Summer School on Multidimensional Poverty Analysis

Transcription

1 Summer School on Multidimensional Poverty Analysis Oxford Poverty & Human Development Initiative, (OPHI), University of Oxford 3 15 August 2015, Georgetown University Washington DC

2 National Multidimensional Poverty Measures Sabina Alkire, Paola Ballon, Adriana Conconi, Christoph Jindra and Gisela Robles A. 5 August 2015 Session III

3 MPI: Two kinds ~ both useful Internationally comparable: - Like $1.25/day poverty measures - Can compare regions, subnational, groups, time - Could track SDG-1: poverty in its many dimensions - Could measure both acute and moderate poverty

4 MPI: Two kinds ~ both useful National MPIs: - Reflects National Priorities - Vital for policy - Measure to Manage ~ Target, Coordinate, M&E

5 MPI: Two kinds ~ both useful Internationally comparable: Example: The Global MPI estimated and analysed by OPHI and published by UNDP s HDRO can be compared across 110 countries. Facilitates lessons learned across countries. - Like $1.25/day and $2/day poverty measures & MDGs - Useful for policy analysis, but limited national ownership Context-Specific: Example: National MPIs reflect national contexts and priorities. They guide policies like targeting and allocation and monitor changes - Like National income poverty measures - Useful for policy but can t be compared internationally

6 Normative Decisions Purpose Unit of Analysis Dimensions Indicators Deprivation Cutoffs Weights Poverty cutoff

7 MPI in Action Official National MPIs Colombia Mexico Bhutan Chile Other national applications underway.

8 - 4 key aspects: 1. Reflects the objectives of social policy 2. Coordinates public policy sectors 3. Monitors public policy 4. Informs Decision making: 1. Geographic targeting 2. Programme composition 3. Graduation from CCTs

9 Our starting point: Improving the instruments and methodologies of poverty measurement Motivation: Designing a strategy for the reduction of poverty and inequality based on a complete approach using income and multidimensional measures

10 Colombia s Multidimensional Poverty Index (MPI) Education Childhood & youth conditions Labor Health Public utilities & housing conditions

11 Colombia 2011, 2013: MPI-Colombia Educational Conditions Childhood & Youth Work Health Housing & Public Services Schooling Illiteracy 0.1 Used to allocate resources in national development plan School Attendance At the right level Access to infant services No Child Labour Absence of long-term unemployment Formal work Coverage Access to health care given a necessity Improved Water Sanitation Flooring Exterior Walls Overcrowdin g

12 Selection of Indicators (Variables) Colombia s MPI Criteria for variable selection Frequent usage (national or international); literature review; discussion with experts; other indicators. IPM-OPHI Internacional, NBI, ICV y Sisbén III. 1. Indicators can be affected by public policies. 2. Availability of information (in the survey of Quality of Life in Colombia). Criteria to validate variables Precision of the sample to estimate the variable estimated coeff of variation <15%. *EL DANE utiliza: 0-7: Estimación precisa 8-14: precisión aceptable ó 15-25: Precisión regular y por lo tanto se debe utilizar con precaución

13 Poverty Cutoff Colombia. The number of MPI deprivations experienced by those who were income poor, and those who perceived themselves to be poor, was compared with the number of deprivations among the non-income and non-subjective poor. Median and Average number of deprivations 2008 Median Average People who perceive themselves to be poor Income poor people Income poor people who perceive self as poor Those who don t perceive themselves as poor Those who are not income poor All people Fuente: Cálculos DNP-SPSCV, con datos de la ECV2008 A non-poor person on average has 3 deprivations, which suggests that a low value of k would capture deprivations that were not related to or sufficient to identify poverty.

14 Pobreza monetaria We are on are the right track Sustained poverty reduction , Pobreza Multidimensioal I P M 0 Pobreza extrema monetaria , Fuente: DNP SPSCV 4 0

15 From 2010 to 2012, 1.3 million persons overcame multidimensional poverty

16 To what do we owe these good results?

17 High Official Commission Monitoring the national strategy for the reduction of poverty using official poverty measures Leaders Presidency (Mandatory presence of the President of Colombia) Department for Social Prosperity National Planning Department Permanent members Ministry of Health Ministry of Labour Ministry of Housing Ministry of Agriculture Ministry of Education Ministry of Finance 18

