Reading this before the long weekend may save your life
Self-sufficient charts

A luxury we can afford to miss

Raghu R. alerted us to this bar chart from, reporting survey results in collaboration with Gallup


The poll showed that most of the most pro-U.S. countries are African, and this chart displayed the levels of approval in Mali, Tanzania and the Democratic Republic of the Congo, all ranked fifth.

This effort is an obvious failure, and the resulting chart, particularly the background image, is too much baggage for the simple data set it contained.



  In our Trifecta Check-up, this chart fails on two fronts: it uses the wrong chart type, and it asks a meaningless question.

  • The percentages are proportions of each country's population, and it is silly to stack them up as if they could be added.
  • Because the three countries are all ranked 5th -- in each country, 89% of the respondents approved of the U.S., there is no point in showing three series of data! Each series say the same thing.

I am not offering up the following profile chart as an alternative but it is here to illustrate why there really are only three data points, and so a chart is a luxury we can afford to miss.




Not everything can be solved with a line chart. Line charts are generally used to show a progression. This is not a progression. Its a comparison. Use bars. Just not stupid ones like in the original.


Totally agree. This is not a progression, so a bar or pie charts would work best and show the data correctly

Nick Cox

Yet another objection to the line chart, at least as implemented here, is the introduction of a legend (key). Legends are at best necessary evils to be omitted where possible. The back and forth -- Mali is this colour, etc. -- is dispensable.

This dataset is also so simple in structure that a table would do fine.

Nick P.

I think the point here is that you might not need a chart at all - since you are showing the exact same data for all three countries...


Nick P. got the point. Yes, I agree the line chart is not the best here; that's why I said I wasn't offering it as an alternative.

Nick Cox

Well, yes and no. If I read your line chart correctly, the disapprovers vary from about 7% to 11%. Whether that is about the same, considering, or an interesting deal is one on which experts in survey statistics and African politics might have something to say.

My instinct is to say with you that it's not a big deal. In addition to the usual problems of sampling error and non-response, I'd add a language question. It is difficult to believe that the surveys were all administered in the same language, and that the words used are exactly equivalent in meaning, although you'd expect Gallup to be aware of that issue.

But the numbers are not identical.

Ronan Conroy

Let's start with what we are trying to tell the reader: that approval is high (89%) in all three countries. The don't know and refusal rates are so small that they really don't matter, so the disapproval rates can easily be imagined, once the approval rates are seen.

Three bars, one for each country, showing the approval rating, make the point: approval is 89% in each country.

Of course, this would be better written in a sentence than shown in a graph, as graphing this information violates an essential principle of charts: it shows less than one piece of information per square centimetre.

The comments to this entry are closed.