Re: Boulder County Commissioners: 12/4 10:30am

[Ralph Shnelvar <ralphs@xxxxxxxxx>]

Dear Kell:

On Fri, 21 Nov 2003 21:01:37 -0800 (PST), you wrote:

>I'd like to know more about how handcounting would work before we accept or reject it.  
> 
>How many people does it take to hand count accurately 140,000 ballots, with, say, 25 different races/issues per ballot?
> 
>If Boulder County already has extreme difficulty recruiting the roughly 1500 people needed to run the roughly 245 precincts while the polls are open during the day, from whence would additonal hand counting labor come?  What would be the cost?
> 
>Could we get enough people to come in at 9:00 p.m. to count until 3:00 a.m. (or whenever) in the morning?  Should hand counting wait until the next morning to start?  Do any other areas our size in the US hand count?  How do they do it?  How does Canada do it?

I don't know how Canada does it but the way I would do it is that you would
feed the paper ballots into the scanners the way it is done now.  This,
then, would be an unofficial preliminary count.

After the newspapers publish the unofficial count then the official hand
count would proceed.  It would be the official (and very transparent) hand
count that the canvass people would certify.

> 
>How about sample batch counting?  Is there a generally recognized minimum statistically reliable hand counted quantity to prove the accuracy of the opscans?

Sort of.  It depends on how well sorted the ballots are.  If the ballots are
completely random (which they pretty much were in this last election) then
...

Statistics tells me that if I watch 1,000 votes the margin of error on a
random sample is less than 3%.  The table is, actually,

	Sample size		Margin of Error
	100				<10%
	500				<4.5%
	1000				<3.0%
	4000				<1.5%

It is a tiny oversimplification but, remarkably, the number of ballots cast
is irrelevant.  It is the absolute number of sampled ballots drawn from the
"population of ballots cast" that is relevant.

> 
>What if there is a discrepancy between the handcount and the opscan?

Then you hand count again.
 
> 
>If handcounting only, does a second set of hand counters hand count sample batches to test the accuracy of the first set of handcounters?

I don't understand.

> 
>Would handcounting be preferable to sets of opscans from different manufacturers, checked against each other?

No.  But it sure would keep the manufacturers honest.

> 
>What about opscans that photograph each ballot, save it on file, and tell exactly how that ballot's votes are counted, and teams of people compare the opscan decisions with the real ballots?  Is there such a machine?

I don't understand the point of photographing each ballot.

> 
>Anyone who has some data please pass it along.
>
>kell

Ralph


>
>Evan Daniel Ravitz <evan@xxxxxxxx> wrote:
>
>Joe, I'd like to make the case for hand-counting as a temporary
>measure (politicians like to put off the hard decisions, and this is
>one case where it makes sense).
>
>Evan
>
>On Fri, 21 Nov 2003, joe pezzillo wrote:
>
>>
>>
>> CVV is scheduled on the agenda of the business meeting of the Boulder
>> County Commissioners on December 4th, 2003 at 10:30am.
>>
>> We will have 10 minutes, including any questions from the commissioners
>> (who may also choose to invite us back).
>>
>> This is a public meeting, so please see if you can attend, and perhaps
>> even bring a friend.
>>
>> I presume our presentation will be prepared by our County Commissioners
>> sub group, please contact me ASAP if you would like to participate.
>>
>> In the meantime, we should also work on as much citizen outreach as
>> possible so that they have heard from many people before we arrive (and
>> hopefully we'll have a lot of signatures on our petitions, too!)
>>
>> Joe
>>
>
>---------------------------------
>Do you Yahoo!?
>Free Pop-Up Blocker - Get it now