Reliability of the Uncertified Ballots in the 2000 Presidential Election in Florida. (Statistical Practice)
Wolter, Kirk, Jergovic, Diana, Moore, Whitney, Murphy, Joe, O'Muircheartaigh, Colm, The American Statistician
The presidential election of 2000 was among the closest and most interesting elections in American history. In the state of Florida, 6,138,120 ballots were cast. At the time the U.S. Supreme Court stopped the vote counting, 175,010 ballots (or approximately 2.9%) remained uncertified for the presidential race, including 61,190 undervotes and 113,820 overvotes. The uncertified rate is comparable to rates in the presidential elections of 1996 and 1992, with 2.5% and 2.3% of the ballots uncertified, respectively. However, the 2000 election was so close that every ballot could potentially have made the difference.
Following the 36-day, presidential-election crisis, a group of the nation's most-respected media organizations (The New York Times; The Washington Post; The Wall Street Journal; CNN; The Associated Press; The Tribune Publishing Company, represented by The Los Angeles Times, The Chicago Tribune, and The Orlando Sentinel; Cox News Service, representing The Palm Beach Post; and The St. Petersburg Times) hired the National Opinion Research Center (NORC) at the University of Chicago to conduct a thorough review of Florida's uncertified ballots. This team recognized the historical significance of the uncertified ballots, and worked together to create a definitive historical archive of these ballots. This formidable task was enabled by Florida's sunshine law (P.L. 101.572), which authorizes anyone to view ballots from a state election.
The Media Group and NORC adopted two main analytic goals. The Media Group took as its principal goal the determination of which uncertified ballots could be assigned as votes to specific presidential candidates, and consequently of which candidate might have won the state of Florida given alternative voting counting scenarios or standards. NORC's analytic goal was to measure and compare the reliability of the various voting systems in use in Florida, with the aim of providing elections officials with a base of information to guide the improvement of future elections. On November 12, 2001, NORC released the archive to the public via its Web site (NORC.ORG) and the Media Group published or aired the results of their analysis of vote totals. This article mainly contains the results of NORC's analysis of reliability.
The state of Florida is divided into 67 counties, which used five different voting systems. The majority of counties (41) used a system by which voters used a specified pen or pencil to fill in arrows or ovals to select candidates. An optical scanner would read the ballots and register votes for the candidates. Fifteen counties used a Votomatic punch card voting system. This system required the voter to insert a computer punch card, containing many prescored chad (a small area of approximately 1/16 inch in diameter with a perforated border), into a device and then to use a stylus to punch out the chad for the selected candidate. A machine--which recorded the number of chads punched out for each candidate--counted the votes. Nine counties used a Datavote system, which was similar to Votomatic technology in that voters were required to insert a computer punch card into a device. For the Datavote system, however, the voter aligned a mechanical punch tool with the candidate of choice and punched a hole into the b allot. There were no prescored chads. A machine then read the punch cards and counted the votes. One county used a lever voting system--a system by which voters did not use ballots. A large voting apparatus simply tallied votes. Finally, one county used paper ballots on which voters used any pen or pencil to indicate selected candidates by marking an X in a box next to the candidate's name. Paper ballots were counted by hand.
None of the voting systems are perfect and uncertified ballots can result from mistakes by the voters, errors by the counting system (machine or canvassing board), or intentional actions of the voters. …