HOW can you say with 50% in, someone wins by double digits!!??

[cavebat2000]

Please explain this. It's irritating the hell out of me! How can you predict something with only half the votes? Not saying Kerry wont win, just wondering how they can say it's a fact he will win by double digits.

[NoMoreRedInk]

1. Exit polling and statistical inferences*****

nm

[dobak]

2. easy

Combine exit polling data with returns from the precints and you can get it right almost without fail.

simple statistics

[EllieDem]

4. 2 words: Exit Polling

[cavebat2000]

7. sounds like bull sh*t to me

[leyton]

3. In Iowa

the percentages barely changed all night, and they haven't so far in New Hampshire. Voting seems be consistent across Iowa and across New Hampshire.

For Dean to tie Kerry, he would have to get about, I think, fifty-five percent of the remaining vote. That's not going to happen if he's only gotten twenty-five percent so far.

[progressivebydesign]

5. Hey. they called Kerry the winner when 27% of them were in!!

I was incredulous... With 102 of 306 reporting, a media person declared Kerry the winner. The media calls the game.. not the voting officials! Yeah.. what is up with the exit polls.. geez.

[Guaranteed]

6. It's highly unlikely that things would change in that, mathematically,

in order for Dean to shave another 3-4 points off of Kerry's lead he would have to do even better than +3 or +4 from where he is now...he'd have to be more like +6 or +8, since it has to average in with the first fifty percent....and the more vote comes in without change, the harder it gets to change anything.

[mlawson]

8. They know which precincts are in and which are not. Then,

they compare exit polling with that data, and also the demographics for the mising precincts. It's really a statistical exercise.

[kiahzero]

9. Standard statistical practice

Did you know most polls are done with roughly 1,200 to 1,500 people, for nationwide polling?

Do the sample right, and you don't need a large percentage.

It's hard to have a bad sample when your size is N/2.

[unblock]

10. are you kidding me?

yes, it's entirely possible that the first 50% counted were all the kerry voters and that the remaining 50% will be 100% for dean.

but that's a bit like flipping a coin 100 times, having it come up 60 times heads and 40 times tails, and then thinking, "well, isn't it possible that if we flip it another 100 times, it comes up TAILS every single time? then the coin would be tails 140 to heads 60 and so COULD BE unfairly biased toward tails!"

sure, it's theoretically possible, but if the coin were unfairly biased toward tails, what are the odds that the first 100 flips show up with only 40 heads?

not very likely.


same deal with the elections. unless you have a very good theory as to why the first 50% of the counted votes somehow came from solidly kerry country and that the remaining 50% are from even more solidly dean country, then you really have to say that the odds of kerry having a big lead with the more-or-less randomly chosen first 50% when dean in fact won state wide is NOT VERY LIKELY.

basic statistics.

[cavebat2000]

12. Uh

I said my arguement was with the "double digit win"

Can you read?


Sorry... im pissed

[benburch]

11. Sample size

That is a MUCH bigger sample size than any polling group uses. As sample size grows, uncertainty drops. If you poll half of all voters, uncertainty nearly vanishes. Compare this with, say, a Zogby poll where they poll 1000 people, but with fairly good accuracy. There is just no way to call the election wrong with 50% reporting as long as those reporting are randomly distributed.

The joker comes when things disrupt the counting in a whole region or demographic. The "Dewey Defeats Truman" mistake came because the urban precincts had reported by press time, but the rural precincts largely had not.

Nothing similar to this happened today in NH.

-Ben

[cavebat2000]

13. Not 50%

When people vote is an issue. I dont think the older people will vote at 6:00 pm. Working families would vote, and working families might vote for Dean over Kerry, therefore leading very heavy on the end of the % reporting.

[cavebat2000]

14. SEE!!

2% difference in 10% of precincts. Dean rose 1 and kerry dropped one. IT can change, and has, and prooves my point

[cheezus]

15. You can laugh at everyone

When all the votes are counted, and you're right ;)

[Indiana Democrat]

16. How many times will this question be asked?

Good God! The propaganda against **all** the media has worked.

Exit polls are a PROVEN comodity. These people are VERY good at what they do...They are professionals.

Just take note of the returns now and then compare them to the final results. You'll never as this question again.