Federal Certification Test for Vote-Counting Accuracy Cannot Determine the Error Rate of the Equipment
by Ellen Theisen. March 21, 2009
We live in a world of complex computerized systems -- medical equipment such as MRIs and CAT scans; flight control software; an automotive fuel injection system; the databases that store, retrieve, and collate data. All these systems undergo stringent testing.
The primary purpose of any testing of computerized systems is to show either that the system works as designed or that it doesn’t. If the system does not work as designed, then the test results are expected to provide information on the nature, and perhaps the cause, of the defective behavior.
Using these criteria, this report evaluates one specific test -- a test designed to confirm or refute that a specific computerized vote-counting system meets the accuracy level required by federal law. The details that follow contain many numbers and many technical details, but the discussion centers on a single, simple question: did the test, which was approved by the federal agency tasked with certifying voting equipment, provide evidence that the accuracy requirement was met, or did the test provide evidence that the accuracy requirement was not met?
As the author demonstrates in this specific instance, the answer is: the test provided neither. The test, which was designed by a federally-accredited test lab for the sole purpose of testing the accuracy of the vote-counting equipment, cannot prove whether or not the system accurately counts votes, nor can the test yield any measure of the accuracy rate.
The author wants the reader to understand that such testing provides false assurance that a federally-certified vote-counting machine meets the accuracy requirements of federal law.
Read the report (pdf)