Sunday, August 23, 2009

Chartjunk: Global Health Council's Chart on Spousal Abuse

The Global Health Council has "interpreted" some data from unicef on the "percentage of women who believe it is OK for their husbands to hit them" from developing countries. I'm going to completely ignore the far greater issue, spousal abuse, and focus on what is perhaps the worst use of graphical data that I have ever seen.

The whole point of using an information graphic is to graphically show how information is related. I'm siding with Tufte on this one - keep it simple and don't use graphics if you don't have to.
  • For this graph, they used some sort of pie-chart hybrid. Pie charts are appropriate when you want to show the relationship between "slices" that together make up 100% of the pie. On this graph we have a whopping 731.8% of data shown. Do us a favor and use a bar or line graph.
  • Is this all the data though? If you click the link provided to the unicef data, there are actually 67 different countries that were surveyed. Why aren't they included? What was behind the decision for the countries that were included?
  • Take a look at Rwanda (48%) and then at Jordan (90%). They should not be represented as more or less the same size.
  • What is the actual survey question? If you click to the unicef page, you see that it is "% 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 (2001–2007)". I'm not going to press on the semantics that much, but the GHC wording makes it sound like the women are referring to themselves rather than women in general.
  • Data is only available from some countries, most notably "developing nations". It is horrifying to read that the majority of women in some countries believe that spousal abuse can be justified, but for me, I need some comparison numbers. What would the survey say for the United States? Italy? I hope 0%, but without this data, the chart has less meaning.
  • To that point, to present this data as "shocking", which the GHC was clearly going for, they should not be including the "lower bounds" of Georgia and Serbia. This is biased opinion based off of my ignorance of Georgia and Serbia, but if they are meant to show what "normal" is, they chose the wrong countries.
So, how could you present this data more effectively?
  • Include all the data or group by distinct categories, like "Sub-Saharan Africa" or "Former Soviet Block" or "Top 10 Spousal Abuse Hot-Spots".
  • Use either a bar chart or no graph at all: the numbers and the subject matter speak for themselves.
  • Use full disclosure about what the chart is representing "% 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 (2001–2007)"
  • The result is something that is much more sobering, and in my opinion appropriate, for the matter at hand. This doesn't, however, resolve the lack of comparison data, but that is a rant for another day.


No comments: