On the eve of the Wisconsin primary, SurveyUSA jumped ahead to my obsession: trial heats comparing how Hillary Rodham Clinton and Barack Obama match up against John McCain in a hypothetical general-election contest in key swing states. This one showed a huge advantage for Obama.

In the theoretical November matchup, McCain beat Clinton by 49-42 percent, whereas Obama clobbered McCain by 52-42.

That's a 17-percentage-point swing for the Dems, from a +10 with Obama as the nominee to a minus-seven with Clinton.

KSTP was a sponsor of this poll and it will be presented on news broadcasts this evening.

Attaching too much importance to a theoretical matchup that is nine months away would border on irresponsible. Polls can measure only the present (if that), not the future. Still, if this kind of result were turning up consistently in swing state trial heats, it would constitute a major electability argument for Obama.

Results not consistent so far
But it isn't. Last week, Quinnipiac University released trial heats for the three biggest swing states — Florida, Ohio and Pennsylvania — and they showed no advantage for Obama in any of them, with a small edge for Clinton in Pennsylvania.

Obama has generally matched up slightly better against McCain in recent polls by various organizations taken among a nationwide sample, but the keys to winning the White House are state-by-state races.

Wisconsin definitely starts the year on the list of swing states. It was the closest state in the nation in 2004, going Dem by 0.38%.

The internal cross-tabs of the SurveyUSA poll were consistent with certain stereotypes of Obama's current electability advantages. For example, McCain carries the vote of self-described independents by 51-35 against Clinton, but Obama carries independents by 47-44. Both Democrats beat McCain among the 12 percent of the sample who said that Iraq would be the No. 1 issue faced by the next president, but Obama beat him by a much wider margin (64-28) than did Clinton (47-37).

SurveyUSA is also a robodialer, which means its polls are fast and cheap, but viewed by polling traditionalists as less reliable. This poll was based on robotic interviews with 537 registered Wisconsin voters. The margin for sampling error was +/-4.3%.

My friend Charles Franklin, a University of Wisconsin political scientist and one of the main contributors to the excellent pollster.com, agrees that trial heats in swing states are the most direct current indicator of the relative electability of Obama and Clinton versus McCain. He's agnostic about robodialers, and would like to see numbers on the same topic from more traditional polling methodology.

RELATED CONTENT: Democratic delegates — by the numbers