade 2. Dropout Dropout and Repetition The primary dropout rate has risen since the 2009 crisis. Between Grades 1 and 4, education statistics estimate the dropout rate at 18.7 percent for 2009/10, 5.5 percentage points more than in 2007/08. Between 2008/09 and 2010/11, the number of children dropping out before Grade 5 has thus increased from 469,000 to 724,000, about 255,000 more children (See Figure 2.6 below). Dropout levels have increased the most for Grade 1, from 19 percent in 2007/08 to 25 percent in 2009/10 (See Annex Table A11). The household surveys tend to confirm this trend, with dropout rate increasing among children aged 9 to 11 years, by 1.4 percentage points on average9. In the South survey districts, the dropout rate has also increased for all primary school-aged children, by two percentage points on average (See Figure 2.7). Figure 2.6: Dropout Rate between Grades 1 and 4, 2001/02-2010/11 (ESY Data) Percent % of Repeaters Dropout Rate 31.4 29.6 35.0 30.0 30.5 25.0 19.2 21.0 20.3 Percent 20.0 15.0 18.0 18.7 10.0 5.0 13.1 8.7 8.1 0.0 Source: Education statistical yearbooks, 2000-11 (See Annex Tables A11 and A12 for the rates by grade). 9 Dropout as defined by the household surveys covers all children having attended school at some point who were out-of- school at the time of the survey, whereas education statistics measure dropout between two consecutive school years. The dropout rate measured by the household surveys is probably underestimated due to the size of the sample. 10 | Figure 2.7: Dropout Rate in the South, by Age, 2009 and 2012 (SS Data - Sample Selection) Percent 30 25.3 2009 2012 25 20 Percent 15 10 15.8 3.3 12.6 5 1.2 0 4.3 3.5 5 6 7 8 9 10 11 12 13 14 15 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. Disparities among regions have intensified since 2009, in particular in terms of primary Figure 2.8: Trends in Primary Dropout Rates dropout rates. The regions of Androy, Anosy between 2006-08 and 2008-10, by Region (ESY and Atsimo-Andrefana in the south are again Data) those where disparities are most striking, their Percentage Points dropout rates having increased the most since the crisis, as well as the regions of Boeny and Melaky in the east (See Figure 2.8 and Annex Table A14). These regions are among the least urbanized and poorest of the country. Conversely, the regions of the center and the north have seen their dropout rates increase by less than the national average. Furthermore, a comparison of the pre-crisis dropout rates by region with their evolution since then shows that disparities appear to be on the rise. Indeed, the concentration of points in the top right-hand quarter of Figure 2.9 indicates that it is for the regions with the highest pre-crisis dropout rates (generally above 15 percent) that the rates increased the most during the crisis (by over 5 percent), whereas the points gathered in the lower left- hand quarter of the graph represent the regions where the pre-crisis dropout rate was lower on the one hand (generally below 15 percent), and has increased by much less on the other (by less than 3 percent for most regions).10 Source: Authors’ calculations based on ESY Data, 2010-11 10 The former regions include Atsinanana, Ihorombe, Atsimo-Andrefana, Vatovavy Fitovinany, Menabe, Melaky, Anosy and Androy. The latter regions include: Analamanga, Alaotra-Mangoro, Itasy, Diana, Sava, Vakinankaratra, Amoron’i Mania, Sofia and Upper Matsiatra. 11 | Figure 2.9: Average 2006-08 Dropout Rates and Gap between 2006-08 and 2009-11, by Region (ESY Data) Percent 10.0% 2006-08 and 2009-11 Dropout Rate Gap 8.0% 6.0% between 4.0% 2.0% 0.0% -2.0% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% Average 2006-08 Dropout Rate Source: Education statistical yearbooks, 2000-11. Survival rates Survival rates have deteriorated over the period due to higher drop out and stagnating repetition. While in 2007/08, 57.4 percent of first-grade pupils achieved primary, this was only 44.5 percent in 2009/10. In addition, out of 100 children who started primary education, less than half reach the last grade of primary, 13 percentage point less than in 2007/08. The decline in survival rate is more pronounced up to CM1, reflecting the higher impact of the crisis on the first three years of primary education. Figure 2.10 : Survival rate by grade, 2007/08 and 2009/10 (ASS Data) Percent 2007/08 2009/10 120 100.0 100 81.2 74.3 68.0 80 100.0 57.4 Percent 60 39.0 75.0 40 65.1 55.5 20 44.5 33.5 0 CP1 CP2 CE CM1 CM2 6ème Source : Authors’ calculations based on education statistical yearbooks, 2000-11. Out-of-School Between 2005 and 2010, the increase in the number of children having never attended school has been higher still than that in the dropout rate. On the basis of household survey data, 70 percent of the 1.4 million out-of-school children aged 6 to 14 years in 2010 had never attended school and 30 percent had dropped out. The enrollment rates by single age having dropped for all children except those aged 14 years, and dropout rates having increased by 1.4 percentage points on average (See Figure 2.7 above), the enrollment gap is explained mainly by those children having 12 | never attended school. Indeed, for all primary school-aged children, the share of those having never attended school has increased by 4.7 percent on average over the period (See Figure 2.11 below). Figure 2.11: Children Having Never Attended School, by Age, 2005 and 2010 (EPM Data) Percent 2005 2010 50 40.6 40 Percent 30 23.1 17.8 15.1 17.9 20 30.8 13.9 11.6 13.8 10.8 10 22 15.5 12.7 0 10.4 9 7.7 10.9 6.3 6 7 8 9 10 11 12 13 14 Age Source: Household surveys, 2005 and 2010 (See Annex Table A6). Between 2009 and 2012 in the South districts, non-enrollment combines with greater dropout rates to increase the level of exclusion from school. Indeed, there is a steep increase in the non- enrollment of children from households having a child in Grade 2 over 2009-12, primarily affecting children aged six and eight years (See Figure 2.12). For this age group, the share of children never enrolled has passed from 11.4 percent in 2009 to 23.6 percent in 2012, more than the double. For children aged 14 years on the other hand, the drop in enrollment rates (from 72 percent in 1009 to 60 percent in 2012 – See Figure 2.4 above) is explained mostly by dropout (the rate is 8.9 percentage points higher), the percentage of children having never attended school having nevertheless gained 2.8 percentage points. Figure 2.12: Share of Children in the South Having Never Attended School, by Age, 2009 and 2012 (SS Data - Sample Selection) Percent 60 51.2 2009 2012 36.8 40 50.9 Percent 19.2 17.6 14.8 15.3 20 7.7 9 8.2 10.1 8.5 17.9 14.9 0 10.4 12.5 5.8 5.9 5.2 5.2 6.6 6.9 5 6 7 8 9 10 11 12 13 14 15 Age Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012 (See also Annex Table A7). 13 | 3. THE VARIOUS CHANNELS THROUGH WHICH THE CRISIS IMPACTED ON EDUCATION OUTCOMES Conceptual Framework Education Supply-Side Factors Affected by Crises In times of crisis education can be directly affected by three supply-side factors, whose relative importance determines the amplitude of the effect. (i) Governance Governance in education is a complex notion that includes the legislative and regulatory framework; the sector’s development plans and strategies; school characteristics and their alignment with the system as a whole; the mode of financial, human and physical resource allocation to schools; school accountability in resource use; teacher union issues; and so on. Crises can have a direct impact on a number of these aspects of governance, affecting the smooth running of the sector and its effectiveness. Indeed, political instability can lead to frequent staff rotations at the central or decentralized levels, or to the interruption of the implementation of reforms. Teacher strikes can be more frequent or longer. National exam programmes can be affected, delaying pupil’s progression through the system. These factors all have a direct incidence on the continued supply of education. (ii) Financing Education sector financing can assume a variety of forms: recurrent and development budgets; schools’ and other institutions’ operating expenses; the payment of teachers and teacher trainers; subsidies to households or disadvantaged pupils; classroom building programmes; and others. A macroeconomic crisis generally has a direct and immediate impact on the government’s capacity to finance social sectors such as education, due to lower national production and tax income, which can translate into budget cuts for the sector and the freeze of capital investment or teacher recruitment and training. (iii) Quality Quality in terms of education is measured by learning outcomes, and can be influenced in particular by the amount of resources (adequate infrastructure and equipment; teacher training, qualifications and numbers; available pedagogical and learning materials and so on) and their efficient use. An economic crisis can have a negative impact on the quality of education supplied if, for instance, it leads to cuts in the number of teacher posts, available textbooks and teacher training budgets, or to an increase in absenteeism or the number of days of unplanned school closure, for instance. Education Demand-Side Factors Affected by Crises A number of education demand-side factors can also be affected by a crisis. Household poverty certainly constitutes the main factor, as family wealth often suffers directly from economic downturns. Some other characteristics, linked to children, their parents or their households can also be affected. Furthermore, even if these characteristics do not vary with the crisis, their relation to enrollment can vary over time, making them more or less important in enrollment decisions. Finally, 14 | exogenous shocks other than the crisis, such as natural disasters, can have an impact on demand for education in times of crisis. (i) Poverty Household poverty is a key factor of education demand. Even for primary school-aged children for whom education is theoretically free, a series of costs must be borne, such as school lunches and supplies. These costs are generally higher still for children in secondary schools, which can be further from home, thus adding transportation costs to the bill as well as local accommodation in some cases. These constraints have an impact that is all the greater that household resources are limited. An economic crisis can have direct and multiple effects on household poverty: one of the bread-winners may lose their job, or be paid late; family production may lose value due to a contraction of demand; or access to loans may be reduced. (ii) Individual Characteristics The importance of individual, parent and household characteristics in terms of children’s education varies according to the context. The factors with the greatest impact on education choices are generally the age, gender and status of the child (birth-rank among siblings and if the child is the household head’s son/daughter) at the individual level; the gender and level of education of the household head; and the area of residence of the household, as well as if it benefits from social programmes. Although crises generally have no direct impact on these characteristics, they can nevertheless sharpen or dampen their importance in enrollment decisions. Thus, it is even less likely that girls, the children of uneducated parents or children living in rural areas for instance be enrolled during a recession than at other times. (iii) Exogenous Shocks In the case of Madagascar, exogenous shocks particularly include natural disasters, such as cyclones and the inundations that follow, or drought in the south of the island. The confluence of any external shock with an economic crisis can have an exponentially great impact on education demand, as families resort to saving whatever resources they can in the face of greater incertitude over the future. The Impact Channels of Crises on Education: A Theoretical Model A simple theoretical model of human capital investment based on the optimization of inter- temporal benefit enables to better understand households’ education choices and behavior. Such a model, valid for a multitude of contexts, can be specifically adapted to a context of macroeconomic crises to explain how they affect households’ microeconomic behavior in terms of education (Becker, 1965; Ferreira and Schady, 2008). Although the model considered is generally focused on education demand, it also takes a certain number of supply-side aspects into account, relying on the factors explained below. These, beyond their direct impact on the sector’s results, have an indirect impact on demand-side factors, and thus contribute to determine the magnitude of the impact of the crisis on the education sector, via the impact channels. The model is based on two assumptions, that agents live two periods in their life and derive a benefit uniquely from their consumption of goods. The first period starts after the beginning of a macroeconomic crisis and ends when children leave school, their time being shared between school and work according to the costs and benefits, present and future, that are related to each. It is assumed that in addition to potential paid work, children’s activities also include non-earning responsibilities such as household chores. For the sake of the modelization, these activities are attributed a value equivalent to the pay of an unqualified worker. During the second period, the child 15 | will have become an active adult, remunerated according to the human capital acquired during the first. The level of human capital is related to the education level attained according to the choices made during the first period, as well as to the exogenous quality of the education received. The second hypothesis assumes that education only increases household benefit through the better employment and income perspectives it provides, that translate into greater future consumption. The Channels of Impact on Education Supply and demand factors will have crossed impacts on household education choices. For instance sector financing has a direct incidence on household education costs. Indeed, a cut in schools’ recurrent budgets can entail schools requesting higher household contributions, whereas a freeze in capital expenditure for classroom construction may leave households with no other option but to enroll their children in private schools that are usually more expensive. On the other hand, the costs that households are able to bear are directly determined by their level of wealth, as a factor of demand. Likewise, the perception of the quality of the education service offered (supply-side factor) will determine households perception of the return on investment in the education of their children, as well as a certain number of individual characteristics such as the household head’s level of education or gender (supply-side factors). According to the model, the factors mentioned above combine within four channels of impact on education choices, whose relative importance will determine the effect of the crisis on education demand. These channels are: (i) Direct and indirect education costs; (ii) Access to loans and household income; (iii) The perceived return on investment in education; and (iv) Direct and relative opportunity costs. Each of these channels is prone to be influenced by the crisis and their effects on household education choices can be contradictory. More precisely, the crisis may affect each of the channels in the following ways: (i) Direct and Indirect Education Costs The first period costs that must be borne by households include school fees (enrollment fees, monthly school fees, parent-teacher association contributions) and the other costs related to school attendance (school supplies, uniforms, transportation, school meals and so on). An economic downturn during a crisis can potentially lead to an increase in direct and indirect costs. For instance, the government may reduce its subsidies to schools, which would increase the fees asked of households in compensation, to cover their running costs. On the other hand, some indirect costs such as supplies, meals or transport may increase, due to inflationary tendencies for instance. A significant increase in direct and indirect costs can deter enrollments and encourage dropout. (ii) Household Income, including loans and transfers This factor includes households’ direct income, as well as their loan eligibility and the existence of a credit market that enables households to borrow during the first period and reimburse during the second, to finance the education of their children and optimize the inter-temporal benefit. 11 This factor has a direct incidence on education demand. If loan access is null or limited, only households with sufficient disposable income available in the first period will be able to enroll their children. This factor also has an indirect incidence on the magnitude of the effect of direct and indirect education costs. Indeed, if household income is sufficient to cover higher schooling costs and other factors are constant, the increase in costs should not affect household investments in education. If households’ disposable income is insufficient however, the magnitude of the effect of education 11 The notion of inter-temporal benefit refers to taking the time factor into account in household and individual decisions. It is the trade-off over time of the budgetary constraint. 