A MANUSH or HUMANS Characterisation of the Human Development Index

Abstract

Proposing a set of axioms MANUSH (Monotonicity, Anonymity, Normalisation, Uniformity, Shortfall sensitivity, Hiatus sensitivity to level), this paper evaluates three aggregation methods of computing Human Development Index (HDI). The old measure of HDI, which is a linear average of the three dimensions, satisfies monotonicity, anonymity, and normalisation (or MAN) axioms. The current geometric mean approach additionally satisfies the axiom of uniformity, which penalises unbalanced development across dimensions. We propose ℋ_α measure, which for α ≥ 2 also satisfies axioms of shortfall sensitivity (emphases on the worse-off to better-off dimensions should be at least in proportion to their shortfalls) and hiatus sensitivity to level (higher overall attainment must simultaneously lead to a reduction in gap across dimensions). Special cases of ℋ_α are the linear average (α = 1), the displaced ideal (α = 2), and the leximin ordering (α → ∞) methods. For its axiomatic advantages, we propose to make use of the displaced ideal (α = 2) method in the computation of HDI replacing the current geometric mean.

Keywords:

• Displaced ideal
• Hiatus sensitivity
• MANUSH
• Shortfall sensitivity
• Uniform development

JEL Codes:

• D63
• I31
• O15

Acknowledgements

Comments by two anonymous referees and one of the associate editors of the journal were helpful in improving the clarity and focus of the paper. The authors have benefited from insightful comments and discussions with Dilip Ahuja, Sabina Alkire, Pradipta Chaudhury, Om Damani, Meghnad Desai, James Foster, Supriya Garikipat, Howard Glennerster, Ruth Kattumuri, Amrendra Kumar, T Krishna Kumar, Sunil Kumar, Banikanta Mishra, Anjan Mukherji, Baseerit Nissah, Prasanta Pattanaik, Rathin Roy, Tirthankar Roy, Amartya Sen, Kunal Sen, Suman Seth, S. Subramanian, Polly Vizard, Yongsheng Xu, and from presentations made at Coventry University (EFA Research Seminar Series, Coventry Business School); Department of Management Studies, IISc, Bangalore; IGIDR, Mumbai; London School of Economics and Political Science (LSE's CASE Social Exclusion Seminars, STICERD); NIAS, Bangalore; National Institute of Public Finance and Policy, New Delhi; Oxford Poverty and Human Development Initiative, University of Oxford; Ravenshaw University, Cuttack; University of Liverpool (Development Studies Seminar Series under DRIVE); University of Bath (Internal Economics Seminars, Department of Economics), University of Manchester (Development Economics Seminar Series, IDPM); and Xavier Institute of Management, Bhubaneswar. Earlier versions have come out as working papers at IGIDR, LSE and NCDS. Usual disclaimers apply. The views expressed in the paper are those of the authors and not of the organisations that they are affiliated to or are associated with.

Disclosure Statement

No potential conflict of interest was reported by the authors.

ORCID

Srijit Mishra http://orcid.org/0000-0002-6429-8198

Hippu Salk Kristle Nathan http://orcid.org/0000-0002-0329-4898

About the Authors

Srijit Mishra is Director, Nabakrushna Choudhury Centre for Development Studies, Bhubaneswar. He researches and teaches on development-related issues. His recent works are on agrarian crises and farmers’ suicides in India and on measurement issues. For his work, see https://works.bepress.com/srijit_mishra/.

Hippu Salk Kristle Nathan is an Assistant Professor at National Institute of Advanced Studies, Bangalore. His research interests include economic measurement, energy, human development and sustainable use of resources. For his work, see https://works.bepress.com/hippu/.

Notes

1 For discussions on this, see Streeten et al. (1981), Sen (1989, 1997, 1999, 2000), Desai (1991), Streeten (1994), and Haq (1995), among others.

2 The HDR is being published annually since 1990 and serves as a cornerstone in terms of philosophy as well as an approach of the United Nations Development Programme (UNDP).

3 For discussions on birth, evolution, measurement, and critique of HDI and its policy discourse, see Anand and Sen (1993, 1995, 1997, 2000, 2003), Haq (1995); Lüchters and Menkhoff (1996); Dutta, Panda, and Wadhwa (1997); Hicks (1997); Noorbakhsh (1998); Sen (2000); Panigrahi and Sivramkrishna (2002); Fakuda-Parr, Raworth, and Shiva Kumar (2003); Jahan (2003); Raworth and Stewart (2003); Ranis, Stewart, and Samman (2006); Grimm et al. (2008); Alkire and Foster (2010); Nathan and Mishra (2010); United Nations Development Program (2010); Klugman, Rodriguez, and Choi (2011); Wolff, Chong, and Auffhammer (2011); Harttgen and Klasen (2012); Ravallion (2012); and Permanyer (2013), among others.

