« Boys, retention, and multiple regression | Main | Deposit insurance, bank runs, international currencies, and the inflation tax »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Frances - I realize these aren't "your" charts, but a re-scaling and truncation of the vertical axes in each case would be nice.

Hey! Another degree or two of global warming and I might move there. :) North Battleford, here I come!

How does Canada's low rank in the education index square with things I've seen lately saying that Canada has the highest percentage of adults with a post-secondary education? For example, this article

I feel I should clarify, I meant "low rank" within the countries used in the graph.

I'm surprised at how poorly Japan performs on the education index. Is it because of an older population?

Dave: I disagree. Having the vertical axes start at zero shows how close together all those countries really are, so that very minor changes in definitions and formulas can make a big difference in the ranking at the top.

Well I think that Canadian performance can be influenced by a fact, that Canada has big influx of immigrants. I feel there can be some correlation between immigration and HDI.

Another thing that grabbed my attention is a position of Ireland. I thought they are doing poor after they were severely hit by financial crisis in 2008. But your graph shows that their economic development since 90s positively influenced Irish population positively in many ways.

Dave, what Nick said. The raw numbers for the 1990s aren't readily available on-line, so you're stuck with these graphs. Free pizza and all that. The more annoying thing about them is that the colour scheme is determined by the order in which the countries are chosen, and I could never manage to to choose the countries in the same order every time.

Randy, this is an excellent question, and I really don't know the answer. A few observations. (a) In Quebec, secondary school ends in grade 11, at which point students move into a CEGEP. Especially in recent years, the proportion of students going onto CEGEP seems to be falling - there was some statscan data on this recently (b) In Ontario, secondary school used to end at grade 13, and now ends in grade 12 - the HDR #s are average years of schooling, which isn't exactly the same as % of people with post secondary degrees. (c) the average is the average of people who go to university and people who don't. I suspect what's dragging Canada down is the low educational attainment of the people who don't go to university - people up North and in rural areas of Canada, and especially our Aboriginal population, who drop out of school at a young age. (d) A lot of the people with university degrees are immigrants. Presumably the calculations about the projected educational attainment of young Canadians are based on the educational attainment of people who grew up in Canada, not people who came here as adults.

But all of this is speculation. I'd have to dig deeper into the data to find out.

Nick, doesn't Japan's university sector run on the old British model i.e. brutally hard to get into because there are so few places, but once you're there, fees are low, and you have three or four years to relax and recover from all of the efforts you put into getting there in the first place?

John B "I feel there can be some correlation between immigration and HDI."

What would be the mechanism through which immigration might affect HDI? I can see that immigrants might raise or lower the average education level of the population, or cause per capita GNI to rise (if they're hard working and enterprising) or fall (more people means less income per person), or cause average life expectancy to rise (Japanese immigrants) or fall (some group with lower life expectancies). But I can't any reason to think that Canada's falling relative HDI score is due to immigration. It might be, but it might also be the case that Canada's HDI score would be even lower without an influx of immigrants. It's just really hard to know.

Fascinating post--I had no idea how this was constructed. Your closing comment makes the most sense to me, namely that this is about expanding the discussion. Even if it turned out the educational and years-lived indices were highly correlated with GDP, perhaps HDI still brings on board those who are perpetually skeptical of GDP.

Of course such a measure has to be taken with a grain of salt. At best, it is a composite of three reasonably OK proxies. Just as we can easily see how years of education is a weak measure of human capital, unadjusted years-lived is not by a long shot a good measure of population health. The question is more about broad comparison. Of course, it's important to point out that, per your analysis, the effect of education on ranking does not necessarily imply that education disproportionately determines HDI; it's just the most weakly measured and hence volatile dimension.

Shangwen Thanks for the kind words on the post.

