WASHINGTON (Reuters) - Despite an embarrassing failure in their forecasting four years ago, political scientists and economists are again predicting the outcome of the presidential election, and most foresee a win for President George W. Bush.

"If this election follows historical patterns, it looks very likely that Bush is going to win," said Ray Fair, a Yale University economist whose model is built mainly around gross domestic product growth and predicts that Bush will take 58.5 percent of the vote.

Current polls show a very close race with many suggesting that Democratic nominee John Kerry may be slightly ahead.

There is a wide variety of election models available. Each takes different combinations of factors to calculate a prediction for the Nov. 2 vote but all rely heavily on economic data -- usually growth, inflation, unemployment, wage growth or a combination of these factors.

Chris Wlezien, a political scientist based at Nuffield College at the University of Oxford in Britain, is predicting Bush will win 52.5 percent of the vote in a model that combines income growth with presidential approval ratings.

Four years ago, Wlezien forecast that then-Vice President Al Gore would win with 54.5 percent of the vote. In fact, he took 50.5 percent and lost the state-by-state Electoral College vote after the Supreme Court settled a bitter dispute over the outcome of the vote in Florida.

"Of course, forecasting tomorrow's weather is a lot easier than forecasting what it will be on Christmas Day," Wlezien said. "It's 60-40 the numbers will move in Bush's direction between now and Election Day but nothing is certain."

Michael Lewis-Beck at the University of Iowa, is predicting that Bush will win 51 percent of the vote, which makes the election too close to call. After predicting an easy victory for Gore in 2000, Lewis-Beck has changed his model to take account of jobs growth over the four-year presidential term. Under Bush, there has been a 1.1 million net job loss.

"Bush has the worst jobs record of any incumbent since the Second World War and that is hurting him," Lewis-Beck said.

History suggests that when economic growth exceeds 2.6 percent, an incumbent president or another candidate of his party will win. Gross domestic product climbed at annual rates of 4.5 percent in the first quarter of 2004 and 3 percent in the second.

The last three incumbents to lose -- Herbert Hoover in 1932, Jimmy Carter in 1980 and Bush's father in 1992 -- all did so when the economy was either in recession, or in Hoover's case, depression.

The big unknown in this election is the effect of the war in Iraq, which most models struggle to capture.

"Our equation says the economy ought to be good enough for Bush to win the election," said Nigel Gault of Global Insights, a consultancy based in Lexington, Massachusetts. "If the election turns out to be not entirely or not mainly about the economy, Bush might be vulnerable," he said.

YOUR E-MAIL ALERTS Follow the news that matters to you. Create your own alert to be notified on topics you're interested in. Or, visit Popular Alerts for suggestions. Manage alerts | What is this?

Another criticism arising from 2000 is that most models do not measure the state-by-state votes that actually decide the election. One that does is compiled by Economy.com, an online provider of economic and financial research based in West Chester, Pennsylvania. Their model looks at national consumer price data and state unemployment data.

"We're showing Bush with 53.7 percent of the vote but over the past couple of months our forecast has come down. We seem to be moving towards a tighter election," said senior economist Robert Dye.

He added: "The caveats are legion. We do not consider any type of voter preference on foreign policy and we are not accounting for noneconomic issues. To the extent they are more important, they could make a difference."