The analysts, most of whom are professors of computer science or mathematics at such institutions as the University of Notre Dame and Southern Methodist University, made their claim after finding that exit polls, which are usually reliable, had diverged 5.5 percentage points from official vote tallies. (This disrepancy was larger than was found in one of two Ukrainian exit polls, which played key roles in overturning the December 2004 election in that nation.) Statistically significant variations between the U.S. exit polls and official results were concentrated in five states, four of which were "battlegrounds," such as Ohio, Pennsylvania, and Florida.
I said in my previous column that the topic of election tampering was so sensitive that I'd print responses this week from at least one Republican, one Democrat, and one Independent. It turns out that my readers didn't, by and large, tell me their party affiliations -- but they had a lot to say, nonetheless. I'm publishing the most insightful comments below.
In Defense Of The Official Tallies
Steve Weeks is an attorney in Ohio, one of the most hotly contested states in the 2004 election. Weeks argues that the exit polls and the official tallies were far apart because Bush voters felt social pressure to say they were Kerry voters. This possibility is distinct from the one I described last week that was offered in a study of the discrepancies by the National Election Pool (NEP) exit pollsters themselves. They theorize that Bush voters were less likely than Kerry voters to participate in exit polls at all (a premise that was then statistically disproved by the Ph.D. group).
"You should be old enough to remember GIGO [garbage in, garbage out]," Weeks writes. "The fanciest statistical analysis ever done is meaningless if it is based on false premises.
"The fundamental premise of any poll is that the respondents are telling the truth about the candidate they prefer (or that falsehood rates are substantially the same for all candidates). In 2004, that was manifestly not the case. Bush was vilified far beyond anything that I have seen in a Presidential campaign in my lifetime (I am 55); even when Johnson toasted Goldwater with the A-bomb ad in 1964, the personal attacks did not approach those seen this year.
"As a result, it became unacceptable in politically-correct society for ANYONE to support Bush. When cornered by exit pollsters, a few percent of Bush voters were too embarrassed to admit it. The pre-election polls were incorrectly biased in favor of Kerry for the same reason."
I'm primarily interested in using computer technology to prevent future voting problems, not to overturn any previous election. I wholeheartedly agree that Weeks' theory is worth testing.
A variation in the exit poll at a particular precinct from the final, official tally is called Within Precinct Error (WPE). The USCountVotes group summarizes this theory by citing NEP's own study, saying, "the required shift toward Kerry in the exit polls must have been 6.5%. They [NEP] note that this number is greater than any WPE from past presidential elections going back more than 20 years, to a time when polling science was less sophisticated and less reliable than at present. They also note that this 6.5% WPE stands out in comparison to an average 1.9% WPE from 2004 state primaries exit polls."
The Ph.D.s behind the latest analysis have called on NEP to release the raw, precinct-level data that were used to calculate the 2004 exit poll figures. As of this date, the raw data that would allow testing of the WPE theory have not been made available.
Officials at the two firms responsible for the 2004 NEP -- Joe Lenski, vice president of Edison Media Research, and Scott Dingman, a partner in Mitofsky International -- declined to comment for this article and referred me to other spokespersons. Those representatives did not respond by press time to telephone calls and e-mail messages seeking comment.
How Vote Tallies Might Have Been Changed
Maribeth McIntyre believes the NEP exit polls may show a pattern of vote tampering. This wouldn't require a massive army of crooked election workers, she writes, but only a hidden routine in the vote-counting software:
"As an IT professional with 25+ years of experience (and, yes, a very partisan Democrat), I have been concerned about the ease with which election fraud can occur with the voting technology in place today," says McIntyre. "It would not require anything like a large conspiracy; one or two programmers for the voting software companies could easily pull it off.
"After studying both the Edison/Mitofsky report and the USCountVotes.org analysis, I am completely convinced that the 2004 Presidential election was stolen. In addition to good old-fashioned voter suppression (lack of sufficient machines in predominately Democratic precincts but an overabundance of them in Republican precincts, among other tactics), there were numerous reports of vote-hopping on touch-screen machines -- a vote for Kerry records as a vote for Bush, but never the other way around."
The provision of extra voting machines in Republican areas of Ohio and "vote hopping" that affected some machines was reported in a Washington Post article last December. An interview with an election equipment worker who indicated that very few people would be required to pull off vote tampering on some of today's machines was published by Christopher Hitchens, no Kerry fan himself, in a March 2005 Vanity Fair article.
Solutions For Future Elections
Finally, reader Lance Franklin (my choice to represent Independent voters) proposes printed records as a way to ensure that vote counts are tamper-proof:
"My solution would have been a device that would have been placed between the voting machine and the printer generating the paper trail," Franklin writes.
"In essence, the device would have displayed the vote that was going to the printer, asked the voter to verify their vote, and then passed the OK/Revote response back to the voting machine, for it to either save the vote or reacquire a new vote."
This sounds like a decent requirement for all elections. In fact, groups seeking reform of the U.S. election process consistently call for a "voter-verified paper ballot." How citizens have voted should be clearly visible to them on their own paper ballots, which should comprose the official tally of an election, regardless of what any "quick-count" electronic devices may say.
The problem, according to USCountVotes.org, is that approximately 30% of America's votes are now cast on equipment that has no paper trail and can't be audited in any way, shape or form.
In my opinion, Americans can either go through this trauma every election ("The count was fair!" "No, the count was hacked!") or they can demand that all election equipment maintain auditable paper ballots. Computer experts, especially those who are asked to work on vote-counting systems, should insist on this. It won't by itself eliminate all vote fraud -- history has shown that ballot-stuffing is all too common -- but it should prevent any partisan programmer from single-handedly goosing the results.
For information on the electoral studies at the heart of the matter, see the March 31 USCountVotes.org analysis and the Jan. 19 Edison/Mitofsky report on the exit-poll discrepancies. (These PDF documents require the free Adobe Reader.)
My thanks to everyone who wrote in about this controversial topic. Readers Weeks, McIntyre and Franklin will receive gift certificates for a book, CD or DVD of their choice for sending me comments that I printed.