msnbc.com
All I know is what I don't read in the papers
Keith Olbermann
November 19, 2004
SECURE UNDISCLOSED LOCATION— I’m beginning to think like Jim Bunning now.
So far in this post-election trip through Alice’s looking glass we’ve had:
—a University of Pennsylvania professor defending the accuracy of exit polling in order damn the accuracy of vote counting;
—a joint CalTech/MIT study defending the accuracy of exit polling in order to confirm the accuracy of vote counting;
—a series of lesser academic works assailing the validity of the Penn and CalTech/MIT assessments;
—and now, a UC Berkeley Research Team report that concludes President Bush may have received up to 260,000 more votes in fifteen Florida counties than he should have, all courtesy the one-armed bandits better known as touch-screen voting systems.
And, save, for one "New York Times"reference to the CalTech/MIT study "disproving" the idea that the exit poll results were so wacky that they required thoroughly botched election nights in several states, the closest any of these research efforts have gotten to the mainstream media have been "Wired News" and "Countdown."
I still hesitate to endorse the ‘media lock-down’ theory extolled so widely on the net. I've expended a lot of space on the facts of political media passivity and exhaustion, and now I’ll add one factor to explain the collective shrugged shoulder: reading this stuff is hard. It’s hard work.
There are, as we know, lies, damn lies, and statistics. But there is one level of hell lower still— scholarly statistical studies. I have made four passes at “The Effect of Electronic Voting Machines on Change in Support for Bush in the 2004 Florida Elections,” and the thing has still got me pinned to the floor.
Most of the paper is so academically dense that it seems to have been written not just in another language, but in some form of code. There is one table captioned “OLS Regression with Robust Standard Errors.” Another is titled “OLS regressions with frequency weights for county size.” Only the summary produced by Professor Michael Hout and the Berkeley Quantitative Methods Research Time is intelligible.
Of course, I’m reminded suddenly of the old cartoon, with the guy saying “I don’t understand women,” and the second guy saying, “So? Do you understand electricity?”
In his news conference yesterday at Berkeley (who attended? Who phoned in to the conference call? Why didn’t they try?) Professor Hout analogized the report to a “beeping smoke alarm.” It doesn’t say how bad the fire it is, it doesn’t accuse anybody of arson, it just says somebody ought to have an extinguisher handy.
Without attempting to crack the methodology, it’s clear the researchers claim they’ve compensated for all the bugaboos that hampered the usefulness of previous studies of the county voting results in Florida. They’ve weighted the thing to allow for an individual county’s voting record in both the 2000 and 1996 elections (throwing out the ‘Dixiecrat’ effect), to wash out issues like the varying Hispanic populations, median income, voter turnout change, and the different numbers of people voting in each county.
And they say that when you calculate all that, you are forced to conclude that compared to the Florida counties that used paper ballots, the ones that used electronic voting machines were much more likely to show “excessive votes” for Mr. Bush, and that the statistical odds of this happening organically are less than one in 1,000.
They also say that these “excessives” occurred most prominently in counties where Senator Kerry beat the President most handily. In the Democratic bastion of Broward, where Kerry won by roughly 105,000, they suggest the touch-screens “gave” the President 72,000 more votes than statistical consistency should have allowed. In Miami-Dade (Kerry by 55,000) they saw 19,300 more votes for Bush than expected. In Palm Beach (Kerry by 115,000) they claim Bush got 50,000 more votes than possible.
Hout and his research team consistently insisted they were not alleging that voting was rigged, nor even that what they’ve found actually affected the direction of Florida’s 27 Electoral Votes. They point out that in a worst-case scenario, they see 260,000 “excessives” - and Bush took the state by 350,000 votes. But they insist that based on Florida’s voting patterns in 1996 and 2000, the margin cannot be explained by successful get-out-the-vote campaigns, or income variables, or anything but something rotten in the touch screens.
It’s deep-woods mathematics, and it cries out for people who speak the language and can refute or confirm its value. Kim Zetter, who did an excellent work-up for "Wired News,"got the responses you’d expect from both sides. She quotes Susan Van Houten of Palm Beach’s Coalition for Election Reform as saying “I’ve believed the same thing for a while, that the numbers are screwy, and it looks like they proved it.” She quotes Jill Friedman-Wilson of the touch-screen manufacturer Election Systems & Software (their machines were in use in Broward and Miami-Dade) as responding “If you consider real-world experience, we know that ES&S’ touch-screen voting system has been proven in thousands of elections throughout the country.”
What’s possibly of more interest to us poor laymen is what isn’t in the Berkeley report.
As I mentioned previously, they don’t claim to know how this happened. But more importantly, they say that they ran a similar examination on the voting patterns in Ohio, comparing its paper ballot and electronic results, and found absolutely nothing to suggest either candidate got any “bump” that couldn’t otherwise be explained by past voting patterns, income, turnout, or any other commonplace factor.
In other words: No e-voting machines spontaneously combusting in Ohio.
“For the sake of all future elections involving electronic voting,” Professor Hout concluded, “someone must investigate and explain the statistical anomalies in Florida. We’re calling on voting officials in Florida to take action.”
Anybody want to belly up to this bar?
Thougths? E-mail me at KOlbermann@msnbc.com