How the Presidential Polls are Skewed

One week from today is the first presidential debate. The first debate will be in Denver at 9:00-10:30 Eastern Time, and will be aired on all the major networks. Both sides are trying to downplay expectations for their candidate, while trying to elevate expectations for the opponent. Nothing new there.

Today we will look at the national polls and some key battleground states. But first, you may remember that a couple of weeks ago we predicted that September would see a lot of new polls spring up. Well that seems to have been an understatement! We have seen more polls this month than at any other point in general election cycle history. There have already been 51 polls this month and counting! This will be the most heavily polled presidential race in history.

Not surprisingly, the poll results are still all over the place, but there seems to be a general skew toward Obama. We use the word “skew” here on purpose because we need to take a hard look at what skewed poll results are and how they get that way. I think you will find this very interesting.

So what is a skewed poll anyway? A skewed poll is one that “oversamples” either Democrats or Republicans, under-rates Independents and/or uses inflated party identification methods that exacerbate the oversample. That’s a lot to think about, but it’s not as complicated as you might think.

This naturally begs the question: How many polls are skewed? The answer is, far more than you would think! All of the polls include their methodologies and assumptions, if you’re willing to drill down into the data. Most voters don’t have time (or the interest) to do that. However, there is a great website – www.unskewedpolls.com – that does it for us, every day.

Spencer and I have to admit that we are fascinated by this website. To date, they have examined most of the major polls, and after unskewing the data, they come away with very different results. So how does one unskew a poll? Well when each new national or state poll is released, it is not simply the net results of the poll, say Obama 49/Romney 45, but the “internals” as well…the nuts and bolts of the poll.

When these internal metrics are examined, oversamples can readily be determined. The folks at unskewedpolls.com then apply the Rasmussen national party affiliation data. They use the Rasmussen model for how many Democrats vs. Republicans there are in the electorate because the Rasmussen data for party identification is widely considered to be the most reliable.

If you go to www.unskewedpolls.com (and we highly recommend you do), you may be surprised to find that almost all of the major polls are skewed in favor of President Obama. Yet the media never disclose this!

Here is a recent example of a skewed poll. This Reuters/Ipsos poll had Obama leading Romney 48 to 43 with a 2.9% margin of error. But if you “unskew” the poll, Romney wins by a stunning 10-point margin, 54 to 44. How can this be, you ask. Because this poll oversampled Democrats by 19% and undersampled Republicans by 20%! And it’s really even worse when you consider that more Americans identify themselves as Republicans than Democrats (see Rasmussen data above on this).

This result is not at all unique. You will see when you visit the unskewedpolls website that every unskewed poll favors Team Romney. Now, we admit that this seems too good to be true. However, the folks at unskewedpolls.com go into a lot of detail on their methods and data, and we could not find fault with it.

For some fascinating historical perspective, here is a long but worthwhile piece on the polling in the 1980 election and how it was glaringly skewed toward Jimmy Carter.

With that in mind, let’s jump over to the most recent round of polls. The RealClearPolitics.com national average has Obama up by 3.6%. Not great for Romney but certainly not insurmountable.  Now let’s look at the key battleground states of OH, VA, NC, FL and CO. The polls in these states are still all over the board, so let’s just stick with the averages.

In OH and VA, Obama has a lead of apprx. four percent. In CO and FL, Obama has a lead of apprx. two percent. And Obama is gaining ground in NC where he now leads by a thin one percent. Are there some skewed polls figuring into these averages? Yes! But we must admit that Romney is struggling.

That’s all for now. Next week we will focus on the first debate and what to expect.