Bob Rossney: September 2009 Archives

Global Health Magazine is a publication of the Global Health Council, an NGO devoted to health issues around the world (particularly women and children's health issues, HIV/AIDS, infectious diseases, and health systems).  

If you poke around in their kicky web site, you'll come across this shocking graphic:

Are you shocked?  I'm shocked.

First off, the purpose of a pie chart is to represent relative portions of a whole, not comparative scalar quantities. If one wedge is made bigger, another is necessarily made smaller, and the whole pie represents 100%. This chart suggests, bizarrely, that one way we could reduce the amount of bad things happening in Jordan would be to collect statistics for more countries, since adding countries to this graph would make Jordan's slice smaller.

This is compounded by the fact that the wedges converge on a point that's not in the center, and the shape of the pie is not circular. So, Serbia and Georgia's wedges should be the same size, since one represents 6.2% and one represents 6.9%. Are they? Well, first off, what's their size, anyway? Is it the angle the two sides describe? Is it the surface area? In a circular pie chart whose wedges converge on the same point, these two things are one and the same; not so in this chart. I think that it may be the angle, and not the area, that represents the number, which is unfortunate because Serbia's wedge is noticeably larger than Georgia's.

Actually, now that I compare Rwanda and Jordan, I think that the person who drew this is just an idiot, because there's no way of looking at those two areas that makes one nearly twice as big as the other.

And now, another issue: What do Vietnam, Laos, Georgia, and Iraq have in common? What about Serbia and India? Nepal and Honduras? Well, nothing, or at least, nothing that makes any sense in the context of this chart. So why are Serbia and India black? Why are Nepal and Honduras red?

Let's not even get into the information-destroying jagged lines and concentric arcs.

And now let's move beyond the terrible, terrible graphic design to ask another question: why aren't Tajikistan, Kosovo, Guatemala, and Nigeria on this chart? Is this a graph of the 16 countries with the highest percentages of women who believe it's OK for their husband to hit them? Or is it just a sampling of third-world countries? I'd suspect the latter - I mean, not that I have it in for Sudan, but given that they're a Muslim country in sub-Saharan Africa, I'd think their numbers would be closer to Ethiopia's or Jordan's than to Serbia's or Belize's. But really, who can fucking say?

And wait, I'm not done yet. There's one more thing that bugs me about this. Why is Laos labelled as "Lao PDR"? Yes, the long form of the country's name is "Lao People's Democratic Republic," but I don't see "HK" in front of Jordan or "SR" after Vietnam. What's up with that?

The answer to that last question, at least, can be found in the source data, which for some reason singles out Laos for this weird special treatment. And while there are dozens of countries with percentages higher than Serbia's (so much for the top-16 theory), there's no data for Sudan. Too bad, Janjaweed wives, no statistical recognition for you!

Also, this data is even sketchier than I thought it is. The source for the chart is an appendix to a UNICEF report. Nothing wrong with that, except that as the data set makes clear, these numbers come from five completely different studies, which sort of makes the already suspect three significant figures even more so. Not that anybody involved with making this chart gave a shit, but would it have killed them to pick numbers from the same study?

Rather damningly, the biggest, most dramatic number on the chart - Jordan's 90% - is the only number in the entire data set whose source is not a study but a preliminary report on a study.

And finally, there's the little matter of changing the meaning of the data. This report doesn't list the percentage of women who think it's OK for their husbands to hit them. It lists the percentage of "girls and women aged 15–49 who responded that a husband or partner is justified in hitting or beating his wife under certain circumstances." It's like the difference between saying it's OK to steal money and saying it's OK to steal money from thieves.

This is a very bad chart.