Morning must-read: The geographic sampling issues with the Quinnipiac poll

in Uncategorized by

If you don’t like the results of a poll, just wait, because another one will come along with which you may agree.

That’s what happened this week when, on Wednesday, a Quinnipiac poll showed President Obama trailing Mitt Romney by seven points in Florida.  Yet, on Thursday, a NBC/Marist poll was released showing Obama with a lead over Romney, even if Jeb Bush or Marco Rubio is his running mate.

There has been considerable discussion about how two polls could come back with such different results.

MSNBC’s First Read attributes the difference to the party identification of people surveyed. “Both pollsters use sound methods to conduct their surveys,” its analysis said, and neither weighs by party ID. In other words, what you get is what you get. For Quinnipiac, there was a 3-point GOP edge; for NBC-Marist, there was an 8-point Democratic edge.

The debate about the veracity of Quinnipiac’s polling became so heated that numbers-whiz Steve Schale and Quinnipiac pollster Doug Schwartz exchanged blog posts about why Quinnipiac prefers to rely on people’s self-described party identification instead of weighting to self-described party registration.

Well, today, blogger Dave Trotter offered this criticism of Quinnipiac’s methodology.

The sample margins are all wrong. First, and the most eye-catching error is that north Florida and the Panhandle was sampled just as much as southeast Florida. 336 of those polled were from north Florida and the Panhandle while 339 were from southeast Florida. And no matter how you look at it, that is a big discrepancy compared to the reflection of the electorate.

But even in this case, the north and Panhandle weren’t oversampled. In fact, south Florida was undersampled. So where is the big gap? It seems to be in along the I-4 Corridor. In the Tampa area, the Quinnipiac poll showed 21% of those interviewed came from that region. In the Marist poll, only 16% came from Tampa; a gap of 5%.

If we continue to Orlando and the Space coast, we even get a bigger gap. In the Quinnipiac poll, 29% of those interviewed were from this region. In the Marist poll, only 21% came from this region. Therefore, most of Central Florida, from one coast to the other, was given a 13% bump in the Quinnipiac poll, which took away from southeast Florida. And no matter how “purple” the I-4 Corridor is, taking away the southeast Florida votes will always, 100% of the time, lower the Democratic numbers in a statewide race. No question about it!

In addition, 50% of all votes cast in Florida do not come from this region.

So at this point, it may be best to take the advice of those who suggest averaging the two horse-race results; that basically gives you a tie, which is probably the most accurate way of looking at Florida at this point.

Peter Schorsch is the President of Extensive Enterprises and is the publisher of some of Florida’s most influential new media websites, including SaintPetersBlog.com, FloridaPolitics.com, ContextFlorida.com, and Sunburn, the morning read of what’s hot in Florida politics. SaintPetersBlog has for three years running been ranked by the Washington Post as the best state-based blog in Florida. In addition to his publishing efforts, Peter is a political consultant to several of the state’s largest governmental affairs and public relations firms. Peter lives in St. Petersburg with his wife, Michelle, and their daughter, Ella.