Wednesday, January 14, 2009

Election imprecision

I've just been thinking about the still undecided Senate election in Minnesota between Al Franken and Norm Coleman. Politics aside, to me the issue is ultimately a measurement issue. The purpose of an election is to measure the preference of the active voting population. When the margin between the lead candidates is large, there is no real issue with deciding who voters prefer. The methods and technology used are accurate enough that we can easily determine that one candidate is preferred by the active voting population over another. But when the election results are close, we can't really make such a determination because the measuring device just isn't precise enough.

This is similar to the problem faced by early geodesy in measuring the circumference of the earth. For example Eratosthenes around 240 BC first calculated the circumference of the earth using measurement of shadows at a solstice and the distance between Alexandria and Swenet. His estimate was probably better than 99% accurate, which is quite good considering the technology he was using. Realistically, he couldn't have done much better with out much improved measurement technology and it was roughly two millennia before much more accurate measurements were made. Now, if we were comparing the size of the earth to the size of some other nearly equivalently sized planet using Eratosthenes' technique, we wouldn't be able to determine which planet was bigger if they were too close in size. We'd need a better measuring stick. Similarly, imagine if you were trying to compare the height of two different trees over fifty feet in height that are within inches of each other's height and all you've got to measure with is a yard stick. For all practical purposes, it would be a tie for which one is taller.

The same goes with the election in Minnesota. There were about 2.9 million votes counted for Senate, and the difference between Franken and Coleman is less than three hundred votes. Which means that there is about a .0078% difference. The vote count is subject to imprecision for many reasons, including mistakes by voters and deliberate fraud. In addition, now that the election is over there have been debates and legal actions about counting certain votes, and politicians and courts fighting over which ones are legitimate and so on. This factors make it impossible for us to know exactly what the active voters' preference was. For all practical purposes, the election is a tie.

This type of thing is inevitable so long as you have elections with relatively primitive vote-counting technology. And the more people you get voting the larger the absolute difference in voters it takes to have an election that's too close to be able to distinguish the preference of the active voters. It happened in Florida in 2000 and it will happen again. The solutions are either to simply use re-votes, or to update the technology, use more electronic technology that's going to be less subject to uncertainty, to let absentee voters vote via phone or internet, so that they know their votes will definitely get counted. Otherwise, you get a situation like you have now in Minnesota and had in Florida where the voters are not the one's choosing who gets elected.

No comments: