Reply #50: Please Do [View All]

I have been arguuing with Blumenthal on Mystery Pollster for some time and he wilts under cross examination.

You obviously have done your homework.

If you look at my posts on this board you will see some of my criicisms of Blumenthal. Many in the last 48 hours or so. On his site as well.

I understand the concept of clustering and the design effect phenomenon. I have said that I can see how the concept applies to clusters of marbles in a jar. But it is a stretch from there to precinct exit polls. I understand the general rule, but I also understand that general rules have exceptions when the facts warrant.

This post is a good example. If somone wants to make the point that the general rule is that people lie about the loser they voted for, I can except that as a general rule. What I can't accept is its application to every situation, especially the 2000 election. Who won? Depends on who you ask.

Of course you have even given me pause to accept the "general" rule, which my nature and training makes me suspicious of anyhow.

But to the point at hand. Since MOE's are so crucial in this debate, it would be nice to know if we have some wiggle room based on underpolling of Democrats and overpolling of Republicans. I can see your argument that democrats move because of poverty and are dropped from voter rolls. Makes sense on a local level. However, I would expect those voters to vote somewhere else. Also, this still would not explain overpolling of Republicans. Chances are the democrats would not be moving out of Republican precincts because they could not have afforded to be there in the first place. They are not upwardly mobil. I could see the opposite happening. Upwardly mobil Republicans moving out of lower class neighborhoods.