Tuesday 3 May 2011

HDI 2010: New Controversies, Old Critiques

This post first appeared on 27 April on the UNDP's Let's Talk Human Development


By Jeni Klugman, Francisco Rodríguez and Hyung-Jin Choi
Director, Head of Research and Consultant (respectively),
Human Development Report Office, UNDP

Since its introduction in the first Human Development Report in 1990, the Human Development Index (HDI) has attracted great interest in policy and academic circles, as well as in the media and national audiences around the world. Its popularity can be attributed to the simplicity of its characterization of development - an average of achievements in health, education and income – and to its underlying message that development is much more than economic growth.

The index was originally conceived by the late Pakistani economist Mahbub ul Haq, in collaboration with Amartya Sen and other scholars, as a response to their dissatisfaction with GDP as the standard measure of development. As Haq noted, “Any measure that values a gun several hundred times more than a bottle of milk is bound to raise serious questions about its relevance for human progress.”
Yet the HDI’s very simplicity prompted critiques from the start, with some contending that it was too simplistic, while others who accepted its self-imposed limitations still questioned its choice of indicators and its computational methodology. This article is a detailed review of key critiques of the HDI, today and in the past, and explains the recent changes introduced to the HDI formula and indicators. The full paper can be read online in the Human Development Research Papers series as well as in an upcoming June issue of The Journal of Economic Inequality.

Key innovations in the 2010 Report
In 2010, for the twentieth anniversary edition, the Human Development Report undertook a comprehensive review of these critiques and introduced several major changes to the HDI. Though the HDI had been modified before, as summarized in the table below, 2010 was the first time that major changes were made simultaneously to the indicators used to measure progress, and to the functional form used to convert those indicators to a single measure of progress.
HDI_2010_table_1.png

Until 2010, the HDI had been defined as a simple arithmetic average of normalized indices in the dimensions of health, education and income: HDI = (1/3)*(Hh + He + Hls). Hi denotes the sub-index for dimension I, with i={h,e,ls} respectively denoting the health, education and living standards dimensions. Life expectancy (le) and GDP per capita (gdp) were the proxies for health and living standards respectively, whereas the education dimension used two indicators: literacy (lit) and the gross enrolment ratio (ger).

The indices were normalized using upper and lower bounds. Thus:
Hh = (le-lemin)/(lemax-lemin), 


He = (1/3)*((ger-germin)/(germax-germin))+(2/3)*((lit-litmin)/(litmax-litmin)),


Hls = (ln(gdp)-ln(gdpmin))/(ln(gdpmax)-ln(gdpmin)).


The 2010 Human Development Report introduced several significant changes in the HDI. The new formula is: HDI = (HHealth * HEducation * HLiving standard)1/3 The indices Hi are still normalized indicators of achievements. Life expectancy (le) remains the indicator for the health dimension, while Gross National Income (gni) replaces GDP, and mean years of schooling for adults (ms) plus expected years of schooling (eys) for children now make up the education dimension.

Hh = (le-lemin)/(lemax-lemin)


He = [((mys-mysmin)/(mysmax-mysmin)) *((eys-eysmin)/(eysmax-eysmin))]1/2


Hls = (ln(gni)-ln(gnimin))/(ln(gnimax)-ln(gnimin)).


The approach introduced in 2010 retains the same three-dimensional structure with equal weights, with several key changes: It replaces the indicators for income and education, it changes the method of aggregation from an arithmetic average to a geometric average, and it redefines the upper and lower bounds used to normalize the index, eliminating the practice of capping variables that surpass the upper bounds.

Three of the four variables that go into the HDI were revised. GDP per capita was replaced by GNI per capita (both valued in PPP US$), while literacy and gross enrolments were replaced by mean years of schooling and expected years of schooling.

To supplement these changes to the HDI, three new complementary indices were introduced in 2010 that were designed to capture the deprivations and disparities: the Inequality-Adjusted HDI, the Gender Inequality Index, and the Multidimensional Poverty Index.

It is worth focusing on the changes related to income. The Human Development Report and its HDI was an explicit challenge to the reigning paradigm of the late 1980s in development policy circles, commonly known as the Washington consensus. The prevailing approach for development economics was best symbolized then in the World Bank’s World Development Report’s practice of ranking countries by per capita income.

