Posts Tagged ‘Voting’

Mathematical Malpractice Watch: Non-Citizen Voters

Wednesday, October 29th, 2014

Hmmm:

How many non-citizens participate in U.S. elections? More than 14 percent of non-citizens in both the 2008 and 2010 samples indicated that they were registered to vote. Furthermore, some of these non-citizens voted. Our best guess, based upon extrapolations from the portion of the sample with a verified vote, is that 6.4 percent of non-citizens voted in 2008 and 2.2 percent of non-citizens voted in 2010.

The authors go on to speculate that non-citizen voting could have been common enough to swing Al Franken’s 2008 election and possibly even North Carolina for Obama in 2008. Non-citizens vote overwhelmingly Democrat.

I do think there is a point here which is that non-citizens may be voting in our elections, which they are not supposed to do. Interestingly, photo ID — the current policy favored by Republicans — would do little to address this as most of the illegal voters had ID. The real solution … to all our voting problems … would be to create a national voter registration database that states could easily consult to verify someone’s identity, citizenship, residence and eligibility status. But this would be expensive, might not work and would very likely require a national ID card, which many people vehemently oppose.

However …

The sample is very small: 21 non-citizens voting in 2008 and 8 in 2010. This is intriguing but hardly indicative. It could be a minor statistical blip. And there have been critiques that have pointed out that this is based on a … wait for it … web survey. So the results are highly suspect. It’s likely that fair number of these non-citizen voters are, in fact, non-correctly-filling-out-a-web-survey voters.

To their credit, the authors acknowledge this and say that while it is possible non-citizens swung the Franken Election (only 0.65% would have had to vote), speculating on other races is … well, speculation.

So far, so good.

The problem is how the blogosphere is reacting to it. Conservative sites are naturally jumping on this while liberals are talking about the small number statistics. But those liberal sites are happy to tout small numbers when it’s, say, a supposed rise in mass shootings.

In general, I lean toward to the conservatives on this. While I don’t think voter fraud is occurring on the massive scale they presume, I do think it’s more common than the single-digit or double-digit numbers liberals like to hawk. Those numbers are themselves based on small studies in environments where voter ID is not required. We know how many people have been caught. But assuming that represents the limit of the problem is like assuming the number of speeders on a highway is equal to the number of tickets that are given out. One of the oft-cited studies is from the President’s Commission on Election Administration, which was mostly concerned with expanding access, not tracking down fraud.

Here’s the thing. While I’m convinced the number of fraudulent votes is low, I note that, every time we discuss this, that number goes up. It used to be a handful. Now it’s a few dozen. This study hints it could be hundreds, possibly thousands. There are 11 million non-citizens living in this country (including my wife). What these researchers are indicating is that, nationally, their study could mean a many thousands of extra votes for Democrats. Again, their study is very small and likely subject to significant error (as all web surveys are). It’s also likely the errors bias high. But even if they have overestimated the non-citizen voting by a factor of a hundred, that still means a few thousands incidents of voter fraud. That’s getting to the point where this may be a concern, no?

Do I think this justifies policy change? I don’t think a web-survey of a few hundred people justifies anything. I do think this indicates the issue should be studied properly and not just dismissed out of hand because only a few dozen fake voters have actually been caught.

The Hormone Vote

Wednesday, October 24th, 2012

CNN has an article up that is … kinda dumb:

While the campaigns eagerly pursue female voters, there’s something that may raise the chances for both presidential candidates that’s totally out of their control: women’s ovulation cycles.

You read that right. New research suggests that hormones may influence female voting choices differently, depending on whether a woman is single or in a committed relationship.

Please continue reading with caution. Although the study will be published in the peer-reviewed journal Psychological Science, several political scientists who read the study have expressed skepticism about its conclusions.

Basically, this new study claims — actually, rediscovers — that women in relationships favor Romney by 19 points and single women favor Obama by 33. Their new claim is that when those women are ovulating, those percentages jump by as many as 20 points.

