Chicago Public Radio has a fun series of reports titled Curious City. The premise of the series is that listeners submit Chicago-related questions (e.g., what’s with 16 inch softballs?) and they go out to find the answer. Some of their questions have an operational bent, although “how do they clean The Bean?” or “how do the reversible express lanes on the Kennedy work?” have seemed a little too Chicago-centric to write about here.
But this past week they hit on a question that has relevance across the country: What are elections authorities doing to protect voting?
Of course, voting security is particularly intriguing in Chicago — a city famous for election day shenanigans. Most of the counter-measures for fraud focus on carefully guarding machines and making sure no one messes with the equipment. Here’s what struck me as interesting:
Even Johnston acknowledges that Illinois does have one important safeguard in place that makes it harder to tamper with votes: The state requires a paper record of every vote — a paper record that voters can verify as they cast their ballots. In addition, state officials randomly pick 5 percent of the precincts in each county for a recount after every election. During these recounts, county officials make sure the paper ballots match the totals tallied by the machines.
Cook County has over 1,600 in the suburbs around Chicago. (Brief detour into local political administration: Chicago is part of Cook County but has its own election system while the county runs the elections for the surrounding towns.) So is recounting 5% of precincts enough?
The first thing to note is that checking 5% of precincts is not the same as checking 5% of ballots since turnout can vary across precincts. Further, early voting can complicate things since the county offers are fewer early voting locations than precincts. Total turnout in a precinct might be very high but relatively light on election day because many voted early at a different location.
Still if the concern is that someone has corrupted a voting machine, checking 5% of precincts may come close to checking 5% of machines. Each voting location will have tens of machines so this will get to be a big job pretty fast. Also, one really needs to check all the votes because a smart way to manipulate a machine would be to switch a limited number of votes to minimize the chance that the voter would notice and to keep the machine from standing out as anomaly. For example, not even in Cook County will everyone vote for Obama. Clearly, if every voter on a particular machine voted that way, someone would wonder.
What I can’t get my head around here is what the statistical test would like. These are machines and they are going to be prone to some kind of error but not all errors are created equal. Having the paper get stuck so the vote on a judicial retention case is illegible and hence unverifiable against the electronic record is very different from the paper saying Romney while the electronic record says Obama. The audio report says that the records agree nearly 100% of the time. I hope that means that the paper is occasionally unreadable. That may be tolerable. Flipping votes, not so much.