Elections Run Smoothly in Surprising Places by Thomas Hargrove

Reproduced under the Fair Use exception of 17 USC § 107 for noncommercial, nonprofit, and educational use.

September 15, 2004 — Despite the many election-counting flaws discovered in Florida four years ago, democracy runs smoothly in hundreds of counties and major cities.

A study of 2000 election returns by Scripps Howard News Service found that errors are least likely when state and local election officials conscientiously check for missing votes in every election.

Electoral excellence can be found in some unexpected corners of America. Few ballots are miscounted in Baltimore's gritty inner city where folks are more likely to be impoverished than college educated. Even fewer votes go missing in the bayous of Louisiana where political corruption is legendary.

Florida drew worldwide attention four years ago when 178,145 ballots were not counted in the presidential race, mostly because of poor ballot design and disputed punch-card votes. Across the country in the 2000 election, more than 1.6 million reported ballots did not register a vote for president.

But these problems are hardly universal. In the Scripps Howard study of the nation's 2,233 counties reporting complete election data, there were 469 counties where more than 99 percent of ballots cast four years ago registered a vote for president — one-third of Florida's statewide error rate.

Fourteen of these counties achieved this level of accuracy using much-criticized punch card ballots.

"That's neat!" said Ted Selker, a Massachusetts Institute of Technology elections scholar and co-director of the Caltech-MIT Voting Technology Project. "I think supervision is the whole story here. People matter."

DeForest Soaries, chairman of the new U.S. Election Assistance Commission created by Congress to solve Florida-like voting problems, said too little attention has been given to successful election administration. "Up until now, we've been compelled to recite a litany of voting failures in this country. But when you have a problem, the way to fix it is by studying success," Soaries said.

The Scripps Howard study found that many state and local election officials who delivered unusually precise ballot counts in the last presidential election say they never assume machines are infallible and never stop looking for missing votes. The lowest statewide error rate was delivered by Louisiana, where 99.4 percent of the nearly 1.8 million ballots cast in the Bayou State recorded a vote for president. Experts say results under 98 percent should be examined.

"That does not surprise me. It was not an accident," said Fox McKeithen, Louisiana's secretary of state. "We established a centralized voting procedure 50 years ago in which the state took over responsibility for voting. That was the result from some really heated elections during the Huey Long era. We had some horror stories back then."

Huey Long dominated Louisiana politics as governor and U.S. senator until his assassination in 1935. He became symbolic of Southern political corruption.

"So all parties agreed after that we needed a really reliable voting system, a single statewide system," McKeithen said. "We haven't had hanging chads here for 40 years. We look for any falloff in voting routinely, and we generate statistics to make sure there isn't something wrong with some machine somewhere."

Also producing a surprisingly low error rate was Baltimore, where 192,404 votes were counted out of 193,793 ballots cast four years ago, a so-called "undervote" of just 0.7 percent. Two-thirds of the city's population are racial or ethnic minorities and nearly a quarter live below the official poverty line.

As in Louisiana, Baltimore officials routinely compare the number of ballots cast against the number of votes counted.

"Yes, we do check that," said Baltimore Elections Director Barbara Jackson. "We know by the end of the election all of the percentages on that. And we try to get people to understand that they should vote for every office on the ballot. We don't pressure people, but we do urge them to try to vote for everyone."

Elections scholars in recent months have been effusive in their praise of Baltimore's low rate of undervoting.

"They've done this because Baltimore is a well-run city," said Henry Brady, a professor of political science at the University of California-Berkeley. "There are a lot of officials who don't care about residual vote (undervote) rates and just assume all of these people decided not to vote for anyone. But the people who have focused on residual votes have been able to lower the rate of undervoting."

More than 99.4 percent of the ballots registered a presidential vote in Seattle in 2000, one of the lowest rates of undervoting for any major urban area.

"Certainly, we review the undercount," said Seattle's elections superintendent, Bill Huennekens. "We are required by law to look at this. It is incumbent on election officials to review the overvote and undervote since it could mean that our tabulating equipment is askew or there are other anomalies."

