Wednesday, March 08, 2006

The 411 on Public Polling

Jim Martin, Democratic candidate for lieutenant governor, has a new poll out that shows him beating his primary opponent, Greg Hecht, by a 3-1 margin.  

You know, I debated putting this thing up on the blog for a couple of days, for reasons that will soon become clear.  Finally, despite the fact that this poll means absolutely nothing, I figured if nothing else, it will fill some space, give me a chance to wax poetic about the inside baseball of political campaigns, and maybe, just maybe, educate a few folks about the mechanics of polling and exploiting those polls for your own benefit.

Exploiting a poll like this for the press hit is kind of a cheap move, which is why we weren’t even going to talk about it.  We didn’t really want to give Martin the publicity, even if it is only on AthPo.  

Anyhoo, the 3 – 1 margin that Martin’s campaign is so proud of is a 23% to 7% lead.  We were told that there would be no math, but that leaves 70% undecided.  Right away, this poll is looking less and less credible.  Of course, Jim Martin’s campaign is going to spin it by saying that Martin is such a strong candidate, people are starting to already make up their minds about him – a trend they will say is likely to continue.  Not so much.  

What else hurts the credibility of this poll?  Well, bear in mind that candidate performance in situations like this is usually measured in two ways – the uninformed vote and the informed vote.  The uninformed vote is usually a question like: “If the election for county assemblyman were to be held today, would you vote for Phineas Q. Windbag, or for Thaddeus D. Windbreaker?”  In most credible polls, the order of the candidates’ names are reversed each time, so that each candidate has an equal number of times where they are first – and yeah, it makes a difference, or at least the big brains that run the polling firms say so.

Now the “informed vote” is more interesting, and it can take numerous forms.  At its most innocuous, the pollster reads a brief, positive biographical description of each candidate (usually straight off their candidate biography) and then asks for a preference.  For instance, you might hear:
     “Phineas Q. Windbag is a democrat running for County Assemblyman.  He has served for 12 years as the county’s overseer of elevator maintenance.  In this election, he is stressing universal health care for all county residents and improving our public schools.  Phineas Q. Windbag has been married to his wife Eileen for 23 years and has three children, all of whom attend public schools.”
     “Thaddeus D. Windbreaker is a local attorney specializing in civil rights law.  He was instrumental in obtaining access for disabled residence to the weekly tractor pull.  He is a graduate of our public schools and his priorities are increased funding for our senior centers and putting more policemen on our streets.”

Sometimes, the test also includes relatively balanced negative information.  Maybe the pollster will tell you about Phineas’ 1992 DUI arrest, balanced with the fact that Thaddeus has run for County Assemblyman three times and lost.  (If you get all negative on one candidate, then that’s a push poll, a tactic everyone deplores, but many candidates still employ.)  

The bottom line on “informed vote” polling is that it is subjective by its very nature.  At this early date, we’re betting the 23% figure is an informed vote percentage.

The final thing we want to say is that there are two types of polls.  You’ve got the kind that are designed to shape campaign strategy, and the kind that are designed to make the candidate look good.  Both types are ethical, both are legitimate uses of survey data, but if you’re reading about a candidate’s poll in the newspaper, know that that poll (or at least part of that poll) was probably designed for the press hit.  

In fact, here’s the biggest thing of all.  The horserace numbers (who’s ahead, who’s behind, and by how much) are about the least useful poll numbers a campaign can have.  The polling data that helps candidates is the stuff you find in a benchmark poll – where the public is on certain issues, and how to frame your positions on those issues in such a way that your message resonates.  To a smart campaign, who’s ahead doesn’t matter one bit until about a week or two before election day, which is when a smart and well-funded campaign runs rolling tracking polls to adjust their media mix.  

Rhetorical question time, then we sum up.  Does anyone out there think that it is likely that Jim Martin is going to beat Greg Hecht on election day by a 75%-25% margin?  If so, we’ve got a vacant lot on Cedar Shoals Dr. we’d like to sell you.  It’s not likely, yet that’s what Jim Martin is wanting to imply to you.  So why give his poll any credence whatsoever?  That’s what we thought.

To sum up, here are some hard and fast rules on interpreting polls.
  • Smart candidates don’t put the useful numbers out for public consumption.  Any poll that isn’t internal is just being used for the press hit.  Smart reporters know this and report on the polls in that way.

  • Horserace polling doesn’t matter until less than two weeks out anyway, and no campaign in their right mind would release their rolling tracking numbers.

  • Here are a few things that an informed political consumer should look out for in any public poll: Margin of Error (not included in the email we got from the Martin campaign); Sample Size; Crosstabs (demographics on which likely voters go which way – also not included in the email from Martin’s campaign); the actual polling questions used (many campaigns keep these internal because insidious media outlets like us would seize on any perceived inequity in the informed vote descriptors – Martin did not release his); most importantly, who did the poll.  (Martin’s firm, Cooper & Secrest Associates, has a good reputation.  But be aware that online polls have no control over who responds, thus the sample is not representative, also be wary of anything by Survey USA or other companies that do automated phone polling, for similar reasons.)

Too much inside baseball?  Probably so, but we’d like to do our part to make everyone out there a more informed political consumer, and while Martin’s polling antics aren’t necessarily bad, it is kind of a cheap move for a press hit.


Anonymous said...

good job on this one! Hecht's campaign has already released a similar statement (although much less geeky) that says something like there's an independent poll (AJC? can't remember) that shows him ahead by a 3-to-1 margin over Martin.

I don't trust Cooper Secrest at all! I've been fortunate enough to have had only second-hand dealings with these guys but, the generaly "inside baseball" on Cooper Secrest is that they will push campaigns to pay them for these feel-good, just for the press releases polls and make sure that they keep their client thinking that their campaign has a shot at winning because, of course, if a candidate drops out, they lose a client.

Pollsters are a necessary evil but you can NEVER leave them in charge of anything but making the phone calls. They can't be trusted without serious supervision.

You are dead-on right that legitimate polls, paid for by the campaign, are never "press released" - this is some of the most carefully guarded and expensive info you can get! You don't publish it in the friggin' newspaper!

Dawg Corleone said...

Pet Peeve:

3 to 1 is not a margin. It's a ratio. Sorry to be anal, but I wince every time I heard that mistake on the radio or TV. That means I wince a lot.

Good, informative post, though. I'm always fascinated by polling: how it works and how it falls short. It's an amazing, albeit inexact, science.