How to Think About Polls

Here's how I think about polls I'm seeing, and how I would argue you should think about them.   I'm going to take for analysis today's Fox News poll, which shows a 46-46 tie.

1. Look at the incumbent's raw number.   First, forget about the tie.   The key number is that Obama is stuck at 46%, just as he's been stuck between 46-48 in nearly every poll for a month.   If an incumbent can't break 50%, that means that he hasn't closed the deal with a majority after four years.   Undecideds will break against the incumbent, perhaps even more so in this election, because there will be an aspect of a "preference cascade"... when they break, they'll break like a wave.

2.   Look at the likely voter screen.   Second, the poll shows a sample of 1230 registered voters, and 1,128 likely voters.   That means their likely voter screen only eliminated 102 voters, or about 8%.   92% of registered voters will not make it to the polls in 2012.   In 2008, in Wisconsin -- a state that traditionally has very high participation -- only 80% of registered voters cast ballots.   Since GOP voters tend to be more consistently "likely" voters, the weak screen undoubtedly overcounts Dems, including young people who may want to say they'll vote, but really won't.

3. Look at voter intensity.   Third, Romney voters have more intensity, with 10% more saying it is "very important" to them that their guy wins.   In 2008 Obama held the edge in intensity by 7% over McCain.   That's a big drop.

4. Look at the issues voters care the most about.   Fourth, on the things people care about in voting, the economy and jobs, Romney is winning (up 5 points on "managing your tax dollars," up 5 points on "reducing the deficit," up 9 points on improving the economy and creating jobs," and up 13 points on "has the right experience to create more private sector jobs in the next four years."   Romney's focus on jobs has been a winner for him.

5. Look at independents.   Fifth, Romney is winning among independents by seven points.   That flips Obama's edge from 2008 by 22 points.   In an evenly divided partisan environment, that means Romney will win.

6. Look at the party ID split.  Lastly, I look at the party ID split.   In 2008, the nation as a whole was D+7.   Nothing I see in terms of intensity and just in terms of the psychology of the electorate today suggests that there is anywhere near the wave for Obama that happened in 2008 at the height of "hope and change" mania.   I expect to see a D+2 or D+3 electorate.   In the Fox poll the sample was D+5.   That's too much.  

Put it all together, and this poll seems skewed toward Obama in terms of party ID and the likely voter screen.   But it does capture an electorate that thinks Obama hasn't done a good job (hence his raw score ceiling) and thinks that Romney would do a better job on the economy and jobs.  

I'm sure the campaigns have better polls with stronger likely voter screens and more reasonable party ID splits, and that they're showing that Obama is behind by 3-5 points.   Hence is scrambling in Wisconsin this week, a state he won last time by 14 points.