Multidimensional Poverty Comparisons: The Issue of Inequality © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 1 Outline 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 1. Introduction 2. The Identification of the Poor 3. The Aggregation of the characteristics of the poor 3.1 Core Axioms 3.1 The Core Axioms 3.2 A new Inequality Axiom 3.3 Two Unique Index Classes 3.4 A Numerical Example 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. 4. Summary 5. © 2011 d·i·e Empirical Application 4.1 Poverty Rates and Country Rankings 4.2 Decomposition Summary Nicole Rippin: Multidimensional Poverty Comparisons 2 Introduction: How should poverty be measured? 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes Multidimensional poverty measures are needed, not necessarily to replace but to amend traditional income poverty measures (e.g. Rawls 1971, Sen 1985 & 1992, Drèze and Sen 1989, UNDP 1997) Different approaches exist to derive multidimensional poverty indices (e.g. fuzzy set approach, distance function approach, information theory approach, axiomatic approach) 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Four main papers dealt with the extension of the one-dimensional axiomatic approach to a multidimensional framework (Chakravarty, Mukherjee & Ranade 1998, Tsui 2002, Bourguignon & Chakravarty 2003, Chakravarty & Silber 2008) Nicole Rippin: Multidimensional Poverty Comparisons 3 The framework 0. Outline i = 1,…,n individuals 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms j = 1,…,k attributes z Z ; Z Rk threshold levels 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings xi Rk k-row vector of attributes of individual i Individual i is deprived with respect to attribute j if xij z j 4.2 Decomposition 5. Summary © 2011 d·i·e S j ( X ) is the set of individuals deprived with respect to attribute j Nicole Rippin: Multidimensional Poverty Comparisons 4 The Identification of the Poor 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms The identification of the poor: Common Methods: Union Method: i is poor if j 1,2,..., k : xij z j Intersection Method: i is poor if xij z j j 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. However, for additive poverty measures both methods are practically inapplicable, leading to either exaggerated or minimised poverty rates Summary In response, Alkire and Foster (2009) introduced a new identification method: Dual Cut-off Method: i is poor if x ij z j for j 1,2,..., k and # j d © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 5 Weaknesses of the Dual Cut-off Method 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms The dual cut-off method implies several weaknesses: Arbitrariness: The choice of the cut-off lacks theoretical foundation but affects country rankings Discontinuity: The cut-off leads to a hardly justifiable discontinuity in the identification of the poor: households suffering deprivations just below the cut-off receive a weight of zero whereas households just above the cut-off enter the index with full weight Laborious: The dual cut-off method renders calculations cumbersome since poverty measures have to be calculated k times 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 6 The Core Axioms 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms Anonymity (AN): For any ( X ; z ) K Z : P ( X ; z ) P( X ; z ) where is any permutation matrix of appropriate