Re: Please Evaluate This Ballot Selection Procedure

From: Joseph Lorenzo Hall <joehall_at_gmail_dot_com>
Date: Wed Jan 17 2007 - 13:30:50 CST

On 1/17/07, Charlie Strauss <> wrote:
> This recount is similar to one that has been widely promoted. In that
> scheme 10% of the ballots are chosen at random and recounted, the
> results are compared statistically to the whole. The procedure below
> is an elaboration on what is meant by random but does not change the
> essential statitical approach.

A blue-ribbon panel put together by the Brennan Center and ACCURATE
have studied, over the holidays, all the proposed audit schemes,
including this flavor (which we refer to as "polling audits").

One thing we point out is that it's important that the source of
randomness be publicly verifiable... for that, we advocate the
10-sided dice method of Cordero, Wagner and Dill. Granted, in some
precincts in some states, there can be as many as 7500 ballots cast...
at 10% sample would require 750 dice rolls, which is pretty
prohibitive (especially considering re-rolls when collisions occur).
One way to increase the efficiency of a dice roll sampling is to break
the ballots down into stacks of 10% (hopefully after a simple, not
thorough, shuffle). Then to counter-stack (toggle portrait and
landscape orientation) these stacks in stacks of 10 ballots. Then
roll 1% dice rolls (75 rolls... actually a bit more for collisions and
rolls of numbers greater than 750). Order the numbers and then select
the ballots from each stack.

This is all pretty intense and convoluted. Granted, the polling
audits have some interesting features that whole-precinct audits do
not (covered in the O'Dell and Simons papers linked to here: ). best,

Joseph Lorenzo Hall
PhD Student, UC Berkeley, School of Information
OVC-discuss mailing list
= The content of this message, with the exception of any external 
= quotations under fair use, are released to the Public Domain    
Received on Tue Jan 1 14:12:45 2008

This archive was generated by hypermail 2.1.8 : Tue Jan 01 2008 - 14:12:51 CST