As I and others have pointed out, the polls are suspect. Most polls that have Obama up by more than the margin of error assume there are far more Democrats that will vote than in any prior election. The polls suggest that Obama wins the Democrats, McCain wins the Republicans and they split independents. The only way for Obama to be leading, then, is for there to be more Democrats than Republicans.
This is important because the poll’s “horse race” number (i.e., who voters will vote for) is adjusted to fit this model. For example, if you poll 100 people and 45 say Obama and 44 say McCain, that may or may not be the reported result. If, out of those 100 people, let’s say 30 identified themselves as Democrats and 30 were Republicans and 40 were independents. If you believe there are more Democrats than Republicans, you weight the poll, meaning the 30 Democrats get counted more than the 30 Republicans.
Historically, there have been about 3-5% more Democrats than Republicans. But many polls this year are assuming far more of a difference than that. The ridiculous Pew poll that has Obama up by 15% is the perfect example. Below is a chart of the party identification for the last 20 years of Pew Polls.
| Year | Republican | Democrat | Independent |
| 2007 | 25% | 33% | 34% |
| 2006 | 28% | 33% | 30% |
| 2005 | 29% | 33% | 30% |
| 2004 | 30% | 33% | 30% |
| 2003 | 30% | 31% | 31% |
| 2002 | 30% | 31% | 30% |
| 2001 | 29% | 34% | 29% |
| 2000 | 28% | 33% | 30% |
| 1999 | 27% | 34% | 34% |
| 1998 | 28% | 33% | 32% |
| 1997 | 28% | 33% | 32% |
| 1996 | 29% | 33% | 33% |
| 1995 | 31% | 30% | 33% |
| 1994 | 30% | 32% | 34% |
| 1993 | 27% | 34% | 34% |
| 1992 | 28% | 33% | 36% |
| 1991 | 31% | 31% | 33% |
| 1990 | 31% | 33% | 29% |
| 1989 | 33% | 33% | 34% |
| 1987 | 26% | 35% | 39% |
| Average | 29% | 33% | 32% |
| Difference | 4% | ||
| Min: 25% | Max: 35% | ||
Late October 2008 |
24% | 39% | 32% |
| Difference | 15% |
So Pew is telling us that it adjusted its actual results to reflect the “fact” that there are 15% more Democrats than Republicans, even though the historical average differential is 4%, and the maximum differential over the last 20 years was 9% (1987). In fact, even the difference between the minimum percentage of Republicans (25% in 2007) and the maximum percentage of Democrats (35% in 1987) is only 10%, only 2/3rds of the alleged difference this year.
I wondered what would happen if we used Pew’s 20 year average and reworked the poll. I assumed Democrats go for Obama 90-1 over McCain and Republicans go for McCain 90-1 over Obama (9% are undecided in this poll), and Independents go 47% for Obama and 44% for McCain (these are the numbers that give us the 53% to 38% lead for Obama). Let’s also lump in “others” with Independents so we have 100%.
If we now re-weight the polls to be 33% Democrat, 29% Republican and 38% Independent (or other), Obama leads 48% to 43%. A 5 point lead, nearly within the polls 3.5% margin of error. A far cry from the quoted 15%.
It occurs to me that we have two scenarios brewing, neither of which are good. The first is that Obama wins next Tuesday. As I pointed out earlier, this is the texbook definition of a parade of horribles. The second option is, with Obama with a (Real Clear Politics average) lead of 6-10% going into election week, Obama loses.
Why is this bad? Because instead of the horrors of four more years of Jimmy Carter, we will have four more years of “You stole the election” along with heaping helpings of “you racist bastards.”