16 | costs will depend on households’ ability to borrow. An economic downturn following a crisis can thus reduce households’ income and their access to credit, again deterring enrollment and encouraging dropout. (iii) Perceived Return on Investment in Education This factor is effectively the expected benefit of schooling to be reaped during the second period. It includes the expected remuneration resulting from the investment in education during the first period. This will depend on the perceived quality of the education on offer, on employment prospects and on the expected wage. The benefits expected from investing in education can also be affected by crises, which may entail a freeze or a drop in salaries and reduce the employability of qualified labor. The expected results in terms of employment will depend on the perception of the long-term effects of the crisis as well as of the quality of education. (iv) Opportunity Cost of Education The opportunity cost during the first period represents the value of children’s work to which families renounce by sending them to school. Theoretically, the value of children’s work depends on the salary of unqualified labor as well as on the number of hours children may be free to work. Unqualified labor wages are generally affected by crises and households adjust their decisions accordingly. A drop in the demand for unqualified labor resulting from a contraction of the economy or an increase in the number of adults available to assume responsibility for household chores usually carried out by children can entail a drop in the opportunity cost of education and lead to greater enrollment. It is worth noting however that in the context of Madagascar, using unqualified labor wage as a proxy for opportunity costs might not be the most appropriate. Indeed, waged labor is relatively rare in this context, especially for children of primary school age in rural areas. The Influence of the Crisis on Household Education Behavior - Supply and Demand-Side Factors and Impact Channels CRISIS (Macro Level) Education Supply- Education Demand- Side Factors Side Factors IMPACT CHANNELS 1. Poverty 1. Education sector 1. Direct and indirect (?) governance 2. Exogenous shocks costs 3. Individual 2. Financing of characteristics schools/teachers/etc. 2. Access to loans and (?) (gender, area of 3. Quality of household income residence, etc.) staff/teaching 4. Others 3. Return on (?) investment in education 4. Opportunity cost of (?) education EDUCATION BEHAVIOR (Micro Level) 17 | The Net Effect of the Crisis on Education Behavior The net effect of a crisis on access to education depends on the balance between the negative effects on the one hand, and potential positive effects on the other. The contraction of the private sector, of employment and of public expenditure generally have a negative impact on education access through the rise in education costs and the drop in household income and the expected returns on education. However, the effect of opportunity costs that generally encourage enrollment could partly counterbalance the effect of lower income, making the overall impact ambiguous. Indeed, the positive effect of economic crises on enrollment through opportunity costs has been confirmed in various Latin American countries, in particular during short-term crises (Ferreira and Schady, 2008; McKenzie, 2003; Schady, 2004). The following sections of this study apply this conceptual framework to the Malagasy context, to identify the direction and importance of each of the effects. Channel 1: Direct and Indirect Education Costs Trends in Public Education Financing The budget cuts that have occurred in education expenditure since 2009 have been significant, the share of GDP allocated to education decreased by almost 1 percentage point. The conjunction of weak economic growth, the drop in national income and the reduction of the share of the national budget devoted to education have jointly resulted in education expenditure dropping since 2009. Indeed, the share of total public expenditure allocated to education has been reduced from 26.4 percent in 2008 to 20.1 percent in 2011. The sector budget now only represents 2.8 percent of national GDP, against 3.6 percent in 2008, a level significantly below the Sub-Saharan African average, of 5.0 percent (See Figure 3.1). Figure 3.1: Total Education Expenditure, 2000-11 Share of Total Public Expenditure and Share of GDP 30 23.7 28.0 26.4 21.2 20.1 20 Percent 23.9 % of Total Public 19.8 19.5 Expenditure 10 2.9 3.3 2.7 3.8 3.3 3.6 2.7 2.8 % of GDP 0 Source: Economic and Financial Reports, MFB and PIE, 2013 (See Annex Table A13). Note: Data for 2009 are provisional; data for 2010 and 2011 are estimations. Capital expenditure in particular have dropped. In constant 2011 prices, expenditure was reduced by close to Ar 176 billion between 2008 and 2011, equivalent to a quarter of resources (See Figure 3.2). Recurrent expenditure has only slightly tightened, from Ar 526 billion in 2008 to Ar 469 billion in 2010, before increasing anew in 2011 to Ar 505 billion. Most of the cuts have affected capital expenditure, which has been reduced from Ar 207 billion in 2008 to Ar 52 billion in 2011, equivalent to three quarters of the budget. The number of enrolled pupils having stagnated over the period, recurrent per pupil spending has in fact been reduced by 15 percent over 2008-11, about Ar 10,000 less per pupil. 18 | Figure 3.2: Total, Recurrent and Capital Education Expenditure, 2000-11 Billion Ariary (Constant 2011 Prices) 800 653 733 584 Total Billion Ariary (2011 522 526 557 600 441 Expenditure 381 Recurrent Prices) 400 526 505 Expenditure 446 469 Capital 419 200 368 318 Expenditure 313 52 63 207 207 0 128 154 165 Source: Economic and Financial Reports, MFB and PIE, 2013 (See Annex Table A13). Note: Data for 2009 are provisional; data for 2010 and 2011 are estimations. Although the overall amount for recurrent expenditure has been maintained, several categories of recurrent spending were affected by severe cuts, including the distribution of school kits and cash transfers to schools (MEN, 2012). Table 3.1: Main Education Spending categories, 2008-10 Thousands of Ariary 2008-2010 2008 2010 difference Primary schools basic functioning 4,574,554 0 -4,574,554 School Grants (Caisse école ) 18,240,088 0 -18,240,088 Subsidies to community teachers salaries 17,105,000 28,706,110 11,601,110 Subsidies to private schools 2,805,924 0 -2,805,924 School equipment 7,821,966 0 -7,821,966 Training 3,522,020 3,237,817 -284,203 School canteens 0 0 Other 4,070,511 10,002,917 5,932,406 Recurrent 58,140,063 41,946,843 -16,193,220 Construction 12,576,191 2,225,410 -10,350,781 Equipment 300,000 110,919 -189,081 Other 309,620 702,598 392,978 Investment 21,717,039 3,038,927 -18,678,112 Total 79,857,102 44,985,770 -34,871,332 Source: MEN Trends in Household Education Spending The significant cuts in the public primary education budgets have translated into a substantial increase in the direct and indirect education costs for households, especially in rural areas. Total household education spending per child has increased considerably, from Ar 13,000 in 2005 to Ar 20,000 Ariary in 2010, by about 36 percent in five years. The difference by type of school is 19 | slight however, considering that the cost of education has increased by 39 percent in public schools and by 41 percent in private schools (See Figure 3.3). Furthermore, per pupil spending has increased slightly more for girls (+58 percent) than for boys (+35 percent) over the period. The difference is however most marked by area of residence. Indeed, household per pupil spending has increased at double the rate in rural areas (+45 percent) than in urban ones (+20 percent), even if the latter remains the higher of the two (Ar 35,000 on average in 2010, against Ar 17,000 in rural areas). It is interesting to note that it is in rural areas that both enrollment has receded and direct costs have increased, the most. The difference is most apparent for private schools, where costs increased by 64 percent between 2005 and 2010, against only 22 percent for urban private schools. Figure 3.3: Household Education per Pupil Spending, 2005 and 2010 (EPM Data) Thousands of Ariary (Constant 2005 Prices) 60 70% 50 60% Thousands of Ariary 58% 40 50% 2005 Growth (%) 45% 39% 41% 40% 30 36% 35% 30% 2010 20 20% 20% 10 2005-10 10% Growth (%) 0 0% Total Public Private Girls Boys Urban Rural Source: Household surveys, 2005 and 2010 (See also Annex Table A15). The distribution of spending shows a general increase in food costs and a marked rise in public school fees. Food costs have increased considerably in public schools, by about 46 percent, against 42 percent in private schools (See Table 3.2 below). Public school fees have also increased significantly, by 25 percent. In the private sector on the other hand, school fees and PTA contributions have dropped by five and two percent respectively. Table 3.2: Average Household Education Spending, by Type of School and Type of Expense, 2005 and 2010 (EPM Data) Thousands of Ariary (Constant 2005 Prices) and Percent Public Private 2005 2010 Variation 2005 2010 Variation Enrollment Fees 3,197 4,274 25.2% 10,934 10,346 -5.7% PTA Contributions 3,622 4,387 17.4% 4,754 4,686 -1.5% Insurance 534 383 -39.5% 1,917 1,840 -4.2% Monthly School Fees 9,315 7,347 -26.8% 30,245 30,557 1.0% Uniforms 2,900 2,939 1.3% 3,754 4,447 15.6% Sports Wear 2,960 3,394 12.8% 6,338 3,913 -62.0% Supplies 4,639 5,032 7.8% 9,465 11,178 15.3% Transportation 13,383 21,901 38.9% 53,434 49,782 -7.3% School Meals 14,175 26,356 46.2% 21,033 36,074 41.7% Other Expenses 4,038 3,026 -33.4% 9,727 9,898 1.7% Source: Household surveys, 2005 and 2010 (See also Annex Table A16). Note: The figures presented are the averages of the amounts spent by households having incurred each type of expense. Enrollment fees are paid to schools, whereas PTA contributions are paid to the FRAMs, occasionally covering teacher salaries when these are not civil servants or government-subsidized. 20 | Since the crisis, a greater number of households with a child in Grade 2 have had to pay enrollment fees, PTA contributions and monthly school fees (See Table 3.3). The share of households paying enrollment fees has increased the most, by seven percentage points, against four percentage points for the share paying PTA contributions and one percentage point for those paying monthly school fees. Furthermore, according to FRAM officers, delays in the payment of parental contributions are more frequent in 2012 than in 2009 (by 14 percentage points) whereas fewer households are exempt. Table 3.3: Share of Households in the South Having Paid Enrollment Fees, PTA Contributions and Monthly School Fees, 2009 and 2012 (Sample Selection) Percent and Percentage Points 2009-12 Gap 2009 2012 (% Points) Households having paid enrollment fees 34.6 41.5 6.9 Households having paid PTA contributions 36.5 40.9 4.4 Share of parents having paid late 71.4 85.2 13.8 Parents exempt from PTA contributions 20.0 14.5 -5.5 Households having paid monthly school fees 12.5 13.8 1.3 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. In addition to be more frequent, the amount paid in school fees has generally increased. Average household per pupil spending increased from Ar 6,561 to Ar 8,277 for the tracer study children, representing a 26 percent increase over three years (See Annex Table A17).12 Whereas the shares devoted to enrollment fees and school supplies have dropped over the period, PTA contributions and monthly school fees have increased, the latter significantly. Figure 3.4: Average Household Education Spending per Pupil in the South, 2009 and 2012 Ariary (Constant 2009 Prices)) 20,000 2009 60.7% 80 Ariary (2009 Prices) 16,000 2012 60 2009-12 Growth (%) 40 Growth (%) 12,000 6.2% -0.5% 20 8,000 -17.3% 0 4,000 -20 0 -40 Enrollment Fees PTA Contributions Annualized… School… Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012 (See also Annex Table A17). Note: Average household spending per pupil for whom the given expense was incurred. Enrollment fees are paid once at the beginning of the school year, whereas school fees are paid monthly, and have been computed for a year here. 12 Annex B explains the difference between the two groups: Sample Selection and Tracer Study. 21 | Impact of Channel 1 on Education: Direct and Indirect Education Costs Rising education costs have translated into a significant increase in the share of household spending devoted to education, according to the household surveys. The share of the family budget devoted to education has risen from 2.1 percent in 2005 to 3.1 percent in 2010 on average, close to a 50 percent increase (See Figure 3.5). By wealth quintile, it appears that the middle classes are those for which the share of the household budget devoted to education has increased the most (Quintiles 2 and 3). Figure 3.5: Share of Education in Household Spending, by Wealth Quintile, 2005 and 2010 (EPM Data) Percent and Percentage Points 7.0 1.4 Share of the budget (%) 6.0 1.2pp 1.2 1.1pp 5.0 1.0pp 1.0pp 1.0 Gap (pp) 4.0 0.8 0.7pp 3.0 3.1 0.6 2.0 2.1 0.4 1.0 0.2pp 0.2 0.0 0.0 Q1 Q2 Q3 Q4 Q5 All 2005 2010 2005-10 Gap (% Points) Source: Household surveys, 2005 and 2010. Since the crisis, direct and indirect costs and financial problems are reasons often mentioned by parents and headmasters to explain children being out of school. In the reasons explaining the dropout of children in the South survey tracer study, direct and indirect costs are often mentioned, by 15 percent of households on average over the 2010-12 period (See Annex Table A18). Likewise, in 2012 school-related costs represented the main cause of dropout according to 22 percent of headmasters, and the second most important reason according to a further 9 percent (See Annex Table A19).13 Household survey data further indicate that financial problems are mentioned more often in 2010 than in 2005 (by 19.3 percent to explain dropout and by 23.8 percent to explain non- enrollment, see Annex Table A25 and A26). This could reflect the effect of Channel 1 (Direct and Indirect Education Costs) and/or that of Channel 2 (Access to Loans and Household Income) on household education decisions since the two are closely linked. Channel 2: Access to Loans and Household Income Trends in Household Income GDP per capita remains five percent below its pre-crisis level and poverty rates have increased significantly between 2005 and 2010. GDP per capita remains low in Madagascar, having barely reached US$ 480 at its height (in 2008). On the other hand, the growth trend noted since 2004 was interrupted in 2009 and has still not been recovered, despite a slight improvement in 2011 (See Figure 3.6 below). Overall, GDP per capita has decreased by 4.6 percent in four years, reflecting a 13 Indeed, only one reason for dropout is mentioned more often, both by parents and headmasters: child work. This is dealt with later in the analysis of Channel 4 (Opportunity Cost of Education). 22 | substantial drop in revenues for most households. In addition, the share of the Malagasy population considered to be living under the poverty line increased from 67.7 percent in 2005 to 76.5 percent in 2010, by nine percentage points on average (See Figure 3.7). Recent estimates suggest that this percentage may have reached 92 percent, indicating that only very few Malagasy leave with more than $2 a day. The analysis of the evolution of poverty between regions nevertheless shows that if the average poverty level has increased, gaps among regions have been reduced (See Annex Table A20). Figure 3.6: GDP per Capita, 2002-12 US$ (Constant 2011 Prices) 600 480.6 465.0 421.8 421.0 458.4 500 386.9 325.0 US$ (2011 Prices) 400 269.1 281.7 299.3 251.4 300 200 100 0 Source: Economic and financial reports of the Ministry of Finance and Budget and PIE, 2013. Note: Data for 2009 is provisional; data for 2010 and 2011 are estimated on the basis of the population growth rate as per the EPM, 2005, of 3%. Figure 3.7: Poverty Rate, 2005 and 2010 (EPM Data) Percent 120 2005 2010 94.5 Below the Poverty Line 100 83.7 % of the Population 76.5 80 67.7 54.4 60 Living 41.9 40 20 0 National Poorest Wealthiest Average Region Region Source: Household surveys, 2005 and 2010. Note: The rates for the two years have been adjusted for comparability (See Annex C). The rise in the price of staple goods continues at an accelerated rhythm, affecting household living conditions further. Since 2006, the general consumer price index has increased steadily, by about 10 percent per year (See Figure 3.8). At a disaggregated level, it should however be noted that the price of staple goods has undergone an acceleration of inflation since 2010. The respective index has increased by 15 percent over 2010-11, and the cost of a food as essential as rice has increased by 21 percent over the period. The rise in the rate of inflation for these basic items means fewer household resources are available for other types of spending, such as education. 23 | Figure 3.8: Consumer Price Index Trends, 2003-12 Price Indexes 350 Consumer Price Index Staple Goods Index Rice 300 Price Indexes 250 200 150 100 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source: INSTAT (See Annex Table A21 for further disaggregated indexes). The South surveys carried out in 2009 and 2012 show that the income of poorest households has been disproportionately affected by the crisis. Indeed, whereas the share of better-off households (those owning 10 cattle or more) has remained grossly stable, the share of worse-of households (those with no cattle) has increased from 37 percent to 50 percent, and the share of households with few resources (those owning less than 10 heads of cattle) has dropped from 31 percent to 15 percent. The evolution of the distribution of households according to the area of land owned or cultivated follows a similar pattern. Whereas the share of better-off households (those owning over a hectare) has remained stable, in 2012, 13 percent more households had less than half a hectare than in 2009 (See Figures 3.9 and 3.10 below). Figure 3.9: Distribution of Households in the South, by Heads of Cattle Owned, 2009 and 2012 (Tracer Study) Percent None 2009 37.3 31.1 22.5 6.4 Less than 10 10 to 30 Heads 31 to 60 Heads 2012 49.7 15.2 21.0 6.4 7.7 Over 61 Heads 0% 20% 40% 60% 80% 100% Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. Figure 3.10: Distribution of Households in the South, by Area of Land Owned or Cultivated, 2009 and 2012 (Tracer Study) Percent Under 0.5 Ha 2009 31.9 38.9 17.2 11.9 0.6 to 1.0 Ha 1.1 to 2.0 Ha 2012 45.4 24.8 17.3 12.5 2.1 Ha and more 0% 20% 40% 60% 80% 100% Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. 