Post Debate Polls versus Kosher Political Polls

I've received a ton of emails, calls, and even my wife gets questioned in the Long Valley NJ Dunkin Doughnuts to talk about polls and politics.  So, let me give you an idea of which to believe, which not to believe, and what you should do with the polling data.  Oh and before you write or say to yourself - "oh you work for Senator McCain's campaign so this is slanted", let me explain to you my background. 

I graduated Rutgers University with a degree in Industrial Engineering (thinks sampling, operations research, efficiency models) and immediately went to The Rutgers Graduate School of Management for my MBA.  Once I graduated I was officially hired by AT&T Bell Laboratories in a group called Business Operations Analysis.  This group was staffed with statisticians and I built control-treatment experiments to measure the value of AT&T's promotions, I built databases, and I created predictive models.  Needless to say (really write) I know a thing or two about statistics and sampling.

ENTERTAINMENT AKA NOT WORTH YOUR TIME TALKING ABOUT THEM POLLS

These post debate polls where people go to a website or text a vote in are completely useless.  I don't pay attention to them and neither should you.  Besides the obvious answers of people voting multiple times, certain campaigns or groups game the system by rushing the vote, their sample universe isn't normalized to the voting population, and there are horrible skews when it comes to which site-news organization is sponsoring the poll.  During the Primary Season I had Google Alerts setup to monitor Ron Paul's supporters when they marshaled their people to "poll rush".  Wednesday night for example, a MSNBC debate poll showed 81% for Senator Government, 15% for Senator McCain, 2% not sure, 2% tie - is anyone REALLY shocked that the liberal MSNBC had these results?  Wow, that's important information; it’s the same with CNN and even Fox (but obviously not liberal).  I would not even waste my time with entertainment polls that allows quasi news people like Keith Olbermann to proclaim that the other guy won the debate.

REALISTIC POLLS BUT YOU NEED TO LOOK AT THEM THROUGH ANOTHER LENS

The polls and surveys run by the likes of Zogby's, Rasmussen, SurveyUSA, InsiderAdvantage, Eagleton, and etc are the more reputable ones.  What you need to look for is who the polling audience is (registered, likely to vote, etc), where they are located (national or regional), date, sample size, and some sort of quality rating of the polling firm; some polling firms are more respectable than others.  After looking at all of the numbers, you want to know what the margin of error is.  If Senator McCain has a 4% lead over Senator Government and the margin of error is 4% than the race is a dead heat.  Of course, some companies will write up a whole analysis based on the lead but in reality it is a dead heat and will change.  You also want to track the polling data over time.  Remember election day is a finite day (yes there is early voting and that only complicates matters) so what you want to see is momentum and positive movement for your candidate over time as you head towards the big day (kind of like betting on a horse that is a good closer).  As a reminder, check out this New Hampshire polling data for the January 8 Republican Primary.

OK SO WHERE DO YOU GO TO GET YOUR PUBLIC DATA?

So, I like to look in a few places to see the latest polling data. First, I have a RSS feed setup for racefor2008.com. I have to admit I spent a lot more time on that site during the primary season but they still do a great job of just publishing opinion polls as soon as they come out.  Next up is RealClearPolitics because I like the easily customizable map and the ability to drill down on a state and look at the polling data over time.  I also have a RSS feed setup for Electoral-vote only because I like the way the map is presented and you can quickly get a status on the data.  I don't like how they layer a state in based on the spread because a lot of time it is within the margin of error.   Finally, one site I pay little attention to these days is Factcheck.org. Sure a lot of people believe they aren’t as unbiased as they would have you believe, but for me I think they are going way too far. The post they lost me on was the one when they critiqued our campaign regarding Joe Biden saying that B and he don’t support clean coal. Biden said it, we caught it on video, and it is fair game. It wasn’t taken out of context and while it contradicts what is on their website, that is FAIR game. It says a lot about their character and whether you trust what they are saying. Biden made that comment with a lot of conviction. Was it a typical Biden gaffe? Perhaps, but that doesn’t mean that Factcheck should ding us for it.

What all this means is vote for the person you want to win (thanks for pointing that out to me NickV) and stop letting polls and main stream media cloud your judgment.

PardonMyFrench,

Eric

 

 

 

 

 

Comments

[Steve Nelson]

What do you think of the methodology of http://www.fivethirtyeight.com as a poll-of-polls?

[PardonMyFrench]

Steve,

I'll check that out. Thanks

Eric

[nickv]

Eric, did you really mean your last statement in this post? Vote for the person you "think" will win" Because if you really stop and "think" about it you might come to the same conclusion Frank Lund came to, that Obama will win. The two polls that struck me this past week had little to do with the "industrial" polling machines.

1. Scholastic ran a poll of a quarter of a million kids and Obama won 57 to 39%. They have not been wrong since 1940.

Results from the swing states were interesting:

Colorado: McCain 61% Obama 36% Other 3%
Florida: McCain 41% Obama 55% Other 4%
Indiana: McCain 51% Obama 47% Other 2%
Iowa: McCain 48% Obama 49% Other 3%
Michigan: McCain 40% Obama 57% Other 3%
Missouri: McCain 49% Obama 47% Other 4%
Ohio: McCain 47% Obama 51% Other 2%
Pennsylvania: McCain 46% Obama 50% Other 4%

2. The Roxbury Middle school, in a town that overwhelmingly voted for Bush in 2004 gave Obama a 27% win. 27% in Republican Morris county.

I agree with you that polls are flawed in many ways. Funny thing about kids is they are usually will telegraph what they learn at home even while their parents might lie to a pollster.

And honestly I know you're voting for the person you "want" to win.

[PardonMyFrench]

Hey NickV,

I actually meant to write vote for the person you want to win. I'm not too interested in the kids polls this year. Obama is pulling in more of a youth vote and I'd expect that to carry down to younger kids too. Thanks for your post.

Eric