What Does the Human Development Index Measure?

There’s been some interesting recent commentary on the Human Development Index. But first, some background. This index is calculated each year by the U.N. Development Program as a summary indicator of “Human Development,” combining data on life expectancy at birth, adult literacy, educational enrollment, and average income (measured as G.D.P. per capita). And earlier this week, Catherine Rampell noted a recent effort by the SSRC-funded American Human Development Project to develop a Human Development Index, for U.S. states. Philosophically, it is an attempt to broaden the development debate beyond G.D.P. But does it succeed?

Andrew Gelman isn’t convinced by this effort. Over at fivethirtyeight.com, he takes a close look at this new state-based index, comparing a state’s ranking in this new state-based human development measure and its ranking on average income. He finds yields an 86 percent correlation. Forget the high-falutin’ language of “human development,” Gelman argues, “You’re pretty much just mapping state income and giving it a fancy transformation and a fancy new name.” Over at Economix, Catherine Rampell responded, noting that the “U.N.’s index was not designed to capture the levels of variation that would occur within a single country. It was designed to make international comparisons.”

Given this debate, I wondered whether Gelman’s critique might also apply to the U.N.’s original cross-national Human Development Index, so I downloaded the latest data. The graph below compares a country’s ranking on the human development index with its ranking on average income. The correlation between the two is even stronger — a massive 95 percent! For all but a handful of countries, your ranking on average income is the same as your ranking on this multi-dimensional index.


For comparison, here’s Gelman’s graph for U.S. states. (The only way in which it differs is that Gelman gives the richest state the lowest ranking, rather than the highest ranking.)


For all the work that goes into the Human Development Index, it just doesn’t tell you much that you wouldn’t learn from simple comparisons of G.D.P. per capita. But you do get the veneer of something broader, with a normatively loaded name for this index.

But there’s another reason to be suspicious of the state-based human development index. Some commentators have been comparing the scores of individual states on the state-based index with the international index, which yields newsworthy bites, like “Mississippi has an H.D.I. level roughly on par with that of Turkey.” But the two indices aren’t comparable. Dig deep into the methods used to construct the AHDP’s state-based index, and you’ll find that not only are the inputs different, but so are the formulae. In principle, one could build indices to compare Mississippi and Turkey, but comparing different indices is comparing apples and oranges.

Economix has a longer discussion on the relevant data issues, here.


Looking at your country comparison, it seems to me there would still be value in the process. With 95% correlation, shouldn't we look at the outliers to the correlation? What is it that GNQ, GAB, BWA, and AGO are doing that puts them so low on mortality/education rankings relative to their GDP? And what is Cuba doing to be on the opposite side of the curve?


Thank you. Neat graph.

When you get into the methodologies of rankings, you find many are garbage with a political agenda. And many don't disclose their methods at all. I recently took apart an annual survey on women's property rights and other gender issues, issued by an institute in Jerusalem. One of the metrics was political stability and the method is described as: "Combines several indicators which measure perceptions of the likelihood that the government in power will be destabilized or overthrown by possibly unconstitutional and/or violent means, including domenstic (sic) violence and terrorism." On this measure, the US ranked 50 out of 115 countries. Tunisia, by comparison, was 59. I'm not saying the results were explicitly biased but rather that the underlying method must be lousy.

As for garbage methodology, look at college rankings. One of my favorite is an engineering ranking that consists of a questionnaire sent to the Dean and 2 senior faculty. What does that measure?

BTW, another good example of a simple relation that gets obscured is earnings of graduates from public schools. By far the main contributor is location, which anyone who reads this post should know. Go to school in a higher income state and you make more money.


Dean Rodrigues

I disagree with the conclusion the HDI is without use. High correlation is to be expected - obviously richer countries (or states) - i.e. those with high GDP - are more likely to provide the education, healthcare, social services etc. needed to produce a high HDI. So it is not in the slightest bit surprising that you develop a strong correlation.