order Any personal characteristics apart from endowments are irrelevant for poverty measurement Continuity (CN): For any z Z , P is continuous on K 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary Monotonicity (MN): For any ( X ; z ); (Y ; z ) K Z if: i) for any h, xhl yhl , where yhl zl , 0, ii) xil yil i h, xij yij j l , i, then P (Y ; z ) P ( X ; z ). The poverty index does not increase if, ceteris paribus, a poor individual experiences an endowment increase in a dimension in which it is deprived m Principle of Population (PP): For any ( X ; z ) K Z ; m N : P( X ; z ) P( X ; z ) where X m is the m-fold replication of X The poverty index does not depend on the size of the population © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 7 The Core Axioms 0. Outline Focus (FC): For any ( X ; z );(Y ; z ) K Z : if 1. Introduction 2. The Identification Step 3. The Aggregation Step i) for any h such that xhl zl , yhl xhl , 0, ii) yil xil i h yij xij j l , i, then P ( X ; z ) P (Y ; z ). The poverty index does not change if an individual receives more of a dimension in which it is not deprived 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Non-Decreasingness in Subsistence Levels (NS): For any X K , z Z , ( X ; z ) is non-decreasing in z j j. The poverty index does not decrease if, ceteris paribus, the threshold levels increase Summary Non-Poverty Growth (NG): For any (Y ; z ) K Z , if X is obtained from Y by adding a rich person to the population, then P ( X ; z ) P (Y ; z ). The poverty index does not increase with the population size of the non-poor © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 8 The Core Axioms 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary Subgroup Decomposability (SD): For any X 1 ,..., X m K and m P( X 1 , X 2 ,..., X m ; z ) i 1 ni nP( X i ; z ) with z Z : ni being the population size of subgroup X i , i 1,..., m and m n n. i 1 i Factor Decomposability (FD): For any ( X ; z ) K Z : P( X ; z ) j 1 a j P( x j ; z j ) with a j 0 being the weight attached to attribute j and k k j 1 a j 1. Normalization (NM): For any ( X ; z ) K Z : P ( X ; z ) 1 if xij 0i, j and P ( X ; z ) 0 if xij z j i, j. Unit Consistency: For any ( X ; z );(Y ; z ) K Z : if P ( X ; z ) P (Y ; z ) then P ( X; z ) P (Y; z ) with being the diagonal matrix diag (1 ,..., k ), j 0j. Poverty orderings are not affected by a change of the unit of measurement © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 9 The Core Axioms 0. Outline Uniform Majorization (UM): For any z Z and X, Y of the same dimension, 1. Introduction if X P BY Pand B is not a permutation matrix, then P ( X ; z ) P (Y ; z ), where 2. The Identification Step 3. The Aggregation Step X P Y P is the attribute matrix of the poor corresponding to X(Y) and B bij is some bistochastic matrix of appropriate order. An equalising operation does not increase poverty 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e But in the multidimensional case, inequality does not only exist within dimensions, but also between dimensions Consider the following situation: i 2, j 3, z 3 3 3 3 3 0.5 0 0 0 0 1 0 0 0 0 X Y ; 1 5 10 7 6 0.5 5 10 7 6 Nicole Rippin: Multidimensional Poverty Comparisons 10 The Core Axioms 0. Outline 1. Introduction matrix Y by a weak inequality increasing switch of attribute l from one poor 2. The Identification Step individual to another if for some