In January 1980, I was a gangly teenager with a disturbing gambling habit. Stopping off at a betting shop one day in my hometown of Leeds, in northern England, to check on some horseracing results, I spotted some odds for that year's U.S. presidential election, which was about to enter its primary season. (In Britain, betting on sports and politics—and most other things—is perfectly legal.) As I recall, there were two favorites for the White House: Jimmy Carter and Ronald Reagan. Each was quoted at 7 to 4, meaning you could bet £40 on either of them to win £70. Longer odds were available on Ted Kennedy, Carter's main opponent in the Democratic primary; George H.W. Bush, who was challenging Reagan for the Republican nomination; and John Anderson, an Illinois Republican who ended up running as an independent.

My knowledge of American politics was perfunctory: I was aware that incumbent presidents were usually reelected; that Carter was an eloquent, God-fearing Southerner; and that Reagan was a former B-movie actor. Surely, I said to myself, Carter was a shoo-in. I vowed to withdraw the princely sum of £20 from my savings account and plunk it down on him to win the election. Mercifully, the banks had closed for the day. The next morning at school, I mentioned my plan to my history teacher, who gave me a quizzical smile and said, "I wouldn't be so sure that the American people won't vote for Reagan." Those words punctured my confidence and preserved my meager savings. Ten months later, Reagan trounced Carter.

Twenty-eight years on, I am still interested in predicting elections, but these days my methods are somewhat more scientific. In recent decades, economists have invented their own statistical voting models, which in some cases have better track records for forecasting outcomes than conventional polling. Built on the basis of data gathered from elections going back many decades, these models take variables relevant to the election—such as economic statistics, polling figures, and how long the incumbent party has held the presidency—and use them to determine which party will win the White House. Like the quantitative models favored by Wall Street firms and hedge funds, these mathematical creations are occasionally subject to embarrassing blowups. But they also provide a useful way of cutting through all the campaign coverage. The key message this time around: With the housing bust and credit crunch, the faltering economy could prove as damaging to the Republicans as the war in Iraq. (Read about how shifts in the economy often usher a new party into the White House.)

Ray Fair, a Yale University professor of economics, developed the oldest and most (in)famous voting equation, which he has used since the 1980 election. His model pays no heed to polling data or the candidates' identity, which, early in the campaign season, when polls are notoriously unreliable and the primaries are still taking place, is a distinct advantage. Fair's two key variables are the growth rate of U.S. gross domestic product in the three quarters leading up to Election Day and the average U.S. inflation rate during the previous 15 quarters. The idea is that voters reward parties that deliver prosperity and punish ones that mishandle the economy.

Fair's system also takes into consideration the historical success of incumbent presidents and the tendency of voters to tire, eventually, of incumbent parties. The model has correctly predicted five of the past seven elections: 1980, 1984, 1988, the disputed 2000 election, and 2004.

In November 2006, Fair said the 2008 race looked "very close" and avoided picking a winner. Obviously, the Republicans face a tough challenge. Their president is retiring, and they started an unpopular war. But until recently, the economy was strong, and inflation had stayed fairly low—both of which favor the incumbent party. Since October, however, the credit and housing meltdown has soured economic prospects for 2008. To see how recent developments have affected the presidential race, I took the latest forecasts from the Federal Reserve Board, which predict that G.D.P. growth could be as low as 1.8 percent this year, and plugged them into Fair's equation. The result: a clear victory for the Democrats, with 51.4 percent of the two-party vote. (While this is good news for the Democratic field, it isn't definitive, as Fair's margin for error is about 2.5 percent.)

Another vote-predicting equation that emphasizes economic conditions is the Bread and Peace model, which Douglas Hibbs, a well-known political scientist, created in 2000. Hibbs uses a different measure of economic performance: the average per capita growth rate of personal disposable income over the previous four-year term. He argues that this is a better measure of voters' well-being than G.D.P. growth and inflation, because it takes into account all sorts of income, including payments from the government, and is adjusted for taxes.

Hibbs claims that his system, when fed the appropriate data, accurately picks the winner of 12 of the past 14 elections—1996 and 2000 are the exceptions. In 2004, following two years of strong income growth, the Bread and Peace model predicted a victory for George W. Bush.

Hibbs hasn't yet published a forecast for 2008, so I used his approach to create one. The latest version of the Bread and Peace model, which he updated last year, assumes that for each percentage-point rise in per capita disposable income that is maintained over the previous electoral terms, the incumbent party's vote will be boosted by 3.6 percentage points above a benchmark of around 46 percent. Over the first three years of Bush's second term, per capita income grew at an annual rate of around 1.8 percent. If this growth rate were to persist for another nine months, then on economic grounds alone, the Bread and Peace model would put the Republican candidate's share of the two-party vote at about 52.7 percent, enough for a comfortable victory.

