Voter turnout: a prettier plot
I replotted Henry’s graph. Here’s my version:
I would argue that this graph is not only prettier and better for communicating to the world, but that it can also help us understand the patterns better. In particular, we now see 2008 in the context of the 1950s-1960s peak, and we can also see how the 2004-2008 jump compares to other recent historical changes.
See here for my discussion of the choices involved in making this graph. (I also posted R code, which I would never do at the Monkey Cage…)
Comments
In the spirit of Andrew himself, who enjoys criticizing other people’s charts, I will criticize this one for (1) not labeling the first and last data points, so the reader has to do the math him/herself (tsk tsk) and (2) truncating the y-axis, which has the effect of overstating the magnitude of the trend; 50-65% looks a whole lot different when plotted on a 0-100 scale than on a 45-65 scale (tsk tsk).
Posted by: Lee Sigelman | November 5, 2008 01:01 PM
Andy, Henry confessed to me that he made that graph intentionally ugly. We then discussed betting about when you’d leave a comment or a revised version.
I agree with Lee on #1. On #2, I’m ambivalent. 45-65 on the y-axis does exaggerate the change, but given the historical “bounds” on turnout (at least post-WWII), there’s a case to be made for a more delimited scale. Perhaps this is where the words that accompany the picture matter: one needs to make an argument about the substantive magnitude of any trend depicted.
Posted by: John Sides | November 5, 2008 01:42 PM
1. I wouldn’t want to go from 0 to 100, but maybe 40-70 would work. But then I’d want to change the aspect ratio and make the whole graph taller.
2. I think the best solution to the labeling of the years is to put the information in the title: “Voter turnout in presidential elections, 1948-2008.”
Posted by: Andrew![[TypeKey Profile Page]](http://www.themonkeycage.org/nav-commenters.gif)
|
November 5, 2008 02:39 PM
As long as you want to use 20-year intervals, why not start at the start and end at the end — 1948, 1968, 1988, 2008?
Posted by: Lee Sigelman | November 5, 2008 03:26 PM
And John wins (he took the under bet on whether you would have a response up by 5pm). Originally, I had the Excel graph with the default label of ‘Series 1’ but I thought that this was going just a little bit too far.
Posted by: Henry | November 5, 2008 04:18 PM
I for one would welcome more R code. :-)
Posted by: jim | November 5, 2008 07:21 PM