HANGING WITH THE CHADS
Our Man at the Great Florida Recount

BY RUSS BAKER

Do the media really need much justification for examining -- at this late date -- the ballots that were cast but not counted during the achingly close 2000 presidential contest in all-important Florida? The postmortem obviously won't affect the outcome. Yet with confidence in the electoral system undermined at home and America's reputation as a paragon of democratic voting damaged abroad, an honest accounting of the public's will may be the right medicine. And no price for starting our national therapy seems too high.

Actually, the price can start surprisingly low, as I found out during a visit to Florida in early February. For under five bucks per hour, for example, I could have purchased from Hillsborough County (Tampa and vicinity) the right to sit at a table and watch as county employees held one ballot after another over a light box, and scrutinized dimpled, hanging, and virginal chads to my heart's content. Instead, being a nonpaying observer, I was made to peer from behind a floor-to-ceiling glass partition as an electoral autopsy, with all the visual appeal of a TV test pattern, unfolded.

Yet exciting or not, something historic was taking place. Ever since December, when the Supreme Court halted a pending recount and gave George W. Bush a 537-vote Florida margin and the presidency, news operations have been fanning out across Florida to examine the record. It's hard to recall when the media have championed the public's right to know with quite this mix of breadth and consequence.

The ballot recount has prompted patient curiosity from the public and understandable wariness from the GOP. Republican operatives, in fact, were riding shotgun on the media recount efforts all across the state. As Governor Jeb Bush told the Orlando Sentinel: "There's been a declared victor . . . . The election's over. So go ahead and do it, but is that going to rewrite history? I don't think so. Should it rewrite history? No. We're a nation of laws, and the rule of law prevails."

DEFENDING THE EFFORT

Participants are quick to defend the operation. "Journalists have done this since the beginning of time," said Bill Rose, deputy managing editor at The Palm Beach Post, an early vote-count participant. "There's a story, it breaks, there are lots of questions. Do you take people's word, or do you examine the documents? There's no doubt the election is over. There's no doubt George Bush won the election in the minds of the people who determine that. But there are doubts in the minds of voters."

The Republicans probably need not worry. If the current exercise reeks of anything, it is the perfumed air of propriety and moderation. None of the various ballot inspection operations will dare to pronounce a "real" winner in Florida. Even if it looks abundantly clear that Al Gore received more votes than George Bush, these news outfits will not be making such a conclusive statement. Instead, the idea is to create -- for the public, for posterity, for those interested in reforming voting system accuracy, for academics and specialists -- a permanent record of what happened in the polling booths. The participants don't call this a "recount," but the building of a database, an effort to "describe" each uncounted ballot as precisely as possible, categorizing them by such factors as overvote or undervote, one-corner-detached chad or dimpled chad with light coming through. Once the totals for each category are in, they say, people can draw their own conclusions. They can, if they wish, take into account the relative merits and liabilities of different types of punch card machines with names like Accu-Vote, Votomatic, and Data-Punch, and of various optical scanning technologies, and ponder the efficacy of Martin County's old-fashioned lever-pull balloting and Union County's good old manually counted paper ballots.

DUELING INSPECTIONS

Some might call this a futile exercise, but the venture nonetheless made enough sense to media organizations that more than a dozen of them, in concert or individually, opted in. Luckily for this purpose, the election debacle unfolded in the Sunshine State, which has one of the country's most liberal sunshine laws, enacted in 1967 and inscribed in the Constitution in 1992 after a Florida Supreme Court decision threatened the law.

The first organizations into the water were, not surprisingly, Florida-based. In November, within days of the election, The Miami Herald began making public-records requests for access to all "undervote" ballots throughout the state -- ballots that were recorded as showing no choice in the presidential race. The Palm Beach Post, meanwhile, launched its own limited effort in nearby counties. The Tribune Company paper, the Orlando Sentinel, undertook its own selective forays. (In addition, the conservative group Judicial Watch began a statewide examination.) The state was a patchwork of media legal challenges, with Herald attorneys, in particular, bouncing from Sarasota to Dixie to Duval counties, seeking injunctive relief to force reluctant county officials to comply with their request.

The Sentinel quickly published its first findings on differences in optical scans, on December 19, seven days after the U.S. Supreme Court halted all recounts. Although twenty-six counties scan ballots right at polling stations and allow voters to immediately correct any errors, fifteen others send the ballots to a central county office where they are tabulated later, preventing voter adjustment of unintended selections. In little Lake County, west of Orlando, the Sentinel found that more than 600 ballots had been disqualified not because of ambivalence but over-enthusiasm: voters had marked the appropriate oval for their presidential choice and registered a second preference for the same person as a write-in. The scanners, unable to distinguish between two votes for competing candidates and double votes for the same candidate, negated the ballots. Conclusion: if the double-entry ballots were counted, in Lake County Gore would have gained 130 votes. In all, the Sentinel found what it calls a "possible" net gain for Gore of 366 votes in fifteen small counties, plus 203 in its Orange County, and an additional thirteen in Seminole. Sean Holton, special projects editor, emphasizes that these are "probable," not definite. Still, by my math, that is 582 probable votes, enough to turn the election.

When The Palm Beach Post jumped into the ballot-examination business it started not close to home -- despite Palm Beach County's globally televised ballot quandaries -- but instead rushed two counties south to Miami-Dade, snatching a story from under the Herald's nose. Fifteen days after the election, Miami-Dade had halted its recount -- from which the Gore campaign had expected to pick up as many as 600 votes -- when canvassing board members were besieged by a noisy demonstration of GOP staffers flown in from Washington. "We decided, heck, we gotta know what happened there," said Bill Rose. Post reporters watched as Miami-Dade workers flipped through more than 10,000 ballots that had been rejected by machines. The Post's story, which ran January 14, reported the astonishing conclusion that a full Miami-Dade count would have shifted by just six votes -- and to Bush, not Gore. And that was with a generous standard preferred by the Gore people, counting anything at all with the slightest indication of preference: partially detached chads, even simple dimpled chads. Two weeks later, The Post ran a story about its own county, where, applying the same standard, it found that Gore would have picked up 682 votes -- enough to reverse the election outcome. Even so, Rose, like all the editors, was reluctant to characterize the findings, noting that many other factors, including whether to accept improperly postmarked or late absentee ballots, could have played a role. "This is an If story," he said. "It explains why the Democrats argued so vociferously to count dimpled ballots."

ENTER THE DRAGON

It wasn't long before leading non-Florida news organizations decided to jump in. Representatives of The Washington Post/Newsweek, The New York Times, The Wall Street Journal, CNN, The Associated Press and other organizations soon joined forces, agreeing during a series of conference calls and meetings to finance an examination with the heft and scientific rigor one might expect from such outfits. After initially sitting in on counting sessions where the Herald was also present, the consortium went its own way. It hired a well-respected not-for-profit organization, the University of Chicago-affiliated National Opinion Research Center (NORC.) The consortium came to include news organizations that had already conducted ballot inspections, including Cox's Palm Beach Post, the St. Petersburg Times, and the Tribune Company's South Florida Sun-Sentinel and Orlando Sentinel (Tribune's Los Angeles Times and Chicago Tribune would also join).

The 900-pound gorilla of Florida journalism, The Miami Herald, also participated in initial discussions with the consortium, but opted not to join. The Herald, with resources from its parent company, Knight Ridder, had already been moving forward by the time the consortium held its first formal conference call. The Herald had rushed to make public-records requests in all counties (an action that would benefit the consortium as well). To handle the ballot inspections, the Herald initially approached the Big Five accounting firms -- and was uniformly rejected for what the bean-counters deemed an unnecessarily "controversial" activity. A second-tier but still sizable accounting firm, BDO Seidman, did agree to take on the job.

Tongues soon started clucking about why the Herald had declined to join up with the others, an act that would have created unanimity and hence added credibility to the consortium effort. But the Herald was dedicated to controlling its own project, and it wasn't about to share information with its main competitor. The Herald told consortium members that it would not participate if the Sun-Sentinel, an aggressive turf challenger based in neighboring Broward county, was included.

"We're highly competitive with them," explains the Herald's executive editor, Martin Baron. "The consortium came to us and asked us what are the terms under which there could be a partnership. We gave them an honest response: We didn't feel we could be partners with everyone in the consortium." (The Herald was ultimately joined in its survey by USA Today.)

KAUS AGONISTES

Soon, though, the Herald faced more criticism. In January, Slate published "The Miami Herald Blows Its Pulitzer." Writer Mickey Kaus chided the paper for not counting overvotes, declaring "any recount that doesn't include the overvotes is an incomplete recount." Kaus had pointed out that although the Herald's count will certainly address the "real world" recount scenario, it won't answer the important question of who really won the election. Kaus quoted The Wall Street Journal's Washington bureau chief, Alan Murray, suggesting that the Herald took the "cheap" way out in counting only the approximately 60,000 undervotes rather than all of Florida's 180,000 disputed ballots, including overvotes. The Miami paper, the argument went, was more interested in a scoop than in getting the whole story.

