|
He's got that dopey antelope's number alright.
But let me give him a hand in one or two spots:
A-The exit polls showed that Kerry won
B- Exit polls are inherently unreliable.
A- But they are used all over the world to detect fraud.
B- They are not designed for that here.
A- Thousands are randomly sampled. The margin of error is very low.
B- It is not a true random sample.
great so far! The bonobo gets it!
A- Why not?
B- The cluster effect, or as the cognoscenti call it, the "Design Effect"
OK, guy, but that's only a bit of the story. The Design Effect is easy to fix. The real problem is with the the sampling of the voters. The ones who refuse are not random, and the voter selection may not be truly random either, especially if the precinct is crowded and the pollster is trying to pick one voter in 10.
A- But exit pols must be more accurate than standard pre-election polls. Many more are interviewed.
B- You forgot the design effect.
and again, more importantly you forgot the non-random selection. You can get a computer to generate random telephone numbers - it's much harder to stick to a strict Nth voter protocol. Believe me, I've tried it. Those bears just don't form an orderly line to be counted and some of them have big teeth.
A- But exit polls must be more accurate than standard pre-election polls because people are not undecided; they know excatly who they just voted for.
B- People forget who they have voted for in the previous election, so the responses cannot be believed.
But the good news is it doesn't matter much who they say they voted for last time. The important thing is that they probably remember who they voted for five minutes ago. So they probably are more accurate than pre-election polls. Much bigger samples too. The antelope's got a point there /
A- But that should apply to Gore and Bush voters.
B- No, only Gore voters forget.
Nah, they both forget. But they are more likely to remember voting for a guy they've seen in the news a bit. Who the heck was Gore?
A- I see, the cluster effect and forgetful Gore voters. What else?
B- Some Gore voters lied and said they voted for Bush.
Give us a break! Voting is HARD WORK! You mean the voters are supposed to remember some dope who didn't even win? I'm sure I never voted for a loser. Now Clinton, I remember voting for him. Dole? Who he?
A- Why would they do that?
B- To identify with the previous winner - Bush.
A - But he stole it from Gore.
B - Makes no difference. People like a winner. They forgave him.
Well, some of them did. Gee, he stood up for freedom and democracy when those Iraqis attacked New York.
A- I read that 16 states deviated beyond the Margin of Error for Bush
and the odds are 1 in 19 trillion.
B- Voodoo math - based on invalid assumptions. Bush exceeded the MoE in only 3 states. Remember, you must multiple the standard MoE formula by 1.6. Design effect.
No, the math is fine. The MoE doesn't matters much, because so many states went the same way. The odds of that happening by chance probably really are in the trillions. That antelope get some stuff right
A- But all 22 Easter time-zone states deviated to Bush, regardless of MoE. The odds of that are 1 in 4 million.
B- Sometimes, shit happens.
Yep, it looks like there's something about those Eastern States. Mostly Democratic territory, no?
A- But Kerry won the three exit poll time lines up to and including 13047 respondents. That's a 1% MoE according to Mitofsky.
B - 1% is a rounded number. It was more like 1.49%.
sounds about right.
A- That's still very low.
B- But Bush voters were reluctant to be polled . Haven't you heard of rBr? Of course Kerry won the exit polls. Bush voters avoided and Kerry voters were anxious to be polled. And these Bush voters were former Gore voters who forgot who they voted for in 2000.
Actually both voters avoided the pollsters - no-one was very keen to be polled. But if Bush voters were even less keen than Kerry voters, or if more Kerry voters were likely to volunteer, that might have caused the problem.
A- I'm confused. Are you saying that Kerry voters were anxious to speak to exit pollsters and when they did, they forgot they voted for Gore in 2000 or lied about it?
B- No, I'm saying that there were Gore voters who forgot they voted for Gore or lied about it and voted for Bush in 2004.
