Still thinking about polls and the almost overwhelming problem of bias in them. Polling seems to work fairly well in very straightforward instances (say exit polling or even predictions of elections based on sampling "likely voters"), but whenever polls require more opinion, they fall flat. I found the article Deception and Public Opinion Polling by Mark Sapir of Retro Poll to be an excellent dissection of the problems with polls.

Retro Poll is trying to conduct polls somewhat differently than the usual news and political opinion polls. Here is their mission statement:

Our Mission: Retro Poll is a citizen based voluntary non-profit organization which aims to contribute to building a free, open and democratic society in the U.S. Retro Poll designs and performs opinion polls that look at the relationship between public knowledge and public opinion. In so doing so, Retro Poll reveals how the government and corporate media distort information in order to manipulate, confuse and disorganize the public's will.

Here is their most recent poll from spring of 2004.

That said, I'm still a bit of a junkie for infomation. So here is a aggregate polling site covering the current Electoral Vote predictions for the 2004 presidential election.