This infuriated Baron, who said the Herald's point was to simulate the original, court-mandated recounts (they revolved around undervotes only, which is all that Team Gore had asked for) and not to predict what might have happened had overvotes been included. And he saw a deeper problem with the criticism: "Since when do people in the media think there should be only one source of information? It's almost unprecedented. I'm not aware of another example where media organizations think it's inappropriate for media organizations to embark on their own projects." Baron argued that the Herald, which won a Pulitzer in 1999 for investigating vote fraud in Miami, was once again out front, publishing a raft of investigative pieces on everything from poor-quality voting machines to evidence that more than 2,000 Floridians voted illegally in November's election.

Still, the Herald found it increasingly necessary to consider the overvote, and, three days after telling cjr that the paper was considering "looking" at those ballots, it ran its own "analysis" on overvotes, supported by a study by a professor from the University of California. The professor found that "as many as 1,700" voters in Miami-Dade invalidated their presidential ballots because they mistakenly punched the chad below the one corresponding to their preferred candidate.

By February, when I visited the first formal day of the consortium's recount, the Herald was far ahead, having already polished off most of the state's undervote. The consortium, though rushing late out of the gates and examining a far larger number of ballots, came with superior financial resources and numerical strength, as well as NORC's vaunted technical expertise.

While the Herald's work involved one accountant and one reporter in each county, NORC would have multiple teams at a site, each consisting of three tabulators, or "coders," plus a supervisor. Dan Keating, The Washington Post's database editor, said that having three people look at the ballots gives the review more statistical validity. "We wanted to create the definitive data source," he said. Requiring more than one person to examine each ballot will also help illuminate the degree of subjectivity involved in manual recounts, Keating notes. "The people at NORC are excited because they will do their own analysis on inter-coder reliability, to see whether two people looking at the same thing see the same thing."

But they were very privately excited. When cjr asked Kirk Wolter, NORC's senior vice president of statistics and methodology, where the consortium planned to be conducting ballot examinations in coming days so that the magazine might drop by, he declined to say. The Washington Post's Keating, similarly echoing concerns of "a media circus," noted that cjr couldn't be prevented from canvassing the individual counties to learn where the consortium might be appearing next (a daunting task, given Florida's sixty-seven counties) before grudgingly letting slip one location -- Hillsborough County -- out of six where NORC would be working that day.

When I arrived at the Hillsborough elections center, in an industrial park surrounded by farmland, the NORC officials did not seem particularly glad to see me, even though the feared media circus turned out to be one man with a pad. From my notes: NORC deployed four teams to examine the punch-card ballots, which had ballot position numbers but no candidate names. The teams assigned codes in categories according to evidence of voter intent (as shown in the box on the facing page.)

NORC'S shyness was more understandable later, when, after reading NORC's press release about "producing the definitive archive of the disputed ballots" and hearing them characterize their teams as highly skilled and trained operatives, I learned that many of their coders were bored-looking locals hired through a temp agency. Although the Sun-Sentinel's Ward had described the methodologies as "mind-boggling," and "quite complete," what I saw amounted to something less than that. I was not permitted into the room where the ballot examination took place, observing it instead from several feet away behind glass, but I could see that although the examiners appeared serious and attentive, they seemed prone to the kinds of mistakes we had come to expect from the original recounts. Several were elderly, and one in particular seemed to have problems with her vision, standing up from her chair each time to lean in for a proper look at the ballots.

In Pinellas County, meanwhile, one of the NORC coders was not up to counting speed. Tina Harris, a GOP volunteer, claims the coder showed up late, reeked of drink, and had trouble functioning. Julie Antelman, a spokesperson for NORC, says she tracked down this GOP report and found that the coder "was not intoxicated." She does say that he was let go "because his job performance was not up to the standard." CJR spoke with four people who worked near the coder, including two county officials. None noticed evidence of drink. One coder said GOP operatives tried to get her to say otherwise.

Rounding out the imperfection of the process, I learned from Hillsborough County staffers (and a GOP observer) that two different NORC teams had accidentally examined the same precinct -- counter to the NORC methodology.

Notwithstanding all this, NORC will end up with three opinions of every single ballot. Presumably, in the vast majority of cases, at least two of these will agree, and will be a fairly good reflection of what the ballot actually looked like. Each consortium member will then produce its own stories from the data.

HARK THE HERALD

The next day, in Brevard County, on Florida's Atlantic coast not far from Cape Canaveral, I attended a Herald count. Here, the atmosphere felt more convivial and relaxed -- but no less rigorous. Michael Galy, a gray-haired CPA from BDO Seidman's Miami office, sat next to a Herald reporter, Phil Long. Four Brevard elections officials, all quite veteran and including Fred P. Galey, the supervisor of elections, sat through the count. While Two GOP observers watched from a distance, I stood directly behind the Herald people.

