Too close to call?
I am interested in objective viewpoints from the statistically literate among you.
The numbers I’m seeing published suggest that punchcard-type counters tend to reject up to a few percent of votes, due mostly to the infamous “hanging chads.” I’ve not seen a study done, and I’d be interested to know if one has been. Optical scanners may be up to an order of magnitude better. This assumes no malfunction on the part of the machines.
Assuming no bias (please spare me your opinions on the biases of the counters), humans can pick up maybe 90% of the missed ballots.
Under the best of conditions therefore, it seems that our most precise counting methods will still miss a few votes per ten thousand. This is pretty good when the contestants are separated by several percentage points.
In Florida however, there are enough lost votes within the statistical noise to swing the race either way. Should there be a law declaring that any certified result where the contestants are separated less than one hundredth of one percent be called a draw?
I am not interested in your political opinions here, I’m interested in your math opinions. Thanks…
That makes sense to me. It shouldn’t be too hard to come up with margins of error for different voting machinery- 0.01% for punch cards, 0.001% for fill-in-the-circle type cards, etc. If the election is so close that it’s within the margin of error, then it’s a draw.
An ATM-type touch screen would eliminate any error, though. The problem with other voting methods, such as punch cards, is that the vote isn’t just being counted by machine- it’s being quantized into a yes/no vote, from an analog signal (i.e. the position of any “chad”, or the amount of darkness in a circle). This quantization (the equivalent of a “hard” binary decision in detecting bits in a communications system) is the non-repeatable part. With a touch-screen computer system, the binary quantization is made instantly, when the voter is still standing there staring at the screen, after they press the screen button. Once they accept the vote, the vote is stored in computer as either a vote for one candidate or another- no possible controversy can exist after that about who the vote was for. All that remains is for the computes to count the votes, which should always give the same answer. This system inherently removes any ambiguity about who the vote was for. Of course, the voter might have accidently voted for the wrong person, but at least the vote itself can’t be disupted.
Arjuna43, what happens when a person bumps the screen, or touches in between the two areas, or the screen gets all smudged and doesn’t work? If people can contest the butterfly ballots, you can be certain that they’d find a reason to contest touch-screen ballots.
As to declaring the election a tie, fine, but what then? Most states, when there’s an exact tie (according to the count) decide it by a coin flip or some other random method. Would this really be preferable to listening to the statistical noise?
The touch-screen won’t solve disputes about a dirty screen, accidental bumps, etc. (although having an “Are You Sure” window would help- you could even pop up a picture of the candidate). The beauty of the system, though, is that once the vote is entered (whether it was the intended vote or not), there’s nothing else that can be disputed. There’s no physical ambiguity as to whether a vote is for candidate A or B- it’s already stored as a binary bit in the computer, instead of being stored as a physical piece of paper, which is subject to debate (pregnant chad, partially filled circle, etc.).
That eliminates the need for recounts, hand counts, etc. If a candidate feels that there’s been a big mistake (such as a confusing computer screen), the only thing that can be done is a re-vote, which is pretty rare.
If it were up to me, and I were sure that the Florida election was truly within the margin of error of the machinery, I’d split Florida’s electoral votes evenly (with the 25th vote either thrown out, or each given 0.5 votes from it, or maybe Nader getting that last one). I’d go with a game of chance on who gets Florida, also, but I’d prefer to split the votes.
Avoid the touch screens – just use a button or lever. It works fine in New York State and I’ve never heard of anyone having the type of problems showing up in Florida. You pull a lever; it’s either up or down and it’s designed that you can’t pull two down (you can change your mind just by putting the lever back in its original position before you complete your voting). Margin of error is next to zero – the only errors occur when the votes are tabulated.
There’s no need to reinvent the wheel for a high-tech solution when there’s a perfectly good one already available. The only real drawback to voting machines are their cost (and, of course, the fact that no one currently manufactures one – though if there’s a demand, it’d be easy to take the old plans and start again).
unfortunately the machines in NY have been out of production for 10 years or so. they are starting to break already and next year some of the ballots in NYC will have as many as 36 races to deal with. The question has already come up as to whether or not there will be enough functional machines to serve the 5 boro’s.
Ah, the same faces, or, at least, names are here.
That’s what I know: in science, if two measurements are compared and they fall within the margin of error, the comparison is deemed unreliabe. It is not necessary “invalid” or something, it means the difference between the two is not certain. I guess, it would be called a “tie” in politics.
The solution: devise a more precise metod. Simple recounting 10^n times may or may not bring the result into the range and may be inaceptable due to time/cost/other constrains. New method accuracy can be calculated prior to its implementation. This solution, apparently, is unapplicable to a political situation, but can be implemented in the future.