Monday, May 5, 2008

Weighting our final NC Poll

Our final North Carolina poll will be released within the next few hours. A big decision for pollsters on this one is how to weight for race. 28% of the Democratic primary electorate in 2000 was black, and in 2004 it was 32%. But over 40% of early voters were.

Obviously how you choose to nail down that figure can have a pretty significant effect on your Presidential numbers when the electorate is so polarized along racial lines.

We settled on 35%. We asked folks who were polled if they had voted early. Taking all of the respondents in our poll, if 40% of those who voted early were black then 35% of the population as a whole was.

I don't know if that's the best or most perfect way to do it but it is at least a rational decision based on the data we have.

6 comments:

  1. Tom:
    Can't wait for your poll :-)
    Do you have an estimate for the fraction of early voters? A comment by Michael McDonald over at Pollster.com says EVs are ~400k.
    http://www.pollster.com/blogs/the_demographics_of_the_north.php

    Thanks.

    ReplyDelete
  2. Hurry up, dagnabbit. I'm tired of waiting. :D

    Thanks for hanging in there and sorry for all the grief some are sending your way.

    ReplyDelete
  3. What is your overall turnout prediction? http://www.sboe.state.nc.us/ has registered Ds + Us at 3,877,000.

    ReplyDelete
  4. I think your turnout model is low for AA (at least I hope so). Here's why: Blacks make up 40% of registered Democrats and blacks in NC also vote in higher numbers than whites.

    In the 2004 general, white turnout was 56% whereas black turnout was 82%.

    ReplyDelete
  5. If you look at the OFFICIAL voter registration statistics for NC, you'll find that 37% of registered Dems. are Black. So this estimate looks to be on the mark.

    ReplyDelete
  6. it looks like it's 32%, so I guess Obama will win by 5-6 points

    ReplyDelete