Surface Map, as a rule of thumb, anything Deborah Orin writes is usually misleading. She's done it again, in this article you've linked to.
The poll had a subset sample size of only
62 individuals!!! The New York Post, as pointed out below, lied when it said the initial sample size was 1,275 participants, when the poll actually had an initial sample size of just 371.
http://mediamatters.org/items/200408190011Excerpt from David Brock's debunking:
Napolitano cited a study by HCD Research/Muhlenberg College Institute of Public Opinion, which was conducted online and surveyed a sample of just 371 non-party-affiliated voters (the New York Post, which published an article about the study the same morning, incorrectly reported a sample size of 1,275 participants). But the particular questions gauging the reactions of those supporting Kerry drew from a subset of only 62 people -- those who were "definitely," "most likely," or "leaning towards voting for John Kerry."
According to the poll:
Before the ad, 41.94% of Independents intending to vote for Kerry felt that they would "definitely" vote for him; 37.10% felt "most like" to vote for him; 20.97% were 'leaning' towards him; and none were "not sure."
Those levels of commitment changed markedly after viewing the Swift Boat ad. Independents "definitely" voting for Kerry dropped to 29.03%. Those "most likely" to vote for him were relatively unchanged at 33.87%. Those "leaning" towards Kerry dropped to 9.68%. Finally, Independents were "not sure" their initial choice of Kerry increased substantially to 27.42%.
Here's the problem with a total sample size of 62. A drop from 41.94 percent (carrying the percentage out to two decimal places is an obvious grasp at the appearance of scientific precision that is nonexistent in a sample this low) to 29.03 percent corresponds to a drop from 26 "definites" for Kerry to 18, or a difference of 8 people. The drop from 20.97 percent to 9.68 percent among the "leaners" is a move from 13 people to 6. Those are among the purported "marked" changes that the poll identified.