Nationwide, about 98 percent of all ballots cast registered a vote for president four years ago, according to official election returns in 40 states that reported the number of ballots cast then. In these states, 83,214,517 ballots were cast and 81,550,810 votes for president reported. That means 1,663,707 ballots didn't record a presidential vote.

Some people chose not to vote for president, ignoring the race for the White House in favor of other races and issues further down the ballot. But experts warn the falloff rate formed many alarming patterns that cannot be dismissed as mere non-voting in the presidential race.

The Scripps Howard study found that only 97 percent of ballots cast upon all punch-card machines recorded a vote for president. But the undervote was half that for ballots cast using newer electronic "touch screen" machines or optically scanned paper ballots.

"All of our own studies indicate the most important thing in having a low error rate is to use good voting systems," said Brady. "But the second most important factor, and one that is often overlooked, is having good election administration."

The size of the undervote varied enormously from state to state, reaching 2.9 percent in Florida and 3.5 percent or more in Georgia, Illinois, South Carolina and Wyoming. Sixteen counties reported 10 percent or more of their voters went to the polls but did not register a preference for president four years ago, officially certified returns many experts say should have been challenged.

Yet the undervote was dramatically small elsewhere, dropping below 1 percent in nine other states besides Louisiana: Alaska, Connecticut, Hawaii, Iowa, Maryland, Minnesota, Nevada, Rhode Island and Vermont.

The Scripps Howard study found several commonly cited management techniques used by election officials in well-run counties and states. Among them:

• Routine county-by-county and even precinct-by-precinct analysis of the undervote, looking for unusual patterns or unexpected spikes in ballots not registering votes.

• Mandatory training, even for veteran poll workers, using "hands-on" experience in mock election booths set up weeks before Election Day.

• Unofficial "do not hire" lists of incompetent poll workers.

• Frequent mass mailings of voter guides before every major election, providing a sample ballot and complete instruction on how to use voting equipment.

Also, successful election officials often demonstrate almost brash confidence in their own competence.

"When it comes to voting, it's the people and not the equipment. To paraphrase Bill Clinton, 'It's the people, stupid!'" said Bob Balink, clerk of El Paso County, Colo. His county counted 200,757 votes out of 201,134 ballots cast for an undervote of just 0.2 percent. "If you only knew how much time we spend setting up logic routines to test our equipment. We know that every vote cast is counted, and we can prove it."

Thirty-six states had at least one county reporting a presidential vote count of 99 percent or more of ballots cast, according to the Scripps Howard study. Fifty-one of Louisiana's 64 parishes reported this level of accuracy, as did 48 counties in Iowa, 37 in Minnesota, 30 in California and 29 in Michigan.

Reach Thomas Hargrove at hargrovet@shns.com.

Doubting Thomas

Top

Dear Mr. Hargrove,

Since becoming an active member of the IEEE P1583 voting equipment standards working group three years ago I have been visiting and corresponding with the El Paso County (Colorado) Clerk and Recorder's election staff. I would like to point out up front that I have found the individuals in that office helpful, courteous, and friendly on all occasions. They run one of the cleanest and most competent elections in Colorado and I constantly consult them on election processes in attempting to help write the national voting standards IEEE is developing. However, they do have some blind spots, Diebold equipment being the principal one in my opinion, and there is always room for improvement in any operation.

In researching vote counting I came across your article above. Near the end of that article you have the following statement and quote from Bob Balink, our county clerk:

"Also, successful election officials often demonstrate almost brash confidence in their own competence.

"When it comes to voting, it's the people and not the equipment. To paraphrase Bill Clinton, 'It's the people, stupid!'" said Bob Balink, clerk of El Paso County, Colo. His county counted 200,757 votes out of 201,134 ballots cast for an undervote of just 0.2 percent. "If you only knew how much time we spend setting up logic routines to test our equipment. We know that every vote cast is counted, and we can prove it."

