A Word about Polls

Or, why do I retweet Rasmussen so much…

Polls are the lifeblood of democratic (small d) politics. It’s how we keep score.

Interim drama.Short for who is winning?

But different polls often give conflicting results.

If you check out RealClearPolitics.com today, you will find a poll to support your fondest hopes for the November election. Unfortunately, you may also find a poll to reinforce your worst fears.

Here is a timely example:

Gallup June 13 tracking poll Obama 46% Romney 45%.

But wait! There’s a Rasmussen poll, same date, that has Romney up by 4 points, 48-44%.

Which one is right?

I seldom admit this, but I don’t actually know. No one does.
But here is what I think, and briefly, why I tend to follow Rasmussen

LIKELY  VS REGISTERED VS ADULTS

First look at the sample.

Gallup polls registered voters. This means, for example, that if you are eighteen you go to get a driver’s license, you check the voter registration box, you show up on the voter registration rolls and you might get called.
But will you actually vote?

On the other hand, Rasmussen polls “likely voters.” Generally, these are registered voters who voted last time. In the last few elections, Rasmussen has been pretty accurate, and this is a big part of his secret sauce. It is more expensive and complicated to poll this more select group.

Many, perhaps most,  of the surveys you see poll registered voters. Some poll “adults” regardless of whether the adult can actually vote.

So what does it mean? Gallup polls most likely include a lot of folks who won’t bother to vote. So they may like Obama better than Romney, but not enough to leave work early, or figure out how to get an absentee ballot or whatever it takes.

TIMING

I also notice that the Gallup poll was done over  seven days. Rasmussen –three days.

So as events/gaffes take place, Rasmussen will be more likely to reflect them. “The private sector is fine” took place on Friday. The impact would have been felt over the next three days. Did this contribute to Romney’s bump? Possibly. On the other hand, Gallup may feel that using a rolling average shows the trend and smoothes out the little blips and gaffes. But do we really care about week old voter intentions?

If you were alive in 1980, you may remember the seismic change in polls in the last few days of the election. Alas, I am not old enough to remember “Dewey Defeats Truman” but something similar happened.

DOES SIZE MATTER?

Perhaps, but not that much when we are talking about samples. Unless the sample is especially small, all the correct statistical analysis would have taken place to calculate the margin of error. So unless the sample size is a few hundred, I don’t worry about it.

SO YOU ARE SAYING, IGNORE ALL POLLS EXCEPT RASMUSSEN?

No. All polls are interesting. There are many other factors to consider also, too deep in the weeds for this blog.

But read the fine print. Some may be misleading, outdated or wrong.

As for NovelPolitics polls, they are the most select, of course. They represent the views of a relatively small sample of highly informed, motivated and intelligent electorate.

So if you haven’t already, please check them out and vote.

1 Comment

Filed under 2012 Election, White House

One response to “A Word about Polls

  1. Bill Sleptzoff

    All reading here should read an old classic. “How to Lie with Statistics”. When I used to be a Mathematics teacher I tried to have a student do a book report on it. It promoted critical thinking and made the students realize that given almost any data could be shaped to mean more of what a person wanted depending on the presentation. We need to teach people how to think again.

What do you think about this subject?

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