Wednesday, December 01, 2004

Nationwide Voting Errors

This from the Boston Globe:

More than 4,000 votes vanished without a trace into a computer's overloaded memory in one North Carolina county, and about a hundred paper ballots were thrown out by mistake in another. In Texas, a county needed help from a laboratory in Canada to unlock the memory of a touch-screen machine and unearth five dozen votes.
In other places, machine undercounting or overcounting of votes was a problem. Several thousand votes were mistakenly double-counted in North Carolina, Ohio, Nebraska, and Washington state. Some votes in other areas were at first credited to the wrong candidates, with one Indiana county, by some quirk, misallocating several hundred votes for Democrats to Libertarians. In Florida, some machines temporarily indicated votes intended for challenger John F. Kerry were for President Bush, and vice versa.
In the month since the election, serious instances of voting machine problems or human errors in ballot counts have been documented in at least a dozen states, each involving from scores of ballots to as many as 12,000 votes, as in a North Carolina county. On Election Day, or in later reconciling tallies of ballots and voters, local officials discovered problems and corrected final counts. In some cases, the changes altered the outcomes of local races. But in North Carolina, the problems were so serious that the state may hold a rare second vote, redoing a contest for state agriculture commissioner decided by fewer votes than the number of ballots lost.

And so on, all stuff that people who read blogs have known about for some time. Still, this is a mainstream newspaper report, and the first one which gives a little credence to the worries that us tinfoilhatters have expressed for a while. Though just a little credence, there are serious warnings embedded in the story about why none of this matters at all. For example:

In addition, minor presidential candidates requested recounts in four states -- a partial one completed yesterday in New Hampshire, and statewide in Ohio, New Mexico, and Nevada.
None of the recounts or inquiries is expected to affect the results of the presidential election, which Bush won by more than 3.3 million votes.
Those who believe that either or both of the past two presidential elections were manipulated by a vague conspiracy to elect Bush have done statistical analyses of voting patterns in Florida and argued that the voting discrepancies were much larger and systemic, but their studies have not stood up to scrutiny from academics and other analysts.

How does the writer "know" that the results of recounts wouldn't overturn the election? What inner knowledge does he have that allows him to predict the outcome this clearly? I suspect that he has no such knowledge; he simply assumes that 3.3 million votes can't all be lemmings.

Also, it's interesting how academic studies are pitted against each other and a conclusion is drawn that some of them beat the others without anything like a link to prove this conclusion. I have followed those studies and their criticisms pretty carefully, and though some have been shown to have other explanations (such as the Dixiecrat phenomenom in Northern Florida), others have not been debunked. It's pretty easy to present some criticism of any study in a way which makes the criticism appear valid to someone who doesn't understand the techniques applied, and most of the criticisms of, say, the Berkeley study don't in fact debunk it at all, if by debunking we mean that the study can now be safely ignored.

Anyway, the article gives a pretty good summary of many of the oddities in the recent U.S. elections, together with a map which shows the states with most errors. Not surprisingly, these tend to be the swing states that got the most attention. The reason is probably that people have been more alert in those states for anything that might be an error. Of course, it could also be the case that those states had more errors to begin with.