Throughout its history, the HDI has effectively pinpointed the differences in rankings that emerge from a comparison between the HDI and per capita incomes. Haq pointed out in 1995 that of the 173 countries included in the HDI rankings, there was no difference between the HDI and per capita income rankings for only four countries, while for more than a third of the sample the HDI and per capita income ranks differed by more than 20 places. These patterns still hold true today. The 2010 Human Development Report systematically explores the links between economic growth and other dimensions of human development and documents the finding that over relatively long periods of time high rates of economic growth are not necessarily accompanied by improvements in other dimensions of human development.

New indicators to measure human development


The 2010 HDI revisions took into account measurement improvements in its three dimensions over the past two decades. Unfortunately, alternative measures that may be conceptually preferable are available for fewer countries than were typically covered by the HDI, and are often not updated frequently.

The best example is perhaps life expectancy. The World Health Organization has reported Health-Adjusted Life Expectancy (HALE) for 192 countries, as the number of years that a person might expect to live in good health. The estimates are based on country life tables, analyses of 135 causes of disability for 17 regions of the world and 69 health surveys in 60 countries. However, the WHO does not plan to regularly update its estimates of HALE.

Similar problems arose in the case of education, where the cross-national assessments of science, mathematics and reading levels that could be used to construct quality adjustments are only available for a limited number of countries. The approach in the 2010 Human Development Report was to adopt mean years of schooling as calculated by Barro and Lee as the indicator to measure the education of adults. This indicator is more frequent, has broader coverage, and better discriminatory power than literacy. The calculations are based on primary data from the UNESCO Institute of Statistics’ Database of Censuses and can thus be replicated and updated.

Literacy – which had up to now carried a 2/3 weight in the education index – has become deeply unsatisfactory over time as a measure of progress in education. The world average literacy rate rose from 60 to 83 percent between 1970 and 2010. Almost half of countries have a literacy rate higher than 95 percent and, indeed, developed countries no longer collect data on basic literacy. Further, many developing countries are poised to attain universal literacy in future years. While literacy was a good measure to evaluate progress in the past, it is now much less informative.

The new HDI also changed its measure of the education of children, replacing the gross enrolment ratio with a measure of expected years of schooling. This captures the average number of years that children today could be expected to attain in adulthood if enrolment rates stay at their current levels. The education indicator is now framed as a measure of years of schooling, with the education of current and future generations receiving equal weights.

There is extensive literature today on the drawbacks of GDP per capita as a measure of well-being and economic progress, most recently summarized by the report of the Commission on the Measurement of Economic and Social Progress led Joseph Stiglitz, Amartya Sen and Jean-Paul Fitoussi (2009). Some of these criticisms relate to its inadequacy as a broader measure of welfare or capabilities, while others examine its ability to measure an economy’s capacity to provide goods and services to its people. Most notable among GDP’s many shortcomings are that it does not measure non-market activity, such as subsistence agriculture or household work; it does not register productivity growth in public goods; it makes no adjustment for the depletion of natural resource; and per capita GDP is often unrepresentative of the typical person’s actual income.

Are all of these criticisms relevant for the income dimension of the HDI?  In order to put some of these criticisms in perspective, it is important to consider why income was included in the HDI in the first place. As discussed in the early reports, income differs from health and education in that it is not a direct measure of a capability, but rather an input into these capabilities. The relevant capabilities include adequate nourishment, shelter, and access to participation in the basic life of society. Attaining these capabilities requires people to have command over a basic set of resources necessary to ensure a decent standard of living.

In other words, the measure of income in the HDI is a proxy for the typical person’s command over resources that they can use to acquire goods and services, and save for the future. Note that this is very different from the idea of GDP as a measure of how much an economy can produce. Gross Domestic Product does not include transfers received from other countries, for example, while it does include transfers sent to other countries, as it is meant to capture the goods and services produced within the national territory. But to the extent that one wants to capture command over resources, one would like to include transfers from abroad and exclude transfers to other countries. The new HDI does this by replacing per capita GDP with per capita Gross National Income (GNI), which is equal to GDP less primary income payable to non-residents plus primary income receivable from non-residents.

Combining achievements into an aggregate measure


Perhaps the most radical innovation introduced in 2010 was the shift to a geometric mean in order to aggregate dimensional indices. This marks a significant conceptual change in the way in which one conceives the relationship between different dimensions of capabilities.

