Anytime you try to measure something, there is always a margin of error. One example would be atmospheric conditions which can warp the wood of a ruler and change how accurate it is in measuring distance. Another example would be the individual differences in biology which effect how people react to medications. When you are prescribed a medication, it is given to you because the vast majority of people tested (when exhibiting certain symptoms) experienced positive affect as a result of the treatment compared to a placebo. Positive results expected, but not guaranteed.
So too is there a margin of error when it comes to actually counting votes after an election. It can’t be helped, it’s part of human nature. We are fallible. (Next time you’re in a car stopped at a railroad crossing and there’s someone else in the car with you, count the number of cars of the train along with your friend, but to yourselves. Then compare your results. It was a game played in our family when I was young and after scores of tests, my mother, father, brother and I could never agree on the number of cars of a single train.)
Society thus tries to remove human error by using machines to count the millions of votes cast in an election. Machines are great because they give predictable and repeatable results trial after trial. Unfortunately, machines also have margins of error but they are generally much smaller than what human observers can achieve.
This information is important in regards to understanding Norm Coleman’s not-even-600 vote lead over Al Franken in the Minnesota Senate race. About 2.9 million votes were cast, so a 600 vote “victory” is about a .02% difference. This could be “statistically significant” or “statistically insignificant” based on the expected margin of error of the counting process.
Good luck finding that piece of information. I’m not privy to any special insider information, and a google/wikipedia search left me with few important numbers. One actual number I did find was here:
FORT LAUDERDALE – Optical-scan machines outperformed touch screens in the general election, although both had a similar error rate, a newspaper reported Sunday.
The touch screen voting machines performed better in the Nov. 2 presidential election than they did in the March presidential preference primary, according to a South Florida Sun-Sentinel analysis. But the touch screen machines were outmatched by older voting devices that use pencil and paper ballots.
The evaluation was based on undervotes cast on each system. Undervotes are instances in which the voter skipped the presidential race or a choice was not tallied for reasons including machine and software error.
Overvotes occur when more than one choice was recorded for a single-candidate race.
Of 2.7-million votes on touch screens reviewed, 11,824 ballots had no vote registered for president, resulting in an error rate of 0.44 percent.
Of 2.3-million votes on optical-scan machines, 6,731 ballots were not recorded or flawed, leading to an error rate of 0.29 percent.
In the March primary, optical-scan machines had an error rate of 0.12 percent, and the touch screen machines had a 1.09 percent rate of undervotes, the newspaper found.
According to this test, the error rate for optical scan systems is about .10%.
This means Coleman’s number of votes is well within the margin of error for the system in place. (I admit, I could be wrong here as I’m not privy to any internal data the MN SOS might have regarding the margin of error of Minnesota’s optical scanning equipment).
We don’t know who “won” this election. Our ruler can’t measure a distance this close. If the .12% number is accurate for MN, then the error rate for this election could be as high as +or- 3000 votes. A recount might find Coleman up by over 3,000 votes, or it might find Franken up 2,500 votes.
It would take a margin of error of less than .01% for there to be anything resembling a significant confidence level (interval) for this race. The rules of this race are already in place and they appear clear as day, but if Franken wins the recount or if there are a high number of challenged ballots, I see plenty of room for creative interpretation.
Statistically, a coin flip will be just as good as a recount in a race this close. Of course, no one would put that into law.
Filed under: Minnesota, Numbers and Studies, Political | Tagged: Al Franken, Norm Coleman | 1 Comment »
Marty Andrade 