FORTH GO

Following up on my instant fame as an OrangePolitics.org guest author, here are two more graphs from the Chapel Hill Town Council election results. The first is a mosaic plot, which is close to what Mark Chilton was suggesting, I think. For each cell, the width represents the votes in that precinct, and the height represents the percentage in that precinct that voted for the candidate, so the area shows the number of votes for that candidate in that precinct. Precincts are alphabetical; candidates are in order of overall finish. Unfortunately, I couldn’t sort each column differently, as Mark suggested, so it’s not clear who “won” each precinct. Click graph for larger version.



The second graph is more statistically relevant (ignoring the fact that the “Absentee” early voting precinct is not really comparable to the other geographical precincts). Each box plot summarizes the percentage outcomes for a given candidate in all the precincts. Assuming a normal (bell-curve) distribution, points outside the “whiskers” of the box plot are likely to be outliers affected by factors other than random variation. (Testing for normality indicates that only the Cutson and Baker results do not follow a normal distribution.)



I labeled the outliers, though the Northside and Lincoln labels overwrite each other in the Thorpe column. The box plot outliers show Mason Farm as significant for Raymond, though it was not noticeable from the overlay line plot since that precinct had so few votes.

Orange Country provides timely online election results, and their HTML is friendly enough that is can be imported into JMP. Below is a graph I came up to show the Chapel Hill town council results by precinct. Precincts, along the horizontal axis, are sorted by number of total votes. [Click graph to see full resolution version.]

I’m thinking there’s a better graph using area, but this overlay is the best I can do easily. At least it shows a few interesting pieces of information:

• In general, not much variation among precincts.
• A few significant-looking exceptions to the general case: Cutson in Cedar Falls, Easthom in Patterson, Thorpe in Northside and Lincoln, Baker in Absentee, Harrison in Durham.
• Absentee was by far the largest “precinct”. [Sorry for chopping off the top of the graph, but otherwise there's too much unused space in the graph; Easthom and Kleinschmidt got over 500 there.]
• There is a significant difference in votes per precinct. Is it worth campaigning in Battle Creek and Country Club?

Looking at the voter turn out statistics, it’s clear the main factor in the votes per precinct is just that some precincts have more registered voters than others (I don’t know how registered voters correlates to population, though). However, there are exceptions. The smallest (in registered voters, that is) precinct, Weaver Dairy Sat. (is that Carol Woods?) had the fifth highest voter turn out with over 68% voting. And two of the largest precincts, Battle Creek and Country Club (are those UNC?), brought up the rear, due to a measly 2.4% turn out. So the graph doesn’t tell enough of the story there; those precincts are worth campaigning in if one can increase the voter turn out.