Wednesday, December 9, 2009


A new poll by Public Policy Polling asked an interesting question: Who would you rather have as president right now, George W. Bush or Barack Obama? Obama won, but barely, 50% to 44%, with 6% either undecided or unable to successfully press "1" or "2" on their phone. Jeez, I thought, that seems really bad for Obama. So I looked at some other results in their poll, and it seems skewed towards the GOP (I'm not accusing of malfeasance in any way, just saying that their data has a bias). For example, the actual results from the 2008 election has Obama winning 53% to 46%, but when this sample of people was asked who they voted for, it was 47% Obama and 45% McCain, with 8% remaining. So in this sample, Obama voters are under-represented by about 6 percentage points. Further, when you look at the breakdown of McCain voters who still choose Obama over Bush, and Obama voters who would now choose Bush, McCain has a defection rate of 10%, while Obama's is less than 5%. Further, the 8% of people who voted for "someone else" in 2008 (the non-Obama/McCain percentage was actually less than 1%), well, those folks/fools prefer Bush to Obama by a 3.5 - 1 margin, which further biases the data.

By my back-of-the-envelope calculations, when you correct for the sampling bias, you find that greater than 55% of people would prefer Obama to Bush -- so definitely at least equal to and maybe a smidge higher than actually voted for Obama. That sounds about right to me, and now that I think about it, is not cause for concern.