Are Senate races really tied? Why new polls may not tell whole story.

Three new polls show statistical dead heats in Senate races in California, Florida, and Kentucky – three races that both parties see as key as they vie for control.

But the polls, conducted by CNN/Time, survey registered voters rather than likely voters, and may not give an accurate picture.

In Florida's Senate race, for instance, the new numbers give Republican candidate Marco Rubio only a tiny edge over Independent Charlie Crist (36 percent to 34 percent, with Democrat Kendrick Meek coming in at 24 percent). But another new Sunshine State poll, of likely voters, gives Mr. Rubio a double-digit lead over Mr. Crist.

Even worse news for Crist: More Democrats – whose support would be crucial to his election – seem to be backing Mr. Meek.

Both Crist and Meek unveiled new campaign commercials this week, with Crist making a very literal appeal to voters turned off by the partisan direction of this election season. In it, he walks between the words “Republicans” and “Democrats” before rearranging the large red and blue letters to spell “Americans.” “At the end of the day,” he says, “there’s only one party I work for.”

But it’s an open question how well that message will resonate; the Sunshine State poll has Crist trailing Rubio even among independent voters.

The poll on the Kentucky Senate race seems like even more of an outlier, showing a tie between Republican (and "tea party" candidate) Rand Paul and Democratic Attorney General Jack Conway. Others recent surveys by Rasmussen and SurveyUSA give Dr. Paul a 15-point edge.

In California, meanwhile, virtually all polls show a statistical tie, or near-tie, between Democratic Sen. Barbara Boxer and Republican challenger Carly Fiorina.

What’s behind the difference in the polls, and how can a layperson make sense of it all?

Those of “likely voters” try to weed out not only the people who aren’t registered and can’t vote, but also those who will probably simply stay home on Nov. 2. According to this explanation from a Gallup pollster, respondents are rated on the basis of their answers to certain questions.

In a year like this, when many voters seem uninspired by either candidate, there’s a big likelihood that many will opt to stay home in protest.

Another interesting phenomenon: Screening for likely voters tends to improve prospects for Republican candidates, due to better turnout.

The results have some commentators wondering why CNN would even bother to do a registered-voter poll this close to the election, when likely-voter polls are considered much more accurate. But perhaps the results point to at least one window of opportunity for embattled Democratic candidates in these states: Now they have to somehow find a way to inspire those backers who, at this point, are planning to stay home.