Unearthing the Undervote   

Florida wasn’t the only state with egregious election errors in 2000, and it’s highly likely glitches will occur in November. Are reporters ready to spot them?

Rural Tyler County in the western hills of West Virginia lost exactly 1,000 votes belonging to Bob Dole during the 1996 presidential election. In fact, a lot of votes in a lot of places were lost in the past eight years because of inaccurate voting machines, confusing ballot designs and faulty accounting methods.

Dole was reported to have lost Tyler County, receiving only 734 votes. He actually carried the county with 1,734 votes. The mistake went unnoticed for years, even though the obviously errant tally (detectable with mathematics no more complex than division and subtraction) was publicly recorded at the county courthouse, posted on the Internet by the West Virginia secretary of state's office and filed with the Federal Election Commission in Washington, D.C. Nobody noticed because no one compared the total number of ballots cast with the total number of votes awarded to the presidential candidates.

Odds are mistakes like Tyler County's will be made in November's presidential election. Will the media notice this time around?

After the embarrassment in 2000 of calling Florida for one guy, then the other, then no one, the press will be under more scrutiny on election night than ever before. Likewise, every county clerk and secretary of state will be expected to make certain their polling places don't become the scene of Recount 2004. Reporters will need to discern voting errors within days, perhaps even hours, after the polls close. We've got to get it right this time or, at least, quickly warn if something seems wrong.

Scripps Howard News Service created a database using information purchased from Election Data Services Inc. of Washington, D.C. When I examined voting in the last two presidential races, many counties appeared to have large numbers of the electorate who did not vote for president. At the time of the 2000 presidential election, 38 states reported the number of ballots cast in each county. Of the 80.6 million ballots, 1.6 million didn't register a vote for president.

The terms used by election experts to describe these missing votes include "undercount," "ballot drop-off," "undervote" and, my personal favorite, "residual vote." While the wire service estimates that Florida had an undervote of 2.9 percent of all ballots cast in 2000, seven states reported a higher ballot drop-off than that: Georgia, Idaho, Illinois, New Jersey, North Carolina, South Carolina and Wyoming.

More than 500 counties nationwide had alarmingly high presidential undervotes--such as Bacon County, Georgia's 11 percent in 2000. After the county traded in its mechanical lever machines for an electronic voting system, the undervote dropped to just 2 percent in the 2002 U.S. Senate race.

That anomaly and many others weren't challenged by reporters or politicians at the time of the last two presidential elections. So what can journalists do in the coming weeks to prepare for November 2? Here's what election experts recommend:

• Start asking questions. Call your county elections supervisor as well as the chief elections officer in your state to make sure you will be given figures for the total number of ballots cast in every county and preferably in every precinct. (When possible, you want to know how many people were turned away because of registration issues, as well. A national poll by the Scripps Survey Research Center at Ohio University found 3 percent of U.S. voters report they've been denied ballots, most often because their names were improperly excluded from the voter rolls or a change of address was inaccurately recorded.

The Help America Vote Act of 2002, passed by Congress to fix Florida-like voting issues, has created a new wrinkle for political reporters. Any voter denied a ballot has the right to receive a provisional ballot. Local officials must then determine if the voter is qualified to cast a ballot, a process that can take several days. When faced with a close election, reporters must determine the number of uncounted provisional ballots to determine the level of uncertainty in the race.

• Get a handheld calculator. Determining the undervote is simple enough: Scan the county's canvass report to determine what race scored the highest number of votes. This usually will be for the offices of president or governor. Add the total number of votes received by the candidates (including write-ins) and divide it by the number of ballots cast. Suppose a county reports there were 1,000 ballots cast, but you notice that only 981 votes counted in the presidential race. Using a calculator to divide these two numbers gives you 0.981. Subtract the number 1 from this and you will get -0.019. That represents an undervote of 1.9 percent.

(Why is there an undervote at all? Americans have the right not to vote for candidates if they don't want to. And it's not uncommon for voters to be confused or undecided, especially in races further down the ballot. A zero undervote should be as suspicious as a double-digit one.)

• Investigate any suspicious undervotes. Kimball Brace, president of Election Data Services and a frequent expert witness in court cases involving disputed elections, advises reporters to start asking questions when the undervote exceeds 2 percent.

Reporters have not always been able to get complete election data. Twelve states did not report the number of ballots cast four years ago. They were Alabama, Arkansas, Delaware, Indiana, Maine, Mississippi, Missouri, Oklahoma, Pennsylvania, Tennessee, Texas and Wisconsin. Indiana reports that that was a one-time glitch, and the state has since provided data. At Scripps Howard's request, Delaware subsequently produced the data as well.

Most of these other states say they will report a ballot count for the first time in November. Says Texas Secretary of State Geoffrey Connor, "I don't think there will be any problem in getting this information." Connor told me that "curiously" the state's Election Night Reporting Committee, made up of members of the press and charged with ensuring the quick and accurate release of voting information, had never asked for the number of total ballots cast. "I assume [the journalists] didn't think about it," he says.

How ready are America's newsrooms to check the vote in November? Major newspapers in Florida and Georgia (two states with a history of chronic undervoting) routinely assign reporters to check for missing votes and plan to monitor the performance of new voting machines. Says Lea Donosky, politics editor for the Atlanta Journal-Constitution: "We have a reporter assigned specifically to look for problems, something we've been checking into because we have all touch-screen voting in Georgia."

And the issue will be monitored carefully in Chicago, site of the biggest single vote-counting error in the 2000 presidential election. A study by the Illinois Institute of Technology concluded that poorly manufactured plastic guide templates used to assist voters in punching out cardboard chad contributed to 122,914 ballots in Cook County not registering a vote for president.

"Obviously we will be looking at a whole spectrum on this issue, including provisional ballots and spoiled ballots," says Dan Haar, a political editor for the Chicago Sun-Times. "But how quickly will we be able to know the difference between the number of ballots cast and the votes actually counted? This is a very good question."

Editors at smaller newspapers in downstate Illinois could become overwhelmed in trying to track missing votes. The Land of Lincoln had the nation's worst undervote four years ago, with 3.9 percent of all ballots cast not showing a vote for president.

"Look, we cover 17 counties with a whole smorgasbord of vote-counting equipment – scanners, punch cards and whatnot. Are we concerned about it? Yes. How are we going to cover it? We are still discussing that," says Jerry McDowell, city editor of the Peoria Journal Star. "Questions like whether the election was fair? Was it an accurate count? I'm not sure how to approach this."