I shouldn't imagine Kerry voters who were anxious to speak to exit pollsters lied about voting for Gore in 2000. In fact I shouldn't think any Kerry voter pretended to have voted for Bush. More likely that Bush voters forgot they had voted for Gore. Or voters who hadn't voted in 2000 forgot they hadn't voted for Bush. I expect a few Kerry voters who'd voted for Bush pretended they hadn't as well. Too embarassing.
A- Is that why the 2000 voter split was 43%Bush/37% Gore.
B. Precisely.
Yup, the bonobo gets it.
A- What about rBr? Isn't there a contradiction here?
B- None whatsoever.
Absolutely. After all, the Bush voters who answered the poll weren't the reluctant ones, were they? And anyway, even if they were, you might as well be hanged for a sheep as a lamb. If you are going to admit to voting for Bush this time, you may as well own up to last time as well.
A- But 43% of Bush voters means that he got 52.57 million votes in 2000. He only got 50.45 million in 2000.
B- You forget, of the 43%, 1 in 14 were actually Gore voters who forgot or lied they voted for him.
That antelope sure has a bad memory. Probably a Gore voter.
A- Weren't the final exit poll weights adjusted to match the recorded vote count?
B- Of course.
easy one that - the pollsters told us that before the election.
A- So was that 43%/37% Bush?gore 2000 voter split an adjustment or was it in fact a true sample?
B- Both.
well it was certainly an adjustment. But if it got the proportions of votes correct, it could also have been an accurate reflection of how people reported voting in 2000. But that doesn't mean that WAS how they voted. People, as the bonobo says, often forget having voted for the loser. Or the guy they think is a loser now.
A- What if the vote count was rigged? Why match to a rigged result?
B- Because that's how it's always been done. What makes you think the vote count was rigged?
Yup, that's how they do it. It's up to you guys how you count the thing. All the pollsters do is tell you who's going to win the count. You mean you don't have a proper audit? And you expect the pollster's to tell you if the right guy won? Sheesh.
A- Assume for a moment that the vote count was rigged. If the FINAL exit poll matched a rigged count, why should we believe the final exit poll?
B- No one says that you should believe it.
No reason at all. I wouldn't believe a count that was done by some ruddy machine running Access. .
A- But if the Final Exit poll is wrong, because it matched to a bogus vote count, then the preliminary poll of 13,047 respondents which showed that Kerry must be close to the truth.
B - Not necessarily.
well, one of those counts was wrong, sure enough. But how can you tell which? Could have been one, could have been the other, could have been a bit of each
A- Why not?
B- Design effect
Nope, wrong answer. Design effect's got nothing to do with it. All that does is widen the MoE. But it's a tricky one to sort out - statistics can't help much as you can't infer causality from a correlation - though if you have a good enough a priori hypothesis to test about what has caused the bias then you can take a good stab at it. But bias is not that easy to measure.
A- So why do an exit poll in the first place?
B- Good question. The answer is simple. The people want to be convinced that their votes counted. The Final Exit Poll confirms that.
Wrong answer again (banana's coming soon, guys). There are lots of reasons for doing an exit poll and one of them is entertainment. Give the pundits something to talk about until the results come in. And you guys are impatient. Presumably that's why you use counting machines instead of counting the ballots by hand, and even then you want to know the answer before all the votes are counted. I'll tell you another reason though. It's a great test of polling technique. It's the only survey you can do when you get to know the right answer afterwards. What do you mean, it's not the right answer? You'd better get those ruddy machines fixed, pronto.
A- But how can you make that statement?
B- People believe the vote count is perfectly accurate - for two reasons.
One, 122 MILLION VOTES ARE TALLIED, NOT JUST 13,047.
Two, if it wasn't you would here about in the media.
Two, If people believe the vote count is perfectly accurate they haven't watched a recount. Some people will believe anything
|
Printer-friendly format
Email this thread to a friend
Bookmark this thread
(1000+ posts)
Send PM |
Profile |
Ignore