24 | Trends in Access to Loans Loan access is relatively weak according to the household surveys, although increasing. Without a loan or any kind of formal insurance, various approaches can be adopted by households to face the ex-post risks of negative shocks on their income and protect their standard of living. These strategies include public and private transfers, savings (in funds or in kind) and the sale of cattle. The household surveys only consider private transfers however. On this basis, in Madagascar as in most developing countries, access to formal credit markets is still limited. In 2010, barely seven percent of households had received a repayable transfer, a minority (See Table 3.4). Nevertheless, the data also show that households had received repayable transfers six times more often in 2010 than in 2005. Table 3.4: Share of Households (with a Child Aged 6-10 Years) Having Received a Transfer, 2005 and 2010 (EPM Data) Percent and Percentage Points Gap 2005-10 2005 2010 (% Points) Received a Transfer (Repayable or not) Urban 25.7 37.1 11.4 Rural 23.1 33.6 10.5 Subtotal 23.6 34.2 10.6 Received a Repayable Transfer Urban 1.0 8.1 7.1 Rural 1.3 6.2 4.9 Subtotal 1.3 6.6 5.3 Source: Household surveys, 2005 and 2010. Between 2009 and 2012 in the South districts, the opposite trend is apparent, as the share of households declaring to have received a loan is waning. Whereas 37 percent of households in the districts of Amboasary and Betioky received a loan in 2009, only 26 percent did in 2011, about 30 percent less. The share of households declaring to have suffered an agricultural shock dropped in the same proportion however, which raises the question of the reason for the more limited access to loans: is it due to lesser household needs, or greater difficulty in obtaining a loan? The data do not provide answers. The 2012 tracer study has on the other hand enabled the study of the share of families having received a transfer. This was lower than that having received a loan, but nevertheless significant, at about 10 percent (See Figure 3.11). Figure 3.11: Share of Households in the South Having Received a Loan, Suffered an Agricultural Shock or Received a Transfer, 2009 and 2011/12 (Sample Selection) Percent 80 0 60 77.9 -5 2009 Gap (pp) Percent -10.9 pp -10 40 54.4 -15 2011/12* 20 36.6 -20 25.7 n.a. 9.6 0 -23.5 pp -25 2009-12 Gap Households Households Households (% Points) who received that suffered an who received a loan agricultural shock a transfer Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. * Households indicated at the time of the survey (in 2012) if they had borrowed over the previous four months (early 2012) and if they had received a transfer or suffered an agricultural shock the previous year (in 2011). 25 | Impact of Channel 2 on Education: Household Income, including loans and transfers On a national scale, disparities in terms of enrollment and dropout according to household wealth have widened over 2005-10. The enrollment levels for children aged 6 to 10 years have considerably dropped among the poorest households (from 70 percent to 59 percent for Q1) whereas they have increased among the wealthiest ones, although marginally (from 89 percent to 91 percent for Q5 - See Figure 3.12). Disparities in the dropout rates of children according to household income have also increased. Dropout has increased for children from the poorest households (from five percent to eight percent on average for Q1 and Q2), whereas it has remained grossly stable for children from middle-income families and has improved for the wealthiest (dropping by almost half). Finally, repetition levels have improved for all households, independently of their income level. This may however only reflect the improvement of the repetition rate in private schools. Figure 3.12: Enrollment, Dropout and Repetition Rates for Children Aged 6-10 Years, by Wealth Quintile, 2005 and 2010 (EPM Data) Percent 100 81.9 86.2 91 76.8 80 69.5 84.8 89.1 Percent 60 74.1 75.7 2005 Enrollment 40 58.8 2010 Enrollment 20 0 Q1 Q2 Q3 Q4 Q5 Source: Household surveys, 2005 and 2010 (See also Annex Table A22). Between 2009 and 2012 in the South districts, dropout is more frequent in poorer households, those having borrowed, suffered an agricultural shock or who did not receive a transfer from abroad. On average in Amboasary and Betioky, 28.4 percent of children enrolled in 2009 had dropped out by 2012. Living standards are the socioeconomic factor among those studied that has the greatest impact on household education behavior. 14 Over 33 percent of children from the more modest households (Quintiles 1, 2 and 3) had dropped out, against 16 percent of those from the wealthiest (Quintile 5 – See Figure 3.13 below). Dropout is also marginally more important among households having a loan (29.8 percent) or having suffered an agricultural shock (29.3 percent). Conversely, it is significantly weaker among households having received a transfer (25 percent), by about four percentage points. 14 In the South survey, the need for labor and the cost of education are the two reasons most frequently mentioned by households for their children being out of school, in 19 percent and 16 percent of cases in 2012, respectively (See Annex Table A18). 26 | Figure 3.13: Dropout Rates in the South, by Habitat and Durable Goods Index Quintiles, 2011/12 (Tracer Study) Percent 40.0 32.1 37.7 32.0 29.2 Habitat 36.0 Index Percent 29.3 28.9 16.7 20.0 Durable 20.0 Goods 14.9 Index 0.0 Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012 (See also Annex Table A26). Note: The Habitat Index is for 2012 and the Durable Goods Index for 2011. Overall, the analysis points out to financial factors being an important determinant of school dropout and non-enrollment. The variations by wealth quintile reinforce the idea that there is a financial explanation to household education choices subsequent to the crisis. Indeed, beyond the general impact of the crisis on enrollment through household income and access to loans, Channel 2 appears to deepen schooling disparities between the wealthy and the poor. In addition, financial problems are the first reason mentioned by families both for dropout and non-enrollment. In 2010, financial problems were the reason most often mentioned by parents to explain the dropout of their child (in 25.6 percent of cases) as well as for their non-enrollment (in 26.2 percent of cases - See Figure 3.14). Household data further indicate that this explanation is more frequently given in 2010 than in 2005 (mentioned by 19.3 percent to explain dropout and by 23.8 percent to explain non- enrollment). Figure 3.14: Main Causes of Dropout and Non-Enrollment (Children Aged 6-17 Years), 2005 and 2010 (EPM Data) Percent REASONS FOR DROPOUT Financial problems Child no longer wants to go to school Child's work Child refuses to repeat Child too old REASONS FOR NON-ENROLLMENT Financial problems Child too young Child does not want to go to school 2010 School too far away 2005 No school available 0.0 5.0 10.0 15.0 20.0 25.0 30.0 Source: Household surveys, 2005 and 2010 (See also Annex Tables A25 and A26). 27 | Channel 3: Return on Investment in Education Quality of Inputs and Learning Outcomes15 According to the education statistics, the quantity of available human and material resources is increasing overall. The number of classrooms and teachers has continued to rise despite the crisis, while the number of operative schools has remained constant (See Table 3.5). The stagnation in enrollment has entailed that pupil-class and pupil-teacher ratios have improved, from averages of 50.9 and 47.9 respectively in 2008/09, to 39.6 and 42.7 in 2010/11. In fact, an improvement in these ratios has been underway since 2004/05; not only has it continued its course despite the crisis, but it has even accelerated. Table 3.5: Material and Human Resources Available to Primary, 2001/02-2010/11 (ESY Data) Number and Percent Share of Pupils per Pupils per non-temporary class teacher classrooms 2001/02 46.1 47.5 88.7% 2002/03 48.6 51.6 89.9% 2003/04 54.7 52.4 89.8% 2004/05 56.3 53.6 89.2% 2005/06 52.3 48.1 89.5% 2006/07 52.5 48.7 88.1% 2007/08 50.9 47.2 86.6% 2008/09 50.9 47.9 85.4% 2009/10 44.6 45.5 86.5% 2010/11 39.6 42.7 87.4% Source: Education statistical yearbooks, 2000-11. The South surveys carried out in 2009 and 2012 confirm this downward trend in PTRs, without it however impacting on instructional time. In the regions surveyed, the number of pupils per class dropped from 30 to 22 pupils, the pupil-teacher ratio having dropped from 61 to 57 pupils (See Figure 3.15 below). On the other hand however, the number of hours of teaching per day and the availability of textbooks have worsened between 2009 and 2012. Figure 3.15: Education Quality Inputs in the South, 2009 and 2012 (Sample Selection) 100 2009 2012 79.5 79.5 75.3 80 71.9 68.5 71.9 61.4 56.7 60 40 30.1 21.9 20 5.3 5.0 0 Pupils Pupils Instructional Schools with Schools with Schools with per Class per Teacher Time G 2 math G 2 French G 2 Malgasy (Number) (Number) (Hours/day) Textbooks (%) Textbooks (%) Textbooks (%) Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. 15 It is important to note that these variables provide a comparatively unreliable measure of the quality of education. Means contribute to quality but do not guarantee it. 28 | The share of civil servant teachers has dropped considerably over the decade, with likely impact in terms of further costs borne by families. This tendency was already marked before the crisis and has continued, the number of civil servant teachers representing no more than 33 percent of the total in 2010/11 (See Figure 3.16). The share of subsidized community teachers has also dropped slightly between 2008/09 and 2010/11. In the South districts, the drop in the number of civil servant teachers has translated into an important rise in the share of FRAM teachers, from 54 percent in 2009 to 70 percent in 2012 (See Table 3.6 below). Unlike the national trend however, the share of them that is subsidized has increased slightly, from 70 percent to 74 percent. The increased number of teachers who are not paid by the Government (either because they are not civil servants, or because they are unsubsidized FRAM) has most likely caused higher costs being borne by families. Figure 3.16: Public primary Teachers, by status 2006/07-2010/11 (ESY Data) Percent 100% Unsubsidized FRAM teachers/Others 50% Subsidized FRAM teachers Civil servant teachers 0% 2006/07 2007/08 2008/09 2009/10 2010/11 Source: Education statistical yearbooks, 2000-11. The rapid decrease in the number of civil servants may also have had an impact of education quality. The South surveys highlight a deterioration in some of the variables that are indicative of education quality, such as the share of trained teachers (from over 27 percent to barely 16 percent), the share of them that offer pupils support classes (dropping five percentage points, to 13 percent) and the share of the teaching staff that speak French (dropping six percentage points, to 23 percent). Table 3.6: Teacher Characteristics in the South, 2009 and 2012 (Sample Selection) Percent 2009-12 Gap 2009 2012 (% Points) FRAM teachers 53.8 69.9 16.1 Subsidized FRAM teachers 70.3 73.7 3.4 Teachers with a pedagogical qualification 27.5 16.0 -11.5 FRAM teachers with a pedagogical qualification 6.6 4.3 -2.3 Teachers that offer pupils support classes 17.7 12.8 -4.9 Teachers that speak French 28.8 22.7 -6.1 Teachers that have a secondary occupation 65.7 53.6 -12.1 FRAM teachers that have a secondary occupation 72.7 57.3 -15.4 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. The quality of learning outcomes has receded since the late 1990s and pupils’ levels of learning at the end of primary are weak. Following a continuous improvement over the 2001-09 period to reach 78.5 percent, the success rate at the end of primary examination (the certificat d’études primaires élémentaires - CEPE) has slightly dropped since the beginning of the crisis, to 74.4 percent for 2010/11 (national average). This drop in primary learning outcomes occurs in a context where the number of candidates has also dropped, from 470,000 before the crisis to 435,000 initially, before increasing slightly in 2010/11, to 461,000 candidates (ESY Data, 2010/11). 29 | Figure 3.17: Evolution of PASEC Results, 1998, 2005 and 2012 Average Scores 80 59.1 60 51.3 50.0 42.6 40.0 43.5 PASEC 1998 Percent 40 31.4 26.8 PASEC 2005 20 n.a. PASEC 2012 0 French Math Malagasy Source: PASEC Reports, 1998 and 2005 and MEN, 2012. Note: Each score is the national average of the results of all candidates, adjusted to a maximum of 100%. This deterioration in learning outcomes is confirmed by the results of Grade 5 pupils in the PASEC standardized international assessment. 16 The results of a survey of learning outcomes carried out in 2012 show that between 1998 and 2012 results in all three subjects (French, math and Malagasy) have not ceased to decrease (See Figure 3.17 above). The national average in French has dropped by 16 percentage points over 1998-2012 and the math score by 19 percentage points. Learning outcomes in the South districts, although in line with the general trend, have nevertheless deteriorated less than the national averages over 2009-12. In French, Grade 3 and 5 results have receded by four and two percentage points respectively, whereas the variation for other grades is slighter (See Table 3.7). In math, the gaps in 2009 and 2012 scores are relatively weak. Grade 4 pupils are those whose average has dropped the most, by almost one percentage point. Table 3.7: Results in Math and French in the South, by Grade, 2009 and 2012 (Sample Selection) Math French (Average No. of right answers) (Number of letters read per minute) 2009-12 Gap 2009-12 Gap 2009 2012 2009 2012 (% Points) (% Points) Grade 1 46.4 46.2 -0.2 5 6 1 Grade 2 64.9 64.5 -0.4 25 24 -1 Grade 3 74.2 73.7 -0.5 48 44 -4 Grade 4 79.0 78.5 -0.5 60 60 0 Grade 5 83.8 82.9 -0.9 70 68 -2 (Number of schools) 149 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. The analysis of quality as a supply-side factor leads to the following findings: (i) according to the education statistics, supervision rates have improved continuously since 2001 and this tendency appears to have even accentuated since the crisis (pupil-teacher and pupil-class ratios); (ii) the South survey in the districts of Amboasary and Betioky shows that instructional time has decreased slightly and that the availability of textbooks (math, French and Malagasy) has deteriorated; (iii) the quality of teaching may have declined, as there are fewer civil servant teachers, the share of trained teachers has dropped and the share of teachers offering support classes is reduced. Furthermore, all learning outcomes (CEPE, PASEC and French and math in the South districts) have deteriorated, although continuously over 10 years. 16 The PASEC (Programme d’analyse des systèmes éducatifs de la CONFEMEN) is a standardized learning assessment that enables the comparison of the performance of participants of 15 countries, over successive years. 30 | Impact of Channel 3 on Education: Return on Investment in Education Out-of-school is rarely explained by reasons related to the return on investment in education, most of which are mentioned less often by households in 2010 than in 2005, with the exception of teachers’ incompetency. Households rarely explain dropout and non-enrollment according to the perceived value of education: the related factors are mentioned by under 3 percent of households, against more than 20 percent who mention financial problems (See earlier Figure 3.14). Furthermore, trends over the 2005-10 period show that most factors related to the perceived value of education are mentioned less often (See Figure 3.18). The difficulty of studies, the relevance of content and the unproductive nature of school are slightly less mentioned in 2010 than in 2005 (up to one percentage point less). Dropout due to the incompetence of teachers is, on the other hand, increasing. Figure 3.18: Channel 3 Related Reasons Indicated by Households for Out-of-School (Children Aged 6-17 Years), 2005 and 2010 (EPM Data) Percent REASONS FOR DROPOUT Study too difficult 1.3 2.3 Not worthwhile 0.9 1.4 Incompetent teacher 1.3 2.0 2010 Inadequate content 0.3 0.8 2005 REASONS FOR NON ENROLLMENT Study too difficult 0.8 1.9 Not worthwhile 1.2 1.5 Percent Inadequate content 0.0 0.7 0.0 0.5 1.0 1.5 2.0 2.5 Source: Household surveys, 2005 and 2010 (See also Annex Tables A25 and A26). In the South districts also, reasons related to education quality and the expected return are rarely mentioned. In the South districts, the reasons related to the expected return of education are mentioned more often to explain children being out of school (by up to six percent of households) than the national average (under three percent), although less often than economic reasons (mentioned by 15 to 20 percent of households). The reasons related to Channel 3 most often mentioned are the child’s difficulty at school,) motivations and teacher absenteeism. Figure 3.19: Reasons Related to the Value of Education Mentioned by Households in the South for Out-of-School, 2010-12 Average (Tracer Study) Percent Child in difficulty at school 6.1 Classes are boring 5.3 Teacher absenteeism 4.3 Child is not keen 3.1 Child too old 2.3 Level attained deemed sufficient 1.6 Quality and quantity of infrastructure 0.9 Knowledge acquired deemed useless 0.5 0 1 2 3 4 5 6 7 Source: South Survey in the Districts of Amboasary and Betioky, 2012 (See also Annex Table A18). Note: Frequency of answers provided by households with a child having dropped out between 2009 and 2012. 