Sunday, October 14, 2012


Amazingly Massachusetts allowed me to vote by e-mail !!! I had to waive my right to a secret ballot.  I still can't show you my ballot.   Actually I swore I hadn't done so (the idea is that I am especially not allowed to show it to someone in exchange for money for having marked the ellipse he or she likes).   But wait, I still have my ballot -- I scanned it and sent the scan.  I'm not sure the law forbids me from showing someone my ballot after voting by e-mail (or fax but I hate faxing and that's a fact).  Not being sure it is allowed, I will just upload the very official electoraltronic transmission sheet.

Amazing no ?

Also just to go back to one of the many bees in my bonnet, what about the Gallup likely voter filter ?
It worked very well for decades then totally messed up in 2010.  Gallup began reporting results for two filters .. the traditional filter and one which counted more respondents as likely voters (if I understand it basically anyone who said they were going to vote was counted as a likely voter by the looser ad hoc invented during the campaign filter).  The traditional filter  congressional generic ballot poll was a disaster overestimating the Republican margin by 9%.  This was definitely one of the types of polls of which Gallup had been most proud as it had an amazing record of accuracy.

WTF happened ?  Basically I would guess that the filter has always lead to oversampling old people vs young. Back in the good old days when I was young, old people voted for Democrats about as much as young people did (and more than middle aged people did).  In 2010 not so much, so maybe a long lasting problem which didn't mess up the results in the past came home to roost and keep government hands off of its Medicare.

But I want to get deeper in the weeds.  The traditional Gallup likely voter filter consists of 7 questions with yeses required to get in the likely voter sample.  One question is IIRC "have you ever voted in your current polling place before?" and another is something like "do you know where it is?".  This last question obsessed me in 2010 as I guessed that it caused a bias which vanished by election day.  If someone doesn't know where to vote and it is election day their claims that they will certainly vote, are very exited and have been following the campaign must be taken with a grain of salt.  In early October not knowing exactly where to vote mostly means you haven't voted there before -- so people who have recently moved get two almost automatic no's as do people who will, in fact, vote for the first time.

The fact that the bias increased as election day approached meant my face shared the egg which was all over Gallups corporate visage.

But I am stubborn so I have a new theory.  The new theory is that people who will in fact vote count on being able to find out where on the day using google and google maps.  So I decided to see how long it would take me to find out where to vote in person were I in the USA.  It took me three whole minutes to find out that I would vote at 28 Sacramento St and get a google map showing me where that is.

I am quite sure that the last time I voted in person (same voting address as now) I went somewhere else.

So look it could be that someone plans to vote during their lunch break, has no clear idea where at 11:30 and actually votes.  I don't think it used to be that way.

Gallup is going to have to gallup much faster to keep up with all this newfangled technology.

No comments: