By: Sharon L. Crenson, AP National Writer
(AP) Eight media companies joined together for a thorough review of more than 175,000 ballots excluded from state-certified election results in last year’s historic Florida presidential election.
About a third of the ballots were undervotes, meaning they did not initially register a clear choice for president. The rest were overvotes, thought to show more than one pick.
The media group hired the National Opinion Research Center (NORC) at the University of Chicago to look at the markings on each ballot and build a database categorizing them. Data editors from the news organizations then programmed computers to identify votes that reasonably might have been counted had election officials performed hand recounts.
Since Republicans, Democrats, and the courts disagreed about what type of markings should constitute a vote, the journalists decided to make several separate tallies. For example, the consortium members tallied votes based on ideas advocated by both Al Gore’s and George W. Bush’s attorneys.
The Florida election review was developed by The Associated Press, CNN, The New York Times, The Palm Beach Post, The St. Petersburg Times, Tribune Publishing, The Wall Street Journal, and The Washington Post. Tribune newspapers include the Chicago Tribune, Los Angeles Times, Newsday, the Orlando Sentinel, and the South Florida Sun-Sentinel in Ft. Lauderdale.
While members of the group worked together in developing the project, each media outlet decided for itself how to present the findings to the public.
The project faced obstacles from the outset, as it attempted to replicate ballot counts reported immediately after the election.
For instance, the consortium was only interested in undervotes and overvotes — not properly completed ballots — so counties had to separate them.
Some counties used counting machines, while others did the job manually, a difference that produced results which didn’t always match the expected totals. Some counties were unable to show NORC as many uncertified ballots as they had reported, while others showed too many.
NORC tried to minimize error at every stage, however, and in the end the consortium viewed 175,010 ballots.
The review process began with NORC screening potential field workers for political bias and disqualifying anyone who worked for or donated money to a state or federal political race in the last decade. Next came a vision test. Afterward, team leaders used sample ballots to teach field workers a uniform system of describing various types of ballot markings.
For example, a “1” represented a chad with one corner detached from a ballot and three corners intact. Other codes were devised to represent dimples, erasures, and other marks on or next to a candidate’s name or party.
Most field workers had worked for NORC before, although others were hired from local employment agencies in some rural counties.
The ballot review began Feb. 5 in Hillsborough, Pasco, Miami-Dade, Pinellas, and Polk counties. NORC team leaders oversaw the process, making sure workers didn’t talk to one another about their decisions.
Because state law allows the public to look at ballots but not touch them, county election workers held the cards. Three NORC workers looked at each undervote, and one worker usually looked at overvotes.
Early in the review, the consortium had three workers review overvotes in Nassau, Pasco, and Polk counties, but that strategy proved unnecessarily costly. The workers agreed with each other’s choices so often — at least 97% of the time — that the media group decided there was little need for three people to review each overvote.
On undervotes, the work of a woman in Baker County varied so dramatically from her colleagues — consistently favoring Bush regardless of ballot marks — that NORC supervisors believed she was misunderstanding instructions or biased.
Her work on the 79 ballots she coded were excluded from the media group’s recounts, although it is available as part of the consortium’s raw data and documentation.
As field workers completed their inspection of the ballots, data-entry clerks entered numerically coded results into a database.
Reporters specializing in data analysis then met to consider how to tabulate the data.
It was the second time this year a media group reviewed Florida ballots. The Miami Herald and USA Today partnered with Knight Ridder newspapers, The Tampa Tribune, the Tallahassee Democrat, The Bradenton Herald, Florida Today, The News-Press of Fort Myers, and the Pensacola News Journal on their own study.
That group hired an accounting firm to examine undervotes, with reporters checking the work. Later, reporters and other staff examined thousands of overvotes visually and recreated others using computer records from the counting machines.
Although both the Herald/USA Today group and AP’s partners sought to answer similar questions, they approached them differently and produced different numeric results.
But both groups concluded Gore’s best chance for success appeared to have been down an avenue he did not pursue — a statewide recount of both undervotes and overvotes.