individuals g, h: 3. The Aggregation Step 3.1 Core Axioms 3.2 New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition Weak Inequality Increasing Switch: Matrix X is said to be obtained from i) y hl y gl z l ; y gm z m y hm ii) x gl y hl , x hl y gl ; x gm y gm , x hm y hm iii) xil y il , xim y im i g , h; xij y ij j l , m, i Through a switch of endowments, the individual that is already deprived in fewer dimensions receives a higher amount of an attribute with regard to which both individuals are deprived. Non-Decreasingness under Weak Inequality Increasing Switch (NDW): For any 5. Summary (Y ; z ) K Z , if X K is obtained from Y by a weak inequality increasing switch between two poor individuals, then P ( X ; z ) P (Y ; z ). Like UM, NDW can only be satisfied by cardinal poverty indices © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 11 The Core Axioms 0. Outline 1. Introduction i 2, j 3, z 3 3 3 3 3 2. The Identification Step 3. The Aggregation Step 5 0 0 0 0 Y ; 1 2 10 7 6 3.1 Core Axioms Consider the following situation: 1 0 0 0 0 X 5 2 10 7 6 3.2 New Inequality Axiom Strong Inequality Increasing Switch: Matrix X is said to be obtained from 3.3 Unique Index Classes matrix Y by a weak inequality increasing switch of attribute l from one poor 3.4 A Numerical Example individual to another if for some individuals g, h: 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary i) y hl z l y gl ; y gm z m y hm ;y gk z k ; y hk z k ii) x gl y hl , x hl y gl ; x gm y gm , x hm y hm ;x gk y gk , x hk y hk iii) xil y il , xim y im , xik y ik i g , h; x ij y ij j l , m, k , i Through a switch, the number of deprivations of the individual that is already deprived in less dimensions is reduced. Non-Decreasingness under Strong Inequality Increasing Switch (NDS): For any (Y ; z ) K Z , if X K is obtained from Y by a strong inequality increasing switch between two poor individuals, then P ( X ; z ) P (Y ; z ). © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 12 Two unique classes of indices 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings Proposition 1: The only family of poverty indices satisfying CN, FC, SD, FD, SI, MN, UM, NM and NDW is of the form P( X ; z ) 1 n i 1 j 1 a j f i xij z j n k with f i : 0, R1 continuous, non-increasing in (.), increasing in di # xij z j , with f(0) = 1 and f (t ) 0t 1. Also a j 0are constants k with j 1 a j 1. 4.2 Decomposition 5. Summary © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 13 Two unique classes of indices 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 New Inequality Axiom Proposition 2: The only family of poverty indices satisfying CN, FC, SD, FD, SI, MN, NM and NDS is of the form P ( X ; z ) 1 n iS j k j 1 a j ij 3.3 Unique Index Classes 3.4 A Numerical Example with 4. i (0,1] increasing in d i # xij z j ij 0 Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Also a j 0 are constants with if if xij z j xij z j j 1 a j 1. k Nicole Rippin: Multidimensional Poverty Comparisons 14 Empirical application: the poverty indices 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step I utilize four different poverty indices for the poverty comparisons which all satisfy an increasing number of sensitivity axioms: MPI 1 n iS j POR 1 n iS k j 1 a j ij 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings j k j 1 a j i ij NDS PFGT ( X ; z ) 1 n j 1 iS a j 1 xij z j 2 k UM j 4.2 