With almost every new political poll leaning Democratic, it's no surprise that President Obama's re-election prospects are suddenly looking brighter.

But when University at Buffalo political scientist James E. Campbell summons more than six decades of historical background and a plethora of current data from all kinds of key indicators, a prediction of "a close race tilted toward Obama" may carry more weight than any poll out there.

"I estimate that there is a 67 percent chance that President Obama's vote will be over 50 percent," Campbell said this week in his quadrennial presidential prediction, "so the forecast is for a close race tilted to Obama.

"The prediction is not so definite that a Romney win is impossible," he added, "but an Obama win is more likely."

In other words - it's close.

Campbell, the chairman of the UB political science department whose presidential election forecasts are highly anticipated by political observers, issued his latest study on Tuesday. Though he acknowledged that the electoral college system will eventually determine the race, he prognosticated that Obama will capture 51.3 percent of the national popular vote.

Still, he noted the rarity of a candidate winning a popular vote plurality but not a majority of electoral votes - with the 2000 Bush-Gore presidential race providing the most famous exception.

Campbell arrives at his prediction by plugging results of the Gallup Poll immediately after the political conventions, the second quarter gross domestic product results, and a host of other factors into a formula that - for the most part - accurately predicts the outcome of the presidential contest.

His formula stumbled in 2008 when he predicted at this same time in September that Sen. John McCain would defeat Obama with about 53.8 percent of the national popular vote. Barring an "October surprise," he said back then, McCain would prevail.

But the catastrophe that afflicted the world financial markets just as he issued his report constituted exactly such a "surprise," and the string of accurate predictions was disrupted. But for the most part, it has proven eerily accurate when applied retroactively all the way back to 1948.

Campbell said the model has successfully called the winner except for extremely close elections like Harry S. Truman's upset over Thomas E. Dewey in 1948 -when it missed by only .05 percent. The formula also incorrectly predicted the outcome in 1968, when Richard M. Nixon narrowly beat Hubert H. Humphrey - by .6 percent - a difference that was still within the formula's margin of error.

His forecasts and how he arrived at them are reported in his new article, "Forecasting the Presidential and Congressional Elections of 2012: The Trial-Heat and Seats-in-Trouble Models" for the journal PS: Political Science and Politics, a publication of the American Political Science Association.

He also edited 12 other forecasting models for the journal's October issue, which will appear online in the next few weeks.

According to information supplied by UB, the 13 forecasting models vary from a modest Obama plurality to a slight Romney bulge, with most predicting the election as much closer than in 2008.

Five models predict an Obama plurality, though three of those are on the cusp of predicting the election to be a toss-up. Five of the models predict a Romney plurality. The final three call the election a toss-up, though each sees a slight tilt to Obama. The median forecast is of a 50.6 percent Obama vote squeaker, UB officials said.