Statistics and the Recall

Before he passed away in 1995, my dad was the Deputy Director of the US Census.  So I spent a lot of time hearing about statistics growing up.  I was pleased to see that someone, in this case the Journal-Sentinel, decided to test a statistically significant sample of the Recall petitions and reached the conclusion that only 15% of the signatures could not be verified.

After randomly selecting petitions and names from the accountability board web site, the newspaper checked available public records including the Wisconsin courts database, a state voter database,, the White Pages and other online search engines.

For most of the 73 signatures that couldn’t be verified, there was a record that the person existed and was of voting age, but no address could be found to match the one listed on the petition.

A name was considered invalid in the newspaper review if no record could be found for the signee at the address listed, if they weren’t old enough to vote or if they were a felon under state supervision.

What’s interesting that the errors discovered were not for the breathless GOP fears of Mickey Mouse and Adolf Hitler, but instead it was real people who, for some reason, got some piece of their information incorrect.  Lassitude and not fraud seem to be the source of these errors.

But even with a 15% error rate, there will be no problem bringing Scott Walker to justice recalling Scott Walker.


Related Articles

10 thoughts on “Statistics and the Recall

  1. I’ve read the post. Thanks Phil.

    Steve, not to be a stalker, another question for you: Do you consider signing a recall petition more than once fraud?

    1. I think we have to allow for some measure of independent invention on this one… It’s not a “non-obvious” idea.

      But your point is taken. I too considered doing it until I realized how many I’d have to verify. I don’t have a staff of 5 people to work on it like the JS does. 🙂

    2. Yep, I too had thought the same thing. Great minds. Perhaps this could be suggested as an official amendment to the process. But then again, we know the conservatives would never go for it, as we all know much they like stats, science etc.

  2. I agree, it is a sort of obvious idea, but a bit more than one person can carry off alone. I’m glad to see it carried out.

    Next, we carry out Scott Walker.

  3. While I’m no statistician I just love statistics. Here are some that I have fun thinking about:

    5.2 million people living in Wisconsin
    2.1 million households
    2.5 person average household
    4.37 million people that are voting age
    49.56% of eligible voters voted in Wisconsin in 2010. So that’s 2.16 million total votes cast.
    283,351 people are public employees
    6.5% of eligible voters work for the government.
    “1,000,000” signatures to recall Walker
    15% or 150,000 deemed to be unverifiable or invalid.
    850,000 valid signatures.

    So 39.3% is the % of likely voters (those that voted in the last election) that legitimately signed a recall petition. Now I know it was cold outside but I have a sneaky feeling that if one felt like signing a recall petition they probably did.

    I would just LOVE to know what % of the recall signatures were from Dane county and Brown/Milwaukee counties. I suppose that information would be available. I’ll look for it. It’s important because over 70% of those votes went against Walker in the Supreme court race, which was seen as a referendum on Walker. So if a largely disproportionate % of the recall signatures are from those counties it skews the data because we know those counties are going to vote for a dem. no matter what.

    Conclusion: Walker wins by a nose! I can’t say I’m as confident about the Senators. I predict a divided legislature again just like most of our history. And then we can go back to getting nothing done.

  4. A bit late to the party, but I had started a reply to James Booth, then set it aside to fill in some details and forgot about it.

    James: I don’t follow your math on this.

    Assuming the same turnout of 2.16 million, and that *half* of the legitimate voters that signed a petition all vote against Walker (425,000), then Walker would need ~62% of the remaining votes to reach 50% of the total.

    2,160,000 / 2 = 1,080,000 [that’s 50% of likely voters]
    2,160,000 = 1,735,000 + 425,000 [425K very likely Democratic votes, 1.74M up for grabs]
    1,080,000 / 1,735,000 ~= 0.6225 [% of remaining votes Walker needs to keep 50%]

    Note: I count public employees among the likely Democratic votes.

    Now assuredly there are a greater % of Walker supporters among those up for grabs (because I have subtracted likely Democrats), but last I checked Walker’s approval is well below 62%. This is sort of a dead end, but we might consider the vote Walker got in the last election.

    Walker received 1,128,941 votes in 2010, and needs to keep more than 95% of those votes to have enough to stay over 50%.
    1,080,000 / 1,128,941 ~= 0.957

    The way I see it, if Walker’s support has eroded by even 5% he is in a dead heat. That’s assuming the 5% simply don’t vote – if they swing Democratic, Walker loses.

    Technically, I am doing arithmetic rather than statistics. Maybe later I can put some confidence intervals on this.

    Obviously I never came back to the confidence intervals, but now I think O’d rather start over with some fresh data.

Comments are closed.