31 | Channel 4: Opportunity Cost of Education Estimated changes in opportunity cost of education Despite difficulties in measurement, evidence points to a decrease in the wage differential between individuals with and without primary education. There are considerable difficulties associated with the measurement of opportunity costs of education in contexts such as Madagascar. Indeed, opportunity costs, the equivalent of the earnings foregone by pupils when they attend school, are influenced by the specificities of the labor market and economy in the regions considered. Proxies used to measure opportunity costs include the wage differential between individuals with various level of education. In Madagascar, the average salary for uneducated adults has dropped from Ar 48,411 to Ar 29,243 between 2005 and 2010 (a 39.6 percent reduction) whereas that of adults having at least completed primary has dropped from Ar 107,137 to Ar 76,691 (a reduction of 28.4 percent). Overall the salary gap between uneducated and educated workers has been reduced from Ar 58,726 to Ar 47,448 which potentially indicates a reduction of the opportunity cost for education as a result of the crisis. Impact of Channel 4 on Education: Opportunity Cost of Education Reasons related to children’s work are often mentioned by households to explain out -of-school, although less so in 2010 than in 2005 according to the household surveys. Households explain the dropout or non-enrollment of their children less often according to work in 2010, when only 13 percent of dropouts and 6 percent of non-enrollments were motivated by children’s participation in family economic activities, against 21 percent and 12 percent respectively in 2005 (See Figure 3.20 below). These motives are nevertheless important, and are largely more significant than the child’s responsibility for a sibling. These findings reinforce the hypothesis that the opportunity cost of education has declined in the aftermath of the crisis. Figure 3.20: Work-Related Reasons for Out-of-School (Children Aged 6-17 Years) Provided by Households, 2005 and 2010 (EPM Data) Percent 25 21.1 20 2005 13.2 12.4 2010 Percent 15 10 6.0 5 1.9 1.4 2.1 1.5 0 Child's Sibling Child's Sibling Work to care for Work to care for Reasons for Dropout Reasons for Non-Enrollment Source: Household surveys, 2005 and 2010 (See also Annex Tables A25 and A26). In the South districts, dropout is more often explained by children’s work and the need for family labor. In 2010, 16.1 percent of children having dropped out of school would have been influenced by opportunity costs, against 13.2 percent nationwide. The frequency of mention of this reason (child’s work and need for family labor) for dropout increased to 26.7 percent in 2011, before readjusting to 19.1 percent in 2012. 32 | Econometric Analysis of the Impact of the Crisis The results of the econometric analysis suggest a strong and increasing impact on enrollment of Channel 1 (Direct and Indirect Education Costs)) and the Income component of Channel 2 (Household Income including loans and transfers). The previous analyses are based primarily on descriptive statistics. To test their robustness (the correlation between the variation in education indicators with the observed variables), the effects of the variables identified in the course of Chapters 2 and 3 on out-of-school are analyzed here through econometric regressions (See Annex D). More specifically: Channel 1: Direct and Indirect Education Costs: On the basis of the analysis carried out in the context of the tracer study of the South survey children, the existence of PTA contributions in 2009 tends to increase the probability of dropout over 2009-12 (See Table 4.4).17 Similarly, the analysis indicates that the probability of dropout increases among children whose schools charge PTA contributions in 2012 when they did not in 2009. Channel 2: Household Income: The results of the estimations on the basis of household survey data indicate a positive and significant effect of wealth quintiles on enrollment (See Table 4.1). Households’ belonging to the wealthier categories is correlated with a greater reduction in out-of- school rates. This effect is noticeable in both rural and urban areas (See Table D1 in Annex D). Furthermore, the values of the coefficients, higher in 2010 than in 2005, indicate that the effect is increasing: belonging to a Q1 (Very poor) household increased the probability of being out of school by more in 2010 than it did in 2005. Data from the tracer study confirms this trend: South district children whose households are better-off in terms of durable goods and housing also faced a greater probability of enrollment in 2012, whereas this variable was not statistically significant for 2009. Conversely, the greater the number of heads of cattle owned in 2012, the higher the probability of dropout (See Table 4.2 below). Channel 3: Return on Investment in Education: In 2009, the probability of being out of school is lower for children in the South districts whose learning environment is less conducive (more community teachers). These education supply-related variables do not appear to be significantly correlated to out-of-school rates in 2012 however, and their decline over 2009-12 has no significant marginal impact on the probability of dropout (See Table 4.2). The only variable available in household survey data on education supply is the pupil-teacher ratio (PTR). Although in 2005 a high PTR is positively correlated to being out of school, this variable is not significant for 2010 (See Table 4.1). 18 Channel 4: Opportunity Cost of Education: The increase in this wage differential between the pay of an uneducated adult and that of an adult having at least completed primary had a negative and significant effect on being out of school in 2005, implying that the perspective of greater future income encouraged the enrollment of children (See Table 4.1). This effect is not apparent for 2010 however. The hypothesis that can be formulated as a result is that in 2010 fewer families were able to consider the return on investment in education over the long-term, having to manage their spending on a day to day basis following the crisis. 17 The analysis is based on a limited sample of children who were at school in 2009 but had dropped out by 2012. 18 It should however be noted that the household survey data orient the analysis towards education demand-side factors. On this basis, it is difficult to fully test the significance of the effect of Channel 3 (Return on Investment in Education). 33 | Table 4.1: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years, 2005 and 2010 (EPM Data) 2005-10 2005 2010 Variation Channel 2: Access to Loans and Household Income Wealth Quintile (Ref. Q1 - Very poor) Q2 - Poor -0.306 * -0.472 *** Q3 - Average -0.303 * -0.610 *** Q4 - Wealthy -0.520 *** -1.011 *** Q5 – Very wealthy -0.705 *** -1.314 *** Suffered climatic shock -0.007 0.366 *** Received non-repayable transfer -0.172 -0.219 *** Received repayable transfer 0.515 0.045 Channel 3: Return on Investment in Education Education Level of Household Head (Ref. None) Primary -0.223 * -0.393 *** Lower secondary -0.908 *** -1.025 *** Upper secondary and above -1.669 *** -1.201 *** Son/daughter of household head -0.148 -0.340 *** Primary PTR 0.007 ** -0.001 Channel 4: Opportunity Cost of Education Adult salary ratio -0.066 *** 0.001 Other Variables Age 0.094 *** 0.007 Boy -0.126 0.130 ** Household head is female -0.138 -0.130 * Rank among siblings -0.071 * -0.080 * Dependency ratio 0.100 * 0.054 * Variation Effect of the observed characteristics 0.020 ** Unobserved effect 0.068 *** Interaction between observed/unobserved 0.002 Total 0.090 *** Constant -2.078 *** -0.290 * Observations 3,800 13,037 Pseudo R2 0.110 0.115 Source: Authors’ calculations on the basis of household survey data, 2005 and 2010 (See also Table D1 of Annex D). Note: Logit regressions are based on the binomial: Does not go to school = 1, if not = 0. The adult salary ratio is the relation between the salary of an adult with primary education and that of an uneducated peer. The dependency ratio represents the relation within households of members under 15 or over 65 to adults aged 15 to 64 years. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. It has to be noted that the unobserved characteristics explain the greatest part of the increase in out-of-school rates. At the national level, the coefficient noted by the model for the variation in out-of-school rates between 2005 and 2010 is estimated at 0.090. The share of this coefficient explained by the observed variables is only 0.020 however. This indicates that the observed characteristics explain barely 22 percent of the increase in out-of-school rates over the period. Similarly, in Amboasary and Betioky for 2009-12, the observed variables only explain 13 percent of the rise in out-of-school. The greatest part (over 75 percent) of the change is due to variables not studied here, meaning that other factors (new, or their evolution over the period) strongly influence parents’ enrollment decisions. These factors could be linked to difficulty to precisely measure changes in opportunity costs. 34 | Table 4.2: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years in the South, 2009 and 2012 (Sample Selection) 2009-12 2009 2012 Variation Channel 1: Direct and Indirect Education Costs PTA contributions 0.000 * -0.000 Channel 2: Access to Loans and Household Income Synthetic wealth indicator -0.120 -0.480 *** Suffered climatic shock -0.928 ** 0.116 Household borrowed over last 4 months 0.078 0.017 Cattle ownership -0.089 0.237 * Channel 3: Return on Investment in Education Education Level of Household Head (Ref. None) Primary -0.626 -0.071 ** Lower secondary -1.491 -0.695 ** Upper secondary and above -0.166 -1.227 Number of classrooms in good condition -0.056 -0.010 Share of FRAM teachers -0.011 ** 0.001 Pupil-class ratio -0.033 ** -0.006 French textbook available 1.054 * 0.010 Other Variables Age 0.564 *** 0.009 Boy 0.509 0.718 *** Rank among siblings -0.008 0.027 Household head is female -1.797 * -0.321 Variation Effect of the observed characteristics 0.023 Unobserved effect 0.178 *** Interaction between observed/unobserved -0.018 Total 0.183 *** Constant -8.167 *** -1.814 *** Observations 1,679 2,113 Pseudo R2 0.253 0.079 Source: Authors’ calculations on the basis of South survey data, 2009 and 2012 (See also Table D2 of Annex D). Note: The household synthetic wealth indicator is computed through the principal component analysis of household scores in terms of the housing and durable goods indexes. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 35 | Table 4.3: Multinomial Logit Regressions on the Activities of Children Aged 6-14 Years, 2010 (EPM Data) Work, chores Work Inactive and school and chores Channel 2: Access to Loans and Household Income Wealth Quintile (Ref. Q1 - Very poor) Q2 - Poor -0.006 -0.454 *** -0.471 *** Q3 - Average 0.216 *** -0.448 *** -0.412 *** Q4 - Wealthy 0.271 *** -0.760 *** -0.956 *** Q4 - Very wealthy 0.201 ** -1.021 *** -1.757 *** Suffered a climatic shock 0.085 0.397 *** 0.370 *** Received non-repayable transfer 0.065 -0.139 ** -0.255 *** Received repayable transfer 0.685 *** 0.654 *** 0.249 Channel 3: Return on Investment in Education Education Level of Household Head (Ref. None) Primary -0.250 *** -0.633 *** -0.395 *** Lower secondary -0.199 *** -1.217 *** -1.050 *** Upper secondary and above -0.626 *** -1.834 *** -1.083 *** Son/daughter of household head -0.078 -0.639 *** 0.309 * Primary PTR 0.002 0.001 -0.001 Channel 4: Opportunity Cost of Education Adult salary ratio 0.005 * 0.006 ** 0.005 Other Variables Age 0.216 *** 0.273 *** -0.200 *** Boy -0.185 *** -0.025 0.074 Rank among siblings 0.075 ** 0.050 0.222 *** Household head is male -0.217 *** -0.021 -0.061 Dependency ratio 0.135 *** 0.167 *** 0.086 ** Constant Observations 13,037 Pseudo R2 0.133 Source: EPM, 2010 (See also Tables D3 and D4 for 2005, in Annex D). Note: The reference for the activity status of children is school only. Fixed region effects and a constant were included in all regressions. Standard deviations are corrected due to the auto-correlation issue between children of a same household. The adult salary ratio is the relation between the salary of an adult with primary education and that of an uneducated peer. The dependency ratio represents the relation within households of members under 15 or over 65 to adults aged 15 to 64 years. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 36 | Table 4.4: Binomial Logit Regressions on Dropout among Children Aged 6-14 Years in the South, 2009 and 2012 (Tracer Study) 2009 Gap Variables 2009-12 Channel 1: Direct and Indirect Education Costs PTA contributions 0.885 * 0.028 Channel 2: Access to Loans and Household Income Synthetic wealth indicator -0.074 -0.030 Suffered climatic shock over 2009-11 0.200 Household borrowed over last 4 months 0.959 *** Cattle ownership (Ref. None) 1 to 5 heads 0.828 * 6 to 10 heads 0.010 11 to 20 heads 0.968 Over 21 heads -0.676 Gap in number of heads over 2009-11 0.000 Channel 3: Return on Investment in Education Education Level of Household Head (Ref. None) Primary 0.385 Lower secondary and plus -1.038 Upper secondary and above -1.838 Son/daughter of household head -0.209 Number of classrooms in good condition -0.163 Share of FRAM teachers 0.008 0.014 * Pupil-class ratio -0.013 -0.015 French textbook available 0.084 -0.688 * Number of letters read per minute -0.011 Channel 4: Opportunity Cost of Education Domestic, cattle-breeding or agricultural activities -0.410 Other Variables Age 0.256 *** Boy -0.396 Household head is female 0.661 Rank among siblings -0.293 Dependency ratio -0.000 Betioky (Ref. Amboasary) 0.615 * Cattle theft in the fokontany (2012) 0.994 * Burglary in the fokontany (2012) -0.024 Other insecurity issues in the fokontany (2012) 0.303 Constant -6.687 *** Observations 443 Pseudo R2 0.171 Source: Authors’ calculations on the basis of South survey data, 2009 and 2012 (See also Table D5 in Annex D with the limited sampl e of children who were enrolled in 2009 but had dropped out by 2012. Note: The household synthetic wealth indicator is computed through the principal component analysis of household scores in terms of the housing and durable goods indexes. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 37 | Conclusion The analysis of data sourced from macroeconomic statistics, education statistics, household surveys and a survey specifically carried out in the primary schools of two districts of the south of the country indicate a considerable downturn in households’ education behavior since the political crisis that Madagascar is suffering. Just as the evidence of deterioration in terms of access is irrefutable, the available data also suggest that equity has been eroded over the past three years. Determinants of Household Education Behavior - Education Statistics, Household Surveys, South Survey and Regressions - Impact on Macro/ESY EPM South education outcomes 2005-10 and 2000 - 2011 2005 and 2010 2009 and 2012 Variables 2009-12 Channel 1: Direct and Indirect Education Costs Public Spending per pupil Decreased (–) Education Price Index Increased (+) Share of education in household spending Increased (+) Per pupil household education spending Increased (+) Increased (+) Share of households paying fees and contributions Increased (+) Negative * Channel 2: Access to Loans and Household Income GDP per Capita Decreased (–) Consumer Price Index – Staple goods Increased (+) Access to loans Increased (+) Decreased (–) ns Poverty (1) Increased (+) Increased (+) Negative * Agricultural shocks Decreased (–) Non-repayable transfers Positive * Channel 3: Return on Investment in Education Pupil-teacher and pupil-class ratios Decreased (–) Decreased (–) Share of trained teachers Decreased (–) Textbook availability Decreased (–) Number of hours of instruction Decreased (–) Share of FRAM teachers Increased (+) Increased (+) ns Learning outcomes Decreased (–) Decreased (–) Channel 4: Opportunity Cost of Education Wage differential Decreased (–) Combining activities (work, chores) with school Increased (+) Increased (+) Source: Authors. Note: (1) Poverty levels, the gap in wealth quintiles and the mention of financial problems as the cause of out-of-school for the household surveys. 