In an election eerily reminiscent of the 2000 Presidential election, in Washington state the seated Governor, Christine Gregoire, won a hotly contested election over the Republican candidate, Dino Rossi. The margin of victory was at most 129 votes, and that was a reversal of what was at one time a victory by Rossi, with a margin (at first) of 261 votes and a recount (by machine) of a mere 42 votes statewide. Cite .
Now the Republicans are challenging the election. We can go through all of the Bush sux stuff now, we can drop the “what’s good for the goose” stuff, because I don’t care about any of that. It’s irrelevant. What I’m more interested in is how we can resolve situations like this without endless court challenges. There has to be a winner and a loser, and with elections this close how are we to determine which is which? We can’t do endless recounts and it’s not practical to have a statewide revote. The courts have no real jurisdiction in cases like this unless there’s obvious cases of fraud, so going to the courts over a close election seems stupid to me.
Tough question, at least I think it is. Self government only works is the people in it are interested in governing. If they, or a significant number of them, only want to haggle over things then the system doesn’t work. The courts only have such power as people are willing to grant them. I would cite as an outstanding example the impeachable offense of Andrew Jackson. The state of Georgia siezed the land of the Cherokee indian tribes who took the matter all the way to the Supreme Court which decided in their favor. Jackson refused to enforce the decision despite his oath to see that the laws are enforced and the Cherokees were forced out.
If the opponents don’t go to court or refuse to accept a court decision were else would they go? The next step is guns isn’t it? As to fraud, someone has to decide whether or not a fraud has occurred and the only reasonable way to decide that is to have an investigation and a hearing on the matter. That sort of describes the process of a court trial.
A margin of error is a statistical thing. For legal purposes, there was a clear result: by 42, 129, or 261 votes, depending on the count. If these counts were accurate, the winner is the person who got more votes, however narrow the margin. The problem arises when one party or the other challenges the accuracy of the count (and obviously all three of these counts can’t be accurate, having produced different results). At that time it becomes a matter for the courts to decide if the basis for the challenge has merit.
In the highly unusual case where there’s an accurate count and an actual tie, the usual means of breaking the tie is some random draw. I’ve heard of this happening in local elections, but not at a state or national level.
In the UK, the candidates may choose some form of random act to determine the outcome in the event of a tie, such as tossing a coin or drawing lots. This does not seem to have happened in modern times. The closest margin in recent years was in 1983 when the Conservative candidate won the Leicester South seat from Labour by just 7 votes [21,424 to 21,417 - total valid votes cast: 53,187]. See here.
You have to be careful applying statistical concepts to an election because your sample is self-selected. I think the best you can do in general is to simply declare the winner to be the person who got more votes.
Naaw, Airman Doors is right. A margin of error means “We don’t know who got the most votes, our methods of counting votes are sloppy enough that the apparent margin of victory is less than the margin of error.” So there was not a “clear result”, just as there was no “clear result” in Prez '00 down in Florida.
I don’t have a brilliant solution to offer. There are a lot of problems with a do-over but I don’t see any options that are an improvement on it. You can’t exactly go into sudden-death overtime. I suppose a round of rock, scissors, paper would be out of the question?
In theory, elections don’t have a ‘margin of error’. An election is to a poll as a census is to a sample; the results are accepted as exact and definitive. If you win by 1 vote, then you win by 1 vote, with a margin of error of +/- 0.
If your question is “when do you stop recounting?” the answer is, per the procedures set out in the appropriate state or Federal laws. In the case of the state of Washington, hand recount > machine recount > original count. And nothing out there trumps a hand recount.
But the education the past two Presidential elections has given us has demonstrated that the belief in the exactitude of balloting is somewhat misguided. The question is, what do you do about it?
I think statistical methods should be used to deal with obvious glitches that can’t be remedied by a recount, such as the effect of the Palm Beach County butterfly ballot on the Gore vote in 2000, or the inclusion of non-felons on felon lists that are used to remove people from the voting rolls, and so forth. But unless the resulting change yields a statistically significant result at a high confidence level (I’m thinking .99 here), I think one must bow to the original, exact-but-quite-possibly-inaccurate result.
And for that reason, it’s useless to introduce the MOE of a particular kind of voting machine into the decision-making process; ALL it will do is expose the uncertainty of the process, without changing the result.
I think that such information should be available for informational purposes, so people can understand the reliability problems of the voting systems in this country, and keep the heat on to demand reliable and (especially) verifiable voting systems. But unless the winner is to be chosen by random number drawing in close elections, I think we’ve got to resolve elections more or less as we do now.
Since the margin of error is unknown in direction, it makes sense to follow the “official” tabulation with the understanding that if the error broke one way, you could have had a different winner and if it broke the other way, the margin of victory was underestimated. Either way, the bell shaped curve centers about the actual result and there is more area under that curve that indicates the “correct” person won than otherwise.
The solution of course is to make the election process foolproof. The technology is there, however at present those in control of the technology are not necessarily interested in conducting fair elections.
There’s no way to resolve election recounts afterwards without making somebody unhappy. The real solution is to spend the money beforehand to make sure that the recounts won’t change the results. Create a voting system that can’t be hacked. Make it work so well that recounts would come up with the same totals. O.K., perhaps it’s not possible to create a perfect system, but it’s possible to come up with one better than what we have now.
I think you’re missing the point. Sure, the actual number of votes in an election are exact and definitive and do not have a margin of error, but there’s a difference between “votes cast in an election” and “votes that our vote-counting mechanisms indicate were cast in an election”.
Since our vote-casting mechanisms do not rely on anything remotely akin to random sampling, I suppose you could make the point that “margin of error” isn’t the correct phrase, but the connotations of coarse graininess make it a reasonable nomenclatural borrowing.
Point is, regardless of the fact that the actual votes are definitive, our knowledge of how many were cast (and how) has a degree of fuzziness beyond which we can’t say with confidence how many votes each party actually received. And the question is what to do about it when the margin of victory is less than the “fuzz factor” of accuracy with which we know the votes cast.
This is an unfortunate interpretation of the margin of error. The margin of error is simply a 9x% confidence interval applied to the standard error, the square root of the ratio of the sample outcomes over the size of the sample.
It is not, not, absolutely not a measure of the “degree of fuzziness”. The only terms under which this could possibly true would be if we believed that the process that turns votes cast into votes counted is entirely random. While we can all agree that there is error, you would be hard pressed to make the case that counting votes is an entirely stochastic process. As such, the “margin of error” of an election is a statistical barbarism and not really a measure of anything. It should not be interpreted as such.
If banks can reconcile their ATM transactions down to the penny*, I don’t see why we can’t come up with a fool-proof vote tabulation system that issues a receipt. If the Federal Govetnment could force teens to schlep to their local post office and register for the draft back in the 70’s, why can’t it enact legislation that adds voting interface to ATM machines**?
As a bonus, Banks could have give-aways like every 10,000th vote cast wins $500, direct deposited into your account in you happen to be voting at an bank where your account is located. Granted it’s not going to make anybody switch banks, but it might get a few million more people to regisister and actually vote.
*Yes I realize ATM machines don’t dispense pennies, but you get the idea.
How so? We have the whole population: the individual voting precincts and their counts. It is certainly true that the statistical concept of the error that results from sampling is in appropriate; however, people and machines do make mistakes and each of the individual precinct reports may be off as a result. So the margin of error in an election is not statistical error, it is the error that arises from counting mistakes.
In this regard we cannot say which candidate garnered the most votes because our measuring stick is not finely graded enough to differentiate between the two. If we can reliably say that the vote count will likely be within 3 percentage points of the real count, and the vote count shows a 1% difference, then we simply cannot declare a winner based on the votes. (This isn’t a failure of the system, it’s just a fact of life that some people don’t want to face.) Because of this, I would be perfectly happy with just using some random method to decide.
You still have the issue of whether the counting mistakes are random. I doubt anyone would argue that they are, and that makes statistical techniques suspect at best.
I’m in agreement with Wendell here: make the count more accurate.
It’s worth noting that Washington State does in fact have a legal method of determining when an election is over and declaring a winner. The election law very clearly spells out how many recounts are allowed, which recount is considered the “final” one. Once the Secretary of State certifies the results of the final recount, and the legislature confirms, it’s a done deal. In Washington, the Republican Secretary of State sided with the Democrats and certified the election, the legislature confirmed, and that was the end of it. The GOP’s court filing is entirely unprecedented (although not unexpected.)
The OP’s question is “How can we resolve situations like this without endless court challenges?” The answer is that you can’t. The nature of our system is that any complaint can be raised in court, even if any objective person sees the complaint as groundless. If Rossi had lost in a landslide, he still could have protested the election in court if he chose to. There’s no way to tell someone in Rossi’s position “No, you can’t take it to court.” If you think about it, we wouldn’t want to prevent him, either - suppose the election landslide had been the result of coordinated fraud? What recourse would Rossi have? The pain we are experiencing as a result of this court battle is the necessary price of a fair democracy.
FWIW, I detested both of these candidates when they were running and didn’t care who won. After the recounts started, I immediately started hating Rossi a lot more - he became really whiny and childish, and to me it showed that he was really after the power and none of the responsibility. Gregoire, by contrast, acted like a grownup, so now I’m in her camp.
Why do we have the issue of whether the error is random? If the error is real, then our measuring stick is only so accurate. This isn’t a question of sampling error. Statistics is not the issue.
People and machines make mistakes. People and machines will always make mistakes. Making the counts more accurate does not address problem; indeed, I’d suggest that it will make the problem even worse when it does surface, because people will be even more convinced in the infallibility of the voting system. Don’t use this as an excuse to throw straw men at me: obviously the more accurate the measuring stick the better; yet it remains that we aren’t going to have a perfect measuring stick.
Instead of plowing resources into small gains that can obtain from searching for greater accuracy, wouldn’t it be more beneficial to simply accept that we shall never have a perfect measuring stick and write in to the law contingency plans for such situations?