18 Sectoral goals For accomplishing the strategy 0%-10% avance 10%-25% avance >25% avance Pobreza Línea Base PND 2008 Dato 2011 Dato 2012 Análisis Goal MPI (Multidimensional Poverty) 34.7% 29.4% 27.0% 22.5% A (1) B (2) C (3) D (4) E (5) Educational achievement ( 15 yrs) Literacy ( 15 yrs) School attendance (6-16) No school lag (7-17) Access to child care services (0-5) Children not working (12-17) Long-term unemployment Formal employment Health insurance Access to health services Access to water source Adequate sewage system Adequate floors Adequate external walls No critical overcrowding *** Change est. significant 58.8% 54.6% 53.1% 52.8% 14.2% 12.0% 12.1% 12.0% 5.4% 4.8% 4.1% 3.5% Pobreza 33.4% 34.1% 33.3% 33.1% 12.1% 10.8% 9.4% 10.6% 5.5% 4.5% 3.7% 2.9% 9.6% 9.1% 10.0% 9.3% 80.6% 80.4% 80.0% 74.7% 24.2% 19.0% 17.9% 0.5% 8.9% 8.2% 6.6% 2.4% 12.9% 12.0% 12.3% 10.9% 14.1% 14.5% 12.1% 11.3% 7.5% 6.3% 5.9% 5.6% 3.1% 3.2% 2.2% 2.1% 15.7% 14.2% 13.1% 8.4% *** *** *** *** *** *** *** *** 19

19 NDP Alerts: -Habitability -School achievement -Early childhood care -Income generation

20 NDP Policies: Free tuition fees in primary and secondary education National strategy for childhood care De cero a Siempre Alerts: -Habitability -School achievement -Early childhood care -Income generation

21 NDP Alerts: -Habitability -Lower rhythm of multidimensional poverty reduction

22 NDP policies: 100,000 free housing solutions for poor households Alerts: -Habitability -Lower rhythm of multidimensional poverty reduction -Low budget execution?

23 NDP Alerts: -Lower rhythm of income poverty reduction - Gaps of rural to urban income poverty maintained

24 NDP policies Improved CCT program Más Familias en Acción (Higher amount and beneficiaries on rural areas) Food Security Programa: Colombia Growyh Without Hunger Alerts: -Lower rhythm of income poverty reduction - Gaps of rural to urban income poverty maintained

25 NDP Alerts: -Social mobility -Youth unemployment

26 NDP policies CCT to increase human capital and youth employment Jóvenes en Acción Alerts: -Social mobility -Youth unemployment

27 Initial distribution of beneficiaries Re targeting program s beneficiaries: approximating to municipal MPI National - MPI

28 New distribution of beneficiaries Re targeting program s beneficiaries: approximating to municipal MPI National - MPI

29 - 4 key aspects: 1. Reflects the objectives of social policy 2. Coordinates public policy sectors 3. Monitors public policy 4. Informs Decision making: 1. Geographic targeting 2. Programme composition 3. Graduation from CCTs

30 Multidimensional Poverty in Mexico Adapted from Hernandez-Licona (2015) and Enrique Gonzalez Tiburcio (2015) available at the Coneval and MPPN websites

31 Normative decisions Purpose: To implement the 2004 General Law for Social Development 2009 first multidimensional estimations. Dimensions and indicators established in the 2004 General Law. Individual as unit of analysis. Old survey (ENIGH) adapted with new module for indicators.

32

33 Urbano=$1, Rural=$868.25

34

35

36

37 55.3 millones 46.2% 2.3 carencias NOTA: Los valores de la línea de bienestar mínimo y de la línea de bienestar económico corresponden a agosto Fuente: estimaciones del CONEVAL con base en el MCS-ENIGH 2014

38 Porcentaje Fuente: estimaciones del CONEVAL con base en el MCS-ENIGH de 2010 a 2014

39 Porcentaje

40 Porcentaje Fuente: estimaciones del CONEVAL con base en el MCS-ENIGH de 2010 a 2014

41

42

43 Porcentaje de población con carencia por acceso a la alimentación I II III IV V VI VII VIII IX X Deciles de ingreso Fuente: estimaciones del CONEVAL con base en el MCS-ENIGH 2012 y 2014

44 Adapted from National Statistics Bureau (2014) Bhutan Multidimensional Poverty Index, Thimphu: NSB.

45 Context Development outlook based on multidimensional Gross National Happiness Index First in domains, 33 indicators (2010) Official poverty estimates since 2000 Consumption-based poverty estimate in 2003, 2007, 2012 Wide recognition that poverty, like wellbeing is multidimensional

46 Normative Decision Purpose Targeting the poor Yardstick to measure progress To complement income poverty measure Compare districts in terms of multidimensional poverty to target resources and policies

47 Normative Decision Unit of Analysis: Household Purpose of measure as a tool to target poverty alleviation policies that were at the level of the household was key

48 Data Sources Data Source: Bhutan Living Standard Survey (BLSS), 2012 Past MPI calculations have used: BMIS (2010) BLSS (2007) All datasets representative at the district level and by rural and urban areas

49

50 Poverty cutoffs and MPI results 2012 Poverty cutoff: approximately one third of weighted indicators (k=4/13 or 30.7% of wt. dimensions) 2nd poverty cutoff: intensely poor if a person deprived in half or more of weighted indicators (k=7/13)

51 Censored headcount ratios & contributions to MPI by indicator

52 There is little overlap between income and MD poor

53 MD and income poverty by regions

54 Development of poverty National MPI and National Headcount Ratio for different values of k in 2007 and 2012 Comparison based on BLSS 2007 and 2012 data

55 Policy Use From the 11 th 5-year plan, using MPI as one of the criteria in resource allocation to local governments Setting poverty eradication target to reduce MPI poverty (HC ratio) to below 10% by 2018

56 Chile Paola Ballon, OPHI Session III, 5 August 2015

57 A country effort Actors: Background 1. Advisory committee: Ministry of Sustainable Development and National Institute of Statistics (technical report). 2. Technical assistance from OPHI and CEPAL. 3. Panel of experts: Validated the technical report in August

58 Indicators Measurement design Education (25%) Attendance Health (25%) Malnutrition in children Dimensions Labor & Social Security (25%) Occupation Housing (25%) Overcrowding School Backwardness Health insurance Social Security Housing conditions Schooling Health care Pension Basic Services CASEN Fuente: Ministerio de Desarrollo Social

59 Measurement Design Dimension Indicator Deprivation Cutoff Weights A hh is identified as deprived if at least one of its members.. Education Attendance is not attending an educational institution. People 4-18 years of age (and 6-26 people with disabilities) School Backwardness Schooling is delayed 2 or more years in their studies of basic or secondary education. Students of primary and secondary education. has achieved fewer years of education than those established by law (according to age). Persons 18 and older. Fuente: Ministerio de Desarrollo Social, Encuesta Casen

60 Measurement Design Dimension Indicator Deprivation Cutoff Weights A hh is identified as deprived if at least one of its members Health Malnutrition in children Health insurance Health care is malnourished, at risk of malnutrition, overweight or obese. Children / as of 0-6 years. is not affiliated to a health insurance system and has no other supplementary insurance. The entire population. did not receive health care in the last 3 months or had no coverage BOOM-GES system. People needing medical care or being treated for arantizada pathology (BOOM-GES) Fuente: Ministerio de Desarrollo Social, Encuesta Casen

61 Measurement Design Dimension Indicator Deprivation Cutoff Weights Labor and Occupation is unemployed. Persons 18 years of age and over social not listed on the pension system. People aged 15 Social Security security and over who are occupied Pension does not perceive a contributory or noncontributory pension. People in retirement age A hh is deprived in each housing indicator if.. Housing Overcrowding Housing conditions Basic services It is crowded (2.5 or more people per bedroom for the exclusive use). Living in poor housing or in a home with walls, ceiling and / or floor in disrepair. Living in a house without basic sanitation (WC, key in the housing and urban or rural water as standard) Fuente: Ministerio de Desarrollo Social, Encuesta Casen

62 Main results Adjusted headcount ratio M0 (k >=25%) 6,7 5,7 4, Unit of analysis: household Fuente: Ministerio de Desarrollo Social, Encuesta Casen

63 Main results Percentage of people living in multidimensional poverty ( ) 27, ,3 20,4 % Fuente: Ministerio de Desarrollo Social, Encuesta Casen

64 Dimensional break-down ( ) ,8 12,2 32,5 27, ,3 11,4 31,2 30, ,4 13,5 32,4 27,8 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% % Education Health Labor Housing Fuente: Ministerio de Desarrollo Social, Encuesta Casen

65 Percentage of households living in monetary and multidimensional poverty ( ) % ,8 8,9 19,7 7,2 12,8 10 4,4 5 13,9 12,6 8,4 Households living in monetary poverty Households living in monetary poverty and in multidimensional poverty Fuente: Ministerio de Desarrollo Social, Encuesta Casen

66 Percentage of households living in MD by income quintile ( ) 40 39, ,5 30 % ,3 29,2 25,8 21,4 21,3 19, ,3 13,6 12,2 9,8 7,3 5 4,5 3,6 0 I II III IV V

67 Chile:

68 Percentage of people living in multidimensional poverty, by region % ,8 20,7 26,2 21,6 18,0 ( ) 28,5 26,1 21,1 21,9 22,4 22,3 18,0 22,9 23,6 20, , Fuente: Ministerio de Desarrollo Social, Encuesta Casen

69 One method: different purposes They: Draw on censored matrix Share a Alkire Foster class of measures, specifically, the M0, and they fulfil the properties or axioms acknowledged. Some countries report H, A and M0, others show only H Decomposable by dimension/indicator, population groups and changes in time. Mix continuous and ordinal data National scope, but subnational and group decomposition Unit of analysis: household Weights: different weights schemes, adapting them to national needs.

