2. This bring up another point. Almost all national discussion on the Bradley Effect centers on non-blacks lying to pollsters because they don't want to sound like racists if they vote for McCain. Up until now, it has been a black and white issue. However, this year we have a new wrinkle. Little has been said about the awesome intimidation of black voters who prefer McCain. The Michigan polling shows the black side of the Bradley Effect very clearly. McCain will get black votes from those who are pro-life, pro-gun ownership. Affluent blacks have the same concerns as affluent non-blacks over taxation. If the poll shows McCain at zero, some people are lying -- and you will see it on Election Day.
3. Gallup just released a poll (Sunday evening) that gives Obama a ten point lead, 52 t0 42. Just four days earlier, Gallup called the race for Obama 49 to 47. This latest would mean that at least 5 percent of those who were voting for McCain a couple days ago changed their minds. I say "at least 5 percent" since it is likely higher to offset some undecideds who have decided for McCain in the meantime. Gallup can call it a shift, but basically, one of these polls is just ... wrong. Maybe both. We'll find out in a couple days.
4. About the same time Gallup was showing Obama breaking away in a romp, the poll that claims to have been the most accurate in 2004, the IBD/TIPP poll, claims the race is closing in with Obama's lead shrinking to 46.7 to 44.6. While Gallup has the undecideds stampeding to Obama, IBD/TIPP has them flowing to McCain. If they are right, the 8.7 percent undecides will put McCain over the top. Stay tuned.
5. They say that there maybe be around 130 million voters this election. The typlical poll registers the opinion of between 600 and 1000 of them. Using the higher figure, this means that each person being polled stands for 130,000 voters. So, when I fib and say I am voting for Obama, but I actually go in and vote for McCain, my impact on the election is a 260,000 vote difference --- the 130,000 I take away from Obama and the 130,000 I add to McCain. (If I use the 600 sampling, the impact is more than 430,0000 vote difference.) If you have a 5 percent error in the sample population (including fibbers, like me), the projected error is between 13 and 22.5 million votes.
6. Wonder why pollsters usually say an election is closing in at the end? Because you can't be wrong in predicting a close election. If you give both Obama and McCain 50 percent, with a margin of error of 4 percent, the election can go to 54 percent to 46 percent either way and the pollster will pat himself on the back for an accurate prediction. And how many presidential elections are outside the 54 to 46 range? Damn few.So ... you can see why I think polls are a bunch of hogwash.