Elections Technology: Recording and Tabulating our Votes Correctly

In brief:  Electronically tabulated election results will occasionally be incorrect despite the best efforts of even the most highly skilled local elections clerk. 

Wisconsin state law provides for a 'canvass' period of a few weeks following each election during which local election officials are to review election results for accuracy. If they find any errors during the canvass, they have authority to correct them. 

Verification of voting-machine output during the canvass period could protect our final election results from the inevitable occasional electronic miscount, and would likely deter all electronic fraud.  

* * *

To understand how any election technology works, we first need to know how it records votes, and then how it counts them.

Wisconsin uses three types of elections technology, depending on the local jurisdiction. Each type of machine has a unique set of advantages and risks.

  1. Opscans (optical scan) count, but do not record, votes. Voters record their own votes on paper ballots, and insert them into the voting machine to be counted;
  2. Automarks record, but do not count. These are are touch-screen machines that enable voters with disabilities to mark ballots which, after being printed, will be inserted into an opscan for counting; and
  3. DREs (direct-recording electronic machines) both record and count. With DRE machines, voters indicate their choices with a touchscreen, and the computer prints a paper ballot that stays inside the machine, although voters have the option to inspect it if they wish.

1. RECORDING our votes

In jurisdictions that use opscans and Automarks, voters can make sure their votes are accurately recorded simply by following instructions carefully and inspecting their ballot closely before they insert it into the opscan machine.

MarkingABallot.jpgIn jurisdictions that use DREs, voters do not mark ballots by hand, but rely on the voting machines to create a "voter-verifiable paper trail" that contains a printed record of every ballot cast.

When it's printed correctly, this record can be used to check the accuracy of the electronically tabulated totals. However, with DREs, certain malfunctions (whether accidental or deliberate) can eliminate any true record of the voter's intent. 

Poll workers must be vigilant to ensure the DRE paper trail is created. During Election Day the paper might jam or tear. It might print illegibly or not at all if the wrong type of paper is used, if the paper was loaded incorrectly, or if the printing mechanism malfunctions in some other way. Voters cannot be counted on to report these problems to poll workers promptly. The machines retain an electronic 'cast vote record,' which allows a paper trail to be reprinted if the paper trail was not printing on Election Day, but the reprinted trail is not a valid record for recount or audit, because the voters had no chance to verify it.

Voters must be vigilant when using DREs, although most are not. Displaying the vote on the screen, counting a vote in the internal tally, and printing a vote on the paper trail are three separate processes. They can be hacked so that they do not all show the same candidate. That is, for every fifth voter for Candidate A, the machine could display a vote for Candidate A on the screen, count a vote for Candidate B, and print a paper trail that shows a vote for Candidate B. The common assumption is that the voters would notice the misprinted audit trail, report it to the poll workers, and the machine would be shut down until the problem is corrected.

However, studies have found that if the paper trail shows incorrect votes, chances are small that the incorrect votes will be noticed. And of the few voters who do notice, most will simply correct their vote and leave the polling place without saying anything to the poll workers. So poll workers cannot count on being able to notice if a DRE has been hacked to flip votes.

Poll workers need continuously to remind voters to check the paper trail before casting their votes, and to take any reports of errors seriously. 

Because of these additional risks with DREs, the Wisconsin Election Integrity Action recommends that jurisdictions purchase only opscan systems for use with voter-marked paper ballots.

2. COUNTING our votes.

Many smaller Wisconsin jurisdictions still rely on hand counts, which routinely include a validation count. Two people simultaneously count votes from separate stacks of ballots, then switch stacks and count again. When each has counted both stacks, they compare their results. If their counts agree, they move on to two more stacks. If they disagree, they must resolve the discrepancy before moving on.  When votes are hand-counted, even preliminary election results are not reported until the totals have been completely double-checked in this way.

When election results are tabulated electronically, our votes are counted only once--by software written in proprietary secrecy by private vendors in other states or by whomever hacked in behind them. 

Wisconsin's local elections officials do a reliably good job of making sure the voting machines counted the correct number of ballots, but if the computer miscounts the votes, the flawed output will likely not be noticed or corrected unless the miscount is dramatic.

Sources of electronic miscounts

The causes of voting-machine miscounts are the same as in other computers: 1) Malfunction; 2) Mistakes; and 3) Malicious interference.

Malfunction. Anyone who uses a computer knows that computers can do unexpected things because of power surges or outages, physical damage, or no apparent reason. Around the nation, voting machines have malfunctioned on Election Day without voters or poll workers noticing. In one case in New York, optical scanners overheated mid-day and lost calibration, but kept counting ballots without counting any votes. Thousands of voters were disenfranchised, and the miscount was not discovered until long after election results had been declared final.

Mistakes. More than 600 voters in Medford, Wisconsin were disenfranchised in November 2004 when their voting-machine vendor neglected to program the machines to read straight-party ticket votes. The problem was discovered in March 2005, when a political party consultant compiling data on registered voters noticed that the number of voters who cast ballots was much higher than the number of votes counted.

Malicious interference. Wisconsin's county and municipal clerks might maintain the tightest security they can, but the elections software security is not fully in their control. Private out-of-state corporations write and maintain the software that counts our votes. Wisconsin's election clerks can do little more than trust that the vendors' employees were all honest and competent, and that the vendors' own security systems successfully resisted hackers. 

3. VERIFYING our election results. 

Although even the most skilled and conscientious elections clerk cannot prevent every electronic miscount, we don't have to worry about flawed preliminary output from determining the outcomes of our elections--IF we routinely audit the machines' Election-Day accuracy before declaring results final.

The results reported on Election Night are only preliminary. Our local elections officials cannot declare the election results final ('certify' them) until they have waited for late-arriving absentee ballots and reviewed the records of the election to make sure everything is in order. Before they certify  the election results, they could verify the accuracy of the electronically counted results, and can correct any errors they discover. 

National elections-administration authorities recommend a verification method known as 'risk-limiting audits'. (Description here, with links to Presidential Commission & American Statistical Association)


More information:

How IT experts tell us voting machines could be hacked, or compromised by corrupt insiders 

How local officials could verify voting-machine output