In Part One we saw there are three kinds of polls:

Public Polls
Internal Polls
Persuasive Polls

It's not always very simple to determine which category a poll falls into or how to distinguish junk polls from quality polling.

Let's take a look at this poll that was commissioned by The Daily Caller. First we must evaluate the source of the poll: it was commissioned by The Daily Caller, a center-right online publication, which suggests that this poll will likely slant Republican. Not all partisan polling is bad but it's best to assume a slant in partisan polling.

Secondly, we should take a close look to see what the data says. If you look on page 11 of the poll, you'll notice it shows 47% of African Americans supporting Republicans Richardson and Wehby. While a shift of African Americans towards Republicans isn't improbable, a voting bloc that normally goes 80-20 or better for Democrats suddenly splitting 50-50 is out of the ordinary. Additionally, consulting other polls of the same race showed African Americans supporting Democrats at traditional margins.

Thus we can come to two conclusions: 1) This poll is probably intended to be slanted towards Republicans based on the source and the dramatic headline that accompanied it. 2) This is a junk poll that is not only biased but just totally useless for much of anything. I would categorize this Daily Caller poll as a persuasive poll.

In part one I mentioned that you will never see a true internal poll in the media. This is a true statement. However, I happen to have gotten my hands on an internal tracking poll that has been anonymized. You can read the topline survey and the full crosstabs. These will give you a good idea of what a real internal campaign poll looks like. I'm not an expert poll interpreter by any means but take a look at the topline survey. It gives you a good idea of what the numbers mean.

There are two kinds of internal polls. The first is a baseline poll. These are longer and more costly polls that test a lot of issues and aspects of a candidate and their opponent. These are done usually once (maybe twice) per campaign.

The second kind of poll is a tracking survey like the one you can read above. These are done more frequently, are shorter, and less costly than a baseline poll. Tracking surveys are useful to test where the candidates are at any point in a race. For instance, a campaign might do a big mailer to teachers and then do a tracking survey to see how that helped them both with teachers and overall.

Polling is complicated. I am not always the best at explaining complicated topics. If you have any questions feel free to get in touch. I may publish a follow up if there's enough questions.