Princeton University computer scientist Andrew Appel offers a clearcut look at why a 1 percent audit of voting machines after an election isn't sufficient to catch fraudulent machines in all elections. His explanation involves marbles and beads.
In this pic, Appel explains, the 6,300 beads in the two tubes represent all of the precincts in an election for the New Jersey governor's race. Of those beads, 10 percent are blue, representing fraudulent voting machines. If you take out a 1 percent sample of the machines to audit (represented by the 63 beads on the side), Appel says the sampling is "extremely likely" to catch at least one fraudulent machine in the mix (the sample here caught 7 fraudulent machines).
But that won't hold true for the audit of a smaller election. Here, 100 marbles represent all the precincts voting in an election for city mayor. Here again the blue marbles (10 percent of the total) represent fraudulent voting machines used in the election. But take a 1 percent sample of these machines (represented by one marble) and Appel shows that it's unlikely the sample will include any fraudulent machines. In sum, he says, while a 1 percent audit "works well for statewide races, it does not suffice for local or legislative-district elections." (Photos: Alex Halderman)
Read Appel's report about what constitutes an effective audit here.