A story from election integrity watchdog Mark L. provides yet anotherexample the stark contrast between current election systems vendors current behavior and products, versus the kind of election transparency that’s needed to inspire trust in election results.

At issue the requirement that election systems product should track “undervotes” (the situation where a valid ballot contains no voter selection in a contest or measure) and report on the undervote rate.

The story starts with an announcement by Premier Election Systems (formerly Diebold), in which they have changed their tracking of undervotes. Going forward, the term undervote only applies to situations in which a voter selected only one candidate in a race where they could have selected more, for example, vote for 3 out of 7 candidates. These “undervotes” the Premier systems will track but not the old-fashioned kind where the voter selected no candidates in a race (now coyly renamed as “blank votes.”)

It looks like an attempt to get off the hook for true undervote tracking, a sore spot since the infamous 2006 race for Sarasota, FL’s 12^th congressional district in which the reported undervote rate in one part of the district was hugely larger than other.

After reading this news, Mark decided to look into the history of Premier/Diebold’s reporting in elections in Georgia. Choosing the 2004 election, he examined one county’s precinct-level reporting and discovered that there was no reporting undervotes (of any kind, include the so-called “blank votes), and indeed no reporting of turnout! Without that, it’s puzzling to see how the state had any "total ballots cast" figures from which toe derive state-wide undercount reporting.

In fact, looking at that reporting, it seems that turnout figures are actually smaller than the number of presidential votes recorded! There are some plausible explanations, but my point is that no one can actually take the numbers reported by the state, and get the figures to add up, even for the presidential race: how many people voted, how many ballots were cast, how many ballot seemed to omit a selection for any candidate.

What should this scene really look like? Well, we can dream (and work towards it) – all parts of an election system record everything, externalize the logs in a form easy to publish on the Internet, and be searched and crunched by anyone who wants to. Anybody who doesn’t think the votes got added up right, can not only tabulate results themselves, but also see records of every persons’ act that affected the total.

Maybe then we wouldn’t again see a situation in which someone like Mark goes looking for some statistics, and finds an election in which it’s not possible for a public-spirited number cruncher to actually get the votes to add up right.

-- EJS