On polling

As I’ve noted before, political polling is really hard. When you’re trying to predict who will win an election there are so many things that you have to worry about that it’s near impossible to adjust poll numbers. That seems to elude some people:

Basically, what UnskewedPolls does is take a poll and look at the  partisan percentage split that the poll reports from the random sample.  This could be, for instance, a 33 percent Democrat, 28 percent Republican, and  39 percent independent breakdown.

Then, they somehow weight the data so that what comes out is what the sample would look like if the  Democrat/Republican/Independent percentages were equal to the  results found by Rasmussen Reports in the single month of August— 35.4  percent Republican, 34 percent Democrat, and 30.5 percent Independent.

This is wrong for a few reasons, such as:

Even the founder of Rasmussen Reports, whose surveys often show higher Republican numbers, cast doubt on Chambers’ methods: Scott Rasmussen told BuzzFeed in an e-mail that “you cannot compare partisan weighting from one polling firm to another.”

“Different firms ask about partisan affiliation in different ways,” explained Rasmussen. “Some ask how you are registered. Some ask what you consider yourselves. Some push for leaners, others do not. Some ask it at the beginning of a survey which provides a more stable response while others ask it at the end.”

So a Republican for Rasmussen might not be a Republican under another poll.

Another problem is that Rasmussen is an outlier as the Business Insider article shows, so how do we know that that’s the right one? Well, we don’t–that’s why we should trust the average of lots of polls more than any one poll.

Finally, the implication is that most pollsters are skewing things towards Obama which doesn’t make sense. Every time a poll is wrong by a significant amount their reputation goes down (and Unskewed is saying they’re off by more than 10%). Their business rests on them being accurate, do you really think they would risk that to show Obama is ahead in their polls (and this would be almost all pollsters)? I guess you might if you’re a Republican.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: