Steve Vancore
Andrew Gillum may indeed be in the lead.
And so may Adam Putnam.
Is it possible that after Donald Trump’s endorsement and the dramatic rise in Ron DeSantis’ numbers, GOP voters are having buyer’s remorse and are reverting back to the more familiar comfort of Putnam?
Maybe. And if so, good for both of them.
But for readers of this website — political junkies, all of you! — please try to ignore many of the recent clickbait polls. Why? Because when you get past the headlines, even those readers without advanced degrees in statistics or polling methodology can and should quickly see that many of these polls are really not worth their (ahem) salt.
To help each of you get through the next few days, use the following guidelines to help you apply your own saltshaker test and decide if a poll is worthy of sharing on social media.
These are a few items that should make you question the validity of a poll:
— Small sample sizes: If you see a statewide poll of fewer than 500 respondents be wary. If it is less than 300, take out your Men In Black neuralizer and block it from your mind.
— Polls of “adults”: Several recent polls actually surveyed “adults.” As if they matter. Well yes, they matter in the grand scheme of things, but not in elections, as only 75 percent of them are even registered to vote, and only about half of those who are registered will cast a ballot in the general election. Of course, for primaries, surveying “adults” is even more of an absurdity.
— Polls of “registered voters”: Better, but still a problem. Polls — especially primary polls — should be taken only of likely voters as determined by the readily available voting history. Past is prologue, and a good pollster will ensure that a large percentage of the survey sample will contain those who have a proven track record of actually voting. And while it is true that about a third of the electorate will be infrequent and newer voters, having a sample of “registered voters” is not a good sign.
— Self-described “likely voters”: Everyone who reads Florida Politics flosses every day, balances their checkbooks every month, calls home to mom at least once a week and never – ever – misses an election. Of course, we know that only about half of registered voters will vote in the upcoming general election — and far less will vote in the primaries. Yet in poll after poll that we have taken, about 80 percent of respondents claim they vote in every election. (Sure you do.) There is a voter history and every pollster should always use it! If a pollster is relying on letting the respondent tell their voting history, this is a clue that the poll also has other problems.
— Online polls: These can be excellent if done right. But in this growing field, getting a valid and reliable online poll (especially of primary voters) is tough. But here’s a clue. If it’s an online poll and has one of the common problems listed above, it probably was not done well.
— IVR or robopolls: Also, potentially good and frankly they are proving to be reasonably reliable if the vendor knows how to manage the voter files (and we do them all the time.) But, as with online polls, they can be tricky, and they have their limitations. Also, consider this; many cannot be taken on cellphones. Do you trust a poll that does not include at least 30 percent of calls taken on cellphones? As a rule of thumb, if a robopoll also has one of the above problems, it likely is not valid.
— Is it an outlier?: Without some dramatic event (A POTUS endorsement, a criminal conviction, etc.) voters’ minds don’t change that quickly. In fact, voters are a stubborn lot, and it takes a lot to change their minds — especially this late in an election. If a poll is very different from the others and defies trends by a large margin, it is probably flawed. (Sorry to use this example because he seems like such a decent guy, but if one showed Chris King surging to the lead, we just have to assume something is wrong with how the poll was conducted.)
Over the past few weeks, we have been asked literally dozens of times to evaluate the validity and reliability of numerous published polls. Use these simple guidelines and decide for yourself.
We hope this helps.