Decomposition 5. Summary PCR ( X ; z ) 1 n j 1 iS a j i 1 xij z j 2 k UM & NDW j i © 2011 d·i·e k j 1 a j ij 1 if and ij 0 if xij z j xij z j Nicole Rippin: Multidimensional Poverty Comparisons 15 Empirical application: dimensions, weights and thresholds 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step The MPI comprises three equally weighted poverty dimensions: education, health and living standards themselves captured by an overall of ten indicators 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 16 Five Indian Households… 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. HH 1 2 3 4 5 Education Years xmi = 0 zj = 5 15 10 9 5 0 Health/Nutrition Health/Nutrition Living Standards Assets BMI w/a z-scores Sleeping Rooms Weighted Sum xmin=14 zj=18.5 xmin=-4.5 zj = -2 xmin = 0 zj = 0.3 xmin = 0 zj = 0.27 30.29 -0.21 0.25 1.00 19.32 2.73 0.32 0.15 17.19 -1.46 0.29 0.25 18.47 -2.02 0.25 0.26 19.17 -5.58 0.06 0.00 Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. As an illustration, consider the following numerical example which is drawn from the Indian household DHS (2005) Summary © 2011 d·i·e HH 1 2 3 4 5 All MPI POR FGT PCR 0.000 Rank 1 0.000 Rank 1 0.600 Rank 3 0.800 Rank 4 0.800 Rank 4 0.320 0.040 Rank 1 0.040 Rank 1 0.360 Rank 3 0.640 Rank 4 0.640 Rank 4 0.256 0.006 Rank 2 0.040 Rank 4 0.018 Rank 3 0.005 Rank 1 0.725 Rank 5 0.146 0.007 Rank 1 0.018 Rank 2 0.047 Rank 4 0.033 Rank 3 0.607 Rank 5 0.128 Nicole Rippin: Multidimensional Poverty Comparisons 17 Empirical application: changes in poverty rates and rankings 0. Outline A Sample of 28 Countries Country MPI 6 Value 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Rank Por Value Rank Change 0.004 Rank 1 0.008 Rank 1 Armenia 0.009 Rank 2 0.012 Rank 2 Moldova 0.021 Rank 3 0.021 Rank 3 Azerbaijan 0.085 Rank 4 0.054 Rank 4 Peru 0.139 Rank 5 0.086 Rank 5 Morocco 0.139 Rank 6 0.090 Rank 6 Ghana 0.174 Rank 7 0.104 Rank 7 Zimbabwe 0.175 Rank 8 0.106 Rank 8 Bolivia 0.184 Rank 9 0.107 Rank 9 Swaziland 0.187 Rank 10 0.113 Rank 10 Namibia 0.263 Rank 11 0.156 Rank 11 Cambodia 0.271 Rank 12 0.159 Rank 12 Congo, Rep. 0.285 Rank 13 0.178 Rank 14 -1 India 0.291 Rank 14 01.87 Rank 16 -2 Cameroon 0.292 Rank 15 0.174 Rank 13 +2 Bangladesh 0.302 Rank 16 0.183 Rank 15 +1 Kenya 0.306 Rank 17 0.193 Rank 17 Haiti 0.326 Rank 18 0.194 Rank 18 Zambia 0.350 Rank 19 0.218 Rank 19 Nepal 0.368 Rank 20 0.248 Rank 22 -2 Nigeria 0.385 Rank 21 0.236 Rank 20 +1 Malawi 0.393 Rank 22 0.240 Rank 21 +1 DR Congo 0.412 Rank 23 0.269 Rank 23 Benin 0.481 Rank 24 0.321 Rank 25 -1 Mozambique 0.484 Rank 25 0.308 Rank 24 +1 Liberia 0.557 Rank 26 0.390 Rank 26 Mali 0.569 Rank 27 0.391 Rank 27 Ethiopia 0.642 Rank 28 0.475 Rank 28 Niger Methodology: Countries sorted in descending order by the MPI The MPI’s neglect of inequality leads to an artificial distortion of poverty rates which are understated in less poor countries and exaggerated in poorer countries Nicole Rippin: Multidimensional Poverty Comparisons 18 Empirical application: changes in poverty rates and rankings 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Table 4: Country Implications for vulnerable population groups (MPI – Por) Rural 7-10 Religion of Minority Areas members Christianity Islam 0.215 0.048 Armenia 0.062 0.037 Moldova 0.092 0.076 0.017 Azerbaijan 0.057 0.003 Peru 0.023 -0.003 Morocco 0.044 0.005 -0.001 Ghana 0.027 0.005 0.004 Zimbabwe 0.042 