Friday, April 18, 2008

Why is Pennsylvania so undecided?

Mark Blumenthal's column yesterday took an interesting look at the Pennsylvania polls.

He notes that there's a relationship between the percentage of undecideds in polls and how Obama and Clinton are performing relative to each other. Basically most polls show Obama's percentage within a small range but Clinton's number is all over the place- greater in polls with small undecideds, worse in polls with larger undecideds. This could be an indication that leaners are going for Clinton, but we don't really prod them to pick one of the two candidates whereas other companies might do more of that. That could help explain why Obama is doing so much better with us than in other polls.

Our double digit undecided level got me to wondering what we showed in our other polls this cycle before the last weekend, and the conclusion is that Pennsylvania is a bit of an outlier. This table shows the percentage of undecideds we've had in our second to last polls so far since the race came down to just Clinton and Obama:













We haven't done it before but since PA is showing such a high level of undecideds I'm going to ask people this weekend who say they're undecided who they're leaning toward, and we'll report both numbers. That should help bring a little more clarity to the race.


jr1886 said...

Thanks guys! I really like the way you take your job and the public seriously. I am interested in seeing how the leaners plan to vote too and maybe we can infer with greater confidence the likely outcome.

One question: it seems you don't plan to factor leaners in your reported numbers? Do you plan to just release the topline numbers and the leaners separately. Or would you say something to the effect: when leaners are factored in, candidate x support increases or decreases by that %?

Again, thanks for your strong work and I look for forward to read your answer to these questions.

Tom Jensen said...

We'll say something to the effect of 'Hillary Clinton leads Barack Obama 48-47 in PA. 45% of respondents said they supported Clinton and 44% said they were for Obama. Undecideds were asked which candidates they were leaning toward choosing and with those factored in it becomes a 48-47 advantage for Clinton.'

DISCLAIMER: I made the numbers above up for the example, please don't anyone take them literally!

jr1886 said...

Thank you! I think you are doing an amazing work with the tools that you have. I like the way you guys are willing to adapt in different circumstances. I will be watching for your poll on Monday.

Web Statistics