First, the bad news. The latest Ohio Poll came out today and suggests that Kasich is sitting on an eight-point lead heading into the election. It claims that Strickland has only 42% of the support, even though the most recently polling has had him at 45% and above.
But, this poll sampled 17% less people than the one in September. The smaller the sample size; the more likely an error. But that’s not the only thing to suggest that the Ohio Poll, which is usually the most reliable poll, is flawed. Remember, any poll of likely voters is only as reliable as it’s turnout model predictions come to fruition.
The partisan breakdown of each candidates’ support in this poll is this:
- Strickland gets 86% of the Democratic vote. Kasich gets 10%.
- Among independents, they’re split with Strickland with an insignificant one-point lead. * (Please note the asterisk, I’ll get to it in a moment.)
- Kasich gets 86% of the Republican vote; Strickland gets 9%.
The poll shows that both candidates are holding their bases the same, and neither has any edge on cross-over support. The candidates also split among independents. So, how does Kasich get an eight-point lead?!?
The only way that’s possible is if the Ohio Poll’s likely voter model predicts that more Republicans vote than both Independents and Democrats in a State where Democrats have nearly a 1 million voter registration advantage and Independents out number Democrats and Republicans. In other words, when you break down the electorate by party (Independent, Democratic, and Republican), this poll is predicting that the smallest registered voter demographic in the group will be larger in voting than the other two votes combined. As we noted yesterday, the early vote data so far suggests the opposite is true. That Democrats are turning out more votes than Republicans. So, not only is the Ohio Poll’s likely voter model staticially improbable–to the point of approaching impossible–it would appear it’s factually impossible already.
* The poll does something with Independents I don’t recall ever seeing in an Ohio Poll, either. The poll admits that of its sample size of 663, only 75 or less were “registered” Independents. That means that it’s sample size in its likely voter model produced a result in which independents make up only 11% (or less) of the sample.
In 2008, exit polling in Ohio indicated that Independents made up of 30% of the electorate in Ohio, essentially tying GOP turnout, which both came behind Democratic turnout which was 39%. So for the Ohio Poll’s likely voter turnout model to be believed you’ve got to believe that despite being a statistical minority in the electorate, the actual turnout will have more Republicans than Democrats and Independents combined as Independents stay home is such large droves that they make up less than a third of total electorate they’ve made in the past. That’s so improbable that it approaches impossible, too.
Regardless, we already know that such turnout is unlikely. Already, we’ve seen that Democrats are turning out more votes than Republicans in early voting. We have no reason to believe that is unique to only early voting. The idea that on Election Night we’ll see the Republicans vote while the Independents and Democrats stay home and not vote at all is about what is required to get the results the Ohio Poll reached. We already know that’s not happening.
In short, what we can discern of the cross-tab data, this poll is based on a likely voter model that confounds the demographics of the Ohio registered voter electorate and would require a far more lopsided GOP turnout than we’ve ever seen, and 180-degrees from what we’ve seen so far in early voting. To see Independents go from being a third of the electorate to potentially less than a tenth? I don’t think that’s ever happened in Ohio in a midterm election.
A rare miss for the Ohio Poll.
[UPDATE:] If you need any more proof that this poll is not accurate, consider this. We’ve learned from reliable sources who spoke to us on condition of anonymity that Kasich held a meeting with major donors/political players yesterday in which a representative of his campaign pollster, Tarrance Group, indicated that their latest polling showed Kasich with a five-point lead. (Ed Goas of the Tarrance Group contacted us via Twitter and insisted that his poll showed a five-point lead, not two as someone else with knowledge of the meeting indicated. I admit I misunderstood my source into thinking it was a conference call, it wasn’t. I haven’t seen the actual poll, but I doubt Mr. Goas would misrepresent his own data.) A review of Kasich’s campaign finance reports verified that Tarrance Group is, in fact, Kasich’s campaign pollster.