Here's the scam in Wenzel polls

[Cicada]

Wenzel Strategies polls for Repubs, including the deeply dishonest Citizens United. They have a poll showing Romney leading in Ohio. Even thr (nice but) Republican Real Clear Politics won't list Wenzel. Here's why. If you don't weight your sample to match census groups - such as by age, race etc - you get mostlly people like me - old white folks. We don't go out much so we're near the phone. And Wenzel polls are ludicrously packed with seniors. They don't lis race - I guess old folks are more likely to be white. In the Ohio poll they have more than twice as many seniors as other polls. And fewer people under 40. I think it was Wenzel who earlier had a very pro-Romney Michigan poll where only 10% of the respondents were under 50!

[geek tragedy]

1. Weighting by demographic groups can be tricky in its own right.

We don't know how many blacks, Latinos, etc will show up on election day, so if you weight by demographics you're introducing an assumption of the pollster into the data.

The weighting has to be done for each respondent, instead of each person who makes it through the likely voter screen.

[Cicada]

5. Not really

You call the right number of whites etc. You often know how many blacks, white, seniors are registered to vote using census figures or sometimes state registration numbers. You weight the sample to get the right mix of registered voters. Then you ask the likely voter questions. All good polls this year have a lot fewer young people than their number in the population - because they are less likely to register and then they are less likely to vote. If you just take the people who answer the phone - it's senior city.

A similar problem arises with hispanics. They come in more than one flavor, such as spanish dominant and english dominant. It so happens the english dominant are more likely to talk to pollsters and they are more likely to vote Republican. That's why Harry Reid massively outperformed all the polls in 2010 except polls done by his pollster Mark Mellman. Nevada has a huge hispanic population but Mellman knew how many prefer spanish and how many prefer english - and he weighted his polls to get that right. He was the only pollster to get it right. He had fewer hispanics than their population numbers - he knew they were less likely to register and if registered less likely to vote. But he knew about the flaw in other polls. So he predicts Obama will out perform polls in Nevada and Colorado because pollsters don't - yet - weight the latinos for language preference.

Yes good pollsters predict - those things they are certain about - age, sex, race etc. But not party ID. That varies from day to day so no one knows the true numbers for that.

[geek tragedy]

6. Is Mellman a NV native or is he just really, really good at his job? nt

[Cicada]

7. Mellman is just great, not from NV

He is one of the leading pollsters. His clients rarely lose. He went to grad school at Yale studying under the awesome great pollster Stan Greenberg and then also taught at Yale. He seems to know what he's doing. see the Las Vegas Sun article "How Harry Reid's pollster got it right."

[VirginiaTarheel]

2. Hilarious that their bogus poll got no publicity

except for people panicking on progressive forums.

[dixiegrrrrl]

3. My phone goes unanswered during elections.

Everyone gets voice mail.
No polls, no "research" questions, no robo-calls, no more Rachel from Card Member services.

You want me to answer a poll question, you can pay me for my time, I figure.

[ejbr]

4. They probably call using the same

exchange xxx-555-xxxx, creating a same county scenario.