Re: Mandatory vote audit law language: A discussion thread-- choosing precincts at random for audit

From: <dr-jekyll_at_att_dot_net>
Date: Wed Jul 12 2006 - 12:32:27 CDT

I still prefer an allocation of choices amongst the stakeholders with the losing candidates' getting the lion's share of precincts to be audited. I prefer the TAR (Targeted Audit Recount) approach for the following reasons:

The stakeholders who choose the precincts to be audited can use these scientific methods to make their selections or they can use other methods.
If better scientific methods for choosing precincts are developed, they can be implemented immediately rather than wait for an act of Congress to allow change.
If the exact means of selecting precincts to be audited is written into law, that can help criminals decide how to circumvent it because they will have knowledge of what their opposition will do.
The stakeholders can also choose precincts to be audited based on evidence of fraud or negligence.
It keeps power in the hands of local people rather than concentrate political power in the central government. This protects from corrupt decision makers as well as from false allegations of corrupt or inept decision makers.
When people who lose elections complain about fraud, they usually have specific allegations. Why not build a system that can address their specific concerns?
It is better human factors for the people to be auditing a precinct where a stakeholder has chosen it rather than one selected randomly.

This email sent using 100% 
recycled electrons.
-------------- Original message from Jerry Lobdill <>: -------------- 
>From the discussion list:
Thanks for broaching this subject. Your story and your concerns are an illustration of why a law should be careful not to use such language as "chosen at random".  Some of the worst "deciders" about randomness are lawyers and election administration officials. :-) 
As a scientist who designed or analyzed simulators that required the generation of time series with specified statistical characteristics that simulate data series originating in natural processes and being processed through various signal processing systems, I have spent many an hour studying and arguing the merits of various simulation schemes.  Even experts in the business of building such equipment frequently make mistakes or become deluded and go down a primrose path. 
The term, "uniformly distributed", as you know,  is a mathematical term that refers to the probability density function that describes a population of samples of a given process.  The probability density function (a plot of "value" of a sample vs the frequency of occurrence of that value in the population) is flat; i.e., the frequency of occurrence is constant regardless of the value.  
For example, suppose that we have a box of marbles of identical size. There are, say, five colors of marbles, and 50 of each color. For this simple case there is a total population of 250 marbles, and the frequency of occurrence of each color is 1/5. The population is uniformly distributed. Suppose I put these marbles into a bingo drum that outputs one marble "at random" with every rotation.  Is this really an effective randomizer? We can test that hypothesis to any desired confidence level by conducting a "sampling with replacement" experiment. 
But the question is, does this model apply to the problem of detecting corrupted precincts in an election? What if the probability of corruption is not constant over all precincts? Then no matter what method we devise to ensure that our selection process doesn't favor any precinct or group of precincts, we are assuming a characteristic of the population of precincts that doesn't exist in our situation. That is, we are assuming that no precinct is more likely to be corrupted than any other. This assumption may ignore advance knowledge that we might have about an election and the county whose precincts comprise our population. 
What is our goal? Is it to be "fair", or is it to detect compromise?  
Stephanie wrote:
What does "random" mean??? 
Posted by: "Stephanie Frank Singer"   sleepy_sfs 
Sun Jul 9, 2006 5:56 am (PST) 
Here's a concern I have and don't know what to do with:
How do governing bodies implement "randomness"? We talk of "routine 
random audit", but how, in fact, would the precincts be chosen? My 
tentative answer (please send more ideas!) is that the precincts 
should be selected by the same physical mechanism used to select 
lottery winners.
We need to protect against enforcers who would protect misdeeds by 
making sure that suspect precincts are *not* chosen for audit.
This is not a merely theoretical concern. I went with the Committee 
of Seventy (a local election watchdog organization) to test 
Philadelphia election machines before the 2005 primary. Because the 
city has about 2000 machines, and there were maybe 10 of us on the 
team, we couldn't hope to test all of them. The person in charge of 
the Committee of Seventy effort, a lawyer, had chosen the machines to 
test in what he considered a random fashion. However, his method was 
flawed. So flawed, in fact, that there were over 1000 of the 2000 
machines that were guaranteed not to be on his list!!!
((OK, here are the details: Philly is divided into 66 "wards", each 
of which has between 20 and 30 precincts. Each precinct has (at 
least) two voting machines. This fellow had a list of wards (Ward 1 
through Ward 66) and had written opposite each ward a number between 
1 and 20. So he had chosen one precinct in each ward. He sent this 
list of precincts to someone in charge of the voting machine 
warehouse. When we got to the warehouse, the technicians had open 
one machine from each of these precincts. But remember, each 
precincts has at least two machines! So the technicians got to 
choose which machine to open. Ask Marian Vos Savant if you don't 
believe me: this really screws up the works. Plus, the technicians 
had advance knowledge of which machines would be inspected, so any 
nefarious ones among them would have a chance to cover any exposed 
By the way, the word "random" has two related but different 
meanings. One is the definition you would find in, say, a graduate 
mathematics text book. The other is a collective general public 
understanding that corresponds to what mathematicians would call 
"random, and uniformly distributed." Here's an example. Suppose I 
have a trick coin, weighted so that, on average, it will land on its 
head 99 times out of 100. Can I use that coin to choose between two 
precincts under the law? According to the standard mathematical 
definition, the results of that coin toss are in fact "random", since 
there is no way to predict when the unlikely event of "tails" will 
occur. However, surely we don't want that kind of mechanism for 
precinct selection -- if we had two precincts to choose from, we 
would want to use a fair coin. (Of course, we're talking about 
picking, say, 200 out of 10,000 precincts, not one out of two. But I 
wanted to distill the idea into its simplest form.)
Cheers from one of your resident mathematicians, 

attached mail follows:

OVC-discuss mailing list

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 Mon Jul 31 23:17:03 2006

This archive was generated by hypermail 2.1.8 : Mon Jul 31 2006 - 23:17:09 CDT