Why selective recounting of ballots is wrong

Paid supporters of Mr. Gore are making the argument that selective hand recounting of ballots in one county is not unfair to voters in other counties where hand recounting is not being done. That statement is not true, and people making it most likely know that is it not true, but they are counting on the fact that most people will not work out any mathematical example that proves that it is not true.

Let me give two sample demonstrations of why it is unfair to selectively hand recount ballots. The first example shows unfairness that benefits Mr. Gore, but the second example shows problems that could cause Mr. Gore to loose votes.

Example 1

Please consider two counties, MOE and JOE. In both counties the voters intended to cast 60 votes for one candidate, and 40 for the other, resulting in a tie because these counties vote in exact opposite ways. There is uniform, consistent, fair, 10% error rate with the counting machines. For every 100 votes cast, 10 are not counted. The result that we actually see after a machine count is still a tie. But what if a hand count rejects fewer ballots, and counts more votes? For example, what if the reject rate is only 5% in hand counting? Then instead of counting 54 votes from 60, the hand count would count would show 57 votes. But what if only one county does a hand recount, because emotions are running highest in that one county. Will the result be fair? Look for yourself:

County MOE County JOE Results

True Intent Machine Count Hand Count True Intent Machine Count True Intent Machine Count Selective Recount
MR. GORE605457 4036 1009093
MR. BUSH403638 6054 1009092

The result of a selective hand-recount is that MR. GORE will be chosen the winner, even though the true result is a tie. That is not correct. That is not fair.

Any hand count is subject to the same untrue result when different standards are used for the hand recounting in different counties. If a more liberal county that voted more for MR GORE has a more liberal standard for reading the intent of the voter by visual inspection, and rejects fewer ballots, and a more conservative county that voted more for MR BUSH has a more conservative standard for reading the intent of the voter by visual inspection, and rejects more ballots as unreadable, then the total result will still be untrue and unfair.

That is why a consistent and enforced standard throughout, such as a straightforward machine reading, is more likely to give a true and fair result. The alternative is to let politics enter too heavily into the counting process. Or as MR. LENIN once said:

It isn't who votes that counts, it is who counts the votes.

It is very bad for our democracy for politicians to carry the selective statements of facts, political talk and general spinning of tales from the time before votes are cast deep into the time to be simply counting votes. The entire process of extended counting and recounts combined with heavy political activity and organized demonstrations in the street for the purpose of influencing the decisions made about the counting methods is not good, not good at all.

Example 2

Consider a different source of errors other than undercount due to the inability of the card reading machines to read a hole (punched chad) in a ballot. Instead, consider manufacturing defects, imperfections, in the paper ballots themselves. If you obtain 400,000 ballots fresh from the manufacturer then you would expect that these would be produced for the lowest possible cost, using the cheapest possible materials, in order to be able to sell them to the government at a profit. You would expect that the ballots would not be so bad as to actually register false votes if fed through a mechanical counting machine, however, they could very well have dimples, minor gaps, bumps, and other extremely small imperfections that are detectable only through careful visual inspection. Consider also that when these mechanical voting machines were originally designed that there was a strong, competitive market for computer punch cards. The market as a whole would not tolerate poor quality in a material commonly used in the computer industry. However, consider that paper punch cards are now a specialty item, entirely out of the mainstream of daily life, and purchased now only by government agencies with limited budgets.

Visual imperfections in the paper cards would not be a defect when cards are read by machine. However, when the cards are read by human visual inspectors these impefections would be a serious source of random noise. Humans have very little ability to filter out this kind of noise. In fact, humans have an ability to find patterns that do not actually exist when faced with random patterns.

If you took 500,000 cards fresh from the manufacturer, and ran them through a voting machine once, you would expect to see no more than 1 or 2 false votes, if the manufacturer of the cards has done a good job. However, if the manufacturer was not told to create visually perfect cards you would not be suprised that several hundred false votes would be seen when cards are "manually" counted through visual inspection. You would expect this false count to increase as the cards are run through the mechanical counting machines multiple times or handled by people. These cards have been put through a machine at least twice, have been handled by a voter, have been handled by multiple other people.

Consider a dimple or visual gap in a punched card. If that is to be interpreted as an attempt by a voter to punch a true hole in the chad, then a dimple could cause a blank vote to show up as a vote for either candidate, or it could cause a vote for a single candidate to be counted as a double vote.

In a place where Gore has more votes than Bush it is more likely that random noise would invalidate a Gore vote then it would be for Gore to pick up a new vote. This is because Gore starts off with more true votes, properly punched out by the voter.

This would also explain why Bush does not want to recount in places that he has fairly one. It would also explain why Gore has not picked up significant votes, and has even lost votes in some places, after almost two weeks of manual recounting via un-aided visual inspection.

It also explains why selective visual inspection of a subset of ballots (only those show to be unreadable by machine) is also subject to generating false results.

Visually imperfect cards would introduce errors of a much greater amount then any reasonable confusion or incompetence of the voters in manual operation of the ballot punching mechanism they were provided with at the polling place.

George J. Carrette, November 13, 2000.