A historic week in Wisconsin has left Democrats asking if the county clerk of Waukesha has committed fraud by artificially raising the vote totals for Republican Incumbent Supreme Court Justice David Prosser two days after having been delivered a narrow defeat of about 200 votes.
Conversely, the Republicans should be asking how stupid can this woman be to forget to report all of the votes for the City of Brookefield? Or should they be worried that members of their party are stealing an election and further enraging an already awoken liberal front?
Either way a friend of mine put it less eloquently, but no less accurately after a few beers “Bitch needs to go to jail, or bitch needs to be fired.”
But what really happened? Is it possible that Waukesha County vote totals were accurately reported before the “fishy” adjustment?
What does the data tell us?
Based on Nate Silver’s interesting blog in the New York Times, the data points to incompetence and not fraud. He comes to this conclusion after applying statistical wizardry and regression models using a bunch of information about potential motivating factors and calculating a 95% confidence interval of how many voters should have turned out in Waukesha. I’m a believer in statistics and can’t find any flaws in his approach, but the math and conceptual tools may be beyond comprehension of many readers.
Using his data and adding some additional population statistics I’ve asked the question: “Is there any relationship between the change in voter turnout (calculated as a percentage of registered voters) and the voter results of the election?”
For example, the November election was a big win for the Republican party in Wisconsin, will they come out and support their Republican candidates as consistently as they did last time?
Take a look, click on the image to open up a larger version of the scatterplot:
The scatter plot shows Voter Support for David Prosser ( percentage of total votes cast in that county) on the x-axis vs the change in voter turnout between the April 2011 Election and the November 2010 election. The data here is the complete original reporting before Kathy Nickolaus came forward with her “omission.”
Voter turnout is calculated as a percentage of total votes divided by the total number of registered voters in that county. The y-axis shows the voter turnout for the April 2011 election minus the turnout for the November 2010 Election. Data points higher on the y-axis show a more consistent voter turnout from November 2010 to April 2011. Conversely, data points lower on the y-axis show that there was less consistent voter turnout at the polls.
The size of each data point indicates the size of the population in each county data point. The color of the data point similar to the x-axis indicates the percentage of voters supporting incumbent Supreme Court Justice David Prosser. A bright red data point (farther right on the x-axis) indicates strong support for Prosser. A bright blue point (farther left on the x-axis) indicates strong opposition to David Prosser.
Additionally, I’ve included a “smushed” scatter plot on the left which shows how all of the data points are separated by the difference in voter turnout. This shows a fairly consistent spectrum of blue point to red points as you move from the top of the voter turnout difference to the bottom of the scale.
What’s going on here?
What I think is most interesting about this plot is the strong correlation of voter turnout difference by support for the Republican candidate. Imagine a line being fit through that data, drawn more heavily to the larger data points and we have a line starting at the top left and working its way to the bottom right. It’s fairly reasonable to say that people supporting the republican candidate were not as motivated to head to the polls again as folks that opposed the Republican candidate.
But what about Waukesha county, the large red point at the bottom right of the plot? It looks kind of lonely down there. Could Waukesha county voters have been so lethargic on April 5th so as to have by far the biggest difference in voter turnouts by about 5 points? Is there something else happening?
By contrast, the bright blue points in the top left are a fairly good distance away from all of the other data point populating the center of the plot? Conspiracy theorist conservatives will likely point at this and claim voter fraud and demand voter id laws and the elimination of same day voter registration laws. They are on the way, trust me.
The “Correction”?
So what happens when we apply the “correction” or “fraud” depending on which side of the wedge you’re occupying?
With the “adjustment” Waukesha moves significantly up the screen (more consistent turnout, but still at the bottom of the bunch), keeping about the same ratio of Prosser supporters. Also notable, it moves much closer to its nearest neighbors, both in the sense of data proximity (voting habits) and geographical proximity.
This is consistent with analysis looking at support for Scott Walker in the 2010 election vs the 2008 presidential election. The link opens up a Tableau workbook with the Walker scatterplot and the Prosser Scatterplots used in this post.
Bottom line, this data doesn’t prove that the voter turnout in Waukesha is abnormally low. Its also shows that the adjusted point is not unreasonable. I’ll leave it to the Government Accountability Board and the US Attorney General to prove that Waukesha’s reporting is on the up and up.
Update: Sep 28, 2011
The GAB has concluded that Kathy Nickolaus did not commit a crime:
“her conduct does not appear to rise to the level of conduct that can be described as willful neglect or a refusal to comply with the law.”
She will however have to start playing by the rules.
“Your actions following the April 5, 2011 Spring election did not conform to the legal requirements imposed on county clerks,” G.A.B. Chairperson Thomas H. Barland said in a letter to Clerk Nickolaus. “When one election official fails to act consistent with those responsibilities, steps must be taken to correct the failure in order to prevent it from recurring, and to restore public confidence and trust in the administration of elections.”