Was it statistically impossible to have a recount in Florida?

Sunday, October 16, 2005
I am an ignoramus when it comes to statistics. However, a few of you might be able to answer this question: was it statistically impossible to have a manual recount of the Florida votes to resolve the Bush-Gore election controversy in 2000. I can well remember a NPR interview with a statistician who went into some detail to explain why such a recount of some five million votes would get a different answer each and every time. Human beings with the best intentions invariably make mistakes. Even an error rate of a one in fifty thousand would prove to be disastrous. People get tired, get slightly confused, knock things over, etc. I did not bother to remember this young lady’s name. I simply expected that her views would be repeated often by others in the next few weeks. Wow, was I ever a fool! That was the first and last time the statistical argument ever came up. Was the very idea of a statewide manual recount inherently a foolish idea?


terrye said...

I doubt if you would ever get the same count twice, not when counting that many votes. I do remember that every count they had looked better not worse for Bush.

I think we will never know exactly who won, but Bush got the job and it is time [as they say] to move on.

I read somewhere that there is always a gap, always a question, the difference is there is usally a large enough lead by someone that the gap would not change the outcome.

2000 was the exception to the rule.

David Thomson said...

“I doubt if you would ever get the same count twice, not when counting that many votes.”

I agree---but what do the staticians say? Why should we return to this question? Isn’t it better to “move on” to more current topics? Nope, we must not let this go. It is also my theory that the leftist establishment is putting enormous pressure on their academic friends to keep their mouths shut if their field of expertise might inadvertently assist Republicans. Nobody tells them to lie. They are only asked to remain quiet. This particular example is not the exception. It is probably the general rule.

At this moment, it is the consensus opinion that the Bush-Gore contest is unresolvable. Only so-called anecdotal evidence is supposedly available to those on both sides of the argument. I don’t think this is accurate. It is my suspicion that there is enormous hard evidence available to prove that a recount was inherently foolish. Am I right?

chuck said...

Bush did seem to have the edge in Florida, but I always figured it was pretty much a coin toss. *Shrug*. Someone has to have first possession. It never bothered me much, and I voted for Gore.

flenser said...

I don't believe that statistics made it impossible to have a recount in Florida. But IIRC there were too many variables at play for statistics to be able to state who "should" have been the winner.

The paper ballots were altered everytime they were handled, let alone counted. Adding that to the very subjective nature of what was to be considered a valid ballot and it becomes clear that multiple counts might very well produce different results.

On the other hand, I think statistics may have something to say about the likelyhood of large changes in the results. After the first count Bush had a lead of several thousand votes. The first recount shrank this to below a thousand. I recall reading at the time that the odds against this happening were quite large, and that it suggested some new ballots were being "found" and added to the mix for Gore.

Rick Ballard said...


Would any statistical outcome of less than 1 be satisfactory to the theft fantasists? I would submit that even certainty would have no effect on the lefties - the '00 election has achieved myth status and is moving into their Articles of Faith. The same goes for the Plame Game. It doesn't matter what Fitzgerald does, (short, perhaps, of indicting and convicting Wilson) the leftie deadheads are going to continue to howl that a great wrong was committed.

The way I look at it, good for them, the more time and effort they put into fighting lost battles the time and effort remaining to them to do present and future damage.

MeaninglessHotAir said...


I have always believed that the whole problem with that vote was exactly that it fell into the statistical noise. We all have the illusion in everyday life that we can make exact measurements--we act as though we make exact measurements--but the reality is that, except in the simplest cases, we can only make approximate measurements. Most of the time it doesn't matter, but in that case it did.

But this is not a provable hypothesis. Any statistical analysis would depend upon assumptions made about the variables involved. You would have to assume that 1 in 50,000 counts would be in error, etc., and those people on the other side would argue with your assumptions. So, bottom line, it is in my opinion 1) statistically impossible, 2) people who don't want that to be true will not believe it, and 3) it doesn't matter anyway because the people still fighting this battle have, as Rick says, completely mythologized the entire incident anyway.

ambisinistral said...

It is irrelevent because Florida law does not allow for endless recounts. A contested election is recounted once.

SneakyFeet said...

It's not impossible to have endless recounts, but you wouldn't get the same numbers twice, and the populace would be in an uproar (even those who understood the concept of noise in statistical measurement, who in this case may have been screaming the loudest for a recount).

It would be interesting to attempt to measure the noise in the presidential election count, or at least to estimate it. I wonder if anyone's ever done it.

Syl said...

I wrote a comment and must have closed the browser while I was only previewing it. sigh.

Anyway, I know the question here is whether we could ever get an truly accurate recount...but...

We are a nation of laws. Therefore it matters that there was not enough time for a recount by the deadline that was dictated by Fla election law.

What has to be accepted re the 2000 election is not who got the most votes, but that who got the most votes doesn't matter.

David Thomson said...

“Any statistical analysis would depend upon assumptions made about the variables involved. You would have to assume that 1 in 50,000 counts would be in error, etc.,”

I think it is abundantly clear that a rational person could easily assume “that 1 in 50,000 counts would be in error.” Not only that, many ballots were not clearly marked forcing the counter to guess.

“...and those people on the other side would argue with your assumptions.”

And there are people even today who insist that the world is flat. The heck with the crazies. What would a reasonable person conclude? I am not a postmodernist who believes that truth is relative. Some things can readily be ascertained as either true or false.

“It's not impossible to have endless recounts, but you wouldn't get the same numbers twice”

No, you almost certainly would not. It is therefore fair to describe those calling for the recount as mentally challenged---if not even mad.

Eric Blair said...

Historically, (staticically?) the conventional wisdom is that any election in the US has an error rate one way or the other of 1% to 3%

Check your local laws, but most places mandate a recount when the difference is less than 2%.

Given that the difference in the popular vote was less than .5%, you can see the issue nation wide. (Especially with the vote fraud in places like Wisconsin, Missouri, New York, and New Jersey to name a few).

So, its not statically impossible, but as has been pointed out, but there'd be a different result every time.