Interesting Polling Trends
Polls, polls, polls everywhere. Some interesting footnotes in recent polls.
1) Sample size
The number of people polled, especially as most people move to cellphones and are harder to reach, increases the costs of a poll. But they also increase the accuracy. That matters.
Here is an interesting trend I have noticed: the larger the sample size, the bigger the Romney lead. Another way to say that, the more accurate your poll is, the bigger the margin for Romney.
Gallup uses 2700 with a 2% margin of error. Last poll - Romney up +5. Rasmussen uses 1500 and Pew uses 1495 with error margin 3.0% (50% higher margin of error). Rasmussen has a Romney lead of 2 and Pew has it tied.
There are a cluster of polls using 1000 - 1300, with margins of error greater than three: 1 Romney lead, 2 ties, and 2 with Obama leads of 1%.
Only two polls are fairly inaccurate (CBS/NY Times and National Journal which use 713 and 563 to poll the entire nation). Both have Obama in the lead.
If this holds over the next few days and Romney wins, just like 2008 when 17 of the 23 polls overstated Obama's vote, there needs to be some thorough analysis by most polling firms and media organizations as to whether they are deliberating over-stating the Democrat vote.
2) Libertarian vote in the Indiana Senate election
In Indiana the core Libertarian vote in any race is around 2-3%. They then add disaffected voters from one side or the other to range from 4-7% in most races. In my congressional races the Libertarians, when there was one, went from a few percentage points to 5% after I voted for TARP in 2008. It was clear from our polling that had those Libertarians been forced to choose they probably would have gone 3-3.5% for me, 1.5-2% for Montagano. It never goes strongly one way or the other.
I have been maintaining that the Indiana Senate race is more unusual than most because the high Libertarian vote is inflated by Republicans and independents who do not like Mourdock but don't want to vote for Donnelly, thus being responsible for giving liberals control of, and even another seat, in the Senate. In makes no sense to vote for Romney and a GOP congressional candidate but then elect a Senator (for 6 years!) who supports Obama. But many of these voters are upset, not just by Richard Mourdock's alleged views on religion but other comments he's made ranging from "inflicting his views on others" to questioning the constitutionality of social security or the department of education.
An internal poll by Joe Donnelly, released to give momentum to his campaign, is the first time I have seen evidence that my point is correct. Without the Libertarian candidate included, Donnelly scores 2% higher than if you include the Libertarian candidate. The more votes the Libertarian gets, in other words, the better for Mourdock because these voters are not going to vote for him anyway. Voters who don't like and will not vote for Mourdock, and there are a few, can either: not vote, skip voting for Senate, vote Libertarian or vote for Donnelly. Suppporters of Mourdock should hope that they choose any of the others rather than voting for Donnelly. I expect a few undecided but loyal Republicans to swallow hard and vote for Mourdock.
The way this breaks down may decide the fate of the Indiana race and the United States Senate control.