Yet, Two More Revisions to the Human Development Index

Abstract

The Inequality-adjusted Human Development Index (IHDI) was adopted in the 20th anniversary edition of the Human Development Reports, in 2010. In using a penalty set-up for the calculations of the IHDI, however, the results overestimated the adjustments. This article suggests a revision to the procedure in order to harmonize the calculations with the underlying attainment set-up of the Human Development Index and minimize the bias in the adjustments. This article also suggests an extension to the IHDI, specifically the inclusion of a subjective measure of inequality in the calculation of the IHDI.

Keywords::

• Human Development Index
• Inequality-adjusted Human Development Index

JEL classifications::

• I31
• D63
• O15

Acknowledgement

I thank an anonymous reviewer for an engaging debate on the issues raised in this article. Of course, the remaining errors are mine.

Notes

¹ An exception is the Human Development Report 1993, which presented IHDI for some countries, and the Human Development Report 1995, which introduced the gender-related development index. There are other suggestions as adjustments to the HDI such as political freedom (Desai, 1994) and sustainability (Sagar & Najam, 1998). Historical discussions on the HDI are available from Ul Haq (1995), Fukuda-Parr and Kumar (2003), and Alkire (2010).

² The standard procedure is to convert income into log value before inputting it in the formula.

³ From 1990 to 2009, the HDI was calculated as the simple average of the three indices. From 2010, the HDI is calculated using the geometric mean of the three indices. Desai (1991) discusses the conceptual difference between the arithmetic mean and geometric mean in the context of the HDI.

⁴ Note that A_i is computed as one minus the ratio of the geometric mean and arithmetic mean of the distribution.

⁵ The expression in (·) does not precisely fit the “same dimension” and the “same description” as Equation (1). The revision in Equation (4) is, in a way, only a correction to a bias in the amounts of adjustment. What this paper does is to take (1–A_i) as its entry point for the suggestions for revisions and, perhaps, a step for rethinking how an Atkinson-based inequality metric should be introduced as an adjustment to the HDI. Of course, the end goal is to achieve a revision that is consistent with the underlying structure of Equation (1).

⁶ The chance that (1–A_i)/(1–A)_max = 1 for all health, education, and income of the same country is small.

⁷ Subjective measure of inequality is an emerging topic in subjective well-being research. See Ferrer-i-Carbonell and van Praag (2003), Kalmijn and Veenhoven (2005), Kalmijn and Arends (2010), Dutta and Foster (2011), van Praag (2011), Clark et al. (2012), Beja (2013), and Gandelman and Prozecanski (2013) for recent findings. Key concepts of subjective well-being are discussed in, for example, Diener (1984), Kahneman (1999), Diener et al. (1999), Ryan and Deci (2001), and Baumeister and Vohs (2002). In this article, subjective well-being is presented as “happiness,” which is how a person considers one's state of being at a point in time.

⁸ As with Equation (5), the chance that (1 − S_i)/(1 − S)_max = 1 for all health, education, and income in the same country is small.

⁹ The figures in Table 1 were calculated using World Values Surveys data because Gallup World Polls data are not available without payment. The proper application of Equation (7) requires information on a person's happiness with one's health, education, and income. Moreover, the scale for their measurement needs to be of the same dimension. Specifically, the measurement must be done on a 0–10 scale with the ends of the scale defined as “0 means 0% happy” and “10 means 100% happy”. See Footnote 12.

¹⁰ Recall that this paper is about the inequality adjustments to the HDI. If the subjective measures are added as extensions to x_i, then IHDI-2 = [IHDI-1_obj·IHDI-1_subj]^1/2, where This specification was not executed in the paper because, as pointed out, the available happiness data were collected not as inputs to the IHDI. For discussions on attempts at finding the common grounds between subjective and objective measures of human well-being, see Alkire (2005), Comim (2005), Schokkaert (2007), Bruni et al. (2008), Anand et al. (2009), Veenhoven (2010), and Binder (2013).

¹¹ Adaptation when applied to happiness is called “hedonic adaptation” but when applied to utility is called “preference adaptation.” See Parfit (1984) and Haybron (2008) for typologies of human well-being.

¹² Measurement plays a crucial role in bridging the information gap and adaptation. Specifically, it requires a 0–10 unipolar scale format with the end-points labeled as “0 means 0% happy” and “10 means 100% happy.” This design normalizes happiness ratings across persons, societies, regions, and periods. Consider the following exercise as proof. Suppose there is a 50% full glass. That 50% full glass is deemed 50% full regardless of the size of the glass, location of the glass, the time the glass was evaluated, or the person who made the appraisal of the glass. Correspondingly, a self-rating of 50% happiness in one instance, location, etc., is comparable to other self-ratings of 50% happiness in other instances, locations, etc. Information on, say, subjective health can then be elicited using a standard question but applying the required format of the scale, as follows: How happy are you as a whole with your own health? Using the scale below, “0” means “0% happy” and “10” means “100% happy.”The queries for subjective education and subjective income have similar phrasing. In addition, multiple questions to cover the different components of, say, health can be introduced provided the scale for each component follows the required format.