0.006 Bolivia 0.014 0.009 -0.008 Swaziland 0.013 0.005 Namibia 0.007 -0.003 0.000 Cambodia 0.009 0.004 -0.002 Congo, Rep. 0.003 -0.007 -0.008 India 0.009 -0.008 -0.014 Cameroon 0.004 -0.004 Bangladesh 0.007 -0.013 -0.013 Kenya 0.002 -0.012 Haiti -0.002 -0.012 -0.009 Zambia 0.003 -0.008 Nepal -0.008 -0.005 -0.043 Nigeria 0.003 -0.008 -0.005 Malawi -0.001 -0.010 DR Congo 0.001 -0.010 -0.019 Benin -0.012 -0.012 -0.009 Mozambique -0.027 -0.007 Liberia -0.015 -0.014 -0.015 Mali -0.003 -0.019 -0.010 Ethiopia -0.015 -0.014 -0.015 Niger Methodology: Countries sorted in descending order by the MPI. The MPI’s neglect of inequality leads to an artificial distortion of the contribution of vulnerable population subgroups: their contribution is overstated in less poor countries and understated in poorer countries Nicole Rippin: Multidimensional Poverty Comparisons 19 Empirical application: changes in poverty rates and rankings Country Comparisons in 28 Countries 0. Outline Country 1. Introduction 2. The Identification Step 3. The Aggregation Step Moldova Armenia Azerbaijan Peru Swaziland Bolivia Zimbabwe Ghana Namibia Congo, Rep. Morocco Cameroon Zambia Kenya DR Congo Nigeria Liberia Malawi Haiti Bangladesh Benin Cambodia India Nepal Mozambique Mali Niger Ethiopia 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Value H 0.008 0.006 0.011 0.045 0.040 0.056 0.051 0.061 0.043 0.059 0.056 0.085 0.076 0.089 0.085 0.104 0.124 0.107 0.117 0.088 0.120 0.106 0.104 0.105 0.130 0.195 0.254 0.259 0.228 0.246 0.360 0.584 0.641 0.663 0.799 0.711 0.633 0.825 0.578 0.785 0.839 0.888 0.916 0.836 0.905 0.951 0.883 0.829 0.841 0.927 0.846 0.903 0.938 0.909 0.971 0.982 FGT I 0.033 0.025 0.030 0.078 0.063 0.085 0.064 0.086 0.068 0.071 0.097 0.108 0.091 0.100 0.093 0.124 0.138 0.113 0.133 0.107 0.143 0.114 0.123 0.116 0.138 0.215 0.261 0.264 Pcr Rank Value H I Rank Rank 2 Rank 1 Rank 3 Rank 6 Rank 4 Rank 9 Rank 7 Rank 11 Rank 5 Rank 10 Rank 8 Rank 13 Rank 12 Rank 16 Rank 14 Rank 17 Rank 24 Rank 21 Rank 22 Rank 15 Rank 23 Rank 20 Rank 18 Rank 19 Rank 25 Rank 26 Rank 27 Rank 28 -1 +1 -2 +1 -3 -3 +4 +3 -1 +1 -2 +1 -1 -7 -3 -3 +5 -2 +2 +5 +5 - 0.004 0.004 0.008 0.028 0.027 0.038 0.031 0.043 0.033 0.042 0.045 0.062 0.062 0.069 0.069 0.087 0.093 0.089 0.099 0.085 0.100 0.094 0.96 0.104 0.119 0.162 0.226 0.237 0.228 0.246 0.360 0.584 0.641 0.663 0.799 0.711 0.633 0.825 0.578 0.785 0.839 0.888 0.916 0.836 0.905 0.951 0.883 0.829 0.841 0.927 0.846 0.903 0.938 0.909 0.971 0.982 0.016 0.015 0.022 0.048 0.042 0.058 0.039 0.060 0.052 0.051 0.077 0.079 0.073 0.077 0.075 0.104 0.103 0.094 0.112 0.103 0.119 0.101 0.114 0.115 0.127 0.178 0.233 0.241 Rank 2 Rank 1 Rank 3 Rank 5 Rank 4 Rank 8 Rank 6 Rank 10 Rank 7 Rank 9 Rank 11 Rank 13 Rank 12 Rank 14 Rank 15 Rank 17 Rank 19 Rank 18 Rank 22 Rank 16 Rank 23 Rank 20 Rank 21 Rank 24 Rank 25 Rank 26 Rank 27 Rank 28 -1 +1 -1 +1 -2 +1 -2 +2 +1 -1 +1 -1 -2 -3 +4 -2 +2 +2 - Nicole Rippin: Multidimensional Poverty Comparisons 20 Empirical application: changes in poverty rates and rankings 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary © 2011 d·i·e Country Rankings in 29 Indian States State FGT Value H I Kerala 0.021 0.557 0.038 Delhi 0.042 0.590 0.072 Sikkim 0.055 0.703 0.078 Goa 0.039 0.626 0.062 Manipur 0.047 0.727 0.064 Himachal Pradesh 0.048 0.729 0.065 Mizoram 0.062 0.754 0.082 Punjab 0.052 0.653 0.079 Tamil Nadu 0.068 0.799 0.085 Jammu & Kashmir 