70 Dimensions Mexico Colombia Bhutan Chile Reflect social rights and income for food basket Reflect goals of the National Development Plan MPI dimensions + access to food, roads, wealth Reflect goals of the National Development Plan Indicators Legal norms that reflect rights Literature; experts and policy makers; availability within LSMS; precision of measure 8 MPI indicators + access to food, roads, land, livestock and housing Additional indicators: school backwardness, health insurance, health care, occupation, social security and pension, overcrowding... Sources of data Existing income survey with new module Existing LSMS Existing LSMS style survey Existing LSMS type survey (CASEN) Thresholds Legal criteria Expert criteria Triangulation and Robustness MPI adapted to local reality Legal criteria, expert criteria Institutions involved Coneval+ Inclusive cabinet + monitoring Poverty committee + monitoring GNH Commission to identify areas for intervention Ministry of Sustainable Development, National Institute of Statistics. Technical assistance provided by OPHI and Cepal.

71 Basic steps to create a MPI

72 Creating a MPI Creation of MPI requires multiple steps and coordination of multiple actors 3 areas: technical correctness, political feasibility, administrative feasibility In general, technical team presents technically rigorous options for a MPI to a political committee First step: Determine purpose and institutional issues: debate with different actors Normative decisions that guide process: dimensions? weights? (often based in legislation, national plans, etc.) 74

73 Creating a MPI 3 main actions, once the normative decisions have been made: Explore potential indicators Create trial measures Analyze trial measures 75

74 1. Explore potential indicators Objective: emphasis in components of the measure, specifically the indicators of each dimension. Understand which information is being added to the index and how it will be possible to disaggregate the information on the MPI. There are different ways to choose/construct indicators, even when the normative decisions are very clear. Steps: Create universe of indicators: consider large set of available indicators (binary 0/1) For each available indicator on the database, create different specifications (e.g. read, write, read&write, read or write) 76

75 1. Explore potential indicators Examples When individual info is aggregated to create indicator at hh level No member, every member, members aged a-b, x% of hh, every woman in the hh, etc. Test different deprivation cut-offs Result: set of available indicators on the data, to be contrasted against normative decisions Product: table with different indicators and proportion of people deprived in each of them. 77

76 1. Explore potential indicators Additional considerations: Understand which is the applicable population (e.g. nutritional info & vaccinations only available for children under 5) Compute missing values among applicable population for each indicator (limit of 15%, for instance) Atenttion when coding: only consider applicable population (e.g. school attendance only for school-aged children) Product: Include in the table column indicating applicable population and missing values Note: applicable population and missing values important to determine weights, so not to overestimate the incidence of a particular deprivation (e.g. vaccinations for children 0-2: applicable population is small % of total population, lower weight??) 78

77 1. Explore potential indicators Additional considerations: Understand association/redundancy among indicators Results must be consider jointly with the normative decisions, the timing in which each deprivation happens and policy priorities Generally, empirical tests are used as source of information but the decision of dropping an indicators is not directly derived from them Product: table with different indicators, the deprivation rates, and indices of correlation, redundancy and association 79

78 2. Create trial measures Objective: assign pre-selected indicators to each dimension, set weights and compute several MPIs, in order to find a final MPI that works and is robust. Assign indicators to dimensions Political considerations: based on legislation, national plan, participatory process, etc. (e.g. water can be a health indicator or a living standard indicator) Technical considerations (weights): for example, if weights are pre-set normatively 80

79 2. Create trial measures Compute trial measures: several adjustments are possible Take individual info to hh level Test different deprivations cut-offs for each indicator Test different weighting structures Test different poverty lines (k). In general, countries report estimations for at least two values of k 81

80 3. Analyze trial measures Objective: compute H, A and MPI for each trial measure, rates (censured and uncensored) and contribution of each indicator. When comparing measures is important to remember the purpose of the MPI Disaggregate by regions, ethnic groups, gender, age groups, etc. Compare trends with monetary income results Don t let the level of H, A and MPI determine the decision of which measure to use Product: table with H, A, MPI, rate, contributions for each trial measure, and break-downs. Also, associated figures. 82

81 3. Analyze trial measures This first set of trial measure gives place to debate and discussion by different relevant actors (experts, political committee, etc.) Next step is adjust measures based on their suggestions and feedback and recompute (sequential process) Prepare non-technical document explaining measure (and each step that led to it) This can be done relatively fast in turn political process can take significant amount of time 83

82 Communication of the MPI throughout the process It is crucial to communicate and be transparent during the whole process of creating a MPI After creating universe of indicators with deprivation rates and missing values, communication is useful to guide team in which ones to keep and which to drop After creating trial measures, communication is essential to check and legitimize 84