Tuesday, March 10, 2009

Statisticians and election systems

One of Colorado Voter Group's objectives is to achieve transparent verification of each election.

To my knowledge, there has been no peer reviewed specification of the election system components. Nor is there, to my knowledge, any study of the means by which each component/process can be verified.

None of the Colorado election systems I am familiar with are transparent. None of the Colorado election system components/processes that I am familiar with are verifiable.

All of the statistical work that I have reviewed to date has insufficient scope to be conclusive. For example, there has been no attempt to determine whether or not eligible electors have been disallowed from voting or ineligible electors allowed to vote. Counting the wrong votes correctly does not mean that an election is fair or accurate.

Without an agreed to specification of the entire election system it is far too early for a professional statistician to seriously undertake to test and make conclusions regarding the fairness and accuracy of an election.

Furthermore, the statistical work that I have reviewed to date seems to give insufficient consideration to the variation in populations being sampled, and treats unlike populations as though they were statistically identical. This would be like testing for the color of fruit by sampling only bananas in a carton of apples and bananas.

I ask that we not relinquish our quest for transparent verification. We must not accept simplistic statistical methodology that does not accomplish the goal.

To be clear, there is a place for statistical methods in election system verification, but the statistical methods used must be comprehensive and correct or they serve to further mask real problems. We must not delegate our voice on this vital matter.

Al Kolwicz
Colorado Voter Group

No comments: