This morning, Tom Jensen of Public Policy Polling writes in, presumingly in response to this morning’s article, “Look at the party ID numbers from our last pre-election poll in NC in November and how close the poll came to getting the final outcome right: 8% more Democrats than the exit poll, right on the mark.”
Their sample is 49 percent Democrat, 38 percent Republican, 13 percent other; the CNN exit poll split 42 percent Democrat, 31 percent Republican, and 27 percent independent.
In PPP’s sample, the Democrats split 81-18 for Obama, the Republicans split 90-9 percent for McCain, and other 50-46 for Obama.
In CNN’s exit poll, the Democrats split 90-9 for Obama, Republicans split 95-4 for McCain, and independents split 60-39 in favor of McCain.
In other words, while the PPP sample had a lot more Democrats than the electorate did, they had enough Democrats for McCain and underestimated the share of independents in just the right proportion to come up with numbers that were quite close to the actual results.
So we’re left with three interpretations. Option one is that party ID is not really a significant enough issue to warrant weighing a poll sample. Option two is that PPP’s sense of party breakdown may be off sometimes, but as long as their sense of how each party demographic is similarly off in just the right proportion, it will result in an accurate reading.
Option three is that they were lucky.