Pollster.com

Articles and Analysis

 

Rendell Gets Lucky

Topics: Bob Casey Jr. , Ed Rendell , Hillary Clinton , Pollsters , Quinnipiac

Back before I shifted to blogging full time, I did fair amount of survey work in a particular locale where, by some odd twist of fate, we always managed to interview some relative or close friend of the client on every project. One reason for this phenomena, no doubt, was that we were calling into a relatively small county. As such, the odds of reaching someone just one degree of separation removed from the client were not that long. But someone who is the object of a question on the survey in a state the size of say, Pennsylvania, involves some very long odds.

Those long odds are what makes this story about the most recent Quinnipiac survey, blogged by Brett Lieberman of the Harrisburg Patriot-News (via Ben Smith), one for the record books:

[Governor Ed Rendell] was hanging out at home last weekend when the phone rang. On the other end of the call was a pollster from Quinnipiac University, who had no idea whom his computer had dialed. "That was funny," said Rendell, who says it was the first time in his life that he's been called by a pollster.

What was funnier was his response to a question about whether the governor's endorsement of Hillary Rodham Clinton make him "more likely to vote for Clinton, less likely to vote for Clinton, or doesn't it make a difference?"

"When they asked if Ed Rendell's endorsement had any impact, I said, 'Absolutely,'" Rendell told us last night.

Of course, the next question asked whether U.S. Sen. Bob Casey Jr.'s endorsement of Barack Obama had an impact on his vote. Rendell and Casey fought a tough 2002 primary and are often referred to as political rivals said. Rendell's response? "I said, 'no difference.'"

Anyone willing to calculate the odds of Quinnipiac's reaching Rendell and how they compare to say, winning the Powerball lottery? Extra credit for factoring in response bias.

 

Comments
FlyOnTneWall:

Well, I know it was offered firmly tongue-in-cheek. But I'm a geek. I can't help myself.

Starting in January of 2007, there have been 50 polls with questions on the Democratic Primary in Pennsylvania, according to the always-reliable data on Pollster.com. For the 42 polls for which data is available, there were 25,400 voters in their samples. Extrapolate that out, and we're talking about roughly 30,000 overall. Let's assume there's some repeat sampling, though, and conservatively knock it back down to a nice, even 25k. And there are about 4.2 million Democrats in Pennsylvania.

So the odds work out to about 1:168. Improbable, but better than winning Powerball by a fair margin. And when you consider that a decent number of the polls are sampling likely voters - a smaller subset than registered voters - it may be that the real odds were more like 1:100. Not to mention (because I won't) response bias. Suddenly, the remarkable thing is not that they reached Rendell, but that no one's called Casey! I smell a conspiracy...

____________________

HESSIAN:

Forgive me if this is too simplistic as it's been some time since I took statistics, but if we assumed that the sampling was random and that all the samples were of 504 voters and those were the only people they had to call, then the odds would be 1-[{(4,200,000-504)/4,200,000}^50] = .00598 or 1:~167 is the chance that he would be called. However, this is really the chance he would be included (assuming again, and clearly incorrectly here, that he was no more likely to be screened in or out or choose to respond than anyone else who was called). I'd guess that each of those surveys had to call ~8K people to get the ~500 sample size. Furthermore, they may have called a broader universe of people than the 4.2MM to search for the recent new registrants and party switchers (who are part of the 4.2MM number, but probably wouldn't show up on any lists yet). I don't know if the pollsters have access to the party registration lists or if they have to call all registered voters and ask about party registration. Furthermore, how would they find newly registered voters who number >200K and are not insignificant; they might have to call all adults and screen from that. All told too many imponderables, but I'd say that Rendell's chance of being polled was actually relatively high:
let's assume that the pollsters called ~8k people each time in a universe of 5MM, and the calling was purely random. In this scenario, Rendell's chances of being called as a member of that "universe" would be ~7.7% or on the order of 1:13. I'd buy plenty of Powerball tickets with those odds.

____________________

cinnamonape:

In addition, given that there are 100 primaries and caucuses...so it's almost inevitable that SOME Governor in one state being polled regarding the primary or caucus in their State at some time this election.

When looked at in this way, rather than the probability of Rendall being polled we see that the odds increase considerably. It's quite simply the odds of any one person in a sample of about 50 individuals in various States being polled at some time in the Primary season.

What's most interesting is that, although this has likely happened a few times we haven't heard of it until Rendall told his tale. Apparently the other Governors are more tight lipped about their participation in such surveys.

____________________

Tom Veil:

Well, what makes this interesting is not that the governor was called by a pollster, but that he was asked questions about himself. That's a considerably rarer event -- but I'd still buy powerball tickets with those odds.

____________________

At the risk of being overly cynical for a cheap gag:

People, remember that this is Pennsylvania. If you want to win the lottery, you have to know which balls haven't been injected with water. Statistics has nuthin' on who ya know.

Of course they got the Gov in on the poll.

:-)

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR