IF IT'S BROKE, FIX IT: THE BROKEN ELECTION MACHINERY OF 2000

Cape Girardeau County Clerk Rodney Miller has defended the punch-card ballot system used in this county (and in a great many others among the nation's 3,051 counties) as an accurate and relatively trouble-free method of recording and counting votes ("Punch cards work well here, clerk says," Nov. 17 Southeast Missourian article)

Miller said the system has been in place and worked well for 20 years, that it is superior to hand-counting of ballots and that of 30,588 ballots cast in Cape Girardeau County up to Nov. 7, election judges found only two ballots that could not be machine-counted and another 346 "spoiled ballots" that voters turned in for a new one before proceeding to vote and turn in the replacement ballots.

I believe punch-card ballots are not accurate enough and should be replaced by optical-scan ballots with superior accuracy. The well-intentioned defense of punch cards by a competent and experienced election official uses criteria that aren't disputed but does not address the true problem that is deeply disputed right now in Broward County and Palm Beach County in Florida. We should replace the punch cards with optically scanned ballots that voters originally mark with a pencil or pen. The hotly disputed Florida county balloting demonstrates exactly this.

The real problem

Punch-card ballots are indicted for failing to optically record all punched holes as votes. This is the source of arcane discussions about hanging chads (paper fragments from the punch site that do not fall entirely separate from the ballot) and its kin such as pregnant chads (indented but still attached to the original site) or dimples (early-month versions of same). The machine's fast optical scan may miss any of these. About 2 to 3 percent of ballots have a marked site that the machine count misses.

How is this discovered? By doing a laborious hand count of the machine-counted ballots and comparing the number of votes for an office at the top of the ticket (such as the presidency), there are 2 to 3 percent more votes in the hand count. Machine recounts do not catch this unless judges take the entire 30,588 votes, comb it for every single case the machine's optical scan might miss, and rerun them all after a clear hole punch is made through that location. Suppose that produces 31,000 votes. Then the error rate is 412 divided by 30,588, or about 1.4 percent.

The two ballots that could not be machine counted and the 346 spoiled ballots are not part of this problem. Neither is the pre-election testing and retesting of the machine. It works as intended but cannot solve a chad problem while doing 1,000 cards per minute.

Who wins?

The object of all voting is to decide who wins. County Clerk Miller and secretary of state representative Jim Grebing both affirm that no disputes have arisen over punch-card use to decide this. There's no reason to doubt that, but in Florida this same characteristic means every disputant on both sides could accurately forecast the result of a recount. Here's why.

Assume Cape Girardeau County records 30,000 votes in a contest of Smith and Jones. Smith wins 15,030 votes (50.1 percent) and Jones 14,070 votes (49.9 percent). A hand recount shows the scanner missed 1.67 percent of ballots, about 500 in all. Will this alter the election outcome? No. Those 500 extra ballots are random, not really different from the 30,000, so they are expected to divide about 251 for Smith and 249 for Jones. So Smith originally won by 60 votes, and now that's boosted to 62. In short, the original inaccuracy won't jump up and bite us by picking the wrong winner. For that reason, a county voting official will not direct a laborious full recount in the first place.

But what if the county's vote result is part of a close and disputed statewide vote with 66 other counties? Suppose it's a diverse state, with heavily Republican counties one place and heavily Democratic elsewhere. And suppose it's a very large county, such as Broward or Palm Beach? Then a small inaccuracy can magnify into a large problem.

That's exactly what every vote counter for both Bush and Gore recognizes, else they would not fight so furiously over the issue of a hand recount.

We have Florida casting 5,822,814 votes so far (counting absentees abroad), with Bush ahead of Gore by only 930. Palm Beach County originally cast 462,657 punch-card presidential votes counted and re-counted by machine. Let's round that down to 450,000. Gore wins two of every three, this being a heavily Democratic county. Now assume a recount by hand shows the usual 2 percent or so error rate. That's 9,000 additional ballots. How much is that worth for Gore and against Bush? Easy. It's 333 votes per thousand extra, or about 3,000 in all. That's 10 times the statewide difference in the secretary of state's interim official count.

Not rocket science

This is not rocket science or anything else highly difficult to figure. Everybody in both the Bush and Gore camps recognizes some version of this. Certainly the secretary of state in Florida does. That is why no compromise is possible or foreseeable here. Either those counties will be counted into the official tally by the secretary of state, or they're hand-counted and published but not included. Either way, citizens everywhere will know that accuracy problems with punch cards are crucial to all this. The legitimacy of the probable 43rd president of the United States is in severe question for this very reason.

And that's why inaccurate punch cards are a problem. This problem cannot be dismissed away as an idiosyncratic bad design of the butterfly ballot can. It does not produce 346 spoiled ballots that are replaced and then filed by the voter. The voter's ballot is not counted 2 or 3 percent of the time, but nobody can (or should) know whose secret ballots these are. Hand recounts only discover that some were not counted. That's no big deal in deciding a Cape Girardeau election. But it could be a very big deal some day in deciding a Missouri one. Or it might even control the entire Electoral College.

Russell Renka teaches in the Department of Political Science at Southeast Missouri State University.