Donkey Cocooning (good analysis of political party ID and its effect/weighting on polls)

Slate ^ | 9/22/04 | Mickey Kaus

[Cableguy]

[snip]

This provides evidence to support both aspects of my hypothesis. First, polls that weight [by party ID] are more similar to one another than polls that do not weight. Second, polling firms that weight show less movement from poll to poll than polling firms that do not weight.

Well, yeah. If you weight every poll to the same PID [party identification], you will get the same vote. How can you not, when about 90% of Dems vote KE and about 90% of Reps vote BC? Imposing a pre-determined PID weight insures consistency, at the cost of repressing changes in public opinion. In fact, why do we bother doing poll after poll after poll at all? PID, and therefore the vote, will be the same as last time, right?

Wrong. PID changes in the electorate, both because Reps, Dems and Indies have differing levels of motivation to vote, and because sometimes one party or the other is temporarily more appealing or less appealing to people. That's why PID changes, and why it is a mistake to impose pre-determined PID on a poll.

For example, according to the world's leading expert on public opinion polling, Ruy Teixiera, "[A]s a campaign progresses, the level of interest among voters tends to change, particularly among those with partisan inclinations whose interest level will rise when their party seems to be mobilized and doing well and fall when it is not. " (emphasis in the original).

So, on the one hand, Ruy says the Proper Analyst will reweight to a fixed PID [derived from the actual 2000 turnout]. Then he says people in different parties will be more or less likely to want to vote as the campaign progresses--and of course, that means Reps and Dems will have differing turnout....PID will change with turnout.

[Cableguy]

More at the website. Read the whole thing if you are interested on polling analysis

[Cableguy]

Take a look at Kaus' analysis. He makes some good points.