The Prague Post | Geoff Staples | December 20, 2000
I'm an election judge in Dallas County (Texas) and I have been in the computer business long enough (early 1970s) to have actually used punch cards.
So, here's my take on the Florida election: Machines are used for counting votes because they are much faster than humans, not because they are more accurate -- they aren't. The machines have a known error rate. In the case of punched-card machines, it is considered to be in the 2 to 5 percent range. For most elections, this is just fine since elections are generally decided by a margin of at least 5 percent. The 14,000 votes in Palm Beach County that were not counted by machine in a previous election were not hand counted for a simple reason: the margin of votes between the candidates was greater than 14,000 votes, so a hand count would not affect the outcome of the election.
In fact, hand counts are only done when they could change the result of the election. In Florida, Governor [George] Bush led by about 1,000 votes out of the about 6 million cast -- only 0.016 percent. In this situation, an error rate of 2 percent (assuming the best case) is well beyond the precision of the machines. This is equivalent to a poll in which the results were 49.0 percent for [U.S. Vice President Al] Gore and 49.016 percent for Bush with a 2 percent margin of error. In other words, the election is too close to call because the margin of victory needs to be much greater than it is now for the public to have any confidence in the result without a hand count.
As for the hand recount, this is a misnomer. The important ballots are not those that were counted by the machines. It is the ballots that were not counted because the machines could not read them that are important. The error rate on the ballots that the machines actually counted is quite low because the machines are designed to reject a ballot rather than count it incorrectly. In a sense, the error rate on the uncounted ballots is 100 percent. Many votes were rejected because the chad was pushed back into the hole when it went through the machine, or the chad was punctured instead of falling out. These ballots have never been counted -- only rejected by the machines.
The chads that are falling out are not ones where all four corners are still attached. The ballots have to be quite sturdy to withstand the punishment of being put through high-speed counting machines. The chads that are falling off are ones where the chad is already partially detached. I can assure you that the people who are counting these ballots by hand are treating the ballots with kid gloves, compared with the counting machines.
In fact, the first time the ballots are put through the counting machine is the time when the largest number of partially attached chads will fall off. But again, the ballots are quite sturdy. Most of the chads that fall off in the counting machines are ones that are already partially detached: the so-called hanging-door chads, in which two of the four perforations are broken through.
So, here's the bottom line: if 10,000 votes were not counted by the counting machines in one Florida county, this is 10 times the margin by which Bush is ahead. If only 20 percent of those votes can be discerned by manual counting, that could easily be enough to change the result of the election.
It is obvious that the ballots should have been hand counted in every county where a significant number of ballots were rejected by counting equipment.
One thing is for sure: A coin toss with a 50 percent probability of winning is more likely to provide an accurate result then a 0.016 percent lead resulting from a machine count.
Geoff Staples
Dallas, Texas
