The Seattle P-I reported yesterday that according to a GOP statistical analysis, Dino Rossi won by about 100 votes, when accounting for ballots cast by felons, dead voters and non-citizens. How convenient.
How very, very convenient.
Indeed, to achieve this stunning reversal of fortune, several extremely convenient events would have to take place. First, the Republicans would have to prove that the number of illegal votes is anywhere near the thousand-ish they have alleged (not likely considering their track record thus far.) Second, the court would have to ignore offsetting errors from pro-Rossi counties… like the 1793 provisional ballots that were improperly counted without matching signatures. Third, the court would have to accept a statistical analysis. And fourth, the court would have to adopt the rather unscientific methodology the Republicans are proposing for their analysis.
As the P-I explains, Republicans are relying on a precinct level analysis to pad Rossi’s numbers.
For example, if Gregoire received 60 percent of a vote in Precinct A and Rossi 40 percent, and there were 100 improper votes from that precinct, Gregoire’s total would be reduced by 60 votes and Rossi’s by 40.
In most cases, the number of improper votes traceable to a single precinct is very small — often, as few as one. In that case, using the same percentages, Gregoire would lose 0.6 votes and Rossi 0.4.
The problem is, a single illegal vote per precinct is not a statistically significant sample. This results in what statisticians call an “ecological fallacy,” a logical flaw whereby aggregate statistics collected for a group are used to make inferences about an individual member.
For example, Democrats have argued that the majority of felons are male, and men tend to disproportionately favor Republicans, thus complicating any statistical analysis of the felon vote. One could also argue that felons tend to be, um… felons, and as such, would be less likely to vote for the state’s leading law enforcement officer.
Republicans, on the other hand, argue that such a demographic analysis would be absurd… that the only accurate means of deducing a voter’s behavior is to simply look at the totals for his precinct. But the absurdity of the Republican’s favored methodology can be clearly illustrated by using it to infer the ballot of a single voter: Stefan Sharkansky. If right-wing poster-blogger Stefan were an illegal voter casting a ballot in his heavily Democratic Greenlake precinct, the Republican analysis would subtract .78 votes from Gregoire, and only .19 votes from Rossi.
Of course, Stefan is not your typical Greenlake voter… but then, neither is a felon. And that’s exactly the point. The fact that Stefan is an angry, selfish, right-wing, extremist prick a Republican, is a helluva lot more predictive of his voting behavior than his precinct.
Thus any court judicially active enough to read statistical analysis into a contest statute that clearly doesn’t envision it, would be faced with a quandary: if statistical analysis is to be used to determine the winner of this election… which statistics should be analyzed?
No court is going to put itself into that kind of a quandary.