Breaking Tampa Bay, Florida and national news and weather from Tampa Bay Online and The Tampa Tribune |
  • Home

About those clashing polls

Posted Aug 14, 2010 by William March

Updated Aug 14, 2010 at 06:21 PM

Politics junkies may have noticed a surprising clash in poll results on the Republican primary for Florida governor this weekend.

A Mason-Dixon poll for the Tribune and other news clients showed Bill McCollum 4 points ahead of Rick Scott, while an Ipsos poll for the St. Petersburg Times and others showed Scott 10 points up – a 14-point swing on polls released within a day of each other, both from reputable pollsters.

Whassup with that?

The answer provides an interesting lesson in the vagaries of political polling.

First of all, Mason-Dixon is the one most likely to be accurate as a gauge of the primary outcome – even Ipsos pollster Julia Clark acknowledged that.

The chief reason for the difference, according to Clark and Brad Coker of Mason-Dixon, is that the Ipsos sample was drawn from registered Republican voters, while the Mason-Dixon sample came from likely Republican voters – those who had voted in previous GOP primaries or said they were likely to vote in this one.

Primaries tend to be low turnout elections, Coker noted. In 2006, barely 25 percent of the 3.9 million Republicans then registered voted in the Charlie Crist v. Tom Gallagher primary for governor.

Those who do vote in primaries tend to be very interested in politics or at least in a candidate – they may even be party or campaign activists.

Those folks are more likely to be behind the “establishment” candidate favored by party leaders – in this case, McCollum.

In addition, the Ipsos poll was done three days earlier, starting Aug. 6, while Mason-Dixon started Aug. 9.

McCollum is probably moving up against Scott – several negative stories about Scott have broken in the past week, while McCollum has been out campaigning with the popular Jeb Bush, and McCollum only recently began saturation TV advertising. So those three days could have altered the result as some voters swung to McCollum.

So why did Ipsos use registered voters instead of narrowing down to likelies?

After discussion, the clients – the Times and Miami Herald – wanted a registered voter poll, she said. Such a poll could be more accurate on questions about the general election.

And drawing a separate sample of likely voters for the questions about the primary election would have been too expensive.

Meanwhile, a John McLaughlin and Associates poll for the Florida Chamber of Commerce showed McCollum up 12 points – a 22-point swing vs. Ipsos. MOP hasn’t talked to McLaughlin about that one, but will note that McLaughlin is McCollum’s campaign pollster and the Chamber has endorsed him.

The Times and Herald will have another installment of the Ipsos poll on the Senate race out tonight or tomorrow morning, also based on registered rather than likely voters.

Watch for that poll to show Jeff Greene looking better against Kendrick Meek, the party establishment candidate in the Democratic primary, than he does in today’s Mason-Dixon, which showed Meek 40, Greene 26.

It also may show Charlie Crist, the best known of the four major candidates, looking better in the general election matchup against Rubio, beloved by GOP insiders, and Meek, champion of Dem activists.