Furthermore, it is even less surprising that the international figures develop an even higher correlation. When samples have a clear underlying relationship with one another (which HDI and GDP clearly do, not least because the former includes the latter as one of its metrics), then increasing the sample size will always serve to increase the overall correlation - much like financial portfolio diversification.

As such, it should be obvious that HDI is not meant to be producing spectacularly different trends - it is meant to be used alongside GDP to illustrate intricacies in the socio-economic structure and institutions of countries (or states) - especially those based on inequality. Notice how lots of the "stray" nations on your comparative graph are those rich with resources? GDP would suggest that they are happily middle-of-the-road developing nations. But HDI illustrates that the wealth is not being translated into social development, which implies dictatorships, corruption, and/or massive inequality. Which stacks up with the evidence.

Applying this to US states, compare Hawaii with Wyoming. GDP would suggest that they should be very similar in development. But look at the intricacies - the real story - and it becomes clear that the institutional structure and or socio-economic makeup of Wyoming is not conducive to applying their extra wealth to social development. Thus it scores an HDI of 12, vs Hawaii's 45.

I whole-heartedly disagree with the conclusion that HDI is just a regurgitation of GDP. In is a tool to be used in parallel with GDP, and as such is very useful indeed.

Dean Rodrigues
Pembroke College, Oxford


Eric M. Jones


So the most liberal states scored the highest....no surprise.

Tom Smith

"nate" asked: "What is it that GNQ, GAB, BWA, and AGO are doing that puts them so low on mortality/education rankings relative to their GDP? "

Presumably HIV prevalence is a big part of the answer here. I don't know what GNQ and GAB are; but BWA = Botswana, ZAF = South Africa, NAM = Namibia, AGO = Angola (?), SWZ = Swaziland. Those countries have the highest HIV infection rates in the world, and consequently have relatively low life expectancies from birth. This explanation deals with many of the outliers.


@Eric Jones

Who do you think is pushing this index, liberals or conservatives?


I think this just shows there are other factors that we feel might be considered in something called a "human development" index. I think we have a gut feeling that GDP shouldn't so accurately measure our humanity, but why? If so, then what are those factors that aren't dependent upon GDP like education is?

Should happiness be included? Stress and anxiety? Crime rates? Freedom of the press?

I have no idea, but if we're not OK with such a correlation to GDP, then why not? There must be a reason.



that is the whole reason this has come about, and the reason why it is garbage. The point is that it was made by someone trying to make red states look third-world.
I'm a liberal and I know people with college educations voted Dem in '08. No duh. Read the 538 post.

The thing to take away from this is how much the other two thirds of the index (education and life expectancy) are dependant on GDP/person.


well, at least now we know what's wrong with missouri


I haven't taken a deep look at these methodologies, but it seems that you are going to incorporate some urbanized state bias by not deflating each state's GDP with it's respective costs (to my knowledge, there's not state specific GDP deflator). So it may be the case that some of the more rural states, which likely have lower costs, get the shaft.

Mark B

@ Dean who said...

"As such, it should be obvious that HDI is not meant to be producing spectacularly different trends - it is meant to be used alongside GDP to illustrate intricacies in the socio-economic structure and institutions of countries (or states) - especially those based on inequality."

But there are measures of inequality that can be used directly and intuitively rather than the HDI to GDP and see where they diverge three-step process. Of course measures of inequality do have some controversy, however I would be surprised if they have an inflated error compared to the three step process you describe.

Has anyone factor analyzed the measures used to create HDI? If they all load well on one factor (which I think is likely) then it would suggest that the HDI is a good measure that just combines several highly correlated variables.

It also isn't surprising that GDP and HDI correlate so well, because one contains the other... Besides the intuitive notion that countries with more money can have better education and medicine the HDI actually contains GDP per capita. The correlations reported above seem to be GDP correlated with a different version of itself. How do those correlations look if you remove GDP from the HDI calculation?

Another further consideration. How well to the individual-level versions of these variables correlate? How do they factor analyze?


