A read of the initial reports from the UC Berkeley Quantitative Methods Research Team left more than a few Wizbang readers scratching their head. The Berkeley team left out any number of potentially statistically significant factors in their analysis, such as the Jewish vote, the African American vote, etc. Several people have written to me and said that they are working with the data provided (and mining for more as needed) to address these issues.
I chose a different tack – attacking their basic premise. They attribute an increase in votes for the president to the properties of the voting equipment as opposed to voter behavior. From their report [PDF]:
In our research we used ordinary least squares and more sophisticated linear modeling approaches to assess the statistical properties of e-voting. In particular we develop models that predict both the percentage of the votes registered for the incumbent – President Bush – and the amount that percentage changed between 2000 and 2004. These models can incorporate adjustments for a large number of factors that we or others thought might help explain the patterns. These include socioeconomic and demographic factors like the typical family’s income or its ethnic ancestry. We also adjust for ecological factors like the size of the county. Most importantly we adjust for its voting history, reaching back not only to the 2000 election but farther to the 1996 election. To this list of factors we add consideration of whether the county’s voting technology was e-touch machines or optical scanning equipment.
Finding
Electronic voting raised President Bush’s advantage from the tiny edge he held in 2000 to a clearer margin of victory in 2004. The impact of e-voting was not uniform, however. Its impact was proportional to the Democratic support in the county, i.e., it was especially large in Broward, Palm Beach, and Miami-Dade.Of course there’s more to their reports, and you are encouraged to study their results on your own.
Couldn’t the percentage of the votes registered for the incumbent – President Bush – and the amount that percentage changed between 2000 and 2004 be explained by the simple fact that more people voted for him in 2004 than in 2000? Certainly that seems like a reasonable hypothesis, but how can we prove it?
Consider the gyrations that the UC Berkeley Quantitative Methods Research Team go through to explain in quantitative terms what is actually a series of choices and actions by over seven million Floridians. Should they bother to vote? Who should they vote for? Are they voting for a candidate or against a candidate (by voting for the opponent)? What factors influence their choice? etc…
Wouldn’t it be great if we had done a survey of Floridians and asked who they voted for and why? If we deployed an army of pollsters to gather that data at polling places throughout Election Day that would be a good idea as well. Amazingly enough that critical piece of missing information is available in the Florida exit polling data. And fortunately for us the data is sufficiently detailed that we can estimate (using the exit poll data for Broward, Palm Beach, and Miami-Dade county) how many votes were cast for each candidate before the votes had even been counted. All this should sound painfully familiar to election junkies…
So what did voters in those 3 counties report about who they voted for?
Exit Polls - Florida Vote By Region
Miami Area
(26% of the 7,446,434 statewide Bush and Kerry votes)
Bush 774,429* 40%
Kerry 1,142,283* 59%
Total Miami Area Votes
(See extended entry for detail data)
Bush 778,316 40.27%
Kerry 1,154,366 59.73%
Conclusion
In Broward, Palm Beach, and Miami-Dade counties the number of people who reported they voted for Bush matched the number of people who actually voted for Bush. In this case the exit polling data was devastatingly accurate.
Additional data available below the break.
]]>< 
I looked at the docs they generated. The Berkeley Summary page claims a Broward County excess vote for Bush of almost 72,000. That is to say Bush should have gotten 5,000 less total votes than he did in 2000.
And why is the Berkeley Summary so much different than what they put in their press release for Broward? (28,000 excess)
The numbers I gathered for Broward are here:
http://thedustyattic.blogspot.com/2004/11/uc-berkeley-studies-stats.html
has it occured to anybody that the reason that Bush’s numbers improved relative to 2000 is that all those 2000 ‘Buchanan’ votes were really for Bush, not Gore, and that the electronic voting systems used in 2004 kept those poor voters from making another mistake?
Steve,
The Buchanan vote was rather small in comparison. In Broward it was ~800, PBC it was ~3,400 and MDC was ~570.
I suspect this case has more to do with lousy programming of a computer model, than it has to do with lousy programming of voting equipment.
Steve,
The Buchanan vote was rather small in comparison. In Broward it was ~800, PBC it was ~3,400 and MDC was ~570.
I suspect this case has more to do with lousy programming of a computer model, than it has to do with lousy programming of voting equipment.
This study is bogus.
The authors claim that President Bush did significantly better compared to 4 years ago in counties with e-voting.
Any way you slice it, he did not. Bush’s strongest gains from counties WITHOUT e-voting.
More here: http://www.patrickruffini.com/archives/2004/11/fisking_the_ber.php
(1) The generation that fought in WW2 and Korea isn’t likely to vote for the “America is the bad guy” candidate.
(2) With the new machines, you can’t cheat in the traditional ways. The Bush bounce is just as likely to be a result of poll workers not being able to defraud the GOP out of votes this year.