Reply #35: OK, let's talk about New York [View All]

[OnTheOtherHand]

There is no tag team; I just get annoyed when people offer dogs as dogmas.

"Exit polls: Where those M-E's, Got that link? Make it a good one, and not, please, Lugar's, again."

I have no idea what that means. (What, you still think that the statement of Bush's representative at the Ukraine election was irrelevant to assessing what the Bush administration was saying about the Ukraine exit polls? but I digress.) Was M-E the same as MOE, and if so, what statistic(s) do you want margins of error for?

If you want to know where the 30 point Kerry win figure for New York comes from, you could take it from Steve Freeman's presentation in Philadelphia (he ended up with 29.7), or you could take it from exit-pollz.org (http://www.exitpollz.org/cnn2004epolls/Pres_epolls/NY_P.html), who in this case apparently took it from the Edison/Mitofsky evaluation report http://www.exit-poll.net/election-night/EvaluationJan192005.pdf at p. 22 (the Best Geo estimate was 31.3).

If you want an MOE for the exit poll result, well, TIA has it at about 3 points with a 20% cluster effect. E/M has a SEDF of 3.7, which is about the same as a 3.7-point MOE.

If you want the MOEs for the polls, you can download the whole file at http://www.electoral-vote.com/2004/info/allpolls.csv . More details below.

"Lever machines are damn near totally illegal." Umm, they will have to be replaced, yes.

"Is it because they can't be recounted and are prone to fraud?" That's a pretty complicated question to squeeze into rather few words. Yes, individual votes can't be recounted. No, the machines aren't especially vulnerable to fraud in the sense of vote stealing (technicians can preset the counters, but poll workers should detect that). The machines are getting old and breaking down (and potentially could be deliberately made to break down), which is probably the main reason they are being replaced under HAVA -- although HAVA does also require a paper record.

So, I guess I will go with No, unless you have something from the legislative record that demonstrates that fraud was a major rationale for moving away from lever machines.

"Skimming 10 percent, statewide, would be easy."

No, it wouldn't. It would require widespread collusion and/or incompetence, and it would tend to stick out like a sore thumb, unless you think that every race is jiggered in the same way. And maybe you do. Please note, however, that paper ballots can perfectly well be jiggered (stuffed, destroyed, miscounted), so once you start postulating that everyone is crooked, you're really not talking about technology problems at all.

"With a MOE of say 6%, the numbers you gave us about the pre-polls could have easily matched up with the smaller MOE's from good exit polling."

Do you realize that you are just making stuff up?

The MOEs on those three polls were 3, 7, and 4. It's not "easy" at all to get all three of those polls to yield a Kerry vote share at least 6 points lower than the exit poll share (corresponding to a 12-point difference in margin, 30 minus 18, which is pretty generous) -- even if you start by assuming that the exit poll was unusually wrong in New York alone. If I spot you three points on the exit poll, the chance of all these polls being wrong by another 3 points in the opposite direction is about 1%. But since that 3-point error (6 points on the margin) in the exit poll is about a one-in-forty shot, the overall probability is actually lower.

So, bluntly, the pre-election polls and the exit poll disagree. Now what?