4 The normalisation used: value = (actual − minimum) / (maximum − minimum).

5 In this paper, attainment refers to the normalised value of the indicators representing the dimensions of HDI.

6 The trade-off across dimensions in this method, as indicated by Chakravarty (2011) and Ravallion (2012) is troubling. Zambrano (2017) has also responded to this concern.

7 The ideal corresponds to the maximum values for all the three dimensions as posited by UNDP for the calculation of the human development index. In this sense, ideal indicates complete attainment. We use distance in the Euclidean sense unless otherwise specified.

8 Chakravarty and Majumder (2008) suggest the use of shortfalls in targets while evaluating the progress of Millennium Development Goals.

9 MANUSH (or manush) means human beings in some of the South Asians languages such as Assamese, Bengali, Marathi, and Sanskrit, among others. Incidentally, the term has HUMANS as its anagram. In this sense, the paper proposes the axiom of HUMANS for the human development index.

10 There are other deserving alternative choices in dimensions, measurement, weights, and normalisation, and these may have an ambiguity in capability space, quite similar to real income space as discussed in Sen (1979).

11 In a measure of deprivation that involves multiple dimensions, the priority of computing individual deprivation has been indicated by Dutta, Pattanaik, and Xu (2003). Our suggested alternative is a method of aggregation that can also be applied to individual data.

12 The HDI calculation involves three dimensions. However, for generalisation purpose, we have considered dimensions.

13 This measure takes the form of Minkowski distance function. Use of Minkowski distance function in the context of human development is not new. Prior to 2010, the Human Development Reports used Minkowski distance function across different dimensions of deprivations to calculate Human Poverty Indices (HPI-1 and HPI-2), see Anand and Sen (1997) and Pillai (2004). Subramanian (2006) has also used the Minkowski distance function to the Foster, Greer, and Thorbecke (1984) poverty measure.

14 In the computation of HDI in the HDRs, the three dimensions are given equal weights. However, if they were to be given different weights, permutation assumes interchanging of values together with weights.

15 In the context of distribution of resources among individuals, shortfall sensitivity will be a stricter condition than that of the equity axiom of Hammond (1976) or priority axiom of Moreno-Ternero and Roemer (2006), both of which draw on the weak equity axiom of Sen (1973, 18–19) to indicate that a greater emphasis is to be given to the worse-off.

16 This lexicographic extension of the difference principle by Rawls (1971, 2001) is the only generalised social welfare function in the context of interpersonal ordering, which satisfies the Arrow conditions, the equity axiom, and Suppes’ grading principle (Hammond 1976).

17 In the space of distribution of income among individuals, level sensitivity refers to society's greater concern for inequity with increase in prosperity, as suggested by Sen (1973).

18 The similarity of ℋ _α measures with constant elasticity of substitution functions is obvious (Rao 2011).

19 For ℋ _α, the minimum bound of α is restricted to 1, i.e., the condition where the optimal path gives equal emphases to all the dimensions in future progress. For α < 1, the optimal paths are such that the emphasis given to the neglected dimension is less than the emphasis given to favoured dimension. Also, note that for α < 1, we will have iso-HDI lines that are concave to the origin.

20 Sen (2000, 2009), while critiquing other aspects, surmise that a major contribution of Rawls (1971) is the invoking of reasonable pluralism. This has larger implications and is different from leximin ordering identified with his difference principle. For an application of Rawls’ reasonable pluralism on conflict resolution, see Mishra (2011). In the current context, leximin ordering is a special case of the Rawlsian spirit; an exacting situation from the multitude of possibilities that satisfy shortfall sensitivity.

21 The formulae for the optimal paths of ℋ_𝒞 and ℋ_ℱ in a two-dimensional situation of 𝒽 < ℯ are 𝒹𝒽/𝒹ℯ = (ℯ/𝒽)^(1−δ); δ = (0,1] and 𝒹𝒽/𝒹ℯ = (ℯ/𝒽)^ϵ; ϵ ≥ 0, ϵ ≠ 1, respectively. The two formulae coincide in the domain ϵ < 1 as ϵ = 1 − δ.

22 In fact, Chakravarty (2003, 104) also points out that ℋ_𝒞 “will attach greater weight to achievement differences at lower level of attainment.” Thus, confirming our observation that it fails hiatus sensitivity to level.