This calculation doesn't account for the possible economic slowdown in 2008. If the economy turns down slightly during the next nine months—as many economists expect—the predictions don't change dramatically. For example, should the annual income growth rate fall to 1 percent, the Republican share of the two-party vote would slip to about 52 percent; if income doesn't grow at all between now and the election, the Republican voting share would be approximately 50.9 percent. Bad news for the economy is good news for the Democrats.

Of course, noneconomic factors will also influence the election, none more so than the war in Iraq. The Bread and Peace model acknowledges that American voters dislike seeing their soldiers die in foreign wars. In the 1952 and 1968 elections, anger at the rising number of fatalities in Korea and Vietnam, respectively, may have helped unseat the Democrats, who took America into both conflicts. This year, Hibbs believes, the Republicans will pay a price for starting wars in Iraq and Afghanistan.

How high a price? According to Hibbs' model, every 1,000 American military fatalities costs the incumbent party 0.3 percent of the two-party vote. Given that the U.S. military death toll in Iraq and Afghanistan is now close to 4,500—and may well reach 5,000 before Election Day—the model predicts that the war will decrease the Republican vote by as much as 1.5 percentage points. In a tight race, that could prove decisive.

At least one eminent economist believes that the Bread and Peace model understates the electoral importance of Iraq. In a paper published in 2005, William Nordhaus, a colleague of Fair's at Yale, argues that the Iraq war almost cost Bush the 2004 election. Given the healthy state of the economy that year and the fact that he was an incumbent, Bush should have defeated John Kerry by about 15 percentage points. But his actual margin of victory was just 3 percent. If this logic holds up in 2008, the Democrats should win comfortably.

All of this leaves me persuaded to put some money on the Democrats. However, when I logged on to WillHill.com, a British betting site, I found that the odds of a Democratic victory had already tightened to 1 to 2. That means if I wagered $100, I stood to win just $50. (The Republicans were on offer at 6 to 4, meaning I could have put down $40 to win $60.) With the candidacies of Michael Dukakis and Al Gore firmly imprinted on my memory, I couldn't bring myself to back the Democrats at such unfavorable odds.

These odds indicate that bookmakers believe the likelihood of a Democrat triumphing in November is 66.6 percent, whereas they see the chances of a Republican winning to be 40 percent. In short, the betting market concurs with the voting models: The Democrats hold a significant advantage, but the race is far from over.

I've decided to wait and see who the nominees will be. When one is making bets, waiting often pays off. The best time to try to predict presidential races is in the summer of an election year, when timely economic information is available and opinion polls are more reliable.

Several political scientists have devised voting models that combine economic statistics and survey data. Alan Abramovitz, of Emory University, uses three indicators: G.D.P. growth rate in the first half of the election year, the president's approval rating in June, and a "time for change" factor if the incumbent party has been in office for two or more terms. Michael Lewis-Beck, of the University of Iowa, uses the same economic variable as Abramovitz, the president's approval rating in July of the election year, and a polling indicator that captures the electorate's feelings about noneconomic issues. James Campbell, of the State University of New York at Buffalo, waits until early September before issuing a forecast. His model has just two variables: the growth rate of G.D.P. in the second quarter and how well the incumbent party's candidate performs in polls conducted on Labor Day weekend.

Each of these model builders has had some success, and details of their work are available online. Campbell, for instance, correctly predicted the popular-vote winner in 1996, 2000, and 2004—a feat that neither Fair nor Hibbs managed. But picking the winner a month or two before an election is hardly the same as forecasting the outcome 12 months in advance. If you wait until Election Day is that close, you don't actually need a complicated statistical model to get decent results. Here are three simple rules of thumb that work almost as well:

1. If the annualized growth rate of G.D.P. in the second quarter is 2.6 percent or higher, the incumbent party will retain the White House. If the growth rate is 1.5 percent or lower, the other party will win.
2. If the president's approval rating in the mid-June Gallup poll is below 45 percent, his party will lose in November. If his approval rating is 51 percent or higher, his party will win.
3. Whichever candidate is ahead in the Gallup poll taken in the third week of September will win the election.

I gleaned these points from an informative paper that Randall Jones, a political scientist at the University of Central Oklahoma, posted online before the 2004 election. Using them as my guides, I hereby predict that the Democrats will carry the election.

But please don't hold me to that forecast, much less bet on it. In case the economy rebounds strongly from the credit crunch in the next six months and the situation in Iraq improves, providing a boost to President Bush's approval rating, I reserve the right to change my mind.