0.063 0.772 0.081 Nagaland 0.091 0.839 0.108 Uttaranchal 0.071 0.769 0.092 Haryana 0.071 0.778 0.091 Arunachal Pradesh 0.101 0.814 0.124 Maharashtra 0.078 0.847 0.092 Karnataka 0.091 0.837 0.109 Andhra Pradesh 0.099 0.848 0.116 Tripura 0.080 0.824 0.097 Gujarat 0.090 0.819 0.110 0.083 Assam 0.837 0.099 Meghalaya 0.126 0.852 0.148 Orissa 0.114 0.858 0.133 Rajasthan 0.131 0.892 0.147 West Bengal 0.105 0.843 0.124 Chhattisgarh 0.099 0.891 0.111 Uttar Pradesh 0.123 0.899 0.136 Madhya Pradesh 0.135 0.905 0.150 Jharkhand 0.130 0.898 0.145 Bihar 0.167 0.925 0.180 Pcr Rank Value H I Rank 1 3 7 2 4 5 8 6 10 9 17 11 12 21 13 18 19 14 16 15 25 23 27 22 20 24 28 26 29 -1 -4 +2 +1 +1 -1 +2 -1 +1 -6 +1 +1 -7 +2 -2 -2 +4 +3 +5 -4 -1 -4 +2 +5 +2 -1 +2 - 0.014 0.033 0.041 0.031 0.037 0.036 0.047 0.045 0.050 0.052 0.069 0.058 0.063 0.083 0.067 0.077 0.083 0.076 0.082 0.083 0.105 0.109 0.117 0.107 0.100 0.113 0.129 0.134 0.170 0.557 0.590 0.703 0.626 0.727 0.729 0.754 0.653 0.799 0.772 0.839 0.769 0.778 0.814 0.847 0.837 0.848 0.824 0.819 0.837 0.852 0.858 0.892 0.843 0.891 0.899 0.905 0.898 0.925 0.026 0.056 0.059 0.050 0.051 0.050 0.062 0.069 0.063 0.068 0.083 0.076 0.081 0.102 0.079 0.092 0.098 0.092 0.100 0.100 0.123 0.127 0.131 0.127 0.112 0.126 0.142 0.149 0.183 1 3 6 2 5 4 8 7 9 10 14 11 12 18 13 16 19 15 17 20 22 24 26 23 21 25 27 28 29 -1 -3 +2 +2 -1 +1 -3 +1 -1 -4 +2 -2 +3 +2 -1 -2 -3 +1 +4 +1 - Nicole Rippin: Multidimensional Poverty Comparisons 21 Empirical application: Indian poverty maps 0. Outline Chattisgarh MPI 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. [0,.05] (.05,.1] (.1,.15] (.15,.2] (.2,.25] (.25,.3] (.3,.35] (.35,.4] (.4,.5] No data Empirical Application 4.1 Rates and Rankings FGT Por [0,.05] (.05,.1] (.1,.15] (.15,.2] (.2,.25] (.25,.3] (.3,.35] No data Pcr 4.2 Decomposition 5. Summary [0,.02] (.02,.04] (.04,.06] (.06,.08] (.08,.1] (.1,.12] (.12,.14] (.14,.16] (.16,.18] No data © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons [0,.02] (.02,.04] (.04,.06] (.06,.08] (.08,.1] (.1,.12] (.12,.14] (.14,.16] (.16,.18] No data 22 Summary 0. Outline 1. Introduction 2. The Identification Step 3. The Aggregation Step 3.1 Core Axioms 3.2 A New Inequality Axiom 3.3 Unique Index Classes 3.4 A Numerical Example 4. Empirical Application 4.1 Rates and Rankings 4.2 Decomposition 5. Summary I introduced a new axiom that removes an anomaly in the way poverty indices account for multidimensional inequality The new axiom solves the exaggeration problem for additive poverty indices by implying a change in the identification of the poor: any household that is deprived is considered poor to a certain degree, i.e. receives an individual weight that depends on the number of dimensions in which it is deprived In the ordinal case, it removes some weaknesses of the MPI as induced by the dual cut-off method: arbitrariness, discontinuity and distortion while at the same time introducing sensitivity with regard to inequality among the poor. At the same time it is much easier to calculate than the MPI as it is only calculated once Due to its specific two-step method, it is the first composite index that allows for a degree of association-sensitivity between dimensions while it is still additive and therefore decomposable © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 23 Thank you! © 2011 d·i·e Nicole Rippin: Multidimensional Poverty Comparisons 24