The U.S. House of Representatives dismissed the contest of 2006 Sarasota’s District 13 election with its missing 18,000 votes. As they should since GAO testing, “obtained increased assurance, but not absolute assurance that the ES&S iVotronic DREs used in Sarasota County’s 2006 general election did not contribute to the large undervote in Florida-13 contest,” (1).
Does this mean nothing’s wrong with the iVotronic system? No. The iVotronics touchscreen system is comprised of a plethora of interacting and reacting parts and pieces that tell the iVotronic what to do, alter the software, the ballot and also interpret what the iVotronic says it counted at the end of election day.
A PEB or personalized electronic ballot device activates/initializes the iVotronic with the appropriate ballot information for each and every voter. Four independent flash memory modules contain program code and ballots. The VOTE button casts the ballot and records the information into flash memory. And a compact flash card is used for updating the machines firmware/program code, and for loading sound files. This same compact flash card is used to receive voting data at the end of the day when ballots and audit information are copied from the machine’s internal flash memory to the compact flash card.
There is no indication that these components were put through rigorous systematic testing. Moreover, it is problematic that a component with capabilities to alter program code is also used to capture data at day’s end. As for the central computer vote tabulation processes and data transfer upload functions, GAO believes these work correctly within the ES&S Unity Election Management System.
GAO determined 100 trillion possible voting combinations were propagated by Sarasota’s 2006 ballot configurations of from 28 to 40 contests, as presented in various permutations. Of the 100 trillion combinations, GAO developed 224 test ballots based on their analysis of 112 common ways voters cast their ballots, (1). Though it would be impossible to test 100 trillion ballot voting patterns, concentrating on “common” configurations is much less likely to expose errors in any system. It proves only that the common “working” ballots get counted. Invariably it’s the uncommon idiosyncrasies that get us.
It would be very interesting indeed to rigorously test District 13 in combination with the statewide Attorney General’s race, the uncontested race where the iVotronic’s undervote rate was more than 5% higher than any other voting machines, even ES&S’s own optical ballot scanner.
1.5 million total ballots were cast on ES&S iVotronic touchscreens in 11 Florida counties. 137K of these electronic ballots, resulted in no votes being counted for Attorney General. An undervote rate of 8.65%. (2)
730K paper ballots were counted on ES&S optical scanners in 21 Florida counties. Of these 730K paper ballots, there were only 22K undervotes for Attorney General. An undervote rate of 3.04%. (Note: Sequoia’s touchscreen undervote rate was 3.0%. Diebold’s optical ballot scanner undervote rate was 2.72%.) (2).
Had these same voters cast paper ballots on ES&S’s optical ballot scanners, their votes might have been counted. In other words, if the 1.5 million ballots cast on ES&S iVotronic touchscreens had been paper ballots cast on ES&S ballot scanners, the undervote rate would have been more in line with 3.04% or 48K undervotes, instead of 8.65 % or 137K undervotes. 89,000 would not have been lost.
It would also be interesting to conduct a more thorough test of miscalibrated touch screens. Instead, GAO tested 10 miscalibration patterns using 39 ballots on TWO properly functioning iVotronics machines. Testing defective machines that registered high undervote counts election day might have produced more conclusive results. Especially since some iVotronics registered 39 undervotes out of a maximum of only 121 ballots per machine on election day. The median number of ballots per machine was 66. 39 undervotes out of 66 ballots cast? A comparison of undervotes by precinct, that’s all the votes tallied on all of the machines within each precinct, ranges from 0 to 41 percent.
Another area of interest is that of iVotronic interactive screen functions, such as page forward and backward. Some of us cringe recalling our own lost data using these functions in the business world of beta inventory, accounting, sales. GAO refers to usage of these functions as "voter behavior." There was no systematic testing of varying combinations of voter behavior in relation to ballot configuration. Instead GAO randomly determined which behaviors to use, such as: one-touch or two-touch method to change a ballot, varying pages forward and backward to review or make changes, casting ballots from the review screen. Rigorous, concise, targeted testing might have produced different results.
After a year’s investigation and speculation, the iVotronics are acquitted of any wrongdoing. Still GAO assures us, “Absolute assurance is impossible to achieve because we are unable to recreate the conditions of the election in which the undervote occurred. Although the test results cannot be used to provide absolute assurance, we believe that these test results, combined with the other reviews that have been conducted by the State of Florida, GAO, and others, have significantly reduced the possibility of that the iVotronic DREs were the cause of the undervote,” (1).
An inconclusive conclusion at best. If you don’t find it, ‘never mind’ it. Now, about those other missing 89,000 votes across the state of Florida….
(1) Elections: Results of GAO’s Testing of Voting Systems Used in Sarasota County in Florida’s 13th Congressional District, GAO08-425T. GAO-08-7T and GAO-07-1167T were also used for information in this article.
(2) We are indebted to Florida Fair Elections Coalition for the extensive research and data collected as presented their comprehensive, “Florida’s Vanished Votes Reports.” The reports can be found on www.floridafairelections.org and www.VotersUnite.org.