In the weeks before the November 2, 2004 Presidential Election partisan pollsters focused on which candidate was getting a bounce from which immediate issue of the day. Meanwhile writers on the AAPOR (American Association for Public Opinion Research) list serve were appropriately concerned more with which likely voter screens might have the highest accuracy at predicting who would vote on election day. During this period I told friends who asked that I thought the outcome would depend less on particular issues than on the size of the turnout. My reasons were not obscure.
Both parties concentrated on "energizing their base". But the Republicans, with the lower proportion of registered voters in key states, being always dependent upon the rural vote for victory, and historically having a higher percentage voter turnout among their registered voters could only improve just so far with increased voter turnout. After some threshold limit where the Republicans could improve their totals and percentages, most of the new votes would come from the urban cores where most people live and would represent strongly Democratic constituencies. These demographic shifts should have a greater impact than immediate issues. Although I did not have numbers from which to even estimate that threshold limit, I arbitrarily guessed it at a total increase of 5% in the electorate over 2000. And I suggested to friends that if the turnout went to 70% nationally Kerry would easily win.
The results defy not only my particular threshold guess, but this modeling, completely. And that will require a serious in-depth investigation. Despite the fact that the Democrats registered far more people in the past six months than the Republicans, and despite a huge voter turnout, with first time voters (according to Warren Mitofsky's poll) giving Kerry a 60:40 edge, President Bush appears to have increased his national vote total by 8 million votes compared with the 2000 election, yet Mitofsky saw no desertion to Bush from 2000 Gore voters (90% of Gore voters stayed with Kerry and 90% of Bush voters stayed with Bush).
How can these contradictory pieces of information be reconciled? They can't if Mitovsky's data is correct. So let's assume Mitovsky is wrong and there was some shift of former voters to Bush. One possibility is that an increased rural vote went for President Bush more heavily percentage wise than it did in 2000. However, there do not seem to be enough rural voters in the U.S. to improve that vote by more than perhaps a few million votes. A second is that perhaps Mr. Bush uniformly made major inroads in the urban-suburban areas and lost them by a much slimmer margin this time, adding vastly to his urban vote totals as well as to rural increases. Reviews of the actual major urban-suburban vote totals will confirm or refute this hypothesis. A third possibility is that Mr. Bush improved dramatically in some urban areas in particular and not in others. If such asymmetrical results were to be determinative in a few states such as Ohio one would have to ask the question "how did it happen?"
To begin with, I'd like to ask Carl Rove, known for his razor precision, how he called Ohio so early for Bush without public data to back his assessment. But the more valuable approach, were there to be significant non uniformity seen across urban areas, would be to carry out a study of results comparing urban counties in key states that had used the Diebold electronic voting machines versus those that had used other methods of voting; to also evaluate the turnout and results of each of these metropolitan areas comparing their 2000 and 2004 experience both controlling for and not controlling for a shift in the methodology to touch screen computers. And thirdly to consider the issue of potential absentee and provisional vote suppression if there are some urban areas with lower turnout, looking at the challenged voter experience (though this last concern is separate from the 8 million vote demographic issue).
During the run up to the election there was an e-mail spoof circulating that showed a Florida ballot with Bush and Kerry's names and the option to click on your choice for president. When you clicked on Bush he got your vote. When you clicked on Kerry the Kerry box moved and you could never catch up to it. Although this spoof was not to be taken seriously, a woman interviewed on network TV from Florida on election night anecdotally reported that although she had voted for John Kerry on the screen, the machine tabulated her vote for George Bush. The major networks were meanwhile praising the faultless experience with the machines. Let us remember that the computer software on these machines is proprietary and protected from public scrutiny. Because neither the polls nor the demographics appear to statistically explain the 8 million vote (16%) surge for Mr. Bush in this election, the 2004 Presidential race can not be declared final, free or fair without such studies. They are, of course, easy to perform for people in the business and could lessen any concerns of fraud.
Marc Sapir Marc Sapir MD, MPH Executive Director Retro Poll www.retropoll.org
