THAT LA TIMES POLL
Last week, the LA Times ran a poll showing a widening Kerry lead (46-42), which it headlined "Voters Shift in Favor of Kerry" The Drudge Report is trumpeting its discovery that the LA Times sample included 38% Democrats and 25% Republicans. It was clear that something was strange because the data showed Bush leading among Republicans (92-4) by a greater margin than Kerry led among Democrats (86-7). Bush also led among Independents (43-41).

Often, when trying to determine who are likely voters, pollsters will adjust their data based on a model of predicted turnout by party. If the turnout is likely to be 38% Democrat and 25% Republican, then the LA Times has it right. But that turnout model is unlikely.

According to a 2003 survey, the Pew Research Center for People and the Press reported that 31% of Americans identify themselves as Republicans and 30% as Democrats. Had the LA Times sample been 31% Republicans and 30% Democrats, then the LA Times would have reported that "Voters Shift in Favor of Bush", with the following results:

Bush 47% | Kerry 43% | Nader 5%

In 1996, Zogby had success with a turnout model of 34.5% Democrat and 34% Republican. Applying those numbers gives this result:

Bush 47% | Kerry 44% | Nader 4%

From 1997-2000, the Pew Research Center reports party identification of 33% Democrat and 27% Republican. Weighting the results based on those values:

Bush 44% | Kerry 45% | Nader 5%

I don't know what a good turnout model is, but I suspect that the LA Times is way off. I can't seem to find more recent data on party identification or even on voter turnout by party in 2000. If you can find it, please add a comment.

**UPDATE**: According to 2000 exit polling, turnout for that election was 39% Democrat and 35% Republican. Applying those numbers to the LA TIMES data gives the following results:

Bush 46% | Kerry 45% | Nader 4%