What's Up With the Gallup Model?

[babylonsister]

http://www.boomantribune.com/story/2010/10/20/222726/61

What's Up With the Gallup Model?

by BooMan
Wed Oct 20th, 2010 at 10:27:26 PM EST


Let me ask you a question. Do you think that Latinos, Asians, Native Americans...basically our non-white/non-black population...are going to vote for the Republicans in this election by a 52%-42% margin? Yeah, neither does Simon Rosenberg, but that is what Gallup's model is predicting.

The non-white, non-black portion of the American electorate went more than 2:1 Democratic in 2008 and 2006. Hispanics who make up the largest portion of this slice of the electorate, voted 70% to 30% for the Democrats in 2006, and 67% to 31% for President Obama over John McCain in 2008. In two recent polls of Hispanic voters, these basic ratios have not changed, and if is any movement to be found in these polls it is a drop in support for the Republican Party since 2008. A Latino Decisions poll has the Latino vote now at 59% Democrat, 22% Republican. The most recent, and highly respected, Pew Hispanic poll had it at 65% Democrat, 22% Republican. The gap between the most respected Latino poll in the nation - Pew - and this recent Gallup model is 50% percentage points.

Much of the remaining portion of this non-white, non-black slice of the American electorate is Asian. This community actually voted more Democratic in 2008 than Hispanics.

The problem here is not differential turnout. It could be true that Latinos and Asians and Native Americans will not turn out in the same numbers as whites, blacks, or the rest of the electorate as a whole. That would change Gallup's prediction on how many people vote, but it would have no effect on their prediction on how they vote. As Rosenberg points out, it defies common sense that Latinos will be more supportive of the Republican Party this year than they have been in the past.

Given that the Republican Party has actually ratched (sic) up its anti-immigrant rhetoric and activities in the last two years, the idea that the most heavily immigrant portion of the American electorate would see the largest swing to the GOP in 2010 of any slice of the American electorate simply doesn't pass the political laugh test.

I'm not saying that all pollsters are liars and that we don't have some pretty big concerns going into the midterm elections, but people need to remember that a bad profile of the electorate is going to yield a bad prediction of the election results. We've come a long way in getting more sophisticated about interpreting polls over the last four years, but we are still influenced greatly by them.

I've mentioned this before but it demands retelling. The only day that pollsters actually pay a price for being wrong is election day. The rest of the time, there is no penalty for jiggering your model to get the results you want. No one can ever prove that you were wrong because there is no objective fact to compare your poll to. Pollsters do want to have a record of accuracy, and that's why a lot of these dubious polling outfits change their models to something realistic just before election day.

One more thing. All these internal polls that are paid for by candidates for office? It's not that they only release them when they have good news. They pay those pollsters to create good news. Any polling outfit that continuously produces polls unfit for release is going to get replaced by an outfit that understands its job description.

[Historic NY]

1. They still can't figure out the last upset in 08 elections.

[fugop]

2. Wow

I'm not a complete poll naysayer, although I've been highly skeptical of the "likely voter" models this year. This kind of thing is the reason why. When I see pollsters who shift groups of voters from one party to the other like this, in spite of all recent evidence to the contrary, I find my eyes rolling almost to the back of my head.

I can't believe Latinos have shifted this much into the Republican column. Seems absolutely ridiculous on its face.

Sure will be interesting on election day to see breakdowns of voters!

[uponit7771]

3. At the LEAST Gallup should be questioning their own numbers on this

[Clio the Leo]

4. kick nt