The shift to the geometric mean addresses the issue of perfect substitutability: This was a problematic assumption of the old formula, in that the level of priority to be given to a dimension was invariant to the level of attainments.

The new HDI attains a compromise by adopting a functional form that is between the extremes of perfect substitutability and perfect complementarity. There is a distinct advantage to the geometric mean – unlike the arithmetic mean or other forms of aggregation with a non-unitary elasticity of substitution, the rankings produced by the geometric mean are invariant to the scale in which each variable is measured.

Consider the comparison between Japan and France in the old functional form: Japan has a higher life expectancy than France, with a higher education index and also a slightly higher per capita income. With the additive functional form, Japan ranks 6 positions above France. But if we were to change the maximum for life expectancy by 10 years, we see that Japan drops by 2 positions relative to France. This is problematic as thresholds are likely to be continuously updated s the world experiences progress. This problem does not affect the new HDI, however. In the same example but with the new functional form, the relative ranking between Japan and France is unaffected by the choice of maxima, with Japan always occupying three positions above France.

The choice of minimum, in contrast, will continue to have significant implications for rankings even in the geometric mean. This makes sense because the minima have a concrete meaning as subsistence indicators – values below which it would be very difficult for a society to survive over time. In the new formula, these subsistence values are derived from the historical and epidemiological literature.

Addressing other critiques


Another concern with the HDI relates to its potential redundancy, given the high correlation of the index with its underlying components. The 2010 Human Development Report addresses this criticism head on, by showing that this criticism is fundamentally unfounded. It explores how the HDI illuminates investigations of advances in human development over time. The 2010 Report argues that the picture of development obtained from looking at the past forty years through a human development lens is vastly different from that obtained through a sole focus on economic growth.

To take just one example, China is by far the economy with the greatest rate of economic growth in that period – its annual per capita growth rate of 7.6 percent significantly exceeds not only the world average (1.8 percent) but also that of the second highest growing economy (Botswana, 5.6 percent). But China ranks 64th out of the same 135 countries in terms of average improvement in non-income HDI, and is one of only 10 countries in the world with a lower gross enrolment ratio today than in 1970. More generally, only three of the top 10 fast growing economies for the period of study are also top improvers in the HDI – the majority of countries that get into the list of top HDI performers do so because of their performance in terms of health and education, not income. If one looks at changes over time instead of levels, the redundancy argument disappears.

Critics have argued further that the HDI categories that used to classify countries into low, middle and high levels of human development are not robust to data revisions over time. This was because the numerical HDI thresholds used to define these categories (0.5 and 0.8) were kept fixed even after changes in methods and data. This valid criticism was addressed in the 2010 Human Development Report by changing the method used to create country categories. Rather than using fixed thresholds, countries are now grouped by quartiles of the HDI distribution. The problem of change in development groups due to data updates or formula revisions is no longer an issue, because countries are grouped by HDI rank, not value.

Some recent criticisms have pointed to “paradoxes” that would arise if a society tries to maximize the HDI. These criticisms are based on the incorrect premise that the HDI is a welfare or utility function. There are at least three reasons why one should not recommend that societies try to maximize the HDI. To begin with, not all elements of human development – such as political freedoms, equity or sustainability - are included in the HDI. Second, the expansion of capabilities is only one of the reasons why people may value the components of the HDI. People may enjoy the luxuries that come from higher levels of income even if these don’t contribute to them leading substantively different lives. Third, the HDI is severely limited by data constraints, and thus one can at best expect it to give a rough approximation of a country’s relative development level and progress.

Enabling debate


The Human Development Reports have always stressed that the HDI is not and was never intended to be a definitive measure of development. By design, the HDI is a simple measure that uses available international indicators (that are inevitably imperfect), that provides a broad vision of the progress of people. The aim is to stimulate discussion and debate, including about the appropriate ways to think about wellbeing and progress. As symbolized by the Human Development Report website’s introduction of online tools that allow users to construct their own indices with variable weights and indicators, the HDI should be understood as a starting point of a global conversation about development, rather than an endpoint.

Read the research paper: PDF Inline (GIF) The HDI 2010: New Controversies, Old Critiques [660 KB]

No comments:

Post a Comment