January 04, 2005
More Ohio Statistical Abuse
In Ohio's Cuyahoga County, you have 1436 precincts.
203 total precincts with ballot spoilage > 3%, or 14.1%
218 total precincts with blacks > 80%, or 15.2%
Call these black precincts.
100 total black precincts with spoilage > 3% of ballots cast.
The conclusion that folks are drawing is that since, statistically, you'd expect only 31 black precincts to have a spoilage rate of above 3%, this is proof of discrimination.
Why is this not necessarily a good conclusion?
Doo doo doo doo doo.
Because of the size of precincts. You'd expect larger precincts to have a higher error rate just because of population density. Black precincts are almost always larger precincts. What you'd have to do is compare black precincts with nonblack precincts of similar density or size. If the error rates were the same, then you could have all the results above and still have it not be discrimination. If they were different, you'd have the evidence of discrimination.
Update: It is true that precincts tend to be around the same size. But they do vary, and there are probably other correlations along these lines to control for as well, before you can stake out a conclusion like this.