38 | Channel 1 All the indicators studied to assess the evolution and impact of direct and indirect education costs have worsened since the crisis: those based on ESY data (public unit costs have dropped and the education price index has increased), EPM data (household education spending and the share of education in the family budget have both increased) or the South survey (family spending and the share of families paying fees and contributions has increased). It is highly likely that the negative effects of the crisis on household education behavior are linked in better part to the cost of education, and it has been proven that schools who charge parents higher fees encourage dropout. The effect of this variable on education behavior is significant, and stronger now than before the crisis: an increase in the share of households paying school fees entails an increase in the dropout rate over 2009-12 that is greater still (based on regressions run on the tracer study data - South survey). Channel 2 All the income variables studies have deteriorated since the crisis. Whereas GDP per capita has dropped, the cost of staple goods has increased, doubly affecting household budgets. Poverty has worsened in every respect, but especially the share of households who mention financial problems as the reason for their children being out of school (EPM data). The correlation of the synthetic wealth indicator with out-of-school is significant, high and on the rise (2012, South survey, Sample Selection). Finally, agricultural shocks suffered over the previous 12 months had a greater impact on out-of-school rates in 2012, almost certainly affecting family production and income. The effect of access to credit on education indicators is insignificant according to the tests run. Furthermore, the evolution of loan rates appears to be positive on a national scale, but negative in the South districts. At any rate, household loans do not appear to be used to finance children’s education. Having received a non-repayable transfer on the other hand has a significant, strong and positive impact on enrollment. Channel 3 The evolution of the expected return on education is difficult to measure, complicating the appraisal of its effect on education behavior. The supply-side factors potentially having a strong incidence on the perception of the value of education have worsened (share of trained teachers, availability of textbooks, instructional time and share of civil servant teachers), with the exception of supervision rates, that have improved (PTR and pupil-class ratio). Learning outcomes have been following a downward trend since 10 years, possibly impacting the perception of the quality of education on offer and the future benefit to be derived from it. It has not been possible however to determine with certitude if the supply-side factors affecting the perceived value of education have an impact on behavior, but the education level of the head of household clearly does. The regressions suggest that the demand-side factors of Channel 3 have a strong impact on enrollment. Thus enrollment rates gradually improve for children whose head of household has studied more, the effect being very significant, and high. The fact that the child is the son/daughter of the household head also improves the probability of enrollment (EPM data). 39 | Channel 4 The effect of the opportunity cost of education is inconclusive. Whereas it is clear that more children must work in addition to going to school following the crisis (South survey and EPM data), since 2009 children’s work appears to have waned on the national level (drop in the number of hours of work per day and in the frequency of mention of work as a reason for being out of school) but increased in the South districts (greater share of Grade 2 children performing agricultural or cattle- breeding activities and frequency of mention of work as reason for dropout). Unemployment and underemployment rates, on the rise, would indicate a drop in the opportunity cost of education for children, but their impact on enrollment is not conclusive. Whereas it is highly likely that the negative impact of the crisis on household education behavior is due in great part to the effect of the cost of education (Channel 1) and household income (Channel 2), the effect of the expected return (Channel 3) is difficult to assess with the available data, and the effect of the opportunity cost (Channel 4) is not conclusive. Recommendations To compensate the impact of the crisis on education, a short/medium-term strategy should consider the following issues: (i) Cost/Income: the effective costless nature of primary education appears to be primordial and non-repayable transfers also have a significant effect on school attendance, but any initiative aiming to reduce the household education budget (distribution of textbooks or school kits, school feeding programmes, and so on) would probably have an impact in the current context. (ii) Demand for education: the individual and household characteristics (education level of household head, but also the parental bond of the child with the household head) having such a strong impact on enrollment decisions suggests that the perception of the value of education is to be cultivated. The next step would consist in an evaluation of the mechanisms, tools or programmes that could influence the issues identified. The evaluation should take into account, among other factors, the feasibility of the approach given the current organization of the sector and the available human resources, their efficiency in the Malagasy context and their cost-effectiveness, reasons for which it may be sensible to perform a financial modelization of the costs and benefits expected of each option considered. The supply issues raised are to be taken into account in the context of more long-term plans and strategies, such as the Interim Education Plan. It will certainly be necessary to implement a series of actions to improve education quality, such as to invest further in the training of teachers, to tend to providing an adequate status to community teachers and ensuring that instructional time is respected, while closely monitoring learning outcomes. 40 | ANNEXES Annex A: Tables and Figures Table A1: Macroeconomic Indicators, 2000-11 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Total population 16.3 16.8 17.3 17.8 18.4 18.9 19.5 20.1 20.7 21.3 21.9 (Million) Rural population 72.3 72.1 71.8 71.5 71.2 70.8 70.5 70.1 69.8 69.6 _ (% of Total) Life-expectancy at 61.6 62.5 63.3 64.1 64.8 65.4 65.8 66.2 66.6 * 67.0 _ birth (Years) Human Devt. Index 0.427 0.437 0.446 0.465 0.425 0.431 0.439 0.436 0.435 0.426 _ (HDI) GDP per capita 269.1 325.0 251.4 281.7 299.3 386.9 480.6 421.8 421.0 465.0 458.4 (US$) Nominal GDP 4.4 5.5 4.4 5.0 5.5 7.3 9.4 8.5 8.7 9.9 10.0 (Millions of US$) Nominal GDP 6,008.4 6,778.6 8,155.6 10,092.4 11,815.3 13,759.7 16,049.0 16,802.9 18,469.0 20,072.5 22,049.7 (Millions of Ariary) GDP growth rate -12.7 9.8 5.3 4.6 5.0 6.2 7.1 -4.1. 0.5 1.9 2.7 (%) Primary sector growth 16.7 0.5 1 0.8 2.1 2.2 2.9 8.5 -3.3 -2.3 _ (%) Secondary sector growth 1.6 1.6 0.8 0.4 3.5 9.7 3.6 -7.4 0.2 2.7 _ (%) Tertiary sector growth -4.0 5.0 2.8 2.9 7.1 7.8 8.2 -7.1 1.7 2.1 _ (%) Gross Investment (public & private, % 13.4 16.2 23.4 22.2 25.3 29.3 44.1 31.6 31.5 _ _ GDP) Source: Economic and financial reports of the Ministry of Economy and Industry, Central Bank of Madagascar and PIE, 2013. Note: Data for 2012 are estimations. * End of period estimation. 41 | Table A2: Executed Budget, 2000-11 (Milliards d’Ariary et pour cent) 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Ressources domestiques 1 193 1 129 733 1 243 1 706 1 377 1 496 1 697 2 264 1 883 2 267 2 295 (Prix courants) (Milliards d’Ariary) Ressources domestiques Recettes de l'État 1 612 1 692 1 777 1 865 1 958 2 056 2 158 2 292 2 456 2 366 2 380 2 446 (Prix constants de 2011) Recettes budgétaires 1 127 1 036 706 1 044 1 288 1 221 1 323 1 608 2 137 1 862 2 249 2 216 Recettes fiscales — — — — — — 1 261 1 573 2 087 1 782 1 981 2 180 Recettes non fiscales — — — — — — 62 35 50 80 268 36 Dons courants 66 92 27 199 418 156 173,2 89,1 127,5 21,0 18,2 72,2 (Aides budgétaires) Dons en capital (Projets) — — — — — — — — 420,8 171,2 155,5 416,9 Dépenses totales de l'État (Milliards d’Ariary) Dépenses de l’État 860 956 883 1 235 2 002 2 058 2 521 2 570 3 138 2 478 2 581 2 769 (Prix courants) Dépenses totales de l'État 2 078 2 203 1 926 2 114 2 225 2 328 2 442 2 596 2 780 2 678 2 694 2 769 (Prix constants de 2011) Dépenses courantes 539 567 594 741 1 018 1 100 1 313 1 520 1 754 1 753 1 848 2 160 Dépenses d'investissement 320 389 289 494 985 959 1 209 1 050 1 384 726 733 609 Source: Economic and financial reports of the Ministry of Economy and Industry and PIE, 2013. Note: Data for 2009 are provisional; those for 2010 and 2011 are estimations. Table A3: Primary Enrollment, by Type of School and Available Resources, 1999/00-2010/11 Primary Access Available Resources Public Private Total Number of (Number) Total New Grade 1 Number of school school number of operational number pupils teachers pupils pupils classrooms schools 1999/00 2,208,321 — — — — — — 2000/01 2,307,314 536,915 — — — — — 2001/02 2,409,082 577,831 1,892,801 516,281 52,206 18,295 50,736 2002/03 2,856,480 778,041 2,274,443 582,037 58,725 18,977 55,309 2003/04 3,366,600 896,570 2,715,664 650,936 61,521 20,150 64,265 2004/05 3,597,800 993,736 2,916,158 681,642 63,919 20,636 67,137 2005/06 3,698,610 999,627 2,983,087 715,523 70,658 22,218 76,831 2006/07 3,837,343 969,749 3,104,521 732,822 73,158 23,050 78,743 2007/08 4,020,322 1,032,657 3,263,066 757,256 78,919 24,387 85,257 2008/09 4,323,981 1,108,642 3,546,113 777,868 84,907 25,466 90,265 2009/10 4,329,576 1,109,523 3,552,237 777,339 97,153 27,748 95,184 2010/11 4,305,069 1,110,709 3,539,331 765,738 108,661 27,719 100,918 Source: Education statistical yearbooks, 2000-11. 42 | Table A4: Annual Growth in Primary Access Rates by Type of School, and Annual Growth in Available Resources, 2000/01-2010/11 Primary Access Available Resources Public Private Total Number of (Percent) Total New Grade 1 Number of school school number of operational number pupils teachers pupils pupils classrooms schools 2000/01 4.5 — — — — — — 2001/02 4.4 7.6 — — — — — 2002/03 18.6 34.6 20.2 12.7 12.5 3.7 9.0 2003/04 17.9 15.2 19.4 11.8 4.8 6.2 16.2 2004/05 6.9 10.8 7.4 4.7 3.9 2.4 4.5 2005/06 2.8 0.6 2.3 5.0 10.5 7.7 14.4 2006/07 3.8 -3.0 4.1 2.4 3.5 3.7 2.5 2007/08 4.8 6.5 5.1 3.3 7.9 5.8 8.3 2008/09 7.6 7.4 8.7 2.7 7.6 4.4 5.9 2009/10 0.1 0.1 0.2 -0.1 14.4 9.0 5.4 2010/11 -0.6 0.1 -0.4 -1.5 11.8 -0.1 6.0 Source: Education statistical yearbooks, 2000-11. Table A5: Primary Gross Enrollment Rates, by Grade, 2001/02-2010/11 (Percent) 2001/02 2002/03 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 Grade 1 188.3 233.7 272.9 222.8 212.7 201.8 211.2 225.2 221.1 218.4 218.8 Grade 2 119.7 130.4 149.9 206.1 196.3 190.6 177.7 181.5 178.1 165.9 163.1 Grade 3 101.8 110.2 122.0 132.3 147.7 156.2 153.2 153.3 150.9 146.6 145.1 Grade 4 67.4 73.8 83.0 81.5 86.2 98.5 110.2 114.0 112.3 109.0 108.6 Grade 5 48.1 55.8 62.8 76.8 76.6 79.2 87.6 98.9 88.4 84.9 83.6 Primary GER 107.2 123.5 141.3 146.6 146.4 147.5 150.0 156.7 152.3 147.1 145.9 Source: Education statistical yearbooks, 2000-11 and EPM, 2005. Note: The estimated population growth rate used here is 3.7 percent. Table A6: Enrollment Status of Children, by Age, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (%) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled 6 yrs 66.8 2.4 30.8 58.4 1.0 40.6 -8.4 -1.4 9.8 7 yrs 75.5 2.5 22.0 74.7 2.2 23.1 -0.8 -0.3 1.1 8 yrs 81.8 2.7 15.5 79.9 2.3 17.8 -1.9 -0.4 2.3 9 yrs 86.2 3.4 10.4 81.8 4.3 13.9 -4.4 0.9 3.5 10 yrs 90.5 3.2 6.3 80.3 4.6 15.1 -10.2 1.4 8.8 11 yrs 84.8 6.2 9.0 80.3 8.1 11.6 -4.5 1.9 2.6 12 yrs 83.7 8.6 7.7 75.2 11.1 13.8 -8.5 2.5 6.1 13 yrs 72.8 16.3 10.9 73.3 15.9 10.8 0.5 -0.4 -0.1 14 yrs 52.1 30.0 17.9 60.8 26.5 12.7 8.7 -3.5 -5.2 Source: Household surveys, 2005 and 2010. 43 | Table A7: Enrollment Status of Children in the South, by Age, 2009 and 2012 2009-12 Gap 2009 2012 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled 5 years 48.3 0.9 50.9 45.5 3.3 51.2 -2.8 2.4 0.3 6 years 80.9 1.2 17.9 61.1 2.1 36.8 -19.8 0.9 18.9 7 years 89.2 0.4 10.4 79.6 1.2 19.2 -9.6 0.8 8.8 8 years 92.6 1.6 5.8 82.3 3.0 14.8 -10.3 1.4 9.0 9 years 94.1 0.0 5.9 89.3 3.0 7.7 -4.8 3.0 1.8 10 years 92.7 2.1 5.2 86.0 5.0 9.0 -6.7 2.9 3.8 11 years 90.5 4.3 5.2 84.7 7.1 8.2 -5.8 2.8 3.0 12 years 90.0 3.5 6.6 79.0 10.9 10.1 -11.0 7.4 3.5 13 years 84.6 8.6 6.9 73.3 18.2 8.5 -11.3 9.6 1.6 14 years 71.7 15.8 12.5 60.0 24.7 15.3 -11.7 8.9 2.8 15 years 69.7 12.6 17.6 59.8 25.3 14.9 -9.9 12.7 -2.7 Total 85.1 3.9 11.0 75.9 7.8 16.3 -9.2 3.9 5.3 Obs. 2,343 2,371 — Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. Note: Data here apply to households with a child in Grade 2 and aged between 5 and 15 years. Table A8: Enrollment Status of Children, by Age Group, Gender and Area of Residence, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled 6-10 Years National 79.6 2.8 17.6 75.0 2.9 22.1 -4.6 0.1 4.5 Urban 86.0 1.7 12.3 83.7 2.0 14.3 -2.3 0.3 2.0 Rural 78.0 3.1 18.9 73.1 3.1 23.9 -4.9 0.0 5.0 Boys 79.1 2.7 18.2 73.7 2.5 23.7 -5.4 -0.2 5.5 Girls 80.1 2.9 17.0 76.2 3.2 20.5 -3.9 0.3 3.5 Urban boys 85.0 1.5 13.5 82.3 2.1 15.6 -2.7 0.6 2.1 Urban girls 87.0 2.0 11.0 85.1 2.0 12.9 -1.9 0.0 1.9 Rural boys 77.7 3.0 19.3 71.7 2.6 25.6 -6.0 -0.4 6.3 Rural girls 78.4 3.1 18.5 74.4 3.5 22.1 -4.0 0.4 3.6 11-14 Years National 73.8 15.0 11.3 72.6 15.2 12.3 -1.2 0.2 1.0 Urban 79.9 11.3 8.8 81.7 11.0 7.3 1.8 -0.3 -1.5 Rural 72.1 16.0 11.9 70.4 16.1 13.4 -1.7 0.1 1.5 Boys 74.0 14.4 11.6 71.5 15.7 12.8 -2.5 1.3 1.2 Girls 73.5 15.6 10.9 73.7 14.6 11.7 0.2 -1.0 0.8 Urban boys 80.2 10.2 9.5 81.9 11.1 7.0 1.7 0.9 -2.5 Urban girls 79.6 12.4 8.0 81.6 10.9 7.5 2.0 -1.5 -0.5 Rural boys 72.4 15.5 12.1 69.0 16.8 14.2 -3.4 1.3 2.1 Rural girls 71.9 16.4 11.7 71.9 15.5 12.7 0.0 -0.9 1.0 Source: Household surveys, 2005 and 2010. 44 | Table A9: Enrollment Status of Children Aged 6-10 Years, by Region, 2005 and 2010 2005-10 Gap 2005 2010 2005 Poverty (Percent) (% Points) Rate (%) E D NE E D NE E D NE Alaotra-Mangoro 85.7 3.2 11.1 85.0 2.2 12.8 -0.7 -1.0 1.7 57.7 Amoron'i Mania 93.8 0.9 5.3 76.7 1.6 21.7 -17.1 0.7 16.4 78.0 Analamanga 93.9 1.0 5.1 91.2 2.3 6.5 -2.7 1.3 1.4 42.9 Analanjirofo 85.2 2.5 12.3 85.1 1.9 13.0 -0.1 -0.6 0.7 79.1 Androy 54.4 5.2 40.4 54.9 0.9 44.2 0.5 -4.3 3.8 83.3 Anosy 76.9 1.5 21.7 54.7 5.8 39.6 -22.2 4.3 17.9 73.8 Atsimo-Andrefana 63.5 1.6 34.9 52.9 2.7 44.4 -10.6 1.1 9.5 83.9 Atsimo-Atsinanana 63.2 2.0 34.9 55.1 8.7 36.3 -8.1 6.7 1.4 75.2 Atsinanana 84.9 4.0 11.0 80.6 2.6 16.8 -4.3 -1.4 5.8 79.0 Betsiboka 73.0 2.6 24.4 66.7 4.5 28.8 -6.3 1.9 4.4 70.0 Boeny 71.2 2.1 26.7 60.4 1.6 38.0 -10.8 -0.5 11.3 48.8 Bongolava 68.9 6.6 24.5 72.9 3.8 23.4 4.0 -2.8 -1.1 64.1 Diana 72.7 3.7 23.6 79.9 0.4 19.7 7.2 -3.3 -3.9 49.2 Ihorombe 75.2 3.7 21.0 74.1 2.2 23.7 -1.1 -1.5 2.7 78.0 Itasy 80.8 4.2 15.0 86.5 3.0 10.5 5.7 -1.2 -4.5 68.7 Mahatsiatra Ambony 88.5 1.6 10.0 76.8 3.4 19.8 -11.7 1.8 9.8 72.1 Melaky 41.7 6.2 52.1 51.3 3.2 45.5 9.6 -3.0 -6.6 62.7 Menabe 78.8 3.5 17.7 64.5 5.1 30.4 -14.3 1.6 12.7 61.7 Sava 83.3 1.9 14.9 83.1 2.4 14.5 -0.2 0.5 -0.4 72.5 Sofia 81.7 1.8 16.5 78.1 0.7 21.2 -3.6 -1.1 4.7 80.7 Vakinankaratra 81.9 3.7 14.5 83.4 2.7 13.9 1.5 -1.0 -0.6 74.3 Vatovavy- Fitovinany 65.7 5.9 28.4 79.7 3.4 16.9 14.0 -2.5 -11.5 80.8 Source: Household surveys, 2005 and 2010. Note: E: Enrolled, D: Dropped out, NE: never enrolled. Table A10: Enrollment Status of Children Aged 11-14 Years, by Region, 2005 and 2010 2005-10 Gap 2005 2010 2005 Poverty (Percent) (% Points) Rate (%) E D NE E D NE E D NE Alaotra-Mangoro 75.5 13.6 10.9 69.7 20.2 10.1 -5.8 6.6 -0.8 57.7 Amoron'i Mania 78.3 15.2 6.5 79.4 12.2 8.4 1.1 -3.0 1.9 78.0 Analamanga 81.3 15.4 3.2 83.5 14.8 1.8 2.2 -0.6 -1.4 42.9 Analanjirofo 83.2 11.8 5.0 78.4 15.3 6.3 -4.8 3.5 1.3 79.1 Androy 67.9 7.0 25.1 54.2 13.4 32.4 -13.7 6.4 7.3 83.3 Anosy 80.8 6.3 12.9 45.0 14.3 40.6 -35.8 8.0 27.7 73.8 Atsimo-Andrefana 81.2 5.4 13.3 54.4 8.0 37.6 -26.8 2.6 24.3 83.9 Atsimo-Atsinanana 63.8 17.4 18.8 51.8 19.6 28.6 -12.0 2.2 9.8 75.2 Atsinanana 75.7 17.0 7.3 81.0 11.0 8.0 5.3 -6.0 0.7 79.0 Betsiboka 65.2 16.7 18.1 60.6 24.1 15.3 -4.6 7.4 -2.8 70.0 Boeny 65.6 18.1 16.3 64.2 13.0 22.8 -1.4 -5.1 6.5 48.8 Bongolava 69.9 19.6 10.5 70.0 23.6 6.3 0.1 4.0 -4.2 64.1 Diana 75.3 9.6 15.2 86.7 10.3 3.1 11.4 0.7 -12.1 49.2 Ihorombe 63.5 14.8 21.7 75.3 10.5 14.3 11.8 -4.3 -7.4 78.0 Itasy 72.0 22.4 5.6 69.5 25.3 5.3 -2.5 2.9 -0.3 68.7 Mahatsiatra Ambony 77.5 16.8 5.7 71.9 23.3 4.8 -5.6 6.5 -0.9 72.1 Melaky 29.1 25.1 45.8 48.0 16.0 36.1 18.9 -9.1 -9.7 62.7 Menabe 69.9 17.4 12.7 60.4 17.3 22.3 -9.5 -0.1 9.6 61.7 Sava 84.2 9.0 6.8 88.8 9.7 1.5 4.6 0.7 -5.3 72.5 Sofia 76.7 12.0 11.3 82.6 10.6 6.8 5.9 -1.4 -4.5 80.7 Vakinankaratra 65.3 23.4 11.3 80.6 14.6 4.8 15.3 -8.8 -6.5 74.3 Vatovavy- Fitovinany 60.5 14.7 24.8 77.9 15.9 6.2 17.4 1.2 -18.6 80.8 Source: Household surveys, 2005 and 2010. Note: E: Enrolled, D: Dropped out, NE: never enrolled. 