27 May 2022

Published in

Warning: Graphic Content

14 min readMay 27, 2022

Open chart surgery

I’ve written a few times about people not using 100% bar charts when they should really think about using 100% bar charts. This week, let’s take a look at the opposite — when using a normal bar chart may have been the better choice.

I happened to spot this example from IPPR a few days ago (although they are by no means alone — there are some examples from my time at IfG I could also point to):

A chart by IPPR, available at https://twitter.com/IPPR/status/1529084451749351428

The chart does make some of the desired points, sort of: that support has been given to all income deciles rather than being targeted at the poorest. (It doesn’t necessarily make the other points in the tweet — such as the number of people in each decile group that received or missed out on support.) But there are a lot of colours, some of which are very similar (not least poorest and richest), it doesn’t really give you much else and just isn’t hugely attractive to my eye.

A remix of the IPPR chart — a regular bar chart rather than a 100% bar chart.

This rough remix (using WebPlotDigitizer estimates rather than the source data) instead uses a regular bar chart. I think it looks better and still makes the point that financial support was distributed to people in all income bands (indeed, that more money went to some other income deciles than it did to the poorest). I’ve highlighted the poorest and richest deciles, though we could also have coloured it from light to dark.

While I think this is better, I’m still not sure it’s the best chart choice. Showing what percentage of people in each decile received/did not receive support would probably be the most powerful visualisation — and the one most clearly making the points in the tweet.

Before Excel let me down (i.e., lost all of my work having switched AutoSave off without me realising), I was also going to try remixing this IfG chart:

A chart by IfG, available at https://twitter.com/instituteforgov/status/1529741631619858432

It shows that the poorest vingtiles will see their household income drop between now and 2022/23, and then drop further — unlike any other group — to 2026/27. With Excel having failed me (THE BETRAYAL!), here are some hand-drawn alternatives.

Remix 2 of the IfG chart — an arrow plot

Remix 3 of the IfG chart — an overlaid bar chart

Remix 4 of the IfG chart — a parallel line chart

First we have a dot plot, with both the 2022/23 and 2026/27 data as dots. We could colour the bars between those dots — and even put lines between the dots of the same colour, like a line chart — to accentuate the fact that only the poorest charted vingtile will see a continued fall. Second we have an increasingly popular chart type — let’s call it an arrow plot — that is sort of a dot plot but with an arrow showing the direction of travel, and colour reiterating the message (red for fall, blue for rise).

Third we have an overlaid bar — the second vingtile stands out because it’s the only one where the narrow bar is not contained in the wider one. And fourth is what I’d call a parallel line chart, with a 2022/23 and 2026/27 line for each vingtile clearly showing whether income falls or rises (and colour again underlining the message).

I’d probably have started with one of these four rather than the chart the IfG used. But there’s something about the original’s use of two different shapes — bars and dots — as well as colours that helps make the point. This is the sort of dataset where there are always quite a few visualisation options with no hard and fast rule for which is best — you’ll have to try a few and find out.

As for everything else: