For analysts and the news media, the dilemma that outliers pose is that a poll with very different results from previous surveys could either be the first sign of a new trend, or -- more often than not -- be flat-out wrong.
That's why watching the averages of polls is better than seizing on the results of one particular survey. News organizations could help the situation by doubling up on their polls more frequently -- producing a sample twice the usual size, which shrinks the margin of error and yields subgroup results that have some degree of reliability.
Two Cook Political Report/RT Strategies polls have been conducted in the past three weeks, the first taken on April 27-29 using 885 registered voters, and the second done on May 11-13 involving 855 registered voters. That yields a combined sample size of 1,740. The results for the combo's 786 Democrats or Democratic leaners have a margin of error of +/-3.5 percent, while the totals for its 662 Republicans or Republican leaners have a margin of error of +/-3.8 percent.
Looking at the entire field of Republican presidential candidates, including former Sen. Fred Thompson of Tennessee, who now appears very likely to run, and former House Speaker Newt Gingrich of Georgia, who is toying with entering the race (though I have my doubts), the contest for the GOP's top spot is very, very close -- and clearly within the margin of error. In the combined Cook/RT Strategies surveys, former New York City Mayor Rudy Giuliani leads with 26 percent (28 percent in the April poll and 25 percent in the May survey) to Sen. John McCain's 24 percent (21 percent in April; 24 percent in May).
There's a big gap between the front-runners and the next two candidates, former Massachusetts Gov. Mitt Romney and Thompson, each with 9 percent (Romney hit 12 percent in April and 9 percent in May; Thompson had 10 percent and 8 percent, respectively). Gingrich grabbed fifth place with 6 percent (6 percent in April; 7 percent in May). When Gingrich is taken out of the mix, Giuliani gains 2 percentage points, and McCain and Romney each pick up 1 point.
While McCain and Thompson do better among men than women, the reverse is true of Giuliani. The former mayor also does very well among 18-to-34-year-olds. Thompson's support, not surprisingly, skews Southern. Interestingly, given McCain's fervent support for the Iraq war and the troop surge, the senator from Arizona performs much better among moderates than among conservatives -- further evidence that his party's Right doesn't trust him. Giuliani does particularly well among voters who only weakly identify with the Republican Party and among independents who lean Republican.
On the Democratic side, Sen. Hillary Rodham Clinton remains well ahead of the rest, scoring 36 percent in the combined survey (36 percent in April; 40 percent in May). Sen. Barack Obama still has a firm grip on second place, with 25 percent (26 percent in April; 25 percent in May). Former Sen. John Edwards comes in third, with 15 percent (18 percent; 13 percent). New Mexico Gov. Bill Richardson is fourth, with 4 percent (4 percent; 2 percent).
All those Democratic numbers are based on a fairly safe assumption -- that former Vice President Gore stays out of the contest. If Gore is included, he gets 10 percent, Clinton dips to 32 percent, Obama to 24 percent, and Edwards to 13 percent.
Clinton exceeds her overall numbers among Hispanics, women, voters without college degrees, residents of the Northeast and West, Democrats (as opposed to independents who lean Democratic), and "strong" Democrats. Obama does especially well among African-Americans, 18-to-34-year-olds, college graduates, Midwesterners, Southerners, and independents who lean Democratic.
What are the most-important conclusions to take from the surveys? First, the Giuliani-McCain contest is tightening, at least a bit. Second, Obama has caught up to, and perhaps passed, Clinton among African-American voters but not in the overall race, and that his popularity is driven by the youngest voters.