45 | Table A11: Primary Dropout Rate, by Grade, 2001/02-2009/10 Dropout Rates Number of Dropouts (Percent and Grade 1- Grade 1- Grade 1- Grade 1- Number) Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Grade 4 Grade 5 Grade 4 Grade 5 Total Total Total Total 2001/02 8.6 4.3 11.8 12.4 24.5 8.7 10.0 191,176 240,747 2002/03 10.3 3.3 8.7 8.8 16.9 8.1 8.9 211,989 252,828 2003/04 27.1 7.6 9.9 17.4 19.6 18.0 18.1 555,629 610,568 2004/05 24.5 18.4 11.6 18.4 23.1 19.2 19.6 623,607 705,075 2005/06 21.8 13.0 12.0 15.9 20.2 16.2 16.6 539,298 612,285 2006/07 21.3 13.2 10.9 16.4 19.8 15.7 16.2 543,607 619,862 2007/08 18.8 8.5 8.4 15.6 32.0 13.1 15.2 469,006 609,736 2008/09 23.2 12.7 12.6 18.2 31.4 17.2 18.9 656,626 816,965 2009/10 25.0 13.2 14.7 19.7 24.8 18.7 19.4 723,622 840,311 Source: Authors’ calculations based on Education statistical yearbooks, 2000-11. Note: The 2008/09 dropout rate is the difference between those enrolled in 2008/09 and those enrolled in 2009/10. Table A12: Percentage of Primary Repeaters, by Grade, 2000/01-2010/11 Percentage of Repeaters Number of Repeaters (Percent and Grade 1 - Grade 1 - Grade 1 - Grade 1 - Number) Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Grade 4 Grade 5 Grade 4 Grade 5 Total Total Total Total 2000/01 36.9 27.2 27.9 23.0 25.0 30.7 30.2 646,599 696,822 2001/02 35.9 29.1 30.3 23.3 22.4 31.4 30.6 692,041 737,400 2002/03 32.5 27.7 29.6 23.5 25.9 29.6 29.3 774,768 837,441 2003/04 35.3 26.1 29.5 23.2 24.0 30.5 30.0 941,692 1,008,722 2004/05 14.7 19.6 26.6 9.3 20.7 18.0 18.3 584,569 657,484 2005/06 12.7 28.3 24.7 7.5 21.5 19.5 19.7 651,916 729,729 2006/07 13.3 27.1 24.5 7.8 18.1 19.3 19.1 664,596 734,491 2007/08 14.4 28.0 25.3 9.6 18.1 19.8 19.7 710,748 790,304 2008/09 16.3 26.2 25.8 12.4 20.3 20.5 20.4 780,683 884,091 2009/10 18.4 25.2 24.8 14.8 13.4 21.0 20.2 811,575 874,752 2010/11 18.5 23.4 23.6 14.4 12.5 20.3 19.4 778,222 836,488 Source: Education statistical yearbooks, 2000-11. Table A13: Education System Expenditure, 2000-11 Estimations (Thousands of Ariary and Percent) 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Total Education Expenditure (Current prices) 152.8 194 163.1 206.0 266.2 388.4 387.9 465.1 582.8 531.5 488.7 557.2 (Constant 2011 prices) 441 522 381 468 529 653 584 639 733 617 526 557 (% of Total public expenditure) 21.2 23.7 19.8 22.1 23.8 28.0 23.9 24.6 26.4 23.0 19.5 20.1 (% of GDP) 2.9 3.3 2.7 3.0 3.3 3.8 3.3 3.4 3.6 3.2 2.6 2.8 Recurrent Education Expenditure (Constant 2011 prices) 313 368 318 356 391 446 419 468 526 507 469 505 (% of Total public expenditure) 21.2 23.7 19.8 22.1 23.8 28.0 23.9 24.6 26.4 23.0 19.5 20.1 Per pupil (000s of 2011 Ariary) 60.0 67.8 56.4 60.6 64.2 70.3 63.8 68.4 73.9 68.6 63.9 63.2 Capital Expenditure (Constant 2011 prices) 128 154 63 112 138 207 165 171 207 110 57 52 Internal (Current prices) — — — — — — 33.8 46.4 46.2 43.2 20.3 37.3 External (Current prices) — — — — — — 75.5 77.8 118.5 51.6 32.3 14.9 Internal (% of Tax income) — — — — — — 2.7 2.9 2.2 2.4 1.0 1.7 Memory Items (Million children) School-aged population (6-17 years) 5.22 5.42 5.64 5.86 6.09 6.33 6.58 6.84 7.11 7.39 7.68 7.99 Source: Economic and financial reports of the Ministry of Economy and Industry and PIE, 2013. Note: Data for 2009 are provisional; those for 2010 and 2011 are estimations. 46 | Table A14: Dropout Rate, by Region, 2006/07-2009/10 Gap between (Percent) 2006/07 2007/08 2008/09 2009/10 2007/08 and 2009/10 (% Points) Alaotra-Mangoro 9.6 8.8 11.5 12.0 3.2 Amoron'i Mania 13.3 12.0 16.9 11.8 -0.2 Analamanga 5.8 6.3 7.1 10.0 3.7 Analanjirofo 14.7 10.0 17.0 14.1 4.1 Androy 26.8 23.9 32.2 30.5 6.6 Anosy 27.5 20.1 29.4 31.8 11.7 Atsimo-Andrefana 23.9 14.0 27.7 27.8 13.8 Atsimo-Atsinanana 33.3 25.4 29.4 32.1 6.7 Atsinanana 18.9 14.8 23.7 21.4 6.6 Betsiboka 17.1 13.6 16.7 17.1 3.5 Boeny 17.3 12.4 19.4 23.3 10.9 Bongolava 10.0 11.1 16.1 16.1 5 Diana 11.3 11.2 12.0 15.7 4.5 Haute Matsiatra 16.6 13.3 13.1 18.7 5.4 Ihorombe 20.3 17.0 24.0 23.4 6.4 Itasy 10.9 10.3 11.8 10.3 0 Melaky 21.0 24.5 25.4 37.2 12.7 Menabe 23.5 18.4 30.3 22.7 4.3 Sava 10.8 11.9 13.8 12.4 0.5 Sofia 15.9 13.6 11.9 17.4 3.8 Vakinankaratra 12.6 10.3 11.2 12.8 2.5 Vatovavy Fitovinany 22.2 17.6 20.8 28.6 11 National average 15.7 13.1 17.2 18.7 5.6 Source: Education statistical yearbooks, 2006-11. Table A15: Total Household Education Spending per Pupil, by Gender and Area of Residence, 2005 and 2010 (Constant 2005 Ariary Girls Boys Urban Rural Total and Percent) 2005 Public 8,202 8,662 9,121 8,302 8,436 Private 29,910 33,490 65,016 14,623 31,768 Total 12,409 13,622 27,926 9,327 13,029 2010 Public 13,698 13,900 16,333 13,310 13,799 Private 51,924 56,304 83,041 40,275 54,204 Total 19,601 20,983 34,918 16,909 20,294 2005-10 Variation Public 67.0% 37.7% 44.2% 37.6% 38.9% Private 73.6% 40.5% 21.7% 63.7% 41.4% Total 58.0% 35.1% 20.0% 44.8% 35.8% Source: Household surveys, 2005 and 2010. 47 | Table A16: Disaggregated per Pupil Costs, by Type, Gender, Area of Residence and Type of School, 2005 and 2010 Public Private (Constant 2005Ariary) Girls Boys Urban Rural Total Girls Boys Urban Rural Total 2005 Enrollment fees 3,069 3,321 3,286 3,116 3,197 11,877 10,071 12,496 6,255 10,934 PTA contributions 3,457 3,783 2,768 4,267 3,622 4,221 5,289 3,443 6,639 4,754 Insurance costs 488 584 594 483 534 2,024 1,820 2,315 323 1,917 School fees 10,228 8,515 14,250 6,388 9,315 30,351 30,149 36,499 14,734 30,245 Uniforms 2,889 2,911 2,972 2,843 2,900 3,805 3,704 4,139 2,941 3,754 Sports wear 2,975 2,945 2,880 3,063 2,960 6,959 5,686 6,441 5,719 6,338 Supplies 4,498 4,778 5,228 4,193 4,639 9,628 9,313 11,601 4,939 9,465 Transport 21,190 4,535 17,794 4,961 13,383 16,402 68,864 55,848 28,490 53,434 School meals 13,627 14,656 16,947 11,330 14,175 17,916 23,839 22,945 13,606 21,033 Other expenses 4,168 3,913 3,558 4,437 4,038 11,179 8,468 9,896 9,136 9,727 2010 Enrollment fees 4,392 4,158 4,925 3,708 4,274 10,532 10,167 13,010 5,716 10,346 PTA contributions 4,274 4,495 3,645 4,902 4,387 4,978 4,391 4,441 4,957 4,686 Insurance costs 421 342 392 377 383 1,525 2,159 2,482 951 1,840 School fees 7,729 6,925 8,726 6,275 7,347 30,163 30,936 36,952 18,191 30,557 Uniforms 2,890 2,989 3,257 2,576 2,939 4,530 4,360 4,837 3,605 4,447 Sports wear 3,539 3,231 3,908 2,469 3,394 3,728 4,116 4,192 3,170 3,913 Supplies 4,999 5,065 5,654 4,519 5,032 11,595 10,779 13,441 7,394 11,178 Transport 24,829 17,108 28,203 14,143 21,901 47,057 52,247 49,827 49,602 49,782 School meals 22,452 30,954 23,208 30,063 26,356 40,061 32,434 34,911 38,897 36,074 Other expenses 2,909 3,145 2,800 3,195 3,026 9,688 10,086 12,505 4,043 9,898 2005-10 Gap Enrollment fees 30.1% 20.1% 33.3% 16.0% 25.2% -12.8% 0.9% 4.0% -9.4% -5.7% PTA contributions 19.1% 15.9% 24.1% 13.0% 17.4% 15.2% -20.5% 22.5% -33.9% -1.5% Insurance costs -15.7% -70.8% -51.4% -28.3% -39.5% -32.7% 15.7% 6.7% 66.1% -4.2% School fees -32.3% -23.0% -63.3% -1.8% -26.8% -0.6% 2.5% 1.2% 19.0% 1.0% Uniforms 0.0% 2.6% 8.8% -10.4% 1.3% 16.0% 15.0% 14.4% 18.4% 15.6% Sports wear 15.9% 8.8% 26.3% -24.1% 12.8% -86.7% -38.1% -53.6% -80.4% -62.0% Supplies 10.0% 5.7% 7.5% 7.2% 7.8% 17.0% 13.6% 13.7% 33.2% 15.3% Transport 14.7% 73.5% 36.9% 64.9% 38.9% 65.1% -31.8% -12.1% 42.6% -7.3% School meals 39.3% 52.7% 27.0% 62.3% 46.2% 55.3% 26.5% 34.3% 65.0% 41.7% Other expenses -43.3% -24.4% -27.1% -38.9% -33.4% -15.4% 16.0% 20.9% -126.0% 1.7% Source: Household surveys, 2005 and 2010. Table A17: Average Household Education Spending per Pupil in the South, by Type of Expense, 2008/09 and 2011/12 Sample Selection Tracer Study (Constant 2012 Ariary) 2008/09 2011/12 2008/09 2011/12 Enrollment fees 1,514 1,372 1,185 1,593 PTA contributions 395 484 367 521 Monthly school fees 1,607 2,199 1,357 1,542 Supplies 3,741 4,111 3,800 4,899 Enrollment fees + monthly fees + FRAM 3,486 4,003 2,892 3,581 Enrollment fees + monthly fees + FRAM + supplies 7,082 7,945 6,561 8,277 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. Note: The data presented here concern households having paid school fees, and the averages are computed for the entire group, regardless of whether they incurred each type of expense or not. 48 | Table A18: Frequency of Mention by South Households of Reasons for Dropout (Grades 1-3), 2010-12 (Tracer Study) 2012 2011 2010 2010-12 Average (Percent) Boys Girls Total Boys Girls Total Boys Girls Total Boys Girls Total Financial Child’s work/ Need for family 20.0 18.5 19.1 40.0 15.4 26.7 16.3 16.0 16.1 25.4 16.6 20.6 labor School-related spending is too 17.1 14.8 15.8 15.6 13.2 14.4 17.5 13.3 15.6 16.7 13.8 15.3 high Financial 1.4 3.7 2.8 1.7 7.4 4.8 1.2 2.7 1.9 1.4 4.6 3.2 problems Quality Classes are boring 10.0 6.5 7.9 3.5 5.1 4.4 1.2 6.7 3.7 4.9 6.1 5.3 No teacher 1.4 3.7 2.8 4.3 3.7 4.0 7.0 8.0 7.5 4.2 5.1 4.8 Teacher 5.7 7.4 6.7 1.7 4.4 3.2 4.7 1.3 3.1 4.0 4.4 4.3 absenteeism Child isn’t keen 7.1 1.9 3.9 2.6 4.4 3.6 3.5 0.0 1.9 4.4 2.1 3.1 School closed 0.0 0.9 0.6 4.3 2.2 3.2 3.5 1.3 2.5 2.6 1.5 2.1 School too far 2.9 0.0 1.1 2.6 2.9 2.8 2.3 1.3 1.9 2.6 1.4 1.9 Poor quality and quantity of 0.0 0.9 0.6 1.7 2.2 2.0 0.0 0.0 0.0 0.6 1.0 0.9 infrastructure Knowledge acquired deemed 0.0 0.9 0.6 0.0 0.7 0.4 0.0 1.3 0.6 0.0 1.0 0.5 useless Others Pregnancy or 4.3 13.9 10.1 0.0 14.7 8.0 1.2 2.7 1.9 1.8 10.4 6.7 marriage Child having difficulties at 4.3 6.5 5.6 10.4 5.1 7.6 4.7 5.3 5.0 6.5 5.6 6.1 school Child too young 2.9 0.9 1.7 0.0 0.7 0.4 7.0 14.7 10.6 3.3 5.4 4.2 Child is sick 5.7 3.7 4.5 1.7 2.2 2.0 0.0 1.3 0.6 2.5 2.4 2.4 Child too old 4.3 3.7 3.9 0.9 3.7 2.4 1.2 0.0 0.6 2.1 2.5 2.3 Level attained 1.4 2.8 2.2 1.7 2.2 2.0 1.2 0.0 0.6 1.4 1.7 1.6 deemed sufficient Food insecurity 0.0 1.9 1.1 0.9 1.5 1.2 0.0 0.0 0.0 0.3 1.1 0.8 Child won’t go 0.0 0.9 0.6 0.9 0.7 0.8 0.0 0.0 0.0 0.3 0.5 0.5 Migration 0.0 1.9 1.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.6 0.4 Danger on the 0.0 0.9 0.6 0.9 0.0 0.4 0.0 0.0 0.0 0.3 0.3 0.3 way to school Other reasons 7.1 1.9 3.9 1.7 2.2 2.0 1.2 0.0 0.6 3.3 1.4 2.2 No answer 4.3 1.9 2.8 2.6 5.1 4.0 26.7 24.0 25.5 11.2 10.3 10.8 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Observations 178 251 161 500 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. 49 | Table A19: Frequency of Mention by Headmasters of Reasons for Dropout in the South, 2012 (Percent) Answer 1 Answer 2 Need for family labor 26.2 29.1 School-related expenses too high 21.5 9.0 Food insecurity 8.7 3.7 Pregnancy of marriage 8.1 13.4 School too far 6.0 3.7 Child in difficulty at school 5.4 8.2 Classes are boring 4.7 5.2 Child too old 4.7 7.5 Poor quality/quantity of infrastructure 2.7 2.2 Education level attained deemed sufficient 2.7 6.0 Financial problems 2.7 2.2 No teacher 1.3 0.7 Teacher absenteeism 0.7 0.7 Poor quality/quantity of teaching materials 0.7 1.5 Knowledge acquired deemed useless 0.7 2.2 Language of instruction 0.0 0.7 Violence on behalf of staff 0.0 1.5 Child is sick 0.0 1.5 Other reasons 3.4 0.7 Total 100.0 100.0 Observations 149 134 Source: South Survey in the Districts of Amboasary and Betioky, 2012. Table A20: Poverty Rates, by Region, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) Alaotra Mangoro 59.6 68.2 8.6 Amoron'i Mania 77.3 85.2 7.9 Analamanga 41.9 54.5 12.6 Analanjirofo 77.8 83.5 5.7 Androy 83.7 94.3 10.6 Anosy 73.1 83.5 10.4 Atsimo Andrefana 73.0 82.1 9.1 Atsimo Atsinanana 82.8 94.5 11.7 Atsinanana 79.0 82.1 3.1 Betsiboka 72.8 82.2 9.4 Boeny 48.4 62.6 14.2 Bongolava 65.0 76.8 11.8 Diana 49.1 54.4 5.3 Ihorombe 78.6 80.7 2.1 Itasy 68.0 80.0 12.0 Matsiatra Ambony 70.4 84.7 14.3 Melaky 62.5 80.2 17.7 Menabe 60.8 64.2 3.4 Sava 70.5 74.9 4.4 Sofia 80.7 71.5 -9.2 Vakinankaratra 73.4 75.8 2.4 Vatovavy Fitovinany 80.1 90.0 9.9 National average 67.7 76.5 8.8 Poorest region 83.7 94.5 10.8 Wealthiest region 41.9 54.4 12.5 Standard deviation 11.4 10.9 -0.5 Source: Household surveys, 2005 and 2010. Note: The rates for each year were adjusted for comparability. 50 | Table A21: Consumer Price Indexes, 2003-12 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Local goods 100.0 114.9 138.0 151.3 168.8 184.8 202.4 220.9 243.0 264.4 Semi-imported goods 100.0 106.8 115.9 130.7 136.2 146.0 161.8 181.8 191.7 206.4 Imported goods 100.0 116.6 140.0 157.9 168.2 183.9 189.3 202.4 219.4 232.9 Untransformed agric. goods 100.0 112.4 139.4 154.2 170.6 195.3 222.9 246.6 276.8 310.1 Transformed agric. goods 100.0 131.1 170.3 170.4 199.6 210.9 219.3 227.4 266.4 289.4 Industry-made products 100.0 110.9 127.6 147.2 158.0 174.8 187.2 205.7 216.9 230.2 Hand-made products 100.0 103.2 107.5 113.4 119.0 126.0 143.6 167.4 184.2 204.6 Public services 100.0 108.1 117.8 133.4 163.5 170.9 179.5 192.8 215.2 239.8 Private services 100.0 109.1 120.5 139.8 149.2 160.3 181.5 203.7 214.2 230.4 Staple goods 100.0 126.1 161.5 167.9 190.6 205.5 212.1 221.5 255.2 280.9 Foodstuffs, beverages and tobacco 100.0 120.0 150.7 157.1 179.2 198.1 214.4 228.3 258.6 282.4 Textiles and clothing 100.0 103.4 105.7 118.0 125.0 133.4 154.3 177.7 184.2 197.5 Housing, water, electricity, gas and other fuels 100.0 114.2 132.4 162.9 172.4 188.3 210.2 236.3 252.1 272.1 Furniture, home appliances and home maintenance 100.0 103.9 113.6 129.4 138.7 146.5 168.0 202.2 200.6 215.0 Health 100.0 107.7 117.2 127.1 139.1 154.2 169.3 201.7 218.1 234.1 Transportation 100.0 111.6 127.0 151.9 162.3 175.3 179.7 185.0 190.7 201.2 Leisure, culture and sport 100.0 99.7 105.6 117.5 126.7 133.6 139.7 158.3 149.2 155.7 Teaching 100.0 107.1 114.9 123.3 135.5 144.5 154.5 176.9 203.0 229.5 Hotels, cafés and restaurants 100.0 101.3 112.0 140.2 148.8 160.8 176.1 206.3 223.1 233.6 Other goods and services 100.0 105.0 115.0 126.8 136.3 145.5 157.8 171.4 188.7 209.3 Rice 100.0 136.5 187.2 175.5 215.8 229.3 232.2 235.4 285.3 305.7 Energy 100.0 117.2 136.3 180.4 186.5 207.6 217.0 242.4 263.3 285.8 Total 100.0 114.0 134.9 149.4 164.8 180.1 196.2 214.4 234.7 247.9 Source: INSTAT-NIPC. Note: Each index represents the yearly average; for the remaining months of 2012 at the time of the analysis, an estimation based on low inflation was made. Table A22: Distribution of Primary School-Aged Children by Enrollment Status, Age Group and Wealth Quintile, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled 6-10 Years Q1 69.5 4.2 26.3 58.8 3.9 37.3 -10.7 -0.3 11.0 Q2 76.8 2.8 20.3 74.1 3.3 22.6 -2.7 0.5 2.3 Q3 81.9 2.6 15.5 75.7 2.5 21.8 -6.2 -0.1 6.3 Q4 86.2 1.9 12.0 84.8 2.2 13.0 -1.4 0.3 1.0 Q5 89.1 2.0 8.8 91.0 1.9 7.1 1.9 -0.1 -1.7 Total 79.6 2.8 17.6 75.0 2.9 22.1 -4.6 0.1 4.5 11-14 Years Q1 72.4 12.1 15.5 56.3 18.6 25.0 -16.1 6.5 9.5 Q2 70.8 15.1 14.1 69.3 18.7 12.1 -1.5 3.6 -2.0 Q3 74.8 16.4 8.8 74.9 14.8 10.3 0.1 -1.6 1.5 Q4 72.8 18.1 9.1 79.8 12.8 7.4 7.0 -5.3 -1.7 Q5 78.8 13.5 7.7 85.9 9.7 4.4 7.1 -3.8 -3.3 Total 73.8 15.0 11.3 72.6 15.2 12.3 -1.2 0.2 1.0 Source: Household surveys, 2005 and 2010. 51 | Table A23: Distribution of the Enrollment Status of Children by Transfer Received (Repayable or not), Area of Residence and Gender, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled Total Transfer 81.7 2.9 15.4 75.3 2.7 22.0 -6.4 -0.2 6.6 No 78.9 2.8 18.3 74.8 3.0 22.2 -4.1 0.2 3.9 Urban Transfer 86.9 1.7 11.4 87.0 1.7 11.4 0.1 0.0 0.0 No 85.7 1.7 12.6 81.7 2.3 16.0 -4.0 0.6 3.4 Rural Transfer 80.3 3.2 16.5 72.5 3.0 24.6 -7.8 -0.2 8.1 No 77.4 3.0 19.6 73.3 3.1 23.5 -4.1 0.1 3.9 Boys Transfer 81.5 2.7 15.8 75.2 2.2 22.6 -6.3 -0.5 6.8 No 78.3 2.7 19.0 73.0 2.7 24.4 -5.3 0.0 5.4 Girls Transfer 81.8 3.1 15.0 75.5 3.2 21.3 -6.3 0.1 6.3 No 79.5 2.8 17.6 76.6 3.3 20.1 -2.9 0.5 2.5 Source: Household surveys, 2005 and 2010. Table A24: Distribution of the Enrollment Status of Children, by Activity of Head of Household and Area of Residence, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled Total Active 83.1 1.2 15.7 79.6 2.3 18.1 -3.5 1.1 2.4 Unemployed 79.5 2.9 17.7 74.9 2.9 22.2 -4.6 0.0 4.5 /inactive Urban Active 90.1 0.8 9.1 86.6 2.3 11.1 -3.5 1.5 2.0 Unemployed 85.7 1.8 12.5 83.5 2.0 14.5 -2.2 0.2 2.0 /inactive Rural Active 75.3 1.6 23.1 74.6 2.3 23.1 -0.7 0.7 0.0 Unemployed 78.1 3.1 18.8 73.0 3.1 23.9 -5.1 0.0 5.1 /inactive Source: Household surveys, 2005 and 2010. 52 | Table A25: Frequency of Mention by Households of Reasons for the Dropout of Children Aged 6-17 Years, by Area of Residence, 2005 and 2010 2005-10 Gap 2005 2010 (Percent) (% Points) Urban Rural Total Urban Rural Total Urban Rural Total Financial Child’s work 9.0 12.9 12.2 9.2 10.9 10.6 0.2 -2.0 -1.6 Child must work 6.9 9.3 8.9 2.6 2.5 2.6 -4.3 -6.8 -6.3 Financial problems 27.0 17.7 19.3 32.0 24.5 25.6 5.0 6.8 6.3 Quality Class content is inappropriate 0.5 0.8 0.8 0.4 0.3 0.3 -0.1 -0.5 -0.5 Education is unproductive 1.8 1.3 1.4 0.5 0.9 0.9 -1.3 -0.4 -0.5 School too far 0.5 1.1 1.0 2.2 3.7 3.4 1.7 2.6 2.4 No teachers 1.6 2.0 1.9 1.7 2.5 2.4 0.1 0.5 0.5 No school 0.2 0.3 0.3 0.4 0.3 0.3 0.2 0.0 0.0 School closed 0.5 0.9 0.8 0.4 0.9 0.8 -0.1 0.0 0.0 Teacher is incompetent 1.0 1.3 1.3 0.4 2.3 2.0 -0.6 1.0 0.7 Others Studies too difficult 4.8 1.7 2.3 2.1 1.1 1.3 -2.7 -0.6 -1.0 Marriage 4.0 3.2 3.4 3.7 2.4 2.6 -0.3 -0.8 -0.8 Child won’t go 14.9 16.0 15.8 20.5 21.2 21.1 5.6 5.2 5.3 Autodidact 0.0 0.2 0.1 — — — — — — Physical handicap 0.4 0.4 0.4 0.5 1.0 0.9 0.1 0.6 0.5 Mental handicap — — — 0.4 0.7 0.7 — — — Language of instruction 0.1 0.3 0.3 0.0 0.0 0.0 -0.1 -0.3 -0.3 In charge of other family members 1.1 2.1 1.9 1.1 1.5 1.4 0.0 -0.6 -0.5 Food insecurity — — — 0.4 0.1 0.1 — — — Child too young 1.0 0.8 0.9 0.2 0.4 0.3 -0.8 -0.4 -0.6 Child too old 16.9 19.0 18.6 4.8 6.2 6.0 -12.1 -12.8 -12.6 No birth certificate 0.1 0.2 0.2 0.0 0.3 0.2 -0.1 0.1 0.0 General insecurity 0.2 0.2 0.2 0.3 0.1 0.1 0.1 -0.1 -0.1 Other reasons 1.3 1.8 1.7 2.0 2.1 2.1 0.7 0.3 0.4 Finished school 0.6 0.7 0.7 1.4 0.8 0.9 0.8 0.1 0.2 Pregnancy 0.9 0.8 0.8 2.8 1.6 1.8 1.9 0.8 1.0 Child dismissed — — — 0.6 0.4 0.4 — — — Abuse or harassment 0.7 0.7 0.7 0.3 0.0 0.1 -0.4 -0.7 -0.6 Child won’t repeat — — — 4.9 7.9 7.4 — — — No answer 3.9 4.2 4.1 4.4 3.5 3.6 0.5 -0.7 -0.5 Source: Household surveys, 2005 and 2010. 