This has, for obvious reasons, caused quite a stir in the blogosphere and Twitter. Unfortunately, the primary reaction is for people to clutch their copies of McKinnon and scream at some Texas professor for daring to suggest that women are nothing but hormone-addled idiots, even though the professor in question says nothing of the kind. And that reaction is kind of unfortunate. Because in their zeal to proclaim that women are completely unaffected by their hormones, people are missing the real reason why the article is dumb and should just be snickered at and then ignored.

First, the number of women we are dealing with is small. I don’t have access to the study and their exact numbers but they studied 502 women total. If by “change of 20 points*” they mean that women in relationships went from 59-41 Romney to 69-31 Romney, that’s a total of about 25 women changing their minds. And a similar number among single women. That … really doesn’t strike me as a statistically significant sample, especially given how volatile polls are known to be anyway and how uncertain the date of ovulation can be.

(*A critical point that is missing from the article is whether that jump is 20 points in differential or absolute (i.e, from 59-41 to 69-31 or 79-21). It’s the difference between 25 women changing their minds — a small number — and 50, a more interesting number. I also note the phrase “as much as 20 points”, which suggests that 20 points is at the outer edge of a very large statistical uncertainty and the actual difference is much smaller. This is why I would like to see the actual study.)

Second, it’s difficult to pin down an a priori reason why a woman’s menstrual cycle might affect her voting. In the absence of clear information, we can only speculate. And this is where CNN and the researchers really flounder badly:

Here’s how Durante explains this: When women are ovulating, they “feel sexier,” and therefore lean more toward liberal attitudes on abortion and marriage equality. Married women have the same hormones firing, but tend to take the opposite viewpoint on these issues, she says.

“I think they’re overcompensating for the increase of the hormones motivating them to have sex with other men,” she said. It’s a way of convincing themselves that they’re not the type to give in to such sexual urges, she said.

It’s true enough that women feel “sexier” when ovulating and are known to change their behavior (more likely to have sex, more likely to wear skimpy clothing, etc.). That’s all well-established biology. How this translates into political behavior isn’t clear at all. It seems that the researchers came up with one half of a dubious idea (“women feel sexier so they want abortion to be legal”) and then had to scramble to find the other half (“um, so married women are … repressing?”). That’s nice spit-balling but it’s no more valid than saying that when women are menstruating, they get mad and say, “Screw that guy, I ain’t voting for him any more!” You can basically shove anything you want into that information vacuum and call it “science”.

Something important jumped out at me on a second reading: no one quoted in the article is a biologist or any other kind of scientist. The study author is a Professor of Marketing. They also quote Professors of Political “Science” and Women’s and Gender Studies. I would hazard that maybe the Professor of Marketing knows something about statistics. But this whole things reeks of the Scientific Peter Principle: poorly done studies are the ones most likely to get attention because their flaws produced amazing results.

Here’s $0.02 from someone as equally unqualified to look into this as anyone quoted in the article. I suspect this effect, such as it is, is small, even smaller than the 10% they are claiming. I also suspect that this study was conducted some time ago when a lot of the voters were undecided and might have been a little torn between the two candidates. Undecided voters have a tendency to sway with every breeze that blows. Under those circumstances, it’s possible that the hormone kick at ovulation and the resulting surge in self-confidence might make women a little firmer in their political convictions one way or the other. Or, conversely, that the effects of PMS and/or menstruation make women a little less confident in their choices. One test you could do? See if “ovulation effect” diminishes as we get closer to the election and more people learn about the candidates and make up their minds.

The gripping hand here is that this entire thing is pointless trivia as far as elections go. You see, women’s menstrual cycles tend to be random. So the percentage of women who are ovulating at any one moment is a constant. So the net effect of this on the vote?

Zero.

Update: I just slapped myself in the head for not saying this in the main text: where the hell was the group of menopausal women used as a control?