Many on this board will remember the
study performed by UC Berkeley sociologists on the vote patterns of Florida counties in 2000 vs. 2004. The data they analyzed seemed to suggest that in most Florida counties, Bush's vote share % improvement from 2000 to 2004 correlated fairly well with his actual vote share in 2000. However, this same study demonstrated that two Florida counties, Broward and Palm Beach, were outliers, and the researchers suggested that that might be indicative of electronic voting fraud-related shenanigans in these two highly Democratic counties.
I have done a similar study on the country as a whole. In the chart below, on the x-axis is what % of the vote Bush captured in each of the 50 states in 2000; on the y-axis is by what vote share % Bush improved in each state from 2000 to 2004.
Most states fall within a certain band with an approximate correlation between what % of the people voted for Bush in 2000 and by what % Bush improved his performance in that state in 2004, similar to what the Berkeley researchers found for most Florida counties. However, the most heavily Democratic states fall outside this band, most significantly the solid blue states of the Northeast - New York, New Jersey, Connecticutt, Massachusetts and Rhode Island - as well as Hawaii. Bush made his most impressive vote share improvements in those states.
Secondarily, Bush also somewhat disproportionately improved upon his 2000 showing in the blue states of California, Illinois, Maryland and Delaware, as well as the red state of Tennessee. Tennessee can be easily explained because of the Gore "favorite son" effect artificially suppressing Bush's vote share in 2000.
It should be noted that Bush's vote share improvement in the heavily blue counties of Palm Beach and Broward seems to follow the same pattern as it did in the most heavily blue states around the country. Thus, the anomalous behaviour of the two Florida counties with respect to the rest of that state, found by the Berkeley researchers, is probably more indicative of the cultural and political similarities between South Florida and the Northeast, than it is of vote tampering.
On the flip side, there are a number of states in which Bush's 2004 performance was disproportionately worse. Bucking the trend of the rest of the Northeast is Vermont, in which Bush actually did worse in 2004 than in 2000. Bush also had disproportionately poor showings in Wyoming, Idaho, Montana, South Dakota (along with Vermont, the only state where Bush did worse in 2004 than in 2000), and North and South Carolina. North and South Carolina's appearance on this list can be attributed to the presence of John Edwards on the Democratic ticket. As for the other states, I would argue that Bush is slowly losing the libertarian elements of the Republican party. This probably also explains Vermont. Democrats take note.
The red dots correspond to states in which an anti-marriage initiative appeared on the November ballot. As I discussed earlier, the presence of these initiatives did nothing to improve Bush's relative performance in that state.
Finally, to the conspiracy theorists, the data for both Florida and Ohio are among the mass of states that occupy the middle band. I don't see anything that suggests that election fraud occurred on a large enough of a scale to affect Bush's performance in those states. Indeed, in the case of Ohio, if Bush did in fact do worse in 2004 as some conspiracy theorists suggest, that would make Ohio an outlier.
(This was originally posted on my blog here).
Update [2004-12-15 15:6:8 by DaveOinSF]:Yes, the title of this diary is provocative. That's marketing for you. And I acknowledge that a general analysis of all the data cannot prove or disprove individual fraud allegations. However, it is indeed possible via statistical analysis to determine whether Ohio and/or Florida are statistical outliers to the rest of the country, something that would indeed be the case if fraud did occur in either of those states on a large enough of a scale to change the outcome of the election. It is my contention that it did not.