Wednesday, September 26, 2012

ALL the Polls are Biased

... The first in our many-part series, "Things Troubled Campaigns Say."

You may have seen this article in one form or another over the last 24 hours.

Republicans, conservatives, and Romney campaign officials say that public polls showing Obama well-ahead in Ohio and nationally are biased. Specifically, they argue that these polls are over-sampling Democratic voters.

Let's start by taking a look at the state of the public polling in Ohio:



Pollster.com lists 14 public polls in Ohio in the month of September. Just one of those polls shows Romney ahead in Ohio. That poll was done by Gravis Marketing and it is the oldest poll in the group. And they've re-polled the state twice since then and Obama has been ahead (slightly) in each of those polls. The two most recent polls of the state, one released today and one yesterday, have Obama ahead by 10 and 8 points respectively.

But actually, the news is quite a bit worse than that for Romney in Ohio. If you go back much further, Pollster.com lists 50 public polls in Ohio since October 2011, a full year. The polling is remarkably stable. Romney leads in 6 of these polls and the two are tied in 2 of them. So, the President has leads ranging from 1 to 12 points in 42 polls of the state in the past year including a string of 13 straight polls in September 2012.

So, the only way the Romney campaign could NOT be behind is if these polls are systematically biased against Romney in some way. The specific charge is that the samples of the polls include too many Democrats. Technically, that's possible. Realistically, it is not happening. Pollsters (remotely competent ones that is) don't weight by party ID but they do sometimes weight by demographic groups (like African-Americans, for instance) whose voting patterns are highly correlated with one party or another. If one over-weights a demographic group that supports the Democrats, you would effectively be over-sampling Democrats. But, for the most part, that's not happening. When you look at the polls that provide sub-sample breakdowns, you simply don't see groups like African-Americans being oversampled systematically.

Now, it is true that we have seen a recent rise in the number of people identifying as Democrats in recent polls. But that's consistent with separate polling indicating a rise in voter enthusiasm among Democrats. As Democrats become more enthused, more are likely to vote and to get through the likely voter screens in these polls.

The tell-tale sign the Romney campaign knows they're behind can be found in the comments of Rich Beeson, Romney's National Political Director. In criticizing the public polls, Beeson said, "We are, by any stretch, within the margin of error in Ohio." The Washington Post Poll released yesterday had the President ahead 52-44 in Ohio. The poll surveyed 759 likely voters and has a margin of error of +/- 4.5 points. This means that Obama's number AND Romney's number could be 4.5 points higher or lower within the 95% confidence interval. Romney could be ahead 48.5 - 47.5. And he could be ahead by even more if this poll is one of the forty likely to fall outside the margin of error on that side of the curve! Of course, it is just as likely that Obama is ahead 56.5 - 39.5 and there's a 2.5% chance his lead is even greater.

The reason we know these outlier interpretations are exceedingly unlikely is that so many other polls are showing similar kinds of numbers. We don't see a lot of polls showing an 8-point Obama lead as the Washington Post poll does. But, as I said, we aren't seeing ANY polls right now showing Romney even or ahead. Romney IS behind in Ohio. The Romney campaign can quibble with the size of the deficit but arguing Obama's lead in Ohio isn't real ... is just silly.

2 comments:

Blogger said...

SwagBucks is a very popular get-paid-to site.

Blogger said...

Get daily suggestions and methods for making THOUSANDS OF DOLLARS per day FROM HOME for FREE.
JOIN TODAY