Polling Gap - Dittman Confirms It's the Biggest

A Dittman poll predicted 50% for and 41% against Prop. 5 a week before the April 3 Anchorage Municipal election.  The actual outcome was 58% no and 42% for - a 17% switch in the 'no' votes.

That seemed like a huge switch in a week.  Polling has gotten to be pretty sophisticated and more accurate in recent years.  When all the other issues in this election began to come out - see Bent and  Mudflats  - this seemed to be another anomaly that should be in the mix to be checked out.

So, when I saw that Dittman would be on a panel with the other two best known Anchorage pollsters, I wanted to take the opportunity to see whether this was, as it seemed to me, the biggest difference between polling outcome and election outcome so close to the election.

I got the chance to ask.  I had to ask three times before Dittman confirmed my suspicion.  I've tried to focus just on that part of the video.


[UPDATE 4/29/12 - Someone suggested I add the transcript from the video.

Steve:  Can you think of an election where there was such a big shift [from the poll prediction to the election outcome]?
Dittman:  No.  The answer is no.  I think if you look back historically, the polls have been pretty darned good.  We have on our website the graphic tracks, we have the final poll and the election, it’s just within a percent. It’s dead on. And I think if you look back over time, the polls have been pretty darn reliable.  And in this case you’ve got like a ten percent shift or maybe even twenty . . .
Steve:  Well, it’s almost twenty for the no people.
Dittman:  Yeah.  When you look . .I think that’s one of the highest I can remember.]

Does this even matter?

Some people I've talked to about what I'm calling anomalies in the election have said, "So what? That doesn't prove anything." But I'm working on the grounds that when you're in a situation and you notice things that seem unusual or abnormal, these are signs that something may be amiss. "How to develop Sherlock Holmes intutition" at WikiHow talks about reading the situation.

Understand how to read a situation. There are three parts to reading a situation:

• See. What do you see that is happening? 

• Observe. What do you notice that is different; a stain, a crease? 

• Deduce. What does this imply?

See:   We had an election that ran out of ballots and people were turned away.
Observe:  This is the step that I and other bloggers are going through.  We're looking for things that are different.  The polling gap is one of five or six 'differences.'  In this case it is the extreme variation between the predicted and actual outcome of the election - a greater variation than veteran Anchorage pollster Dave Dittman can remember.
Deduce:  The 17 point shift among 'no' voters - from 41% no to 58% no - that something is different here from other elections.  The two obvious implications are:

1. The polling was inaccurate or
2. The vote counting was inaccurate.

[NOTE:  Never be taken in by such either/or options.  There are others (i.e. it just was an odd set of circumstances, like winning the lottery), but for now these are the key ones and most of the others are subsets of these.

The panelists in the video, all pollsters, focused on their area of expertise (polling) and discussed what could have gone wrong with the polling.

• Polls asked wrong questions - focus on discrimination v. focus on freedom;  use of term gay v. homosexual - will all result in different responses 
• Voters changed their minds - negative campaigning - the panel suggested the no campaign might have frightened voters and the yes campaign may have irritated voters
• Size of turnout -  Craciun said that the low percentage of people who vote in general is a serious problem.  Moore said this was the highest turnout election.  (In the end, the turnout edged out the 2006 Mayoral race by 240 ballots, and Prop. 5 had the most people voting.)

Nevertheless, all these possible polling errors can happen at every election and these are experienced pollsters who can explain the ways polls go bad and are paid to avoid these types of bias in their polling. At this election, Dave Dittman agreed this was the highest deviation between what was predicted and the outcome in a week's time [that he could remember.] 


The other option: there was an irregularity in the election itself and how the votes were counted. 

This all came to light when polling places ran out of ballots.  We also know that Jim Minnery sent out emails to evangelical Christians telling them they could register on the day of the election, even though he'd previously sent out an email telling them to get registered 30 days before or they wouldn't be able to vote.
We also have reports of broken seals on ballot boxes and the deputy clerk in charge of the elections telling a polling worker not to worry about the broken seals.

Back to the Sherlock Holmes webpage:

"It is the highest importance in the art of detection to be able to recognize out of a number of facts, which are incidental and which are vital."[7]. And don't dismiss the smallest details - Holmes made it clear that "The little things are infinitely the most important."[8]

There are lots of individual pieces.  Some, like the broken seals and the missing ballots, should call for an investigation on their own.  Elections are simply too important to not make sure that any problems this time, won't happen again.

This time, the vote was not close in any particular race.  Thus, the investigators need not worry about the political consequences of their findings.  Though there are potential consequences for individuals who may have fallen down on the job.  Higher level public administrators are not merely supposed to do things the way they've always been done, they are expected to be aware that the world is constantly changing and they have to be adapting to those changes.  Based on the video I did with former elections head, Lupe Marroquin, some good things from the past weren't done and some new bad things were.

In any case, I'm adding this long explanation so that people understand that the video, which raises the issue of the gap between the predicted outcome and the actual outcome is simply one of many clues that, all together, raise serious questions about the integrity of the election.

When there is an accident, the police can conclude it was simply an accident and move on, or they can notice clues that suggest that perhaps there was more to it.  I'm just saying, along with others, that there are enough clues, little things that are out of place out there, that suggest there is possibly something more than just an accident.

The Assembly Chair seems to agree, because he's going to appoint an investigator.  I'm just adding this bit of evidence into the mix so that it's considered during the investigation.