Reply #102: yes, election data are more useful

[OnTheOtherHand]

(For the benefit of anyone else following along: this isn't to say that exit poll data are utterly useless. It's sad that I may need to say that, since it's implicit in the prior statement.)

The move from descriptive to inferential comes when one tries to figure out how a correct vote tally would have turned out. One can indeed get a long way with descriptive stats alone.

It certainly isn't nonsense that the largest exit poll discrepancies favored Bush (i.e., Bush received more votes in the official count than the exit poll interviews indicated). What's nonsense is the link to electronic voting machines. Whether or not one thinks that the DRE and/or op-scan counts were accurate, they didn't evince the highest exit poll discrepancies.

I agree: in the Ohio 2004 context, "electronic voting machines" have been a huge distraction. This has nothing to do with thinking the machines are safe or reliable.