53 | Table A26: Frequency of Mention by Households of Reasons for Children Aged 6-17 Years Never Enrolling at School, by Area of Residence, 2005 and 2010 2005-10 Gap 2005 2010 (Percent) (% Points) Urban Rural Total Urban Rural Total Urban Rural Total Financial Child’s work 2.9 3.7 3.6 2.9 2.8 2.8 0.0 -0.9 -0.8 Child must work 6.0 9.2 8.8 2.5 3.2 3.2 -3.5 -6.0 -5.6 Financial problems 23.4 23.9 23.8 27.7 26.0 26.2 4.3 2.1 2.4 Quality Class content is inappropriate 0.3 0.7 0.7 0.1 0.0 0.0 -0.2 -0.7 -0.7 Education is unproductive 1.5 1.5 1.5 1.9 1.2 1.2 0.4 -0.3 -0.3 School too far 6.4 7.4 7.2 5.3 9.4 8.9 -1.1 2.0 1.7 No teachers 3.6 2.9 3.0 1.1 1.5 1.4 -2.5 -1.4 -1.6 No school 8.6 4.0 4.7 4.1 4.2 4.2 -4.5 0.2 -0.5 School closed 1.5 1.8 1.7 1.7 0.2 0.4 0.2 -1.6 -1.3 Teacher is incompetent — — — 0.0 0.2 0.2 — — — Others Studies too difficult 1.6 1.9 1.9 0.7 0.9 0.8 -0.9 -1.0 -1.1 Marriage 0.5 0.5 0.5 0.6 0.5 0.5 0.1 0.0 0.0 Child won’t go 9.3 9.8 9.7 13.1 13.8 13.7 3.8 4.0 4.0 Autodidact 0.2 0.0 0.0 0.3 0.1 0.1 0.1 0.1 0.1 Physical handicap 2.8 1.4 1.6 2.0 1.6 1.6 -0.8 0.2 0.0 Mental handicap — — — 2.2 1.3 1.4 — — — Language of instruction 0.1 0.4 0.3 0.1 0.1 0.1 0.0 -0.3 -0.2 In charge of other family members 1.2 2.3 2.1 0.2 1.6 1.5 -1.0 -0.7 -0.6 Food insecurity — — — 0.1 0.5 0.4 — — — Child too young 17.1 18.6 18.4 25.3 24.2 24.3 8.2 5.6 5.9 Child too old 2.3 1.5 1.6 0.9 0.8 0.8 -1.4 -0.7 -0.8 No birth certificate 1.6 3.5 3.2 3.3 3.8 3.7 1.7 0.3 0.5 General insecurity 0.5 0.4 0.4 0.0 0.1 0.0 -0.5 -0.3 -0.4 Other reasons 8.8 4.5 5.1 1.3 1.0 1.1 -7.5 -3.5 -4.0 No answer — — — 2.7 1.3 1.4 — — — Source: Household surveys, 2005 and 2010. Table A27: Enrollment Status of South Children, by Housing and Durable Goods Quintiles, 2012 (Tracer Study) (Percent) Out-of-school Enrolled Total Habitat and Comfort Index Quintile 1 32.1 67.9 100.0 Quintile 2 37.7 62.3 100.0 Quintile 3 20.0 80.0 100.0 Quintile 4 28.9 71.1 100.0 Quintile 5 14.9 85.1 100.0 Durable Goods Index Quintile 1 29.3 70.7 100.0 Quintile 2 36.0 64.0 100.0 Quintile 3 32.0 68.0 100.0 Quintile 4 29.2 70.8 100.0 Quintile 5 16.7 83.3 100.0 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. Note: Here the synthetic indexes are calculated through the principal components approach, normed and defined by quintiles. 54 | Table A28: Enrollment Status of Children Aged 6-10 Years, by Level of Education of Household Head and Area of Residence, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled Total No education 65.9 2.9 31.1 61.8 3.2 35.0 -4.1 0.3 3.9 Primary 81.5 3.4 15.2 79.1 3.3 17.6 -2.4 -0.1 2.4 Lower Sec. 91.1 1.8 7.1 88.5 1.5 10.0 -2.6 -0.3 2.9 Upper Sec.+ 96.5 1.1 2.4 92.5 1.1 6.3 -4.0 0.0 3.9 Urban No education 70.4 2.1 27.4 68.3 3.0 28.8 -2.1 0.9 1.4 Primary 86.0 2.2 11.9 83.8 2.3 13.9 -2.2 0.1 2.0 Lower Sec. 93.5 1.2 5.2 92.2 1.7 6.1 -1.3 0.5 0.9 Upper Sec.+ 97.3 0.9 1.8 96.5 0.4 3.1 -0.8 -0.5 1.3 Rural No education 65.1 3.1 31.8 60.8 3.3 35.9 -4.3 0.2 4.1 Primary 80.7 3.6 15.7 78.2 3.5 18.3 -2.5 -0.1 2.6 Lower Sec. 90.2 2.0 7.8 87.0 1.4 11.6 -3.2 -0.6 3.8 Upper Sec.+ 95.7 1.3 3.0 89.1 1.8 9.1 -6.6 0.5 6.1 Source: Household surveys, 2005 and 2010. Table A29: Enrollment Status of Children Aged 6-10 Years, by Parental Bond with Head of Household and Area of Residence, 2005 and 2010 2005-10 Gap 2005 2010 (% Points) (Percent) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled Total No 77.8 3.2 18.9 66.9 2.9 30.2 -10.9 -0.3 11.3 Yes 79.9 2.7 17.3 76.4 2.9 20.7 -3.5 0.2 3.4 Urban No 86.8 2.0 11.2 80.2 2.9 16.9 -6.6 0.9 5.7 Yes 85.8 1.7 12.5 84.4 1.8 13.7 -1.4 0.1 1.2 Rural No 75.1 3.6 21.2 63.2 2.9 33.9 -11.9 -0.7 12.7 Yes 78.6 3.0 18.4 74.7 3.1 22.2 -3.9 0.1 3.8 Source: Household surveys, 2005 and 2010. Note: “Yes” indicates that the child is the son/daughter of the head of household. 55 | Table A30: Enrollment/Out-of-School Rates in the South, by Education Level of Household Head, Father and Mother, 2012 (Tracer Study) (Percent) Out-of-school Enrolled Total Education Level of Household Head No education 33.7 66.3 100.0 Primary 29.1 70.9 100.0 Lower secondary 16.9 83.1 100.0 Upper secondary and above 6.7 93.3 100.0 Education Level of Father No education 32.9 67.1 100.0 Primary 25.8 74.2 100.0 Lower secondary 17.5 82.5 100.0 Upper secondary and above 6.1 93.9 100.0 Education Level of Mother No education 33.9 66.1 100.0 Primary 26.6 73.4 100.0 Lower secondary 6.2 93.8 100.0 Upper secondary and above 0.0 100.0 100.0 Source: South Survey in the Districts of Amboasary and Betioky, 2012. Table A31: Distribution of Children Aged 6-10 Years by Activity Status, Gender and Area of Residence, 2005 and 2010 Work - School Work - Chores - Work - Work Chores (Percent) Chores - Inactive only School School Chores only only School 2005 Girls 40.0 3.5 32.6 4.0 4.3 2.7 4.8 8.1 Boys 41.4 4.4 29.5 3.7 4.0 3.7 5.3 8.0 Urban 49.9 2.4 31.2 2.5 2.8 1.9 3.2 6.2 Rural 38.6 4.4 31.0 4.1 4.5 3.5 5.5 8.4 Total 40.7 4.0 31.0 3.8 4.2 3.2 5.1 8.0 2010 Girls 25.3 0.6 44.7 5.7 3.6 0.5 10.5 9.2 Boys 26.4 1.2 41.3 4.8 4.2 2.2 9.8 10.1 Urban 31.8 0.7 46.8 4.4 2.3 0.5 7.0 6.5 Rural 24.5 0.9 42.2 5.4 4.2 1.6 10.9 10.3 Total 25.8 0.9 43.0 5.3 3.9 1.4 10.2 9.6 2005-10 Gap (% Points) Girls -14.7 -2.9 12.1 1.7 -0.7 -2.2 5.7 1.1 Boys -15.0 -3.2 11.8 1.1 0.2 -1.5 4.5 2.1 Urban -18.1 -1.7 15.6 1.9 -0.5 -1.4 3.8 0.3 Rural -14.1 -3.5 11.2 1.3 -0.3 -1.9 5.4 1.9 Total -14.9 -3.1 12.0 1.5 -0.3 -1.8 5.1 1.6 Source: Household surveys, 2005 and 2010. 56 | Annex B: South Survey in the Districts of Amboasary and Betioky In 2009, the districts of Amboasary and Betioky underwent a preliminary survey in the context of an impact evaluation of school feeding programmes (often referred to as the South survey in this report). In the framework of Education for All (EFA), the Ministry of National Education (MEN), in collaboration with the World Food Programme, has indeed implemented since 2008 a series of actions to provide the pupils of schools from five districts in severe food insecurity in the south of Madagascar with nutritional support. In 2008-09, the MEN considered extending the coverage to the two districts of Amboasary and Betioky. An impact evaluation project, in partnership with the World Bank, was implemented. It was to include three survey phases, to be conducted over three school years. The political events of March 2009 and the political and economic crisis that followed entailed the suspension of the activities of most donors however, including the evaluation of the impact of school feeding programmes. The preliminary 2009 survey was nevertheless completed and the database provides detailed information on schools and household characteristics. This database is used as a pre-crisis baseline for the analysis of the data of these two districts. Sample of 299 schools, chosen randomly among those offering Grade 2, in both districts. In each school 3,401 pupils were randomly selected: five for each of Grade 1, Grade 3, Grade 4 and Grade 5 and ten for Grade 2. Finally, the households of six of the ten Grade 2 pupils selected were also surveyed. In April-May 2012 a new survey was carried out, whose main objective was to trace the education of the children surveyed in 2009, as well as the evolution of education supply in the aftermath of the crisis. Target sample of 155 schools (65 in Amboasary and 90 in Betioky). Two methodologies were adopted: (i) A survey of a new sample of 3,197 pupils, selected according to the same approach as in 2009 (Sample Selection); and (ii) The follow-up of 1,746 children surveyed in 2009 (Tracer Study). Furthermore, a sample of 778 households with a child in Grade 2 was surveyed, six per school, to collect information on children’s learning environments. Sample Selection The database of the sample selection includes 149 schools. Indeed, once in the field eight schools could not be reached due to insecurity, six were closed when the inspectors visited, four appear to not have been visited in 2009 and three were removed from the study database due to data inconsistencies in the records. Fifteen new replacement schools were therefore selected randomly from the 2009 sample, in the same areas (zones d’appui pédagogique - ZAP). 57 | Tracer Study For the tracer study of the panel of children, research focuses on the Grade 1, Grade 2 and Grade 3 pupils surveyed in 2009. For the Grade 2 sample, research was nevertheless limited to the children whose households had been surveyed in 2009. The base sample contains 2,131 pupils after the data cleansing of those that appear to have been fictitious in 2009. The surveyors were able to trace 1,746 children (attrition rate of 18.1 percent). Seventy five pupils could not be reached due to local insecurity. Surveyed in 2009 Surveyed in 2012 Attrition Rate (%) Grade 1 736 598 18.75 Grade 2 804 675 16.04 Grade 3 591 473 19.97 Total 2,131 1,746 18.07 Source: South Survey in the Districts of Amboasary and Betioky, 2009 and 2012. Note It should be remembered that these surveys are not representative of the national situation, or of those of the regions of Anosy and Atsimo-Andrefana to which the two districts belong. They provide a picture of the situation prevailing in these two districts, that are among the most vulnerable of Madagascar. 58 | Annex C: Adjustments to Household Survey Data, 2005 and 2010 The household surveys (Enquêtes permanentes auprès des ménages - EPM) carried out by the national statistical institute (Institut National de la Statistique - INSTAT) in 2005 and 2010 enable the analysis of the situation at the household level. The EPM are surveys on a broad sample of households distributed throughout the country and representative at the national level. Since 2005, the EPM are also representative at the regional level. The main objective of these surveys is to collect information on the living conditions of households. In addition to household characteristics, they supply detailed information on enrollment, but do not cover many education supply-side factors. The 2005 data collection exercise was held in October-November, whereas that of 2010 lasted four months, from June to October. For 2005, the database contains 11,780 households and a slightly more than 54,000 individuals, covering 561 of Madagascar’s local authorities. The 2010 sample on the other hand includes 12,460 households and 59,300 individuals, covering 623 local authorities. Some changes and adjustments were made to the data for the analysis carried out in this report. They relate to the coherence of children’s age and grade, rectifying outliers. An adjustment was also made for comparability purposes. Indeed, the 2005 survey sample was based on the national census of 1993, whereas in 2010 the random selection of 561 enumeration areas was performed on the basis of the third general population and housing census. An adjusted weight was thus constructed that takes the change in the sample basis into account. The information on self-consumption collected in 2005 and 2010 are also different, which has affected the comparability of the two databases, entailing that a certain number of extreme values were observed for 2005. These were corrected. Finally, regional and temporal deflators were used so that household consumption in 2005 and 2010 could be comparable in constant prices. It was firstly assumed that inflation is not uniform across different regions. On the other hand, the second assumption is that intra-regional inflation is uniform. For each year, regional deflators were constructed, the capital city being used as a benchmark. Secondly, a temporal deflator was introduced. This is defined as the ratio between the poverty line in 2010 and that of 2005. Price indexes are therefore relative to prices in the capital, reflecting regional disparities. Temporal Deflator = Poverty Line of 2010 / Poverty Line of 2005 = Ar 468,800 / Ar 305,344 = 1.5353 59 | Annex D: Econometric Analysis Methodology The econometric analyses of Chapter 3 are based on an extension of the decomposition model of Oaxaca and Blinder, to: 19 (i) analyze with simple Logits the significance and importance of the variables previously identified in the explanation of out-of-school for children aged 6 to 14 years; and (ii) determine which factors are correlated to the gaps in out-of-school for the respective periods. Then a multinomial Logit is used to explain the fact that children attend school, perform productive or domestic tasks, or a combination of these options. Regressions on Household Survey Data Firstly, the probability of being out of school is estimated for each year k (2005 and 2010). The probability of being out of school POOS is then defined as: POOS = f(Xik, βk) Where the dependant variable has the value 1 if the child i, with X characteristics doesn’t go to school, or the value 0 if not, and where β represents the vectors of the coefficients. Secondly, the standard deviation of the out-of-school rate for the period is disaggregated into: (i) a component explained by the variations in the observed characteristics, using a reference group; and (ii) a component explained by the variations in the coefficients of these characteristics (pertaining to unobserved characteristics). The decomposition of the gap ΔOOS is defined as: ΔOOS = [P(Xi2005, β2005) - P(Xi2010, β2005)] + [P(Xi2010, β2005) - P(Xi2010, β2010)] Where the first part of the equation represents the differences in the observed characteristics and where 2005 is used as the reference year, and the second part represents the variation in the coefficients of the characteristics (the effect of unobserved characteristics). Regressions on the South Survey Data Initially the same approach was adopted with the South survey data, the analyses being run on the children in Grade 2 (Sample Selection) to determine the probability of being out of school in both 2009 and 2012.20 For the analysis of the variations in the observed characteristics, 2009 is used as the reference year. 19 The original Oaxaca and Blinder decomposition model was only applied to linear regressions. An extension of the model applied to non-linear regressions was later developed by various authors (Yun, 2004; Fairlie, 2005; Bauer et al., 2008). 20 The South survey includes data on the household characteristics for some Grade 2 children (of the sample selection), including the enrollment status of other children in the household as well as information on school-level supply. These data are used to determine the factors that explain the drop in the enrollment of children aged 6 to 14 years belonging to these households between 2009 and 2012. 60 | Secondly, a binomial Logit model is used to analyze the data of the tracer study children, to determine which variables are correlated with dropout between 2009 and 2012 (for those children that were enrolled in 2009). The estimated model is defined as: Ai = βXi2009 + ρXHi2009 + λXSi2009 + ΔXH + ΔXS +ε Where: Xi represents the characteristics of the child i in 2009 (age, gender, rank among siblings, number of letters read per minute, performance of domestic, agricultural of herding duties); XH represents the characteristics of household H in 2009 (gender and education level of the household head, if climatic shock suffered, if loan taken out over four months, synthetic wealth indicator, cattle ownership); and XS represents the characteristics of school S in 2009 (share of FRAM teachers, number of classrooms in good condition, average number of pupils per class, availability of French textbooks, PTA contributions). Some variables were then introduced to seize the difference Δ between 2009 and 2012, in particular for household characteristics XH and school characteristics XS. 61 | 62 | Table D1: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years, 2005 and 2010 (EPM Data) Total Urban Rural 2010 2005 2010 2005 2010 2005 Child Characteristics Age 0.007 (0.01) 0.094*** (0.02) 0.017 (0.01) 0.090*** (0.02) 0.001 (0.01) 0.103*** (0.03) Boy 0.130** (0.04) -0.126 (0.09) 0.081 (0.07) -0.104 (0.13) 0.155** (0.06) -0.144 (0.14) Child of head of HH -0.340*** (0.09) -0.148 (0.19) -0.249 (0.14) -0.197 (0.26) -0.459*** (0.12) -0.180 (0.28) Rank among siblings -0.080* (0.03) -0.071 (0.07) -0.107 (0.06) 0.006 (0.10) -0.064 (0.05) -0.139 (0.10) Household Characteristics Female head of HH -0.130* (0.06) -0.138 (0.12) -0.226* (0.09) -0.102 (0.16) -0.027 (0.08) -0.207 (0.18) Education Level of Head of HH Primary -0.393*** (0.05) -0.223* (0.11) -0.288*** (0.08) -0.059 (0.16) -0.438*** (0.06) -0.363* (0.16) Lower Sec. -1.025*** (0.08) -0.908*** (0.15) -0.895*** (0.12) -0.644** (0.20) -1.046*** (0.10) -0.986*** (0.27) Upper Sec.+ -1.201*** (0.12) -1.669*** (0.23) -1.164*** (0.17) -1.510*** (0.29) -1.046*** (0.18) -1.432*** (0.40) Dependency ratio 0.054* (0.02) 0.100* (0.04) 0.059 (0.03) 0.027 (0.06) 0.044 (0.03) 0.134* (0.07) Suffered climatic shock 0.366*** (0.05) -0.007 (0.11) 0.197* (0.09) 0.180 (0.16) 0.400*** (0.06) -0.282 (0.17) Wealth Quintiles Q2 - Poor -0.472*** (0.06) -0.306* (0.14) -0.665*** (0.11) -0.299 (0.20) -0.379*** (0.08) -0.304 (0.21) Q3 - Average -0.610*** (0.07) -0.303* (0.15) -0.735*** (0.11) -0.452* (0.22) -0.525*** (0.08) -0.282 (0.22) Q4 - Wealthy -1.011*** (0.07) -0.520*** (0.16) -1.115*** (0.12) -0.752*** (0.22) -0.895*** (0.09) -0.547* (0.24) Q5 - Very wealthy -1.314*** (0.08) -0.705*** (0.17) -1.598*** (0.13) -1.287*** (0.26) -0.872*** (0.12) -0.303 (0.25) Non-repayable transfer -0.219*** (0.05) -0.172 (0.12) -0.268** (0.08) -0.297 (0.18) -0.282*** (0.06) -0.103 (0.18) Repayable transfer 0.045 (0.11) 0.515 (0.43) 0.190 (0.18) 1.142* (0.54) -0.023 (0.14) -0.579 (0.81) Other Characteristics Adult salary ratio 0.001 (0.00) -0.066*** (0.02) -0.013 (0.01) -0.022 (0.03) 0.000 (0.00) -0.089** (0.03) Primary PTR -0.001 (0.00) 0.007** (0.00) -0.001 (0.00) -0.007 (0.00) -0.001 (0.00) 0.006 (0.00) Constant -0.290* (0.14) -2.078*** (0.34) -0.422 (0.23) -1.518** (0.47) -0.073 (0.19) -1.361* (0.64) Observations 13,037 3,800 5,925 2,320 7,112 1,480 Pseudo R2 0.115 0.110 0.132 0.124 0.107 0.118 Decomposition (Reference group: 2005) Variation 0.090*** 0.065*** 0.089*** Effect of characteristics 0.020** 0.025*** -0.000 Child -0.016 -0.003 0.000 Household 0.278 0.019*** 0.000 Other -0.242 0.009 -0.000 Unobserved effect 0.068*** 0.056*** 0.062*** Child -0.125*** -0.104* -0.134** Household -0.044 -0.039 -0.023 Other -0.009 0.066 0.036 Interaction 0.002 -0.016 0.027 Child 0.000 0.002 -0.000 Household -0.000 -0.005 0.001 Other 0.002 -.0013* 0.026 Source: Authors’ calculations based on EPM, 2005 and 2010 data. Note: Logit regressions are based on the binomial: Does not go to school = 1, if not = 0. The reference groups are: No education for the education level of the head of household and Q1 - Very poor for the wealth quintiles. The adult salary ratio is the relation between the salary of an adult with primary education and that of an uneducated peer. The dependency ratio represents the relation within households of members under 15 or over 65 to adults aged 15 to 64 years. A regional variable is included in the other characteristics. