[ Last updated: Noon ET on 02/16/08 ]

39 polling firms have polled one or more of the 41 Presidential Primaries and Caucuses to date in 2008. Here’s how the pollsters stack-up, ranked from smallest error to largest error (where smaller error is better, and zero is perfect).

First, sorted by median error:

Supporting detail here:

http://www.surveyusa.com/Scorecards/All_2008_Primary_Report_Card_Thru_021308_median_error.pdf

Next, sorted by mean (average) error:

Supporting detail here:

http://www.surveyusa.com/Scorecards/All_2008_Primary_Report_Card_Thru_021308_mean_error.pdf

All information drawn from http://www.realclearpolitics.com/ , from http://www.pollster.com/ and, if disagreement on those two sites, from individual polling firm’s website. Error is calculated using the “Mosteller 5″ Measure, which works this way: If “Pollster A” says Smith beats Jones by 8 points, and in fact Smith beats Jones by 5 points, then Pollster A has an “error” of 3 [8 minus 5]. A 3 would be recorded in the table above for that contest. There are limitations to this and all other measures of pollster accuracy. For a complete discussion of the different ways that election poll accuracy can be measured, including a summary of all 8 known measures, see this paper presented at the 2006 conference of the American Association for Public Opinion Research. All pollsters and timely polls of likely voters included in this table; no timely polls of likely voters intentionally excluded.  Report errors or omissions to editor@surveyusa.com .   Additional pollster Report Cards, from 2008 and from earlier elections, using this and other error criteria, are prominently displayed elsewhere on SurveyUSA’s website, www.surveyusa.com. 

As of noon ET 02/18/08, 100% of votes have still not been counted in all recent contests. SurveyUSA will continue to monitor, and will update all Report Cards if late returns affect any error calculations.

A separate SurveyUSA post, elaborating on and providing additional context for this post, is newly available here.