Friday, April 08, 2011

Here's A Good Project for A Statistician. Or on Wisconsin Elections

The project is this:

Go out and study all those American elections where earlier opinion polls and/or exit polls were seriously wrong, compared with final results. Do that for the last twenty years, say, and see if the way the results were wrong turns out random or not over the long haul. In other words, do these errors favor one party over the other or not, on average? If the errors are truly random we should be able to answer that question with a resounding no.

And of course this idea came to me because of what is happening in Wisconsin. Waukesha County Clerk Kathy Nickolaus found lots of extra votes in the heated Supreme Court race in Wisconsin. The incumbent, David Prosser, is seen as one of governor Walker's Ringwraiths. He was challenged in the race by JoAnne Kloppenburg.

The race was very tight, with Kloppenburg appearing to have narrowly won it. But suddenly, after the elections are over, Nickolaus finds 7000+ extra votes for Prosser:
Waukesha County Clerk Kathy Nickolaus said Thursday that she failed to save on her computer and then report 14,315 votes in the city of Brookfield, omitting them entirely in an unofficial total she released after Tuesday's election. With other smaller errors in Waukesha County, Prosser gained 7,582 votes over his challenger, JoAnne Kloppenburg, leaving the sitting justice significantly ahead for now amid ongoing official counting.


But Kloppenburg supporters reacted with alarm, pointing out that Nickolaus had worked in the Assembly Republican caucus during the time that Prosser, a former Republican lawmaker, served as the Assembly speaker and that Nickolaus also had faced questions about her handling of elections as clerk.

"Wisconsin voters as well as the Kloppenburg (campaign) deserve a full explanation of how and why these 14,000 votes from an entire city were missed. To that end, we will be filing open records requests for all relevant documentation related to the reporting of election results in Waukesha County, as well as to the discovery and reporting of the errors announced by the county," Kloppenburg campaign manager Melissa Mulliken said in a statement.

Assembly Minority Leader Peter Barca (D-Kenosha) raised the possibility of an independent investigation over the recovery of the votes.

"This is a serious breach of election procedure," he said. "We're going to look further. She waited 24 hours to work this. And she waited until after she verified the results, making it that much more difficult to challenge and verify the results."
Just a big oops? So argues the Democrat on the Waukesha County Board of Canvassers.

Nickolaus certainly runs an odd and tight elections ship, as an article about her from last summer tells us:
Waukesha County Clerk Kathy Nickolaus' decision to go it alone in how she collects and maintains election results has some county officials raising a red flag about the integrity of the system.
Nickolaus said she decided to take the election data collection and storage system off the county's computer network - and keep it on stand-alone personal computers accessible only in her office - for security reasons.
"What it gave me was good security of the elections from start to finish, without the ability of someone unauthorized to be involved," she said.
Then there is this (via Firedoglake):
On Tuesday, shockingly-large turnout suddenly emerged from Waukesha County, which did not comport with either the results of previous spring elections, or even internal estimates from city officials mid-day. In fact, a Waukesha City Deputy Clerk said at 1:18pm that turnout was very typical, predicting somewhere between 20 to 25 percent. As Tuesday night wore on, reporting in Waukesha County stopped altogether for hours, leaving observers to wonder what was going on. Then suddenly, results suggesting massive turnout started to pour in rapidly with Prosser adding dramatically to his total by a 73-27 percent margin.

One Wisconsin Now estimates put overall turnout near 38 percent, a wild outlier to historical data and the earlier mid-day estimation of Waukesha’s own officials. In April 2009, turnout was 20 percent; April 2008, turnout was 22 percent and in April 2007, turnout was 24 percent. All of these elections had hotly-contested Supreme Court races as well.
On the other hand, Nate Silver compares the turnout figures of this election to the presidential election in 2008 and the gubernatorial one in 2010 and argues that the turnout rate looks about right from that angle.

The question clearly merits more analysis. Was this election like other April elections? Or was it more like the presidential election? And can we assume that the rest of the Wisconsin turnout rates are the same as before (as Silver seems to assume)? One could make a case that Democratic counties might have much higher turnout rates than in the recent past, what with the events in Madison.

Whatever happened in this particular Wisconsin election, the Wisconsin Republicans have shown us a very nasty side of the game they play in the recent months. For that reason alone I think an impartial investigation would be a very good thing. That Nickolaus seems to have run a tiny monopoly over the actual election results is another very good reason. Without something like that the rest of us will not have faith in the corrected results.