Miami Herald: How the study was done

NEWS

E-Mail this story to a friend

Published in the Miami Herald

EXCLUSIVE: Without voter mishaps, Gore would have won

How the study was done:

Herald reporters contacted the election departments of each of Florida's 67 counties to get the vote returns for each of Florida's 5,885 voting precincts, plus the absentee returns. In addition, they gathered the total number of ballots cast, whether counted or not, in each precinct. This information was then assembled into a statewide precinct-level database and checked against the official countywide totals certified last Sunday by the Florida secretary of state.

To estimate the intent of those who cast ballots that weren't counted, The Herald started with several assumptions:

Everybody who cast a ballot, whether in person or absentee, actually intended to cast a legitimate vote for one of the several presidential candidates.

Both undervotes (failing for whatever reason to successfully mark a ballot or punch out a chad) and overvotes (selecting more than one candidate for whatever reason) should be counted as attempts to vote in the presidential race.

The group of voters in each precinct who cast uncounted ballots are a microcosm of their neighbors in that precinct, so that their collective choices of candidates would mirror the choices of those who successfully cast ballots. In other words, the assumption is that political party preference does not make voters more or less likely to cast an uncounted ballot.

Given those assumptions, The Herald analysis simply calculated the percentage of the successful votes that Bush, Gore and the collective other candidates each received, and then divided up the uncounted votes based on those percentages.

For instance, consider a hypothetical precinct in which there were 500 votes for Bush, 400 votes for Gore, and 100 votes for all other candidates combined.

If this precinct also had 100 more ballots that weren't counted because of overvoting or undervoting, then The Herald analysis would give Bush 50 of them (for a total of 550), Gore 40 more (for a total of 440) and the others the remaining 10 (for a total of 110.)

Of course, it is possible that an unknown number of those uncounted ballots were cast by people who, for whatever reason, actually didn't want to vote in the presidential election. To explore the effect of this, The Herald repeated the above analysis assuming that various percentages of the uncounted ballots in each precinct were turned in by voters who intentionally expressed no presidential preference.

Under this method, Bush maintains a lead only if one assumes that virtually all of the 185,000 uncounted ballots were meant to be uncounted.

Finally, to measure the impact of optical versus punchcard voting systems, The Herald coded each precinct by the type of voting system that was used, and then calculated how the two groups differed.

The Herald analysis was done by Stephen K. Doig, the Knight professor of computer-assisted reporting at the Walter Cronkite School of Journalism and Telecommunication at Arizona State University. Before joining ASU in 1996, he was the research editor for The Herald. During a 19-year career at The Herald, he worked on major projects that included studies of the South Florida criminal justice system and of the damage patterns from Hurricane Andrew. He also has long experience in analyzing election results and demographics.

Information for the analysis was gathered by Herald staff writers Anabelle de Gale and Lila Arzua. Herald staff writer Curtis Morgan discussed the analysis with other researchers and commentators and wrote the story.