This post includes three updates. I have added to the original material to show sequence of my thinking in the chart development based on reader input and charts.
Robert Kosara at EagerEyes.org wrote a post on the Electoral College and Second Terms in which he compared the winner’s percentage of presidential election electoral and popular votes. Based on readers comments, Robert revised his original chart slightly to produce his final version, shown below:
Robert used a stacked bar chart. He says “..stacked are bars quite flexible… They are also available in virtually any program that can draw charts, so this method should work with practically any program.”
Robert is comparing two values (% popular vote & % electoral vote) for each presidential election. With stacked bar charts, there is overlap between the two values, the bar for the larger value will hide the smaller value bar unless the user compensates for this. Robert uses a clever technique to make sure both bars are visible.
Dot Plot Approach
Dot plots provide an alternative to bar charts which work well in this situation. Here’s Robert’ election results in a dot plot that shows both popular and electoral results as dots on the same line. I developed the chart in Excel in a few minutes using a simple XY plot and a custom axis.
This is my 2nd version of the plot. Trond and James submitted a comment suggesting tracer lines to connect the names with the dots. I agreed and added them in this version 2.
Robert Kosara wrote a comment…
I’m not convinced. Even with the lines, the dots are just too disconnected. What you mostly see is two vertical structures, the horizontal connection is minimal. If the interesting thing here were the timeline, I’d agree. But to show the effect per election, I don’t find this plot effective.
Respecting Robert’s views on his chart, I went back to my Excel workbook, hoping to improve the dot plot approach. What was I really trying to show?
- Cases where Electoral votes were less than Popular votes! Why note add a marker for those cases?
- Overall comparison of Electoral & Popular vote results! Why not rank the results by Electoral vote rather than chronological?
I like my third version much better than versions 1 and 2. Sorting by electoral vote results highlights the structure more clearly. The three elections where Electoral results were less than popular resuls show up more clearly, the red diamonds are just there to make sure everyone sees these elections.
Robert, Trond and James have helped to improve the dot plot. Thanks!. Are these improvements enough to match/exceed the bar chart? Again I’ll ask, what do you think?
What Agout Floating Bar Chart?
Derek at I-Ocean.org suggests a floating bar chart in his comment. Here’s a copy of his chart.
Derek’s chart is cleaner than Robert’s original stacked bar chart. Derek’s bars show the difference between the electoral and popular votes, however, the scale takes a little getting used to. I’d prefer a scale that shows the actual % difference (Electoral – Popular) between the 2 votes.
11 responses so far ↓
derek // October 23, 2008 at 3:15 AM |
Thanks for the kind words. I’m concerned that your suggestion would lose information I want to keep. A floating bar with one of two colours, such as the ones in a stock market “candlestick” graph, gives three pieces of information, with two degrees of freedom: the “open”, the “close”, and the spread. Showing just the spread would lose the open and close values.
As a compromise position, I’ve gone back to Robert’s original stacked bar concept, and just given the lower bar a light colour. The result is presented in my blog..
dkodpe // October 21, 2008 at 5:23 PM |
Derek:
I liked your floating bar chart so much, I added it to the post. Thanks.
I have one suggestion, you may want to change the X axis to actual % difference between Electoral and Popular votes. For example, if Electoral was 60% and Popular was 52%, I’d plot 8% rather than plotting bar from 52 to 60%.
Robert Kosara // October 21, 2008 at 9:08 AM |
I like how you’re updating this in response to criticism, but I think you’ve lost sight of the original idea – at least my original idea. I wanted to point to these instances (electoral < popular) as part of displaying the data. You can always put some kind of marker in a chart, and in this case it’s especially easy because there are only three instances.
The lines actually work for this nicely, because you can see where they cross. But lines also imply connections, and this data isn’t really continuous. I also wonder if this wouldn’t be a lot more confusing without sorting the values (which brings out other structures though, no doubt).
Perhaps a compromise would be to place markers at the data points, and connect them with fairly thin lines – i.e., a combination of your two designs. That would put less emphasis on the continuity, but perhaps provide enough cues to see where the numbers cross over.
derek // October 21, 2008 at 3:18 AM |
Sorry about that, my ISP (recently acquired by Cable & Wireless) seems to be failing more and more often these days. Try this link instead.
derek // October 21, 2008 at 3:04 AM |
I produced a floating (up-down) version of Robert’s original bar chart, as the original had too many active edges for me, and I find blue and green hard to distinguish.
I think the floating version better emphasises the difference the electoral college makes in percentage, the occasions when the difference crosses the 50% line, and the occasions when the electoral college vote is a lower percentage than the popular vote (different color).
I kept the hue difference the same, but rotated both hues a little way round toward the red. I asssumed we want to avoid the red and blue colors that are common in graphing, for the party significance they have.
jonpeltier // October 19, 2008 at 10:35 PM |
I think Robert’s stacked bar chart displays more clearly that only three of the elections had lower electoral college tallies than popular vote percentages. The dots do seem disconnected, and believe it or not, I cannot clearly see the difference between the green and blue dots without concentrating, while in the bars, the same colors somehow seem more distinct.
A bar chart version of Excel’s up-down bars might be effective. It would easily show the three elections with lower electoral than popular vote, and reduce the data-ink ratio compared to the stacked bar chart (although I think this ratio is a bit overrated).
Robert Kosara // October 18, 2008 at 8:38 AM |
I’m not convinced. Even with the lines, the dots are just too disconnected. What you mostly see is two vertical structures, the horizontal connection is minimal. If the interesting thing here were the timeline, I’d agree. But to show the effect per election, I don’t find this plot effective.
James // October 18, 2008 at 6:34 AM |
The bar chart has one advantage, since the colour of the left-most bar immediately tells you which point comes first. You can scan down the left of the bar chart and spot the cases where %electoral is less than %population.
In the dot plot, colouring the tracer in the same colour as the dot, would help convey the same hint.
Trond // October 16, 2008 at 10:55 PM |
Nicely done — much easier to read!
dkodpe // October 16, 2008 at 10:19 PM |
Trond –
I like the tracer lines. Thanks!
Trond // October 16, 2008 at 9:07 PM |
I tend to agree that the dot plot here works better, but would like to see a tracer line following from the name at the left to the dots themselves. Without them I get lost figuring out which dot goes to which president.
On the bar charts, it took me longer than I was happy with to figure out why the bottom stack was green those few times.
Another option that would work nicely would be to combine the two approaches. Be consistent with the popular vote as a bar and overlay the electoral vote as a dot. That would solve the eye tracking issue and make it immediately understandable.
Just a few cents worth…