What did we learn from this election?
- Internal campaign polls that make their way into the public domain are highly suspect
- Independent - or unaffiliated - voters don’t vote in non-presidential years.
Actually, we already knew both those things. Unfortunately, this sometimes gets forgotten in the quest to report something “sexy.”
So, when next year’s legislative races roll around, here are two rules for New Jersey media to live by.
1. Don’t report internal polls. Or to be more accurate, don’t report numbers from “interested party” memos claiming to be the results of internal polls.
In this past cycle alone, we had one “poll” from Greenberg Quinlan Rosner for the Adler campaign showing a so-called Tea Party candidate with 12% of the vote (he got less than 2%). And another instance where the governor announced that his pollster, National Research, had Anna Little in a dead heat with Frank Pallone (Pallone won by 11 points and the gap was never smaller than 7 points in the Monmouth University Poll).
Campaigns only release internal polls for one reason – to drive the media narrative. Reporting them is akin to giving in-kind campaign assistance.
Moreover, internal campaign polls “released” to the public consistently show a bias in favor of their candidate. Scot Reader analyzed 136 internal polls released publicly and found that 70% of those polls showed the pollsters’ client outperforming the actual results.
That appears to extend to partisan pollsters’ on-the-record public polling as well. For example, the firm headed by Mark Penn (the Clintons’ pollster) conducted polls for The Hill website in 20 House races during the last two weeks of the campaign. Of those, they overstated the Democratic candidate’s performance in 16 races – including 5 cases where they miscalled the eventual winner. They got the victory margin right in 3 races and overstated the Republican candidate’s performance in just one instance. The average partisan bias in those 20 polls was 6.4% Democrat.
For the record, Monmouth University issued 7 House race polls in the closing two weeks with no overall partisan bias - 0.1% Republican, to be exact (3 of our polls understated the Democrat’s performance and 4 understated the Republican’s edge).
Furthermore, internal campaign poll memos may claim to represent accurate poll results, but give absolutely none of the information necessary to judge whether the poll is valid. In a huge bit of irony, a number of campaign pollsters recently issued an open letter decrying the media for reporting independent polls that “contain inadequate information on how they were conducted.”
Putting aside the astounding hypocrisy, they have a point. The media should be equally critical of independent polls. A legitimate polling organization should be willing to reveal the full question wording, description of likely voters, and basic information on the demographic composition of the sample for every election poll it releases.
Of course, that may open a poll to criticism on whether its sample’s partisan composition is accurate. And that’s certainly a debatable point. But I find that such critics routinely overestimate how many independents should be in a sample of likely voters (especially if their preferred candidate is winning the independent vote). Which leads us to rule #2.
2. Stop reporting the large number of unaffiliated registered voters in New Jersey as if it means something.
All too often, media reports describing the electorate include a statement to the effect that the “largest number of voters, though, are not affiliated with either party.”
So what? These people don’t vote unless it’s a presidential year! It would be much more accurate to say that the vast majority - typically about 75% - of voters who will show up in any given off-year election are registered as either Democrats or Republicans. That’s why elections tend to be about turnout more than about winning over undecided voters. Only in very competitive races do truly unaffiliated voters make a difference.
It’s a poorly managed campaign that focuses on all unaffiliated registered voters. So the media shouldn’t either?
That’s just my two cents.