Unfortunately, I think Mr. Balink is making the common mistake of confusing precision with accuracy. I am not sure what election Bob is referencing in his quote but I would call your attention to the results of the August 10, 2004, primary election in El Paso County. The summary report and the canvass board statement of votes cast present the results.

A fundamental point here is that in Colorado primaries a voter can only vote for candidates running for offices in the party they are registered to vote in, although unaffiliated voters can declare a party at their precinct when they come in to vote. Thus, Republicans can only vote for Republican candidates, and the same for Democrats. Nor can one change from Democrat to Republican, or vice versa, on election day.

Thus, the errors begin on the first line of the canvass board statement. One needs to know that the US Senate race between Mike Miles and Ken Salazar is a Democratic race, which isn't stated.

Secondly, the number of registered voters in the US Senate race for precinct ...001 is given as 538. But, because this is a Democratic race, only registered Democrats can vote. The 538 figure includes Republicans, unaffiliated, Greens, and Libertarians in this precinct. A visit to Balink's election staff got me a report listing Total Registered Voters by Precinct and Party for a modest fee. That report shows precinct ...001 had 115 registered Democrats as of the August 10th election and that is the correct value for registered voters in the Democratic US Senate race for that precinct. It would also be useful for parties, poll watchers, and candidates if the percent turnout were given by party in primary elections.

Also, since the number of ballots counted is always 100% in the canvass board statement that is a useless figure, as well as being inaccurate. The summary report of election results indicates there were 120 provisional ballots counted in this race but doesn't break them down by precinct or tell us how many were cast but not counted. Should any provisional ballot have been cast in precinct ...001, but not counted, the percent counted must be less than 100%. It would be truly extraordinary if all provisional ballots cast were counted. The same logic applies to the 4,056 absentee ballots counted in the US Senate Democratic race. Almost always some absentee ballots are found invalid. But we have no record of the absentee ballots cast versus those counted.

The number of "cards cast" is thus the sum of both Republican and Democratic ballots counted and bears little discernible resemblance to the number of ballots actually cast. It also makes it impossible to discern if the basic function of a canvass board has been carried out. That is to determine that the number of ballots counted does not exceed the number of ballots cast in this or any other race on the Democratic or Republican ballots.

As a result one is unable to form any opinion as to the accuracy, or even the precision of the Democratic U.S. Senate race in precinct ...001 or any other precinct in El Paso County. The same problems carry through the entire report for every race and candidate of both parties.

You've also noted that the percent of undervotes is a useful measure of accuracy both in individual races and as a check on individual voting machines. Here again we cannot obtain any measure of election accuracy because the number of Republican and Democratic ballots cast and counted is unavailable, and not kept by the clerk's election department although the election judges for each precinct produce this number for ballots counted at a precinct. Again the problems here probably arise with absentee and provisional ballots but there is no way of knowing. Without having the total number of ballots counted for Democrats or Republicans it is impossible to calculate the undervote in any race or precinct and that simple check on election accuracy is lost.

Early in my research career I found it was quite wise not to make claims I couldn't substantiate with hard data. I know that Bob Balink and his election department are trying to do a good job but we are never in so much danger as when we become unjustifiably overconfident or brash.

The comments above are intended to help the El Paso County election department improve the way they calculate and display election data and to help them catch the inevitable errors that occur in any complex process. I think they have the necessary data and simply need to improve the formatting and include a bit more information. Doing so would improve everyone's confidence in the election results.

I hope these comments and review might also help you in presenting election issues. You are quite correct in your reporting that far too often votes go uncounted and the reporting by election officials is all too frequently inadequate, as it is here.

Charles E. Corry, Ph.D., F.G.S.A.

President, Equal Justice Foundation

Top

| Chapter 2 — Essays On Voting Problems |

| Next — Diebold Insider Alleges Company Plagued By Technical Woes by Miriam Raftery |

| Back — E-Vote Machine Certification Criticized by Bill Poovey |

Last updated 6/14/09