"the effect of education on ranking does not necessarily imply that education disproportionately determines HDI; it's just the most weakly measured and hence volatile dimension"

There are a couple of issues. One is the variance of the education ranking, that is, how spread out the countries are. The other is the volatility, that is, the amount that it changes over time. Hhe differences in the health index values in the very high human development countries are mostly are in the 0.02 to 0.04 range. The differences in the education index scores are more like 0.04 to 0.08. The fact that the education scores are more spread out means that it's hard to compensate for a low education score with a high life expectancy score. If the measure used was log of years of education all of the countries would be clustered up together, and education would make almost no difference to a country's HDI ranking.

The volatility of the education numbers could be an artifact of measurement, but it could also be that the state of knowledge can be changed more easily than the state of health. If for some reason it was decided that every Canadian needed how to use a yo-yo, there could be publicly mandated yo-yo training, and Canadian yo-yo knowledge would rapidly increase. It's much harder to budge life expectancy numbers. Of course one could argue that knowledge is different from education, and like health is not easily changed. In this case, yes, the volatility does reflect measurement error.

Turns out the Education stats in the UN HDR are based on UNESCO numbers from here http://stats.uis.unesco.org/unesco/TableViewer/document.aspx?ReportId=121&IF_Language=eng&BR_Country=1240&BR_Region=40500. Two things worth noting. First, they count pre-primary education, i.e. child care, where Canada does poorly by international standards. Second, as far as I can tell they haven't been updated since around 2002. The lack of updating would explain why Canada's education index score hasn't changed in about 7 years.

John B: 23% of Australians were born somewhere else. So, we have very high immigration and it does not stop us moving ahead of almost everyone on HDI rankings.

Of course, we are world champions at cherry-picking migrants, being just about the only developed country where migrants do better on education tests (on average) than locally born.

After Lorenzo from Australia came in,

let me add my German 2 cents:

Our data in

are in one case as old as 1991, the school life number is from 1997,
no data available for Literacy ? For Germany?

This completely disfunctional organisation should be completely dismantled. Now.
That a database like that is in such a state, is evidence of the general intellectual corruption there. A script kiddy before finishing his secondary education can do a better job, just mailing the questionaire to the national contacts for filling out, with a weeks work per year.

These folks actually tried to overrule the will of the people here, about building a bridge.

Beyond that, I do not believe that x numbers of years is above a certain threshold of maybe 12 years any kind of proxy of social and education status.

We are cutting here the number of years to university from 13 to 12. The "Main" Hauptschule should be done in 9 years, and after that people go on to paid / working 3 year apprenticeships. Some of them later do their "Meister" if they want to run their own shop. How is this reflected in these numbers?

With the European Bologna Process Reform, we actually reduce the number of years people spent in universities, and expect better results with lower HDI rankings : - )

We think that life long learning is much more important, and it is not us synchronizing to these K-12 + x more years school model. We export our very successful apprenticeship model.

“The life quality goalpost is $89,990.52 per capita, according to the HDI documentation, which is roughly the income level of Qatar…”

So…Country X creates a highly efficient economy that provides every resident with easy access to all the goods-and-services needed to enjoy a decent, satisfying life. Plus a short workweek, long vacations and early retirement. Plus a generous social safety net to protect each resident against life’s vicissitudes. But all to no avail. Because, when it comes to human development, the only thing that counts income-wise is keeping up with the Jones - or, in this case, the al-Thanis. If they get richer but the Xians don’t then the Xians become less humanly developed. At least, so say the UNDP’s statisticians.

Utterly depressing…apply the same reasoning to individuals and I will always be a zygote compared to Warren Buffett.

Giovanni - the Human Development Index gives 1/3 weight to life expectancy and 1/3 weight to education, so countries with strong social safety nets dominate the top of the rankings. In terms of giving weight to things other than money, it does far better than measures such as per capita GDP.

The UN HDR also calculates many other measures of human development, e.g. a gender-equity-adjusted HDI and an inequality-adjusted HDI, that address some of these concerns about life quality. Unfortunately, the idea that there are many alternative ways of ranking human development, and it's complicated doesn't make for good newspaper headlines, so these other indices get far less coverage.

There are many flaws with the HDI. But compare it to the alternatives!