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 63 | Table D2: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years in the South, 2009 and 2012 (Sample Selection) Logit: Does not attend school 2012 2009 (1 if Yes, 0 if No) Child Characteristics Age 0.009 (0.02) 0.564*** (0.08) Boy 0.718*** (0.11) 0.509 (0.29) Rank among siblings 0.027 (0.07) -0.008 (0.18) Household Characteristics Female head of HH -0.321 (0.18) -1.797* (0.76) Education Level of Head of HH (Ref. No education) Primary -0.071 (0.12) -0.626 (0.34) Lower secondary -0.695** (0.22) -1.491 (0.77) Upper secondary and above -1.227** (0.42) -0.166 (0.70) Household suffered climatic shock 0.116 (0.13) -0.928** (0.34) Household borrowed over last 4 months 0.017 (0.13) 0.078 (0.31) Synthetic wealth indicator (habitat and durable goods) -0.480*** (0.11) -0.120 (0.21) Cattle ownership 0.237* (0.12) -0.089 (0.33) School Characteristics Number of classrooms in good condition -0.010 (0.03) -0.056 (0.07) Share of FRAM teachers 0.001 (0.00) -0.011** (0.00) Number of pupils per class -0.006 (0.01) -0.033** (0.01) French textbook available 0.010 (0.12) 1.054* (0.48) PTA contributions -0.000 (0.00) 0.000** (0.00) Constant -1.814*** (0.35) -8.167*** (1.17) Observations 2,113 1,679 Decomposition (Reference group: 2009) Variation 0.183*** Effect of the observed characteristics 0.023 Child 0.000 Household 0.001 School 0.022 Unobserved effect 0.178*** Child -0.299*** Household 0.081** School 0.032 Interaction -0.018 Child -0.000 Household -0.000 School -0.000 Source: Authors’ calculations based on South survey data , 2009 and 2012. Note: The household synthetic wealth indicator is computed through the principal component analysis of household scores in terms of the housing and durable goods indexes. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 64 | Table D3: Multinomial Logit Regressions on the Activities of Children Aged 6-14 Years, 2010 (EPM Data) Multinomial Logit (Ref. School only) Total Urban Rural Work, Work, Work, Work + Work + Work + chores + Inactive chores + Inactive chores + Inactive chores chores chores school school school Age 0.216*** 0.273*** -0.200*** 0.234*** 0.306*** -0.183*** 0.206*** 0.258*** -0.219*** Boy -0.185*** -0.025 0.074 -0.158** -0.060 0.055 -0.220*** -0.024 0.076 Child of head -0.078 -0.639*** 0.309* -0.076 -0.519*** 0.168 -0.133 -0.821*** 0.345 of HH Rank among 0.075** 0.050 -0.222*** 0.049 0.004 -0.212** 0.118** 0.102 -0.222*** siblings Male head of -0.217*** -0.021 -0.061 -0.097 0.223* -0.057 -0.391*** -0.305** -0.162 HH Education Level of Head of HH (Ref. No education) Primary -0.250*** -0.633*** -0.395*** -0.325*** -0.543*** -0.487*** -0.174** -0.654*** -0.334*** Lower Sec. -0.199*** -1.217*** -1.050*** -0.299*** -1.152*** -1.055*** 0.002 -1.085*** -1.001*** Upper -0.626*** -1.834*** -1.083*** -0.565*** -1.721*** -1.232*** -0.624*** -1.645*** -1.007*** Sec.+ Dependency 0.135*** 0.167*** 0.086** 0.113*** 0.164*** 0.070 0.142*** 0.152*** 0.092 ratio Wealth Quintiles (Ref. Q1 – Very poor) Q2 -0.006 -0.454*** -0.471*** -0.058 -0.689*** -0.642*** 0.024 -0.342*** -0.384*** Q3 0.216*** -0.448*** -0.412*** 0.048 -0.672*** -0.621*** 0.312*** -0.293** -0.274* Q4 0.271*** -0.760*** -0.956*** 0.223 -0.891*** -0.936*** 0.264** -0.650*** -0.940*** Q5 0.201** -1.021*** -1.757*** 0.030 -1.465*** -1.718*** 0.472*** -0.302* -1.942*** Suffered a 0.085 0.397*** 0.370*** 0.007 0.190 0.139 0.137 0.414*** 0.465*** climatic shock Non-repayable 0.065 -0.139** -0.255*** 0.057 -0.218** -0.260* 0.065 -0.208** -0.295** transfer Repayable 0.685*** 0.654*** 0.249 0.818*** 0.761*** 0.896*** 0.589*** 0.571*** -0.088 transfer Salary ratio 0.005* 0.006** 0.005 -0.019** -0.038*** 0.000 0.014*** 0.014*** 0.014*** Average Primary PTR 0.002 0.001 -0.001 0.002 0.001 0.000 -0.004 -0.003 -0.009** (Local schools) Observations 13,037 5,925 7,112 Pseudo R2 0.133 0.148 0.146 Source: EPM, 2010. Note: Fixed region effects and a constant were included in all regressions. Standard deviations are corrected due to the auto-correlation issue between children of a same household. The adult salary ratio is the relation between the salary of an adult with primary education and that of an uneducated peer. The dependency ratio represents the relation within households of members under 15 or over 65 to adults aged 15 to 64 years. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 65 | Table D4: Multinomial Logit Regressions on the Activities of Children Aged 6-14 Years, 2005 (EPM Data) Multinomial Logit (Ref. School only) Total Urban Rural Work, Work, Work, Work + Work + Work + chores + Inactive chores + Inactive chores + Inactive chores chores chores school school school Age -0.350*** -0.377*** -0.172 -0.338*** -0.297* -0.283 -0.361*** -0.451** -0.234 Boy 0.115 -0.394 0.679** 0.205 -0.356 0.596 -0.216 -0.752** 0.911* Child of head of 0.019 -0.009 -0.192 -0.038 0.021 -0.153 0.101 -0.001 -0.241 HH Rank among siblings 0.312*** 0.383** 0.240 0.108 0.094 0.302 0.584*** 0.753*** 0.302 Male head of HH 0.014 -0.328** -0.047 -0.239 -0.244 -0.056 0.421** -0.293 0.215 Education Level of Head of HH du ménage (Ref. No education) Primary -0.073 -1.007*** -0.866*** -0.233 -0.744*** -0.776** 0.529** -0.767** -0.777 Lower Sec. -0.258** -2.219*** -0.951*** -0.433** -2.115*** -0.874* 0.129 -1.673*** -0.889 Upper Sec.+ 0.209*** 0.302*** 0.030 0.210*** 0.198** 0.020 0.328*** 0.461*** 0.001 Dependency ratio 0.309** -0.103 -0.314 0.564*** 0.031 -0.100 -0.000 -0.285 -0.206 Wealth Quintiles (Ref. Q1 – Very poor) Q2 0.200 -0.182 -0.316 0.341* -0.116 -0.563 -0.030 -0.431 0.146 Q3 0.335** -0.167 -0.942*** 0.972*** 0.042 -1.174*** -0.584** -0.968*** -0.629 Q4 0.388*** -0.349 -1.284*** 0.986*** -0.597* -1.557*** -0.406 -0.515 -0.740 Q5 -0.165* -0.365** 0.571*** -0.070 -0.123 0.793*** -0.309* -0.709*** -0.028 Suffered a climatic 0.112 -0.110 -0.073 0.107 -0.151 -0.279 0.143 -0.137 0.378 shock Non-repayable 0.028 0.670 -0.245 -0.365 0.913 1.248 1.020 0.254 -14.944 transfer Repayable transfer -0.033*** -0.074*** -0.104*** 0.031* 0.043 -0.162** -0.084*** -0.172*** -0.064 Salary ratio 0.004* 0.013*** 0.003 0.007* 0.007 -0.018** -0.004 0.003 0.006 Average Primary -0.350*** -0.377*** -0.172 -0.338*** -0.297* -0.283 -0.361*** -0.451** -0.234 PTR (Local schools) Observations 3,800 2,320 1,480 Pseudo R2 0.134 0.177 0.203 Source: EPM, 2005. Note: Fixed region effects and a constant were included in all regressions. Standard deviations are corrected due to the auto-correlation issue between children of a same household. The adult salary ratio is the relation between the salary of an adult with primary education and that of an uneducated peer. The dependency ratio represents the relation within households of members under 15 or over 65 to adults aged 15 to 64 years. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 66 | Table D5: Simple Logit Regressions on Dropout of Children Enrolled in Grade 2 in 2009, 2012 (Tracer Study) Logit (1 no longer enrolled in 2012, 0 if not) Child Characteristics Age (2009) 0.256*** (0.12) Boy (2009) -0.396 (0.36) Son/daughter of head of HH (2009) -0.599 (0.75) Dependency ratio (2009) -0.000 (0.23) Rank among siblings (2009) -0.293 (0.26) Performed agricultural, cattle-breeding or domestic activities (2009) -0.410 (0.39) Number of letters read per minute (2009) -0.011 (0.01) Household Characteristics Female head of HH (2009) 0.661 (0.48) Education Level of Head of HH (2009) (Ref. No education) Primary 0.385 (0.37) Secondary -1.038 (0.79) Suffered climatic shock over 2009-11 0.200 (0.40) Household borrowed over last 4 months (2009) 0.959** (0.38) Synthetic wealth indicator (habitat and durable goods) -0.074 (0.28) 2012-09 Gap -0.030 (0.29) Cattle Ownership (Ref. None) 1 to 5 heads 0.828* (0.43) 6 to 10 heads 0.010 (0.54) 11 to 20 heads 0.968 (0.96) Over 21 heads -0.676 (0.99) Gap in number of heads over 2009-11 0.000 (0.00) School Characteristics Number of classrooms in good condition (2009) -0.163 (0.16) Share of FRAM teachers (2009) 0.008 (0.01) 2012-09 Gap 0.014* (0.01) Average Pupil-class ratio (2009) -0.013 (0.02) 2012-09 Gap -0.015 (0.02) French textbook available(2009) 0.084 (0.54) 2012-09 Gap -0.688* (0.37) PTA contributions (2009) 0.885* (0.48) 2012-09 Gap 0.028 (0.40) Betioky (Ref. Amboasary) 0.615*** (0.36) Cattle theft in the fokontany (2012) 0.994* (0.55) Burglary in the fokontany (2012) -0.024 (0.44) Other insecurity issues in the fokontany (2012) 0.303 (0.94) Constant -6.687*** (1.96) Observations 443 Pseudo R2 0.171 Source: Authors’ calculations on the basis of South survey data, 2009 and 2012. Note: The household synthetic wealth indicator is computed through the principal component analysis of household scores in terms of the housing and durable goods indexes. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 67 | Table D6: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years with some variables interacted with income, 2005 and 2010 (EPM Data) (1) (2) (3) 2005 2010 2005 2010 2005 2010 Wealth Quintiles Q2 - Poor -0.336* (0.17) -0.450*** (0.07) -0.149 (0.18) -0.468*** (0.08) -0.270 (0.15) -0.447*** (0.07) Q3 - Average -0.307 (0.17) -0.572*** (0.07) 0.023 (0.19) -0.670*** (0.08) -0.321 (0.17) -0.602*** (0.08) Q4 - Wealthy -0.438* (0.18) -0.964*** (0.08) -0.251 (0.19) -1.066*** (0.09) -0.526** (0.17) -1.047*** (0.09) Q5 - Very wealthy -0.703*** (0.19) -1.229*** (0.09) -0.617** (0.21) -1.450*** (0.10) -0.575** (0.19) -1.373*** (0.11) Female head of HH -0.104 (0.22) -0.040 (0.10) Female head*Q2 0.150 (0.32) -0.046 (0.15) Female head*Q3 0.068 (0.32) -0.125 (0.17) Female head*Q4 -0.412 (0.35) -0.160 (0.18) Female head*Q5 0.001 (0.36) -0.343 (0.22) Suffered climatic shock 0.472* (0.21) 0.244** (0.09) Shock*Q2 -0.420 (0.28) -0.016 (0.12) Shock*Q3 -0.872** (0.30) 0.174 (0.13) Shock*Q4 -0.790* (0.31) 0.182 (0.15) Shock*Q5 -0.239 (0.32) 0.533** (0.17) Non-repayable transfer -0.033 (0.28) -0.285** (0.10) Transfer*Q2 -0.212 (0.38) -0.023 (0.13) Transfer*Q3 0.069 (0.37) 0.041 (0.14) Transfer*Q4 -0.034 (0.37) 0.170 (0.15) Transfer*Q5 -0.505 (0.39) 0.242 (0.16) Observations 3827 13068 3827 13068 3827 13068 Pseudo R2 0.110 0.119 0.112 0.119 0.110 0.119 Decomposition (Reference group: 2005) Variation 0.091*** 0.091*** 0.091*** Effect of characteristics 0.020** 0.023*** 0.021** Unobserved effect 0.068*** 0.071*** 0.069*** Interaction 0.003 -0.003 0.002 Source: Authors’ calculations based on EPM, 2005 and 2010 data. Note: Logit regressions are based on the binomial: Does not go to school = 1, if not = 0. The reference groups are: Q1 - Very poor for the wealth quintiles. All variables – child, household and others characteristics are included in each regression but not shown. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 68 Table D7: Simple Logit Regressions on Out-of-School Children Aged 6-14 Years in the South with some variables interacted with income, 2009 and 2012 (Sample Selection) (1) (2) (3) 2009 2010 2009 2010 2009 2010 Synthetic wealth indicator -0.101 (0.22) -0.476*** (0.11) -0.732 (0.40) -0.362* (0.17) -0.562 (0.31) -0.377** (0.12) Female head of HH -2.174* (1.09) -0.334 (0.20) Female head*synthetic wealth -1.066 (1.56) -0.035 (0.27) Suffered climatic shock -0.981** (0.34) 0.065 (0.14) Shock*synthetic wealth 0.848 (0.44) -0.179 (0.19) Borrowed over last 4 months 0.102 (0.33) -0.187 (0.17) Credit*synthetic weatlh 0.921* (0.38) -0.517* (0.25) Observations 1679 2113 1679 2113 1679 2113 Pseudo R2 0.254 0.079 0.262 0.080 0.266 0.081 Decomposition (Reference group: 2009) Variation 0.183*** 0.183*** 0.183*** Effect of the observed 0.023 0.027 0.021 characteristics Unobserved effect 0.178*** 0.178*** 0.174*** Interaction -0.018 -0.023 -0.012 Source: Authors’ calculations based on South survey data , 2009 and 2012. Note: The household synthetic wealth indicator is computed through the principal component analysis of household scores in terms of the housing and durable goods indexes. All other variables – child, households and others characteristics are included but not shown. The significance levels are: * p<0.1, ** p<0.05, *** p<0.01. 69 | Annex E: Schooling Age Primary school children are slightly younger since the crisis, both when they access Grade 1, and on average. The education statistical yearbooks show that between 2008/09 and 2010/11, the average age of primary access (Grade 1 pupils) decreased slightly, from 6.85 years to 6.77 years on average, a difference of about a month (See Table E1). On the other hand, Grade 5 pupils are younger by about four months. Thus the average age of primary school pupils has dropped by almost two months. Table E1: Average Age of Primary School Pupils, by Grade, 2008/09-2010/11 (Years) Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Total 2008/09 6.85 8.44 9.83 10.99 12.13 9.02 2009/10 6.81 8.43 9.80 10.89 11.88 8.91 2010/11 6.77 8.36 9.75 10.84 11.79 8.86 Source: Education statistical yearbooks, 2000-11. The household survey results confirm this trend. According to the surveys, the average age of first primary enrollment may have dropped by close to a year between 2005 and 2010, from 7.3 years to 6.3 years (See Table E2). This decline is more pronounced in rural areas and for boys, but the lowest age of first primary enrollment is nevertheless found in urban areas. These results can be explained by the fact that some children who previously entered primary late are now enrolling earlier, or by the opposite fact that they are more numerous to never access school. Children who drop out of school are slightly older in 2010 than in 2005 (See Table E3), whereas the average age of those having never been to school is slightly younger, especially for girls. Table E2: Average School Access Age, by Area of Residence and Gender, 2005 and 2010 (Years) 2005 2010 2005-10 Gap (% Points) Urban 6.7 6.0 -0.8 Rural 7.5 6.4 -1.1 Girls 7.2 6.3 -0.9 Boys 7.4 6.3 -1.2 Total 7.3 6.3 -1.0 Source: Household surveys, 2005 and 2010. Table E3: Average Age of Primary School-Aged Children, by Enrollment Status, 2005 and 2010 2005 2010 2005-10 Gap (% Points) (Years) Dropped Never Dropped Never Dropped Never Enrolled Enrolled Enrolled out enrolled out enrolled out enrolled Girls 9.06 11.37 8.964 9.191 11.83 8.766 0.131 0.46 -0.198 Boys 9.27 11.39 8.903 9.357 12.27 8.836 0.087 0.88 -0.067 Total 9.17 11.38 8.932 9.275 12.05 8.804 0.105 0.67 -0.128 Source: Household surveys, 2005 and 2010. Note: The average ages for dropout and children having never enrolled are computed for the 6-14 years age group. 70 Bibliography Bauer, T. K, M. Hahn and M. Sinning. 2008. “The Blinder-Oaxaca Decomposition for Nonlinear Regression Models.” The Stata Journal, Vol. 8, No.4, pp. 480-492. Becker, G. S. 1965. “A Theory of the Allocation of Time.” Economic Journal, Vol. 75, No. 3, pp. 493-517. Fairlie, R. W. 2005. “An extension of the Blinder-Oaxaca Decomposition Technique to Logit and Probit Models.” Journal of Economic and Social Measurement, No. 30, pp. 305-316. Ferreira, F. H. G. and N. Schady. 2008. Aggregate Economic Shocks, Child Schooling and Child Health. World Bank Policy Research Working Paper No. 4701. INSTAT (Institut National de la Statistique de Madagascar). 2006. Enquête périodique auprès des ménages 2005: Rapport principal. Household Statistics Department. Antananarivo, Madagascar. INSTAT (Institut National de la Statistique de Madagascar). 2010. Le marché du travail dans l’agglomération d’Antananarivo en 2010: une mise en perspective décennale. Policy Brief. In collaboration with Institut de Recherche pour le Développement (IRD), Développement Institutions & Analyses de Long terme (DIAL) and Université de Paris Dauphine. INSTAT (Institut National de la Statistique de Madagascar). 2011. Enquête périodique auprès des ménages 2010: Rapport principal. Household Statistics Department. Antananarivo, Madagascar: Pal Prod. Jacquet, L., P. Runner and S. Menard. 2012. Évaluation de l´appui budgétaire and revue de la gestion des finances publiques and des secteurs santé and éducation - Madagascar: Actualisation de la gestion des finances publiques, du cadre macro-économique and des analyses sectorielles santé and éducation, revue des critères d’éligibilité. Rapport phase 2. Rapport final. June. McKenzie, D. 2003. “The Consumer Response to the Mexican Peso Crisis.” Economic Development and Cultural Change, Vol. 55, No. 1, pp. 139-172. McRam III (UN Multi-Cluster Rapid Assessment Mechanism). 2010. Évolution de la Situation Socioéconomique des Ménages de la Ville d'Antananarivo (Madagascar) durant la Crise Sociopolitique – Novembre 2009. PIE (Plan Intérimaire de l’Éducation). 2013. Plan Intérimaire de l’Éducation 2013-15. Ministry of National Education. PASEC (Programme d’analyse des systèmes éducatifs de la CONFEMEN). 2005. PASEC (Programme d’analyse des systèmes éducatifs de la CONFEMEN). 1998. Samuelson, P. 1937. “A Note on the Measurement of Utility.” The Review of Economic Studies, Vol. 4, No. 2, pp. 155-161. 71 Schady, N. 2004. “Do Macroeconomic Crises Always Slow Human Capital Accumulation?” World Bank Economic Review, Vol. 18, No. 2, pp. 131-154. UNDP (United Nations Development Programme). 2011. Rapport sur le développement humain 2011. UNICEF (United Nations Children’s Fund). 2011. Analyse de la pauvreté des enfants to Madagascar. World Bank. 2012. Madagascar Economic Update, 2012. World Bank Madagascar Country Office. October. Yun, M-S. 2004. “Decomposing differences in the first moment.” Economic Letters, No. 82, pp. 275-280. 72 |