Adam S

> For all but a handful of countries, your ranking on average income is the same as your ranking on this multi-dimensional index.

I'm not sure if this is consistent with the data in the HDI report. The final column on page #229-232 of the 2008/2008 report lists the GDP per capita rank minus the HDI rank.

A quick glance shows only 3 or 4 countries out of 177 for which the HDI rank and GDP rank are exactly the same. For some (Norway, Switzerland) the differences are so small that the meaning of your post is not significantly changed, but for others at the top of the list (Australia, the US) the difference is significant.

Not to mention some countries (Cuba, Botswana, South Africa) which range from 40 to 70 positions difference between GDP and HDI ranks.

Are we talking about the same data?

S. Burd-Sharps, K. Lewis

Although income is an important part of well-being, it fails to tell the whole story of how people are faring. Our work in the U.S., using official government statistics from the U.S. Census Bureau and the CDC, shows why. Let's compare two congressional districts, Vermont's only district and Nevada's First District. Median personal earnings in both congressional districts are nearly the same: about $26,300 per year. However, the two districts are separated by 223 places on the American Human Development Index (on a ranked list of the country's 436 congressional districts). Why? A baby born in Vermont today can expect to live, on average, three and a half more years than a baby born in Nevada's First District. And while about one in ten adults in Vermont today did not graduate from high school, in Nevada's First District, that number is about one in four. For people to flourish in our globalized, information-intensive world, completing high school is a bare minimum. Of course in order to fully understand why these two districts with nearly identical income levels have such different outcomes in health and education, one would need to analyze a fuller range of indicators, as well as historical circumstances. But the data make clear that money is buying neither a better education nor a longer life for the average Nevadan. This is one value of the American Human Development Index.

The mere fact that income is responsible for one third of the human development index guarantees that there will be a strong correlation. In addition, income is strongly correlated to pretty much any social indicator one can think of.

While The Measure of America: American Human Development Report 2008-2009 is a first-ever human development report for an advanced economy, human development reports have been written for nearly two decades in over 150 countries around the world. They tell the story of the power of this tool. With the human development index as the centerpiece, they help stimulate a more reasoned dialogue on difficult issues based on objective evidence-- from post-apartheid racial disparities in South Africa, to Russia's role in regional development, to the roots of conflict in Colombia. In the Philippines, the HD Index is an official government statistic, and the annual ranking helps to build healthy competition among the country's provinces to translate GDP and wealth into improved opportunity and freedom. In the best of cases, as happened with the AIDS epidemic in Botswana, these reports can actually change the course of history, save lives, and bring greater peace and prosperity.

Why has the human development index been so successful in challenging GDP as an alternative? One of its strongest attributes is its simplicity. It uses only three variables to capture well-being: a long and healthy life, access to knowledge, and a decent standard of living. Most people would agree those three are the basic ingredients of a life worth living. But its simplicity is also its weakness. It certainly does not capture the full range of capabilities central to a fulfilling life-things like the ability to live in a clean environment, to live free from violence, to participate in the decisions that affect us, and more. No summary index can tell the full story. But it can be a great way to get a conversation started.

----Sarah Burd-Sharps and Kristen Lewis, co-authors of "The Measure of America: American Human Development Report 2008-2009"



I think one of the most interesting parts about the HDI are the outliers.
What we see there is that
1) the oil states (the little cloud off the line on the bottom left) aren't doing great.
2) HIV rates, as Tom Smith notes, are a major issue, in a way especially for mid-income countries (GNQ is Equatorial New Guinea and GAB is Gabon, these are ISO country codes
(I find it very amusing that there is an int'l standard for country codes
3) Cuba does exceptionally well - which is in line with 100% literacy and developed country like life expectancy there, but also shows that HDI does take into account neither political liberty, nor middle-class prosperity.


Wasn't Iceland number one last year?


Black-Sholes formula for Human? No wonder college is going to down size from 4 year to 3 year program in order to help student in the hard time... HDI acceleration... cute.


