Guest AuthorThird in a series about polling.
Every cycle some pollsters seem to get it right. Some get it wrong.
Why?
To answer that question, it’s vital to understand that predicting what the electorate will look like matters — it really matters. And perhaps the single most important variable that pollsters must properly balance, especially in a general election contest like the governor’s race, is party affiliation. How many Democrats will be voting? How many Republicans? How many “NPA/other” voters will be voting?
In a race where the vast majority of partisans are likely voting with their guy (most Democrats voting for Charlie Crist and most Republicans voting for Gov. Rick Scott), getting that balance right is essential to make sure your poll is as accurate (valid) as possible. If you oversample Republicans, for example, your poll will artificially favor Scott. The opposite, of course, is also true.
Because every pollster has to decide — before a single call is made — what the electorate will look like, we must decide what the percentage of Democrats, Republicans and “others” will be in our final sample.
And this is where error can creep in.
Pollsters (at least the ones I know) don’t have time-travel devices and we simply cannot KNOW what the electorate will look like nor can we see the future.
But we can see the past.
As pollsters, we must build models of what we think the electorate will look like and base that on recent elections. This is a simplified turnout model by party and you also need to balance your poll by other things like gender, geography, age and phone type.
Here’s how it works for party…
- In 2012, Democrats made up 40 percent of the electorate, Republicans 39 percent, and “other” was 21 percent. We call this “Dem +1,” meaning that Democrats, as a percentage of those who actually voted, were 1 point higher than Republicans.
- In 2010, we saw something entirely different. In that year, Democrats made up about 39 percent of the electorate while Republicans came in around 46 percent — for consistency sake, let’s call that “Dem -7.”
That’s an 8-point swing from 2010 to 2012.
- If we go back to 2008 and 2006, we see a very similar pattern: “Dem +1” in the presidential election, and “Dem -5” in the gubernatorial election.
That’s a 6-point swing from 2006 to 2008.
In short, it appears that in 2014 — an “off-year” election — it is fair to expect that the turnout ratio will be similar. Therefore, it is fair to conclude that there will be more Republicans voting (probably by 4-8 points) than Democrats. And while you also need to consider current voter registration trends and the ever-growing tide of “NPA/other” voters (those who are not of either major party and are, by the way, growing steadily every year) using this method will help you not just balance your own polls correctly, but help non-pollsters be better consumers of polls. Of course, this only applies to statewide elections. You must repeat this exercise for each district or county or whatever geography you are polling.
As such, here is what we suggest you look at when reading a statewide poll:
- Be sure the pollster spoke only with likely voters, not just registered voters.
- For 2014, if the poll has more Democrats than Republicans, we strongly suggest backing that down a bit. Based on the past four elections, there is almost no scenario where Democrats will outnumber Republicans in an off-year election. After all, in the last two presidential elections, with Barack Obama himself on the ballot, Democrats beat the GOP in turnout by only a single point.
- Based on past gubernatorial turnout models, GOP turnout is likely to be at least 4 points higher than Democratic turnout.
- “Other” voters (NPA plus other parties) should come in around 19 percent to 22 percent of the electorate (and thus the poll) as they were 16 percent in the last governor’s race and are growing by a about a point a year.
Does this turnout skew mean Rick Scott will win?
Not at all. Recall that, even with such a huge deficit in turnout in 2010, Democrat Alex Sink came very close to beating Scott. There are so many other factors at play. But this formula is simply a way to make sure the poll you are reading is properly balanced with the best tools we have at our disposal.
That is, until we get a Tardis*…
* A Tardis is a time-travel machine and an uber-geeky, sci-fi shout out to “Doctor Who” fans.
Steve Vancore is the President of VancoreJones Communications and Clearview Research and has been conducting polling and voter research in Florida since the mid-1980’s. He can be reached at [email protected]. Column courtesy of Context Florida.
