Ideas reach governments most easily when they come with a number. In 1990 the United Nations Development Programme published its first Human Development Report, and with it the Human Development Index, designed by a team led by the Pakistani economist Mahbub ul Haq. The index set out to rival gross domestic product as the headline measure of national progress, and Sen was one of its main collaborators.
A crude number on purpose
Sen was at first skeptical. He doubted that a single number could capture something as rich as human capability, and he said so to ul Haq. Ul Haq's reply, as Sen later recalled it, was that only a simple number could rival GDP for the attention of politicians and the press, while the detailed tables in the report would carry the subtlety. Sen came to agree that a crude measure of lives beat a crude measure of output.
Three dimensions
The HDI combines three dimensions. A long and healthy life is measured by life expectancy at birth. Knowledge is measured by two schooling figures: the average years of schooling of adults, and the years of schooling a child entering school can expect to receive. A decent standard of living is measured by gross national income per person, adjusted for differences in prices between countries. Income enters in logarithmic form, so each extra dollar counts for less as income rises, reflecting the idea that money matters most to those who have little of it.
Each dimension is turned into a score between 0 and 1 by comparing a country's value with a fixed minimum and maximum. Since 2010 a geometric mean replaced a simple average in combining the three scores. The change matters. With an average, a very low score in one dimension can be made up by high scores in the others. With a geometric mean, a weak dimension pulls the whole index down more sharply, which fits the capability idea that one kind of freedom does not fully substitute for another.
A family of indices
Over time the report added companions to the HDI. The Gender Development Index, introduced in 1995, compares human development for women and men. The Inequality-adjusted HDI, introduced in 2010, lowers scores where gains are unequally spread, discounting each dimension by its inequality, so that a country with the same averages but deeper inequality scores lower. The Gender Inequality Index, also from 2010, tracks reproductive health, political empowerment and gaps in the labor market.
The most direct application of capability ideas is the Multidimensional Poverty Index, also launched in 2010 and developed with the Oxford Poverty and Human Development Initiative, led by Sabina Alkire, using a counting method she devised with James Foster. Instead of placing people above or below an income line, it records each household's deprivations across ten indicators in health, education and living standards, such as nutrition, school attendance, clean water, electricity and cooking fuel. A person counts as poor when deprived in a third of the weighted indicators or more. Because it is built up from households, the index shows not only how many people are poor but which deprivations they face together.
Criticisms
The HDI has drawn steady criticism. Its weights are a choice, not a discovery: why should health, schooling and income count equally? The economist Martin Ravallion showed that such indices carry hidden trade-offs. Because of how income and life expectancy are scaled, the HDI implicitly values a year of life less in poor countries than in rich ones, a result few of its designers would endorse if it were stated openly. Others point to what the index leaves out entirely: it leaves out political freedom, as well as personal safety, the environment, and the quality, not just the quantity, of schooling.
Defenders answer that every index simplifies, and that the HDI changed public debate by putting health and schooling beside income in national rankings. Countries with similar incomes can sit far apart in the HDI tables, which invites the question Sen wanted asked: what are some governments doing that others are not?
Worked example: averaging versus the geometric mean
Imagine two countries, A and B, with these dimension scores. Country A scores 0.7 on health, 0.7 on education and 0.7 on income. Country B scores 0.9 on health, 0.9 on income and only 0.3 on education, perhaps because most girls are kept out of school. A simple average gives both countries the same result, 0.7. The geometric mean instead multiplies the three scores and takes the cube root. For A, the result is still 0.7. For B, 0.9 times 0.9 times 0.3 is about 0.243, and its cube root is about 0.62. B falls well below A. The arithmetic expresses a moral judgment: wealth and long lives do not fully make up for a population largely denied education. Whether that judgment should be built into a formula is exactly the kind of question the capabilities approach asks us to settle in public, not leave to statisticians.
