The Knight-Ridder newspapers hired BDO Seidman to tally the overvotes and undervotes. BDOS prepared a complicated set of tables setting forth all the possible ways of counting the votes. Some systems result in Bush winning, others result with Gore winning.

Undervotes

An "undervote" means that the ballot was not marked enough to be counted, as opposed to an "overvote," in which the voter voted for more than one candidate. The "undervote" tables were printed in USA Today on 4/4/2001.

BDOS analyzed 17,312 punch card ballots with undervotes. Of these, 14,940 had some indication of the voter's intent. Here is a table of the possibilities:

Punch Card Undervotes in Florida 2000
Description	Additional Bush Votes	Additional Gore Votes
Dimples: minimal mark on ballot	6,718	5,530
Pinpricks: chads have minor border punch	596	503
One Corner	71	50
Two Corners	248	138
Three Corners	476	215
Clean punch	193	123
Hand-Marked (pencil)	37	42

My analysis: I would not count dimples, pinpricks, or one-corner chads. The clean punch ballots were not initially counted, but apparently the chads fell off afterwards. They were probably all 3-corner chads initially. I normally would not count hand-marked, but in the cases in the table the intent of the voter was clear so I would count them.

As far as the two-corner chad ballots go, I would not mind counting them if the decision were made before any ballots were counted. Otherwise I would not count them. The standard needs to be set before the counting begins.

BDOS analyzed 3,549 undervotes on optical-scan ballots. Of these, 1,861 had some indication of the intent of the voter. Here is a table of these undervotes:

Optical Scan Undervotes in Florida 2000
Description	Additional Bush Votes	Additional Gore Votes
Oval incorrectly marked	246	341
Name of Candidate Circled (oval not colored in)	182	330
Wrong pencil type used	189	139
Oval Circled	94	136
Underlined Candidate Name	32	40
Miscellaneous	32	50

Deciding on the optical scan ballots is more difficult. Gore wins more votes in most categories. These votes were not counted because the voter did not follow the directions. I would normally recommend that using the wrong pencil and circling the bubble not be counted.

Overvotes

The study analyzed 111,261 overvotes. Of these, only 3% could possibly be considered valid votes. The complication: if a voter chooses Gore and Buchanan (the most common combination), which was the voter's preference?

Overvotes Per Candidate
Candidate	Overvotes	% of Overvotes
Gore (D)	84,197	76%
Bush (R)	37,731	34%
Buchanan (Rf)	36,756	33%
Browne (Lbt)	32,495	29%
Moorehead (WW)	28,797	26%
Hagelin (NL)	28,686	26%
Phillips (Cst)	28,396	26%
Nader (G)	28,303	25%
Harris (SW)	28,147	25%
McReynolds (Soc)	23,958	22%
Write-ins	5,010	5%

The following table sets forth the 20 most common types of overvotes. Since many of these overvotes were cast on the "butterfly" ballot, I have grouped them: the left column shows one vote for a candidate on each leaf, while the right column shows votes for two or more candidates on the same page of the ballot.

Types of Overvotes, Florida 2000
One Candidate on each page	Two Candidates on Same Page
Gore-Buchanan 10,234	Gore-Browne 9,321
Bush-Buchanan 4,957	Bush-Gore 8,956
Gore-Moorehead 3,989	Gore-Nader 3,405
Gore-McReynolds 3,976	All Candidates except Bush 3,007
Gore-Phillips 3,868	Gore-Harris 1,782
Gore-Hagelin 3,015	Bush-Browne 1,711
Bush-Hagelin 2,442	Bush-Nader 1,529
Bush-Phillips 2,047	Gore + write-in 1,292
Bush-Moorehead 1,771	All candidates (!) 957
	Gore + 4 others 911
	Bush + write-in 790
[Total of above: 36,299]	[Total of above: 33,661]
All Other Overvotes: 41,301

This table shows that the common complaint about the ballot is somewhat overrated. The 20 most common overvote errors are about equally divided between voting for multiple candidates on the same page and voting for one candidate on each page.

Note: The left page of the ballot had Bush, Gore, Browne, Nader, Harris, and Hagelin. The right page had Buchanan, McReynolds, Phillips, Moorehead, and the write-in option. Some counties had Hagelin on the second page.

Overvote details: Voters in Duval County cast 21,188 overvotes, or 1/5 of the total. Of these, 55% voted for one candidate on each page. Palm Beach County's butterfly ballot was the recipient of 18,748 overvotes. Gore received 80% of the overvotes there to 20% for Bush. Half of the Gore-Buchanan overvotes in the state were cast here.

The following is a table of the BDOS account of the 3% of ballots where the voters' intents were somewhat discernable:

Overvote Breakdown
Type of Overvote	Clearly for Bush	Clearly for Gore
Attempted Corrections	530	863
Write-in Corrections	606	921
Write-in Blank	53	87

Of the three in the table, I would not count the third column.

Result

Deciding how to adjust the total involves making two decisions: how to count hanging chads and how to determine the intent of a voter who did not follow directions. Should a vote be counted in the latter case or not? I would favor either position if the decision is made before counting begins.

You can see that there is a number of ways in which either Bush or Gore could have won Florida in 2000. If my preferences are followed, Gore wins by a margin of 2,915,088 to 2,915,060.