In both states, polls generally underestimated Mitt Romney's performance and over estimated Rick Santorum's.
Also worth noting, is that I picked the wrony winner in Michigan. Ultimately, there were two factors that were the root cause.
First, Public Policy Polling released 2 (relatively) small sample size polls in the last two days with diverging results. On Sunday night, PPP showed Mitt Romney ahead by 2. Monday night, PPP released another poll showing Rick Santorum ahead by 5. Because I only allow one poll from any pollster in my results, Monday night's release was included.
Second, as part of my methodology, I include all polls in the last two weeks. Two weeks ago, Rick Santorum was generally leading in Michigan polls. As polling services scrambled to make final prediction in the closing days of the campaign, most og the older polls were excluded. However, there were a couple of notable exceptions. ARG last released a poll on 2/22 and the Detroit Free Press last released a poll on 2/21. Both showed Santorum ahead by 3 or 4 points. NBC News/Marist also had a poll included in that was taken on 2/20, but had Mitt Romney ahead by 2.
By excluding last week's polls and normalizing the PPP problem and my projection would have MItt ROmney ahead by about 1.5%. Lesson learned.
No comments:
Post a Comment