Follow the feelings beyond the numbers.

ONE of the most popular election-year sports among U.S. political junkies is divining a set of magic indicators to predict the presidential election.

The best-known predictor is Yale economist Ray Fair, whose economic model accurately produced the winning party of the popular vote in 18 of the last 22 presidential elections.

In 2000, Fair's model predicted Vice President Al Gore would win the popular vote by 51 percent to 49 percent for George W. Bush. That proved an inaccurate indicator of who would end up in the White House, yet it was almost dead-on in the popular vote outcome.

Fair relies predominantly on the number for gross domestic product growth (or more precisely, the growth rate in per-person GDP, adjusted for inflation), while taking into consideration incumbency, inflation and a few other minor variables. His current model has Bush winning the election in November with 58.7 percent of the popular vote.

Yet in the 1992 election, Fair's model incorrectly predicted that George H.W. Bush would win the popular vote over Arkansas Gov. Bill Clinton by more than 10 percentage points. Clinton won by more than a 6-point margin. Did an ordinary statistical error cause the incorrect '92 prediction? Does it point to some larger economic and political realities that could be relevant to the current White House race?

While overall economic growth is important, Fair's incorrect prediction illustrates the danger of relying on a major economic indicator to gauge voter sentiment.

By 1992, GDP growth had returned to respectable, though not spectacular, levels. Underneath those figures the labor market was still extremely weak. As voters were making up their minds in '92, unemployment stud at 7.3 percent, and in some regions was substantially worse--8.7 percent in New York, 9.7 percent in California and 11.3 percent in West Virginia. Consumer confidence stood at one of its lowest levels ever. Those realities didn't show up in the GDP numbers; they did in the election results.

Ironically, the same factor that helped Clinton in 1992--the lack of good feeling about the economy--also worked against the Democrats in 1994. Even though Clinton had fulfilled a number of his economic campaign commitments--there wasn't enough good feeling about the economy to keep a demagogic and well-disciplined Republican political machine from a historic Congressional victory.

Today, the Bush administration has two quarters of healthy GDP growth behind it and one strong month of job growth with the March number--an increase of 308,000 in non-farm payroll jobs, the biggest monthly rise in four years. Yet like a decade ago, those numbers don't reflect what average people are feeling.

Even with that boost from a lone month of good job growth, the public is still reeling from the worst loss of jobs since the Great Depression--2.6 million in the private sector since Bush took office and nearly 1 million lost even in the so-called recovery. In his 2004 State of the Union address, Bush erred by failing to even once acknowledge this anxiety and pain. It wouldn't have been a sign of weakness to use the single best opportunity given every year to a president to talk directly to families in their living rooms to show that he understands many of them are still struggling.

For Democrats, and especially Sen. John Kerry of Massachusetts, the real issue is to show an understanding that people are worried about health costs, education and the future of good middle-class jobs, and that an agenda to meet those needs that's far better than Bush's.

While I'm an adviser to Kerry, my advice to Bush or anyone running for public office is the same: Beware of the headline economic numbers. And if given a choice between a town meeting and a clever economic model to gauge voter sentiment, take the town meeting every day of the week.

Gene Sperling, an adviser to Democratic presidential candidate John Kerry and Jormer President Bill Clinton's top economic advisor, is a columnist for Bloomberg News.