This post originally appeared on the Urban Institute’s Urban Wire blog
I had a hard time visually picking out the story in a simple chart published by the Financial Times earlier this week. There was nothing special or particularly complex about the chart—it’s a simple clustered column chart—but the fact that I couldn’t easily see the story made me think there must be a better way.
The FT article focused on how average household income in the United Kingdom is higher than in many other European countries, but that average household income in the bottom 20% of the income distribution is much lower in the UK than in the rest of Europe. The FT story mirrored the message in this report from the High Pay Centre, which used data from the Organisation for Economic Co-Operation and Development’s (OECD) Better Life Index.
The main graphic in both the FT story and in the High Pay Centre report was a simple clustered column chart:
Financial Times | High Pay Centre |
Notice the slight changes FT made to the High Pay Centre’s chart: they aligned the country names at a 45 degree angle instead of at 90 degrees; used a blue color scheme instead of orange and grey; reduced the number of zeros on the y-axis by converting the data to thousands of dollars; and made some other minor changes to the legend, gridlines, and title. Aesthetically speaking, I prefer the FT approach because it does a better job of reducing clutter and highlighting the data. I did, however, like the way the High Pay Centre helped focus the reader’s attention on the United Kingdom by using a different color for the country label on the x-axis.
But both versions of the graphic fail to effectively visually communicate the information.
Here’s the passage from the FT article that discusses the figure:
The High Pay Centre found that, with average household incomes of $53,785, the richest 20 per cent in the UK were the third richest in their bracket of all EU countries measured by the OECD, lagging behind Germany and France.
However, a very different story applied to the bottom 20 per cent in the UK whose incomes were much lower than in other, more equal countries with a similar average income.
The OECD calculates the average income of the bottom fifth of UK households at just $9,530, much lower than the poorest 20% in France ($12,653), Germany ($13,381), Belgium ($12,350), the Netherlands ($11,274) and Denmark ($12,183).
I found it difficult to navigate the text to FT’s graph (and the High Pay Centre’s version as well) for two main reasons. First, with 39 columns, it was a challenge to compare the different measures across the different countries. I couldn’t, for example, easily see how average incomes in the bottom 20% for the UK compared with the other countries. Second, the columns share one color palette, so values for the United Kingdom don’t stand out.
Instead of a clustered column chart, I tried a different approach. In this “dot plot,” I first rotated the graph—this rotates the country labels so that they are easier to read. I then plotted each data series as dots instead of columns and linked them with a thin line. In this way, I think it’s easier to track differences in each data series across countries and to make comparisons of the different values within countries. Finally, to better highlight values for the United Kingdom—the central facet of the FT story—I used shades of red for the dots and red for the text country label. One could imagine doing even more to highlight those values, such as brighter or different colors, or maybe some annotation.
For those of you interested in the technical aspect of constructing this version, there’s nothing too fancy going on here. This was all built in Excel using three scatterplots and some data labels. (There was a little bit of trickery with the axis to get the country labels to fit on the chart; if you’re interested, drop a note in the comments below or contact me directly.)
I think this remake shows that even simple charts can be redesigned to more effectively communicate information. In this case, especially when a single country is of interest, highlighting those data (not just the label as in the High Pay Centre version) can be especially important.
UPDATE (9/23/2014): Here is the Excel file containing the data
I agree that your version is better, but I wonder if it could be improved a bit further. I would move the labels to the left a bit (and maybe align them), and then add a vertical line through both the minimum and the maximum point for the U.K. That would show how much lower the lowest quintile is than many of the others, and the same thing with the top quintile.
Another option might be to align the bottom points on the x axis axis, so that you could compare the range. That would require changing the x axis from absolute numbers to percent above the minimum, though. That might make for an interesting comparison, though, since it looks like the U.K. has the largest range between the bottom and the top.
Thanks for your thoughts, Robert.
I think left-aligning the labels separates them from the first circle a bit too much; plus, I think having them next to the circle helps reinforce the different values. A vertical line through the UK values is an interesting idea, though I wonder if it would just clutter the graph–worth toying around with.
The alternative you suggest is certainly an interesting idea and would probably show the spread a bit better, but would it be too complex for the average FT reader? Straight dollar amounts are obviously easy to comprehend, so does moving to a percent change make it harder for people to understand? I’m not sure. (One could obviously do both.)
I don’t leave a response, however I looked at a few of the comments on A simple remake
for a simple chart | PolicyViz. I do have a few questions for you if
you tend not to mind. Could it be just me or does it look
like like a few of the comments come across like they are coming from brain dead
individuals? 😛 And, if you are writing on other places, I’d
like to keep up with you. Could you post a list of all of your
public pages like your Facebook page, twitter feed,
or linkedin profile?
Hi Jon,
I’d love to know how you got the country labels to align with the first data point.
Thanks!
Dana,
There is an invisible bar chart (empty fill) behind this whole thing with the labels at the end of the bar. In Excel 2013, I believe you can add the labels by selecting the appropriate cells.
Hi there! Big fan of the site.
On the excel building: so you have three series dot plot…and did you use error wiskers on the middle element to draw your lines? It’s very slick!
mike
Yep, horizontal error bars are in there. You can obviously do that any number of ways–go both ways from the middle series, to the right from the left series, etc. Create a new data series that measures the gap and use that in the Custom section of the error bar menu.
Thanks for reading!
Jon
thanks Jon for this great post. Though I noticed you left some instructions in the comment section, I still don’t know 100% how to create this one. Do you have an instruction? Thanks
Hi Lu,
Please download the Excel file linked in the post. You should be able to pull out how to do it from there. Basically, it’s three scatterplots with horizontal error bars.