If the election were held today, which candidate would I be wrong about you voting for?
Here is a ranking of the accuracy of a couple of hundred of the most common political polling organizations. This list might be helpful to you if you see different tracking polls which ask the same questions but have different results.
This list was compiled by using data from an article at Pollster.com, but the information was created in a roundabout way.
The data was originally collected and assembled by Nate Silver at his polling blog at FiveThirtyEight.com. I didn’t like the criteria that Silver used when he rated pollsters and fortunately Pollster.com adjusted that data in a way that I prefer. They removed Silver’s transparency criteria and just rated the pollsters on how well they have historically performed.
I simplified the information that pollster.com created and just turned it into an easy to read pollster ranking. If you want to see the source data that was created at pollster.com then here is an excel spreadsheet.
This list should only be used in a general way– see my qualifications at the bottom of this post.
Here is the list:
Pollster Rank
Qualifications:
Statisticians could find fault with many of the assumptions of the ranking system used here, so I would suggest that you use the ranking pollsters in a simple way. In other words; you shouldn’t assume that the 5th ranked pollster is better than the 6th ranked one. However, you can probably assume that the 5th ranked pollster is better than the 106th ranked one.
Note that most polls are conducted to a 95% confidence level. This means that even a perfect pollster will be completely wrong once every twenty times. And none of these pollsters are perfect.
Also, it would be reasonable to assume that those involved in creating and analyzing this data would consider the table listed above too simplistic. This data rank was initially created by Nate Silver at FiveThirtyEight.com and he factored in transparency criteria, so his ranking of pollsters is different than the above ranking.
Mark Blumenthal at pollster.com reinterpreted Silver’s ranking into a more sophisticated version of what is shown above. But he did it to show that there isn’t a great deal of difference between pollsters and not to rank them like I did.
I don’t think that these qualifications matter if you don’t over-construe the rankings. However; as an example, if Survey USA (top rated) has one set of numbers and the Humphrey Institute (#258th best) comes up with different numbers on the same race, then you would be smart to use the Survey USA numbers.
——————-
How the pros read the polls:
If you want to read a good article about how the political pros read the polls then go to OhmyGov.com. Here is a quick summary of the article.
1. Ignore non-scientific polls.
- They aren’t reliable and may be conducted by an organization that has a conflict of interest.
2. Timing is everything.
- Pay close attention to the timing of polls since early polling is not as indicative as polling closer to the election.
3. Know the poll’s margin of error.
The margin of error might show that a certain poll is unreliable, subject to sampling error, too close to call, or is being misinterpreted.
4. Law of averages.
- Several polls are better than a single poll since the effects of bias and polling errors will be decreased
5. The quality of the polling organization.
Members of the general public not know whether the pollsters are high quality when reading poll results, but the professionals do.
—
Check out this article for a straight-forward way to read political polls:
http://tooldtowork.com/2010/09/a-few-tips-on-reading-political-polls/
or this article which attempts to explain the widely divergent poll results:
http://crosscut.com/blog/crosscut/19796/Explaining-big-gaps-in-this-year-s-election-polls/
Related posts:
Great site. A lot of useful information here. I’m sending it to some friends!