« Bush's Mighty 2 Point Win | Main | The Struggle for DNC Chair »

Hispanic Revisions Update

The indefatigable Mark Blumenthal over at Mystery Pollster has yet more on the revisions of the the NEP exit poll Hispanic figures. Read his whole post, but here's the essence:

1. The initial TX Hispanic figure of 59 percent support for Bush, according to the NEP, was the result of a "tabulation" error that improperly weighted telephone interviews with early/absentee voters that were conducted to accompany the election day polling place interviews. (OK--but how'd that happen? And why did it take so long for them to figure it out? And what about the TX white vote for Bush now--doesn't that have to be higher as a result?)

2. The 40 percent figure for Bush's national Hispanic support issued by NBC, based on aggregating all the state polls, was not a "correction" of the NEP national poll data, but simply a different, (though better, according to NBC) estimate, of Bush's national Hispanic support. The NEP national exit poll figure of 44 percent for Bush's Hispanic support still stands uncorrected by Edison/Mitofsky, the actual exit pollsters. (OK--but if we needed a better estimate than the national poll estimate because the sampling was screwed up--NBC's story--why is Edison/Mitofsky sticking by their national estimate? If NBC is right, doesn't it need to be corrected? If not, why not?--isn't it a problem that the national poll estimate and the state poll-based national estimate don't matchup (they did almost perfectly in 2000)?

Clear? I thought so. I eagerly await, as I'm sure you do, more "clarifications" from the good folks at the NEP, Edison/Mitofsky, NBC and whoever else is getting into the act.

Comments

looks like for mitkofsky its hard to admit a mistake most probaly when he corrected his poll that kerry was leading by 3% he came up with this number now he cant figure it out

but after all votes are in a real hard look at sample hispanic precints would give us a decent sample how the hispanic vote went