One thing to remember with all the polls we’re seeing is that the result on Tuesday could depend more on differential regional turnout than anything else. And the polls have to guess about turnout. The last time Pennsylvania had a competitive statewide primary in the Democratic Party was in 2002, when Bob Casey and Ed Rendell faced off in the gubernatorial race. It pays to look at what determined the outcome.

Casey led in all but 10 of the state’s 67 counties, but those combined margins weren’t enough to close the huge gap Rendell built among the voters who knew him best…

Rendell was able to capitalize on a relatively heavy turnout in the Delaware Valley — 35 to 40 percent in Philadelphia itself, according to preliminary estimates — while Democrats elsewhere in the state were more inclined to stay home.

Mark Wolosik, who heads Allegheny County’s elections division, estimated the Democratic turnout would be approximately 28 percent. In the Democratic primary for governor four years ago, by contrast, Allegheny County’s turnout percentage was double Philadelphia’s.

Rendell amassed a 150,000-vote margin in his hometown — a margin roughly equal to his overall edge over Casey. Casey led in 57 of the state’s 67 counties, but not by enough to overcome Rendell’s Delaware valley totals. The Philadelphian won the surrounding suburbs by one double-digit margin after another — Montgomery by 42,000; Bucks by 34,000; Delaware by 32,000, and Chester by 17,000.

Rendell won because turnout in the Philly area was 7-10% higher than in the Pittsburgh area, even though four years earlier turnout had been twice as high in the Pittsburgh area as in the Philadelphia area. I don’t know what turnout model the different pollsters are using, but all the anecdotal evidence I am hearing is that the black community in Philadelphia is going to turnout like gangbusters. Will the pro-Clinton areas around Pittsburgh and Scranton have comparable turnout? I kind of doubt it.

There is a good reason to believe that Clinton’s areas will not turnout. Clinton has been running negative advertising in the areas of the state where she is strongest and using more neutral or positive ads in the areas of the state where she is weakest. Bob Casey wound up making a similar mistake in 2002, although for a different reason. The Rendell-Casey campaign was relentlessly negative, but the Philly market is so expensive that almost none of the negative ads were aired there.

Because the two candidates focused their TV spending in less expensive markets, Philadelphians saw less of the acidic exchanges that filled the airwaves elsewhere. It is axiomatic among political strategists that negative campaigning tends to suppress turnout. For the last three months, the barrage of negative ads was heaviest in the areas where Casey needed the biggest turnout.

Not all the signs are bad for Clinton. There are more undecided voters in her areas than there are in Obama’s areas. And that could lead to a late break to Clinton in the undecided vote.

In the so-called “T” region of the state (i.e., almost everything between Philly and Pittsburgh), Clinton leads 51-37 with 11% undecided; this is one of the few demographic groups sporting double-digit undecided.

Two other interesting cross-tabs with high undecideds also indicate the potential that undecided vote will break for Clinton. Among bowlers (24% of the electorate) and gun owners (38% of the electorate), Clinton leads big. She’s up 54-33 among bowlers and 53-28 among gun owners; There were 13% undec. among bowlers and 17% undec among gun owners.

I wouldn’t put a tremendous amount of stock in these crosstabs, as the poll has a sample of only 625 likely voters. But it shows a potential area of strength for Clinton. The important thing to remember is that no one knows how differential turnout is going to go, and it can be decisive.

Karen Walsh, Casey’s press secretary, said she wasn’t sure what to make of the numbers.

She said Casey had anticipated a Philadelphia turnout of roughly 30 percent. Reports of a 35 to 40 percent turnout confounded those hopes.

As Rendell’s lead reached 90,000 votes, Tom Gilhooley, a Scranton city councilman, declared, “That’s a wrap.”

I’m not quite ready to predict an Obama upset. What is concerning me is the consistency of polls showing Clinton ahead and Obama trapped at or below 45% of the vote. To be more accurate, I’m concerned that Clinton is polling so close to 50%. But…

I predict that Obama will benefit from extremely high turnout in the African-American community, very high turnout in Philly and its suburbs, a superior ground operation, and much greater enthusiasm among his supporters. I have been telling friends privately that if Obama is polling within three points he will win. Two of the last three polls out show him down by three, and the other has him down by five. In other words, it’s too close to call. I feel modestly more confident that Obama will win narrowly than I do that Clinton will break 10%, but the most likely scenario is that Clinton wins with a 1%-8% margin of victory.

In any case, Clinton’s hopes for a large popular vote victory seem unlikely to materialize.