Visions compete to forecast gubernatorial election

It now appears that at least one pollster of the 2015 Louisiana gubernatorial contest figures a different electorate than recent trends suggest on which other pollsters base their samples. If he is correct, the contest’s dynamics differ from what commonly is believed.

Market Research Insight has polled monthly on behalf of a small group of subscribers. A portion of the proprietary information gets leaked from time to time and made some news last month when it gauged a neck-and-neck race between Sen. David Vitter and Public Service Commissioner Scott Angelle. This was contrary to every other poll from several other outfits that consistently have Vitter and state Rep. John Bel Edwards, the only Democrat in the race, leading the pack considerably over Angelle and the other Republican, Lt. Gov. Jay Dardenne. The MRI poll basically replicates other polls’ results on the placements of Edwards and Dardenne when assigning 90 percent of the total black vote to Edwards and the remainder proportionally to the others.

That July poll truly resided as an outlier as polls immediately before and after it showed clear Edwards/Vitter and Angelle/Dardenne tiers. It was speculated previously that the differences could come only from two sources, one being that particular MRI poll suffered from an “unhappily randomized” sample. Simply, it could have been an instance, with most pollsters choosing to risk this degree of inaccuracy, where the five percent chance of drawing an unrepresentative sample of the population actually did occur.

But the August results ended up fairly close to those of July for MRI and again deviant from all others. This means the unhappy randomization scenario is highly unlikely, as the odds of that occurring are 400 to 1. Instead, it points to the other possible explanatory reason, the actual sampling frame used.

As previously noted in this space, MRI’s July poll sample disproportionately picked up people registered as Democrats and those who registered under that label, probably decades ago, but who do not vote consistently Democrat, if even a majority of the time, and thereby perceive themselves as independent. It overweighed Democrats by six percent among current registrants and underweighed by a couple of points no party (which includes people who call themselves “independent” or another party label) registrants. This creates a sample on face characteristics more likely to vote for Angelle, at the expense of Vitter.

While the internal numbers of the August poll are not yet in the public domain, pollsters rarely make more than minute shifts in their sampling frames during the course of a campaign, so it’s probably pretty much the same as in July. If so, this means MRI is making a radically different bet on the composition of the electorate in October, to the favor of Angelle and disfavor of Vitter.

And if this is the case, then it’s a controversial decision. Recent history suggests that fewer Democrats and slightly more no party adherents and Republicans will turn out than the MRI model assumes. That doesn’t necessarily mean it’s wrong – we won’t know that answer until shortly after election day upon releasing of turnout statistics – but it definitely goes against the grain.

In the meantime, this information may make a difference in the campaign dynamics. Candidates and groups supporting or opposing them plot strategy relative to where they think their preferred candidate is in the field, such as if they engage in negative advertising, how much of it, and against whom. It likely will not affect much, if any, the actual candidates’ campaigns as undoubtedly they all do their own internal polling and base their decisions on those. But organizations unaffiliated with them may not have the resources to do this, and so may take cues from polling information in the public domain.

So this divergence of opinion adds a little to reporting of a contest that, to date, has lacked drama. Depending upon what the final polls of the respective organizations turn up, somebody’s going to be right and somebody’s going to be wrong in their educated guesses about the electorate’s composition in the general election.