Most people consider all overvotes to be unintentional, although I imagine there is an extremely small percentage of the population who casts an overvote as a symbolic gesture.  The percentage of undervotes that are unintentional is certainly up to debate, although everyone wants to eliminate or reduce the chances of unintentional undervotes.

After the 2000 election, Congress passed the Help America Vote Act.  That law required states to take steps to reduce voter errors.  Two options were given to the states.  One was to purchase equipment that notifies voters of errors on their ballot prior to the casting of their vote.  The other was to engage in voter education campaigns that provide voters with information about how to vote and how to correct errors in their ballot.  Most voters in America operate under the first option and are using equipment that notifies them when they cast too many votes.  As I noted in an earlier post, only New York has a statutory provision similar to the Illinois provision of notifying voters of undervotes.

Because new election equipment has changed the dynamics of casting a vote, a one size fits all analysis of residual votes just isn’t sound.  In her book, the Democracy Index, Heather Gerken applauds the residual vote rate for its “elegant simplicity”.  This is a simplistic appraisal that I don’t think can be supported in the post punch card era.  Thereafter she states “we’d expect the number of people who don’t vote for a presidential candidate to be roughly constant.”  Our research in Champaign County and my intuition from decades of political activity suggest the opposite.  For example, to learn that the undervote for Governor in 2006 in Champaign County was highest in Penfield makes perfect sense based on the demographics in that precinct.  In examining the ballots it is apparent that the undervotes for the Governor’s race were generally Democrats who voted for their party candidates down the ticket, but
refused to cast a vote for the Democratic Governor who was later to be indicted and impeached.  Under the “roughly constant” standard set out in the Democracy Index, we’d be examining our voter education efforts in Penfield and examining the equipment.
So does the residual vote rate have any value?  I believe it does, but its value is far less than it was in examining a punch card system.

For optical scan systems, the value is minimal.  There is scant evidence that a voter looking at a typical optical scan ballot will have difficulty in determining what races he or she is to vote on.

In fact, relying too much on the residual vote rate may mask real problems or at the very least, give election officials a crutch to avoid positive changes.  For example, in the Better Ballots analysis from the Brennan Center, they make the argument that Douglas County, Illinois’s optical scan ballots are better than Franklin County, Illinois’s optical scan ballots based on the residual vote rates for the Attorney General and Secretary of State races in 2002.  Those rates were 3.6% vs. 3.1% for the Attorney General race and 4.0% vs. 3.0% for the Secretary of State race.  In Champaign County, our residual vote for the Secretary of State race was just 2.6% using punch card.  Would the suggestion then be that punch cards are better than optical scan ballots?  As proud as I am of the work of our office in getting reliable, accurate, and complete counts of punch card ballots, even I wouldn’t suggest that.  Clearly other factors lead to the differences in residual vote rates and the rates themselves are of more interest to political scientists analyzing voting trends than to election administrators analyzing the efficacy of their equipment and ballot design.

Of course, as I noted in another post, we have incorporated many of the Better Ballot recommendations in our ballot design, not because our residual vote numbers were too high, but because our experience and intuition convinced us that the changes were positive.   Similarly, we have incorporated changes in our ballots based on voter suggestions and complaints.  Even with an excellent residual vote rate in the last election of under .5% for President, we have at least one change we are planning to incorporate for the ballots next year.

I am less familiar with touch screen systems.  My impression is that the residual vote rates here also have less value than might be implied in some of the research.  The “roughly constant” standard is just not valid in any jurisdiction.  At the same time, Better Ballots points out common sense changes that intuitively look positive and which probably need to be implemented.  Additionally, if residual vote rates can demonstrate that machines are not properly calibrated, that would be good.

Finally, for jurisdictions that rely on voter education for reducing voter errors, the residual vote rates probably need further examination.  Many of these jurisdictions are using hand counted paper ballots where they can ascertain voter intent, or they have optical scan ballots and actually have mandatory hand counts of ballots that are under or over voted.  Whether residual vote rates are of value to those jurisdictions I’ll leave to people with more experience.

The residual vote rate was a wonderful tool to point out the flaws of outdated election equipment and the need to have ballots that are clear and understandable.  But its purpose has largely been served and using it to analyze the work of election officials is unfair and unsound.

First, and of greatest importance to Champaign County voters, we learned that our machines are doing an extraordinary job of counting votes, even if voters are not properly marking their ballots.  As most voters are aware, an optical scan ballot in Champaign County should be voted by filling in the oval next to the person you want to vote for.  That particular instruction is given to voters by the election judges when they receive their ballot.  The instruction appears on the ballot itself.  It appears in the voting booth.  It appears on the privacy cover given to each voter when they vote.  It even appears on the pen that is given to voters to mark their ballot.  Unfortunately, a small percentage of voters do not follow those instructions.  Nevertheless, we found that almost all votes that are cast with an X or check mark still properly counted.

Second, regarding voter notification of undervotes on the above noted races, we found that notification would have almost no impact in preventing uncounted votes.

Our method was to examine each ballot and find those where there was an undervote.  We then compared the counts of the undervoted ballots from our examination with the undervote numbers from the final canvass.

In Champaign County, there were 53,869 ballots cast.  Of those, 3,749 (7%) had undervotes for one or more of the statewide constitutional officers.  The total number of undervotes was 6,729.  Our analysis shows that 6,716 of those undervotes were intentional, meaning the computers correctly determined voter intent in all but 13 circumstances, a rate of 99.81% certainty when taken against all the identified undervotes.  When taken against all ballots, the certainty is 99.9952%.

You can see the precinct by precinct results including the intentionally undervoted ballots here.

Precinct by Precinct results.

Put another way, if the notification of undervotes had been in effect in 2006 General Election, 3,749 people would have received a message on the vote tabulator that they had undervoted in a race for which they had no intention of casting a vote.  As discussed before, it’s all but certain that a high percentage of those people would have the secrecy of their ballot compromised.

No more than 13 people would have received an error message that would have allowed them to correct their ballot to make sure that their vote counted correctly.

Perhaps most important, there were 46 undervoted ballots cast using the Automark Voter Assistance Terminal.  For each of these ballots, the voter received a notification during the Automark voting process each time they decided to leave a race blank as well as a summary notification at the end of their Automark voting process.  Once their ballot printed out, they then inserted the ballot into the vote tabulator.  Each of these people would have received a visual message notifying them of their undervote.  In the case of those who are visually impaired or unable to manipulate the buttons on the machine due to a disability, they by necessity lose their right to a secret ballot.  Interestingly, the undervote rate on the voter assistance terminal was higher than the rate for all voters. (46 out of 483 identified Automark ballots.)

You can review the Automark Voter Assistance Terminal ballots here.

In addition to undervotes, we also looked at overvotes.  Except for those people who vote absentee through mail, overvotes result in a notification to the voter and a chance to correct the ballot.  There were 58 overvotes that we found where the voter decided not to correct the error. In 34 of those, voter intent could be clearly determined.  In 24 of those ballots, no clearly discernible voter intent could be determined.

In our examination we ran across a number of other interesting items.  In most precincts we were able to easily identify the undervotes.  By deduction we were then able to determine that the ballots in those precincts with X’s or check marks had been counted by the machine.  We also found eight ballots where voters overvoted because they wrote in the same name that they filled in the oval for.  We also found a number of ballots where it appears that the voter filled in the oval after making check marks or X’s to start with.  Some of these voters may have received an unreadable mark or blank ballot message from the vote tabulator.

X’s and Check Marks That Were Fully Counted
Voted and Wrote-in the Same
Filled in Oval after Using X’s

But this ballot is the poster child for the absurdity of this new law.  After leaving all but the advisory referenda blank, the voter wrote:

Today, Tuesday, November 7, 2006 I am exercising my right to vote by not voting for anyone.
Personally, I am appalled as all the candidates should be ashamed, at how childish they have all been acting.  I don’t want any on this ballot representing me or the State of Illinois.

This year, the Illinois Association of County Clerks and Recorders put the elimination of this provision high on their legislative priorities.  I helped craft a one page summary of arguments for removing this provision.  Unfortunately, the legislative leadership in Springfield doesn’t see this provision as a problem.

I am now putting together the information necessary to demonstrate that this provision violates U.S. law, the Illinois Constitution, and Illinois law.  One piece of that information is the law regarding notification of undervotes in each of the other 49 states. This summary demonstrates that the legal requirement to notify voters of an undervoted optical scan ballot exists in only one other state, New York.  Like Illinois, New York has yet to conduct an election using this undervote provision.

Our right to a secret ballot is fundamental to a democratic process that is open and free from coercion.  I’ll continue to post information as it is gathered and we now have a page on the website devoted to this issue.

In the good news/bad news category, the County sent out their property tax bills today.  So expect it in your mailbox tomorrow or Monday.  The revenue for the County General Corp fund from property taxes is about $6.8 million.   That will allow the County to essentially tread water for the next few months.

Champaign County is one of the few counties in the state to get their tax bills out on time.  It’s a combined and highly cooperative effort between township assessors, the Supervisor of Assessments, our office, the County Treasurer, and the County’s IT department that makes that happen.

With those caveats, I can tell you that my observation is that the hand count of ballots in Bondville was the same as the computer count reflected in the final canvass.  WDWS was here and interviewed Karl Kennicker.  You can listen to them to find out what he is going to do from here.

Champaign County will have two discovery recounts this year.  One is for the Bondville Village President’s race (April 30 at 1:00) which was decided by a single vote.   The other is for the Champaign School Board race (May 1 at 8:30) which was decided by two votes.   We have had a number of discovery recounts since I’ve been County Clerk.  None has revealed any discrepancies between the hand count and machine count of ballots.  Two of those discovery recounts occurred since we implemented the new optical scan voting equipment.

Any results of the discovery recount will not change the results of the election.  Rather, the discovery recount is just a gathering of information that a candidate can bring to a judge to request an election contest be conducted.  An election contest, if granted by a judge, would normally result in a full recount of all ballots in the race.

Our retabulation of votes showed just two ballots where voters used an “X” or other mark instead of filling in the oval.  That is an error rate of under .2%.  Even then, there is a significant chance that the machine will count those ballots that a voter marks incorrectly.  While the error rate is small, it obviously could have an impact in either of these contests.

Personally, I would like Illinois law to mandate a redundant recount in races this close.  I know the machines count accurately and I know that almost all voters mark their ballots correctly.  But in races this close, it would be good to examine each ballot to make sure that the actual will of the voters has been expressed in the final result.

The Green and Republican Parties did not send representatives to the retabulation.  As such, members of my staff selected the races to be hand counted for them.

The results came out largely the same, except for a single vote discrepancy in Champaign 5.  You can see the two ballots in that precinct that were oddly marked.

As in the past, the hand count of ballots demonstrates that our voting machines are counting accurately.

And an examination of the ballots shows that voters are doing a great job of following the instructions and filling in the oval.    Out of 1,108 ballots cast in these precincts, just two marked their ballots with an X or check mark.

Early and absentee voting will commence as normal tomorrow (complete with error detection for your ballot) and I have every confidence that the election will proceed without a hitch.