Plotting rank vs rank is a pretty poor way to show correlation between an index and a variable. If you want to show that the index is primarily dependent on one factor you need to scatterplot. You're likely to find (as seen on fivethirtyeight) that the neat correlation disappears - because the index has a lot more subtlety than just a GDP ranking.

Eduardo Martins

There are two separate issues that must be addressed here.

First, there has been a lot of confusion regarding the U.S. states HDI rankings. As has now been clarified on Economix, The Map Scroll, and a few other places, the set of state HDI data that was used in the international comparisons was NOT produced by the American Human Development Project, nor was it published in our book. It is of an unknown source, and is not in any way endorsed by us.

Our index, called the "American Human Development Index", was designed to capture the human development variations within the context of an industrialized nation, and, while based on the UN's HDI, uses different indicators that more accurately reflect the conditions of a nation like the US. Additionally, the goalposts used in the normalization of each of the three components of the index were chosen so as to maximize the variations inside the U.S. We explicitly state that this modified index is not comparable to the UN index, and in order to avoid this sort of comparisons we decided to use a scale of zero to ten, instead of the UN's scale of zero to one.

Justin Wolfers states that "there's another reason to be suspicious of the state-based human development index. Some commentators have been comparing the scores of individual states on the state-based index with the international index, which yields newsworthy bites, like 'Mississippi has an H.D.I. level roughly on par with that of Turkey.' But the two indices aren't comparable. Dig deep into the methods used to construct the AHDP's state-based index, and you'll find that not only are the inputs different, but so are the formulae."

If he acknowledges that the two indices are not comparable, and that the inputs and formulae are different, it is beyond my comprehension how the fact that someone wrongly tried to compare the two indices is "another reason to be suspicious of the state-based human development index." If someone compares states' median personal earnings to countries' GDP per capita, should I be suspicious of the accuracy and validity of the states' median personal earnings figures, or of the inadequate and unsound analysis that is trying to compare items that are not meant to be compared? The answer to this question should be obvious to anyone with the most elementary training in scientific methods, let alone to someone with an extensive background and impressive academic credentials as Wolfers.

The second, and more general, point is about the usefulness of the HDI against simply using an income-based measure such as GDP per capita. My co-authors and several other readers have addressed this. I will just add that too much attention is being given to a (bogus) map, and the human development framework is much broader than the HDI itself. As Amartya Sen, winner of the Nobel Prize in Economics and one of the creators of the human development framework, said on his paper "A Decade of Human Development":

"it would be a great mistake to concentrate too much on the
Human Development Index, or on any other such aggregative index. (As, perhaps, the principal author of Human Development Index, I say this with some hesitation, but no less firmly for that reason; it is not a case infanticide any way, since the infant has now grown up and can take the rough with the smooth.) These are useful indicators in rough and ready work, but the real merit of the human development approach lies in the plural attention it brings to bear on developmental evaluation, not in the aggregative measures it presents as an aid to digestion of diverse statistics."

Eduardo Martins, co-author of "The Measure of America: American Human Development Report 2008-2009", and Statistics Director, American Human Development Project.



So obvious that it is invisible, the creators of these indices committ two arrogant errors:
1) Calculating ONE MEASURE with very few indicators, and 2) RANKING across localities with the same measure.

Rather than seeing themselves compared unfavorably to others, residents of a country, state or electoral district would much rather see their own community compared to itself over time - these indices do not capture true trends or provide a guide to action.

The development tools I take as worthy models bring together hundreds of indicators and work with the neghborhoods living under the measured conditions to interpret and take action. See for example http://www.thehdmt.org/master_list.php

Listen to Hans Rosling: "I have a neighbor who knows 200 types of wine. He knows everything. He knows the name of the grape, the temperature and everything. I only know two types of wine -- red and white. (Laughter) But my neighbor only knows two types of countries -- industrialized and developing. And I know 200, I know about the small data."



You are using rank correlation!