Hi Chitral,
The bad machine errors should be fixable. If the machine is working properly and the cards are not mis-punched, there should be a highly accurate reading.
It's true that bad data (cards) won't work. If you have a huge stack of misshaped & mis-punched cards, they just will not go thru the machine. It should be easy to test basic hardware/software before each run with a test deck filled with mis-punches, upside-down cards etal. If the test deck is processed to expectations, you should be able to make a run for real and meet those expectations.
As for dups. It's true that cards can be read 2-at-a-time, but they would probably both be rejected (mis-alignment).
I can imagine that in any run of punch cards you would get some rejects. If the percentage is small, they can be dealt with by hand, or re-submitted. However, in my mind, accuracy means that you can account for what happened. Faith isn't really part of the equation.
I, for one, am not convinced that a hand-count of 1million or so votes will be within the error margin of a well-operating card reader. If it is at a variance of 2-5%, then what do you do???? |
