At the risk of drawing a baseball bat to the head from the mathematical guidos who run the numbers racket over at (un)Sound Politics, I thought I’d comment again on their wholly unscientific enterprise of predicting the outcome of our ongoing hand recount.
In an effort to prepare his “fellow Republicans for some possible bad news”, Jim Miller presents the somewhat plausible (though totally unsupported) premise that random counting errors are not random at all, positing that “people tend to make mistakes that favor themselves.”
And so I wondered if the current results — an admittedly small sample representing only 20% of total ballots cast — might support Jim’s thesis. Early returns have tended to come from Republican strongholds where Rossi won handily, thus he currently leads the recount 56% to 41%. If Jim is correct, that some of the newly found votes are the result of nonrandom bias, then Rossi’s lead in these new votes should be somewhat greater than his lead in the overall totals.
Out of 338 aggregate new votes, Rossi has gained 190, while Gregoire has gained 144. Pull out your spreadsheets if you wish, but off the top of my head that comes to about 56% for Rossi, 42% for Gregoire… almost exactly the same percentage as the overall totals. Apparently, these new votes have been distributed randomly between the candidates.
Of course, Jim might attribute this lack of discrepancy to the fact that Republicans are honest and Democrats are not. (I suspect he probably will.)
But more interesting to me is what this says about the statistical contortions his fellow numerologist Stefan is using to reinforce the notion that Rossi has already won… whatever the final vote count. Stefan knows how to write, so I can only assume that his long-winded recitations of obscure Excel functions are intended to confuse the reader, while fulfilling an apparently urgent need to constantly reassure himself of how smart he is.
So rather than torturing you with pages of calculations, permit me to propose a mathematical shortcut, based on the premise that the majority of these newly counted votes are not the result of human error during the recount, but rather, random tabulation error during the machine count. Since the overall vote count is roughly a 50-50 split, can’t we assume that statewide, the newly found votes should also split roughly 50-50?
Add in the 561 new votes in King County, and one would think that Gregoire stands a fair chance of winning the election.
Of course… all of this is complete bullshit, because there are so many factors that could screw up any projection in such a close race. Did King County already vet much of its discernible “under vote” during its more thorough recount? Are there more surprises to be uncovered in King or elsewhere? Will the King cavass board re-examine rejected ballots due to court mandate or public and party pressure? Will the rest of the new votes break as predictably as the first 20% have?
I’ve got no idea. And you know what? Neither does Stefan or Jim.