About Those Democratic Polls: Potential sources of error [View All]

[DeepModem Mom]

WP: The Pollster
About Those Democratic Polls
By Jon Cohen and Jennifer Agiesta

Yesterday's Democratic result is sure to fuel debate among poll-watchers about the accuracy of polls in contests with African American candidates. In several well-known, but long-ago examples, pre-election polls underestimated support for the white candidates in such campaigns. But in the 2006 elections, a strong showing by polls in elections with black candidates seemed to finally put that notion to rest, and there was no apparent problem with reliable pre-election polls in Iowa.

A more likely culprit than the role of race in the New Hampshire election was the "likely voter" modeling, with pollsters perhaps over-counting the boost of enthusiasm among Obama supporters following his victory in Iowa. Another possibility is that independents opted at the last minute to participate in the Republican primary, depriving Obama of crucial voters.

A further potential source of error stems from New Hampshire ballot rules. In previous contests, the state rotated candidate names from precinct to precinct, but this year the names were in alphabetical order, with Clinton near the top and Obama lower down. Stanford Professor Jon Krosnick, a survey specialist and expert witness in a lawsuit about ballot order in New Hampshire, has estimated a three percentage point or greater bounce for a big name candidate appearing high on the ballot. Therefore, if pre-election polls randomized candidate names, as most do, they would have underestimated Clinton's support by at least three points.

Regardless, there were no immediate clear answers, and lots of data analysis ahead.

http://blog.washingtonpost.com/the-trail/2008/01/09/about_those_democratic_polls.html