So, given the controversy over the recent election and the lack of ballots for some voters, I thought that it would be helpful to get some technical points clarified by someone who knows how this is supposed to work. The video is long. Almost 20 minutes. I decided not to edit it, but rather to leave it all there and not decide for you what is important and what is not. There are some breaks where I stopped the video and then started again. Mainly I didn't want to have such long sections that take forever to download. And sometimes we chatted while the camera was off. But everything I videoed is there, even the part where she suggests I cut it because she didn't know the answer. She's agreed that can stay in.
Here's what I got out of the interview.
- There should have been a lot more ballots sent out to the polling places. Lupe said when she used to do it, she'd look back three to six years and figured the highest turnout then add some more ballots for each polling place. The highest turnout in the last two mayoral election was 35% in 2006. If they had sent that many ballots out, there probably wouldn't have been a problem since only about 27% of voters went to the polling places to vote on Tuesday.
- While the folks that came to register and then vote or who weren't Anchorage residents because of Minnery's email, while problematic in their own right, were not the reason the polling places ran out of ballots. She estimated that Minnery's email wouldn't have gotten more than about 300 people to the polling places. We did the interview on Tuesday and Wednesday's Anchorage Daily News said there were 609 rejected ballots compared to 63 in the last mayoral election. So that's a little higher than her estimate, but not enough to run out of ballots if they had enough to cover 35%. There were 5,756 questioned ballots, three times, the number in 2009. These include people who go to the wrong polling place. But if half of the extra challenged ballots were the result of Minnery's email, the impact was higher than Lupe estimated, but still well under the 35% range of the 2006 election that they should have planned for. (Since there are about 204,000 registered voters, 1% would be about 2000 people. So even 4000 more people showing up would still only be in the 28 or 29% range.)
- There used to be two boards that monitored the elections.
- The Accuvote Testing Board which is made up of people working at the polls who test ballots to see if the machine counts them right.
- The Data Processing Review Board - this board has been eliminated. They used to test the machines before and after the election. They also sealed the cards into place to prevent tamperingl. Now one of the issues is that the seal was broken on a number of the ballot boxes. I did ask if there were ever any problems when they tested the boxes and she said no. Maybe that's why the Board was eliminated. But it seems that this is so fundamental to democracy that it's worth it to test.
- There's two kinds of programming
- The Deputy Clerk in charge of elections programs each ballot box. What this means is that she punches in information about what will be on the ballots for that particular polling place. Different polling places have different candidates and issues on the ballots, so each box has to be programmed separately.
- There is also programming of the card which enables it to accept the data from the Deputy Clerk and also tells it how to count the ballots. Lupe was not involved with this part, but this is the part that would seem to be the most invisible and be most vulnerable to someone with good IT skills tampering.
- Because Anchorage does not use touch screens and we have paper ballots, if there are questions, the ballots can be counted.
- This is probably an ideal time to have an investigation.
- Because none of the races is close, the investigation is unlikely to impact any race, so the pressure to impact a race will not be a factor in the investigation. Unless things are much worse than I suspect.
I would also mention that in some of our conversation when the video camera was not on, Lupe expressed respect for the professionalism with which the Municipal Clerk Barbara Gruenstein does her job. My own limited experience with Barbara over the years has also been very positive.
Note: I've used Viddler's onscreen comment feature to try to mark where I asked different questions. Roll over the little white dots on the bottom of the screen to see where different questions were asked. I'm afraid I was a little incoherent in some questions.
OOPS - posted under wrong story. This is where it belongs.
ReplyDeleteThank you, Steve! It looks like the information on how to run the election properly was either not properly passed on to the people running it this last election, or that they really had no understanding of the importance of each step. This interview should be broadcast on all the local news programs so that the general public can better understand what the correct procedure is and how is can be corrupted either by ignorance or by intent.
Friday, April 20, 2012 10:09:24 AM AKDT
Great interview,Dr. Aufrecht. Something odd, indeed.
ReplyDeleteThank-you very much Steve and Lupe for taking the time to record this interview.
ReplyDeleteLupe, I really appreciate your description of the method you used for determining the number of ballots to distribute to each precinct. It appears to be different from what they did this year.
I've done a preliminary analysis of the number of ballots that were originally distributed to the precincts at the beginning of election day. The number of ballots distributed to each precinct at the beginning of the day was approximately equal to the number of registered voters in each precinct in March 2012 multiplied by the percent voter turnout in the April 2011 election plus an additional 150 to 250 ballots. This rule works for about 90% of the precincts.
The curious thing about this allocation rule is that the difference between the number of ballots allocated to the precinct and the expected number based on registered voters and voter turnout last year. For most of the districts, this additional "buffer" amount is about 150 to 250 ballots. So for small precincts that have less then 100 registered voters, they received somewhere between 150 to 250 additional ballots at the beginning of the day above and beyond what would be expected based on past turnout and registered voters. And for large precincts with 500 or more registered voters, they also received about 150 to 250 additional ballots above and beyond the number expected based on last year's voter turnout and registered voters.
This method is flawed because it does not increase the number of additional ballots proportional to the size of the district. For small districts, they added as much as 100% additional ballots above the expected turnout level. For large districts they added less than 10% than the expect turnout level.
This creates a problem for large precincts. I looked through the list of precincts that either ran out of ballots before the end of the day or received additional ballots sometime after the poll opened. These precincts did not have sufficient ballots _at the start of the day_ to serve all of the voters that arrived to vote. Some were able to make up the deficiency by special deliveries from city hall and some were not. But it is important to note that three quarters of the precincts did not start the day adequately stocked with ballots.
Notably, nearly all of the precincts that did not start the day with adequate ballots were large districts. This is likely a direct result of the flawed allocation method that added a certain _number_ of additional ballots to each district instead of an additional _percentage_ of ballots to each district. The large districts did not receive a proportionally larger buffer of additional ballots at the start of the day.
A more reliable method would have determined the base amount of ballots based on the expected turnout as determined by past turnout rates and the number of registered voters. To this base amount, the method should have added a certain _percentage_ of that base amount to the ballots allocated at the beginning of the day. For example, a small precinct with 100 voters expected to show up would get another 80%, or 80 additional ballots as a buffer. A larger district with 500 voters expected to show based on the base calculation up would get an another 80%, or about 400 additional ballots as a buffer. This is not what happened this year. Both the small and the large precincts got somewhere between 150 and 250 ballots. The large precincts were shorted. Some made up for it with special deliveries and others were not so lucky.