Long, a thirty-two-year Herald reporter, exhibited a preternaturally keen interest in every piece of paper, even taking frequent if brief opportunities to quiz county officials about what he saw on the optical scan ballots. He had good reason for caution: he didn't even know for certain whether the putative undervote ballot he was examining had or had not been counted in the first place, since undervotes had not been separated from other ballots during the original election tabulation. At the Herald's request (and on the paper's dime), Brevard County clerks had had to examine 220,000 of them by hand, trying to find the 277 ballots that computer records showed had not registered a presidential choice. This was something of a guessing game. If county officials knew that a precinct showed six undervotes, they could do their best to find six ballots that appeared to be undervotes, but they would not be absolutely certain they had the right ones. Indeed, in some precincts, they came up one or two ballots short.

Beyond noting which candidate the voter had appeared to prefer, the Herald people were trying to figure out, by looking at the entire ballot, whether questionable or stray marks in the presidential section were part of a clearly intended vote. Their tabulation sheets offered the following choices: "No mark; Underlined candidate; Circled candidate; Circled Bubble; Marked X, Wrong pencil; Other." One column read: "Errors in other races: Mark Yes or No" and left room for additional comments.

Here's a typical couple of minutes looking at ballots where voters did not, as instructed, completely color in the oval "bubble" by their candidate of choice: a ballot surfaced with an "X" in the oval for BUSH. Then a ballot with a check mark in the oval for GORE. Then one where the voter had written in "H Beckman" -- perhaps the voter's own name. Next, a voter had circled the GORE oval. "Look, there again, the only error is in the presidential race," Long commented. "They've done it right in the rest of them." Another ballot had a write-in for "J Lennon." "Heh heh," said Long, making a quick notation. "All you need is love," deadpanned Galy, the accountant. Next ballot.

The Herald's effort, on the surface, might have seemed less "scientific" than the consortium's, with Long potentially polluting the process by talking with officials and interacting with the accountant. However, whenever he did so, he seemed only to increase the probability that he was describing the ballots with an understanding of how the county's machines actually work.

After the day's counting was done, Long described himself as thrilled to be part of the effort, despite day after day of what seemed grueling and numbing work. "I've been involved with a lot of big stories," he said. "But this is an important one."

WHAT CAN WE CONCLUDE?

When the Herald announces its findings (probably in March) and the consortium completes its work and lifts its news embargo (most likely in April), the findings may chiefly excite afficionados: professors, probability freaks, C-SPAN buffs, party activists, all of whom may enjoy the gaming and number crunching from so many different variables. But rather than ending the national discussion, the examination is likely to spur further debate. Determining who really "won" will not be possible without agreement over what constitutes a legitimate vote -- something the counties themselves couldn't agree on, and on which the media aren't likely to take a firm, unified position.

The exception will be if the least controversial ballot description categories (such as chads with three corners detached) produce a decisive shift to Gore.

Citing human fallibility, the Herald's Baron predicts differences between his paper's counts and that of the consortium even when comparing undervotes in the same county. "The notion that any one review is the definitive review doesn't make any sense," he said, warning of the risks inherent in relying on a single source -- as happened to the networks on election night.

In truth, the more journalists who can examine the business of voting, the better for everyone. Perhaps the most useful thing to come out of this historic media effort will not be numbers but technical observations that might prove useful as the country goes about trying to reform the way it records votes. Personally, I learned lots of things that hadn't been clear from the wave of coverage right after the election. Like how, GOP protestations to the contrary, few absentee punch card ballots had actually been altered with scotch tape or other means. And I got to play an interesting little game with Earnest Williams, the manager of the Hillsborough County Elections Service Center. After numerous unsuccessful attempts to simulate a dented but not punctured rectangle, I finally did it: created a pregnant chad. "Notice how hard you worked at that," said a grinning Williams.

Postcript: as CJR went to press, the Herald printed a scoop that put Gore's strategy in doubt. Based on undervotes in Miami-Dade, Volusia, Palm Beach, and Broward counties -- the only places Gore had contested -- the former vice president would still have lost. The Herald -- by choosing not to wait for its compete statewide results, and by headlining the kind of definitive judgment that it had seemed to want to avoid -- could claim a competitive victory. Still, that left the consortium with the bigger job (and perhaps the bigger story): deciphering the will of the people.

Russ Baker is a contributing editor to CJR.