jackdied.com RSS feed
   Men are not fit to rule men and saints don't run for office.

My Histogram of 2004
[jackdied]
Nov 07 2004, 17:46 EST [updated Jan 03 2005, 22:40 EST]
I liked the idea of the histogram below but obviously it had some serious data errors. I downloaded all the state-by-county information from USAToday (which you can examine yourself (40k gziped)). I only saw seven counties with no data, so I don't know what data set the other guy was using. Instead of counting counties I add up all the voters. So we add up all the voters from counties that voted 50% to 50.999% for bush as one data point. The line graph has 100 points, the bar graph is smoother because we lump 50.0% through 54.999% into one bar labeled "50." It is not color coded, all the bars are red just because that is the default color for the program.
First the more raw line graph

Next the same thing put into smoother "buckets." Keep in mind the 50% bucket means 50,51,52,53,54 so it is a leaning Bush bucket.
Error: the line between Pro-Kerry and Pro-Bush on the X axis should be one bucket to the right more than it is.

It is easier to see in the line plot, the pro-Bush righthand side is much smoother and the tail is shorter. The pro-Kerry lefthand side drops off less sharply but there are more people out on the tail.

Overall it looks similar to the other guy's below even when we count people instead of counties. So I would say his "blue spike" was a data error. The rest of it looks about right.

Update: I'm a geek so I added a Y2 axis with "net votes" on it. It is low in the middle because no one picks up many votes when the county is split 51-49. For the Repubs the all-Red counties don't add many votes, for the Dems the all-Blue counties do. This doesn't tell us much other than people who voted for Bush are more likely to live in a county that is more evenly divided.

Font: Bigger Smaller Screen Width: Narrower Wider
0.10 seconds
jackdied.com 2003-2007