...it's not dark yet, but it's gettin' there...

August 13, 2004

i Say Again, Don't Believe The Polls

Regular listeners to Professor Hewitt's radio show are already aware of this story, but i thought i'd reiterate it with some links.

In the recent Colorado Senate primary, the pre-election buzz was that the GOP candidates, Pete Coors and Bob Schaffer, were in a statistical dead heat. In fact, AP repeated this assumption on the day of the election.

Then, Pete Coors won by a margin of 61% to 39%!

Twenty-two percentage points is a pretty decent margin of victory, and while the press avoided calling it a landslide (the Democratic candidate won his primary with 73% of the vote) i would not hesitate to call it just that.

How did the pre-election polls get it so wrong? Were the pollsters biased? Maybe not, the primary was between two Republicans, after all. Were the polling methods faulty? i don't know the details of that particular Colorado poll, but in my opinion, most polls are screwy and inaccurate by nature.

The only polls i put any stock in are Zogby's exit polls, because they've been shown to be the most accurate after the last two presidential elections.

Another problem with poll accuracy is that people who do vote are increasingly less likely to pick up the phone, thanks to telemarketing abuse. i don't think this problem necessarily favors one party over the other, but it does make the raw data suspect. And that requires the pollster to make assumptions about who is being underestimated when the pollster adjusts the numbers for "accuracy."

The point i want to make is this: i think there's a lot more support for the GOP, and specifically for Bush-Cheney, than the pollsters and the media are willing to recognize or admit. Most of the presidential polling is deliberately skewed in favor of the Democrats, in my opinion. (Dick Morris explains how the media accomplish and justify thier biased polling in his book, Off With Their Heads.) i'm not saying the pollsters are lying. i just think they overestimate the amount of Democratic support when they adjust the raw data.

The Coors election shows how wrong the polls can be. The lesson i'm hoping to extrapolate from Colorado is that in this post 9/11 era, polling and voting are two vastly different things. i think people are a lot more serious about their vote when they actually get in the booth. They may support any number of candidates during pre-election polls, but when it's time to pull the lever, i think there's a newfound tendency to lean towards the conservative side.

i'll be very interested to see if my theory holds true in November.

Posted by annika, Aug. 13, 2004 |
Rubric: annikapunditry



As usual, you are on point. An additional thought is the interpretation of polls.

By way of illustration, take the national polls. They reflect (or hold themselves out to reflect) the overall popular vote at any certain moment in time. But to properly interpret them, you'd need to see the state by state raw numbers, as several states are just so lopsided that the electoral vote is often times nowhere near the way the polls appear to read.

I suspect that right now, Bush is pretty far ahead in the electoral college vote, while even or a bit behind in the popular vote.

Posted by: shelly s. on Aug. 13, 2004

Annika: I hope you're right.

Shelly: If she's not, you're probably not, either.

Posted by: Xrlq on Aug. 13, 2004

Allah willing. I would absolutely wet myself with happiness if Bush wound up winning by a decisive margin. (actually I'll probably still wet myself if he simply wins)

Posted by: Bill from INDC on Aug. 13, 2004

So Bush makes us save when he goes to war so that we can defeat an enemy -terrorism- that he
now admits can not be defeated? Are you people for real? I guess the fact that most Republicans never graduated high school or at most attended but a single year in college is to blame for your
complete lack of logic and reason.

Posted by: Peter R. Green on Sep. 9, 2004

Nice spelling and grammar, idiot.

Posted by: annika! on Sep. 9, 2004