On education, I wonder if there are 2 issues:

1. immigration - like the US and australia, we do have high immigration - and we do we let in economic immigrants based on years of education, BUT refugees, spouses, and sponsored family members (parents and grandparents) likely have low years of education

2. aboriginals. this is where we genuinely have a problem - a growing aboriginal population but high levels of drop-outs.

Thank you. These sorts of posts help me understand current issues.

In terms of measuring quality of "education, isn't the switch from an outcome based measure (literacy) to an input based measure (years of education) an improvement that makes things worse? Outcomes are what we care about, and are often poorly correlated with inputs.

The is obvious from the health component of the HDI - no one would ever suggest replacing longevity (an outcome-based measure) with spending-per-capita or doctors-per-capita (input-based measures). The fact that the US would instantly rocket to the top of the index, apart from being off-putting to many proponents of the HDI, would starkly illustrate the problem with that approach, given their relatively poor health outcomes. Ditto for living standard (we care about GDP per capita, not capital-stock or workers per capita). So it's strange that they've taken a different approach for education.

I have no doubt that the shift was, at least partly, motivated by a lack of differentiation between developed countries in terms of literacy rates (as Frances notes, there isn't a heck of a lot of difference at the top. Not sure why that would be a problem, but I can see it bothering people who make a living ranking things). But surely the better approach would have been to try to find better measures of educational output (is there some international measure of effective literacy or numeracy - i.e., can this person comprehend the editorial page of whatever the national equivalent of the Globe and Mail is? Even secondary school graduation rates might be better). This strikes me as a "well, this is what we can measure, so this is what we'll use" kind of approach.

Finally, at least part of the blame for Canada's "declining" educational score could be attributable to Ontario's decision to slash publicly funded years of schooling by 8% (14 to 13, including senior kindgarden) when it abolished grade 13 in 2003. That the previous curriculum had managed to squeeze 4 years of secondary education into 5 years, such that the cutting off a year didn't neccesary result in less education, wouldn't be reflected in the HDI measurement. Again, that example emphasizes the uselessness of input based measures.

Bob: regarding the appeal of input measures. We still have some vestigial ability as a society to divorce health from health-care, which is why life expectancy is probably not disputed much as a component. But I don't think you could say the same for education (or learning, or human capital). Of course it's true that a lot of human capital change is carried out in institutions, and extra-institutional improvements (teaching yourself to code, on-site training, relocating to a better job market, etc.) are hard to observe. But I think there is also a very strong bias still that schooling=learning.

Shangwen: "But I think there is also a very strong bias still that schooling=learning."

I have no doubt that there is a perception that that is true, which is probably why policymakers are always eager to adopt that logic ("Whee, we've hired more teachers, and we're paying more, aren't we the education government?" umm, can Johny read?). But whether that's true or not is a different question. Did educational outcomes in Ontario decline by 8% when the government did away with grade 13? Will spending another couple of years pursuing a near-useless degree from a third-rate suburban French university (because they can't get a job) improve the educational outcomes of an unemployed french youth? I'd be willing to bet no. But the HDI says otherwise.

Bob "it abolished grade 13 in 2003"

Remember the HDI numbers are based on UNESCO figures that haven't been updated since 2002, so that shouldn't be it. Also all day kindergarten plus increasing enrollments at universities and colleges plus expansion of Quebec's $7/day child care program should have boosted up years of education enough to offset the grade 13 effect.

Since my livelihood depends on the equation schooling=learning I'm not going anywhere near this discussion.

btg: I don't think #2 (aboriginals) amounts to much impact. First, the official aboriginal population is replaced every 2-4 years by immigration. Second, if you look at the definition of the components (http://hdr.undp.org/en/statistics/hdi/), the education input measure has two parts: "mean of years of schooling for adults aged 25 years and expected years of schooling for children of school entering age". If you read further it looks like detecting drop-outs would be virtually impossible. Lots of aboriginal kids are enrolled; fewer aboriginal adults go to post-secondary ed.

