The Difficulty of Measuring "Likely Voters" and Why Obama may do better than the polls suggest

I read all the polls suggesting a Clinton victory in PA, and they just don't seem right to me.  Having followed the race very closely the last few weeks, I believe momentum is on Obama's side heading into today, which is very different from what happened in Ohio.  Below are the possible outcomes in descending order of likeliness, imho:

1. Clinton wins by 1-4%

2. Clinton wins by 5-7%

3. Obama wins by 2% or less

4. Clinton wins by 8-10%

5. Clinton wins by double digits

6. Obama wins by more than 2%.

So, I predict Clinton will win a close race with Obama having a better shot of winning outright than Clinton winning by 8% or more. How can I make that prediction given what the polls are saying?  I think the polls may turn out to be wrong because they all may have made faulty assumptions in determining who's likely to vote and therefore underestimated turnout for Obama.  More analysis below.  

First, let me say I'm not an expert on the science of polling  But, I did a bit of research into a little discussed aspect of polling - determining which voters will be "likely voters."  How accurate a poll predicts likely voters is very crucial to the reliability of the poll, because it's not enough to determine what voters think about a candidate, it's more important to know which of those voters will actually vote for their preferred candidate.  The question is, how do pollsters figure out what voters are likely to vote?  From Slate:

Likely voter screens. All pollsters identify likely voters by gauging each respondent's stated level of interest in the election. Some pollsters also weigh the respondent's stated voting history. If you're really excited about this election but you failed to vote in the last one, some polls may screen you out, thereby excluding your opinion and failing to anticipate your vote. We tell you who those pollsters are.

Demographic weighting. Some pollsters count you as one voter no matter what your race, sex, or age. Others "weight" your vote more or less heavily depending on which group you belong to. A pollster who assumes that the election turnout will have a certain percentage of blacks, women, or people aged 18 to 29 may find that her sample has too many blacks or not enough women to match that projection. If you're a black respondent in that sample, your vote may be weighted more lightly so that blacks don't exceed their quota. If you're a woman, your vote may be weighted more heavily so that women don't fall short of their quota. We tell you who weights which traits.

You can see that trying to determine likelihood of voting and voter turnout among social groups is no easy task, and the pollster will have to make some assumptions that may or may not pan out.

For example, if a polling company takes into consideration past voting behavior in determining likelihood of voting, that may skew the results if more new first time voters vote in a particular election.  Also, if a polling company weighs lightly votes of African Americans based on assumptions of low black voter turnout, that will skew the result if blacks turn out in much higher numbers.

What does this all mean for the PA primary today?  My theory is that the polls all suggesting a Clinton win may be off the mark with their determination of likely voters and determination of voter turnout for specific sub-groups like African Americans.  Why?  Pollster Mark Blumenthal explains:

My view, and the view of most of my colleagues who poll for political candidates, is that we need to impose controls to keep the composition of the likely voters as constant as possible. However, those controls require making subjective decisions about what the likely electorate will look like on Election Day. Some weight by party (like Zogby and others). Others stratify their sample regionally to match past vote returns (like Greenberg/Democracy Corps and Fox/Opinion Dynamics) – an approach I prefer. However, supporters of the Gallup model argue that both alternatives pose a greater risk of imposing past assumptions on an unknown future.

In other words, in determining who will likely vote in an upcoming election, they have to first make assumptions about who they think will likely vote in the first place.  For example, if you assume that someone who's voted in the past is more likely to vote in an upcoming election, then you'll use that criteria to determine likely voters.  Additionally, if you assume that, based on past elections, you think there will be low turnout among young voters, then you may give the opinion of young voters less weight in the poll.  

As Mark Blumenthal points out, making such assumptions carry the great danger "of imposing past assumptions on an unknown future" and therefore completely misreading what the eventual electorate will actually look like.  

Typically, though, the conventional assumptions and wisdom about who are likely voters usually work fairly well in typical elections.  But, what happens if those conventional assumptions just don't fit the current mood of the electorate?  

And that's my argument for why Obama may do much better than most polls suggest.  I suspect most polls concluding a Clinton victory are using conventional assumptions about what the electorate will look like today (more older, conservative, rural and white electorate versus an electorate that is younger, more liberal, more urban, and less white).  

I want to suggest that such polls are completely misreading the mood and state of the PA electorate over the past few days.  Ignoring conventional wisdom and looking hard at the present reality, what I sense is enormous energy and excitement among Obama's core voters - young, black, urban, and liberal voters.  This excitement is in part due to voters thinking that Obama's nomination is imminent, and in part due to backlash against Clinton's negative scorched earth tactics.  

I also sense a shift towards Obama in suburban voters.  On the other hand, I sense less energy and a sense of defeatism among core Clinton supporters, especially among white men (the bowling and hunting group).  I don't think they'll vote for Obama, but voter turnout among white men Clinton supporters may be lower than expected.  

Although turn-out for Obama will be high and turn-out for Clinton a bit on the lower side, I think there will still be very strong turn-out for Clinton in the western parts of PA, especially among white women and older white voters.  So, that's why I still think there's a strong chance that Clinton will win the race by a small margin.  But, I also hold out realistic hope for an Obama win, depending on how big the turnout will turn out to be in SE PA.  

The one poll that I think may more accurately reflect the mood of the PA electorate today - the PPP poll predicting a 3% win.  

Now, I could be totally wrong about all of this and the conventional wisdom will turn out to be right and I may just be engaging in wishful rationalization.  But, if I believed in conventional wisdom, I wouldn't be supporting a black man named Barack Hussein Obama for President of the United States.  We'll find out soon enough!