The accuracy of elections.

greenspun.com : LUSENET : TB2K spinoff uncensored : One Thread

We are told that the accuracy of polls is 4% (my recollection). I don't know how anyone can prove this, but for sake of discussion let's say it is true. My question is: how accurate are the actual elections? Has this ever been studied?

In the recent election, we are seeing how innaccurate can be the vote itself. I am beginning to think that polls may be more accurate that vote tallys. I am becoming suspect of any candidate who wins by a vote margin of less than, say, 5%.

No, I am not saying we should hold elections via poll. I do say that we had better implement a more reliable voting technology, and fast.

No matter how the FL recount turns out, the election is actually a tie. We just don't know enough to say more. How about a coin toss for President at the halftime of the Super Bowl?

-- Lars (lars@indy.net), November 13, 2000

Answers

Lars, you ignorant slut. Your blatherings are irrelevant.

-- (who_let_the_dogs_out@who.who), November 13, 2000.

Lars:

I read a published number that depending on the type of ballot, machine counts could be off anywhere from 2-5%. However, the variance among the Florida counties that did machine recounts was nowhere near that large. True, only a minority of counties got the same result precisely, but few if any counties varied by more than 1% between the first and second machine counts.

The greater "accuracy" of a hand count also depends on the type of ballot, but much moreso on exactly who gets to interpret ambiguous ballots. A 5% or greater swing between a Republican and Democrat counting the same votes wouldn't surprise me a bit (assuming both are honest).

Nationally, the popular vote differed by about 0.1% and the Florida second machine count differed by 0.005%. Both of these counts are well within both mechanical and manual error limits. I'd estimate that if we did a nationwide machine recount 100 times, Gore would win about 60 and Bush about 40 of these, all of them of course so close as to be within the limits of error. In Florida, you can flip a coin, and Bush winning twice in a row is much like flipping heads twice in a row - about a 25% probability.

So you are right, we have a flatfooted tie, a photo finish whose winner depends on which camera provides you the perspective that matches your bet. Whoever wins this thing, it will NOT be based on any "real" vote count, but rather on whose count some court ends up accepting (assuming Gore's specially selected count produces the results it is designed to produce - a foregone conclusion).

-- Flint (flintc@mindspring.com), November 13, 2000.


Flint--

Thanks for your thoughts. There should be a better way to resolve a tie than by these power struggles. I truly would prefer a coin toss.

Of course then there would be wrangling over the balance of the particular coin, the technique of flipping, the elasticity and finish of the surface on which the coin lands, yadda, yadda.

-- Lars (lars@indy.net), November 13, 2000.


Moderation questions? read the FAQ