On the age-25 cut-off, that of course doesn't capture the large amount of education that can happen after 25, much of which is highly specialized and high-value. In a country where many people get advanced training into their 30s, that's a pretty weak measure even of just the input.

"Since my livelihood depends on the equation schooling=learning I'm not going anywhere near this discussion"

I'm sure in your case, the correlation strongly postive! :)

Interesting HDI critique from the World Bank here:


The change in education measurement hit Zimbabwe even worse...

Shangwen: just read the World Bank paper. Fascinating. Equating a gain of $ 0.51 with a loss of one year of life? Essentially meaning that oil at $ 200 (meaninig moving the United Arab Emirates goalpost) mean we are all living in a cave? If we can take as a country the UAE ( a couple of buildings plus some oil wells), why not counting two blocks of Bay St. and compute its HDI?
I always believed that such rankings were arrant nonsense ( don't start me on the Times Supplement on Education...), but they eally take the prize...

"I always believed that such rankings were arrant nonsense ( don't start me on the Times Supplement on Education...)"

The real problem with the HDI is that it's subject to the same caveats and problems as the indicators it tries to replace or supplement (such as GDP), with the added twist of being impenetrable and/or arbitrary. Ditto for even more ambitious (or ridiculous, depending on your take) schemes for measuring "Gross National Happiness".

The whole Human Development Report exercise is about making a whole lot of data readily available and bringing it together in one place. It's not the people who put together the Human Development Report who insist on the rankings, it's the journalists who are looking for a story, and readers who are more interested in satisfying nationalistic pride than understanding complex analysis.

And I'd go out on a limb and say the HDI is a better measure of human development than, say, per capita GDP. Given a choice, I'd much rather live in Norway or Australia than in Qatar, even though Qatar's per capita income is much higher than either of those two countries'.


Unfortunately, ranking is inherent in the HDI exercise, since the HDI scores are only meaningful relative to other HDI scores. Saying that a country has a GDP per capta of $40,000 dollars (or that life expectancy is 80 years or that literacy is 100%) is a meaningful statement - we can translate that into something that people can wrap their minds around, even if they know noting about the GDP (or life expectancy or literacy rates) of other countries. Saying that a country has a HDI score of 0.8936 isn't the least bit meaningful to anyone on its own. That newspapers and politicians jump on the "rankings" is a just a function of that characteristic of the HDI.

As for the usefulness of the HDI score, I'm not sure your example proves the point. After all, if Qatar had higher life expectancy and educational inputs (as well as GDP per capita) than Norway or Australia(and therefore a higher HDI) would you prefer to live there then? Yeah, me neither, so it's not clear that the HDI gives you much in the way of useful information in that case. In any event, its hard to argue against the proposition that looking at education (however imperfectly), life expectancy and GDP will give you a better sense of human development than looking at GDP alone, if for no other reason than that you have more information. More data will always give you a better picture of the underlying reality. Had they stopped at that, I wouldn't have any qualms. My problem is trying to bodge all that good stuff into one number and then pretending that the result has any meaning.


In my previous comment I was thinking the primary purpose of HDI was to track changes in a country's well-being over time. UNDP's approach makes much more sense if the purpose is to do cross-country comparisons for a particular period. In particular, the "Qatar problem" I complained about above has no bearing on such comparisons. I also agree that if policy and media people insist on using a catch-all measure to compare well-being across countries it is better that they use HDI rather than GDP.

However, a caveat. You noted in your analysis that UNDP's use of logarithms in dealing with income tends to discount the effects of income differences. I think this may lead to misleading results in comparing lower income countries. Suppose Country 1 and Country 2 have per capita incomes of $5000 and $7500, respectively, but have the same life-expectancy and education values. My calculations say that the difference in their incomes will produce only a small second decimal place difference in HDI, whereas I suspect that in the real world, at this level of income, the extra $2500 enjoyed by Country 2 residents would have a very big impact indeed on their relative well-being.

Yet more evidence that credential based measures of education leave things out:


The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad