I received a few e-mails and posted comments over the weekend from staff at the Secretary of State's office with answers to some questions I had posted about the statewide voter database. I appreciate them taking the time to communicate directly with the Sound Politics audience. Their comments are worth reading, and I include them in their entirety in the extended entry --
From Asst. Sec. of State Steve Excell (also posted in a comment at an earlier entry):
Most of the duplicate registrations occur for two reasons:In response to my follow-up question wondering whether they're also looking for intracounty duplicates, Excell added:(1) Voters move around the state and do not notify their old county that they have moved. They simply register like a new voter in their new county. The new Voter Registration Database solves this problem. Voter Registrations are now transferable like your car title.
(2)Women voters who get married do not notify their county election official of their name change, but fill out a new form. Most remember to get their driver's license changed, but do not think about their voting registration. The new system does not yet deal with this problem, but we hope to be able to track maiden name changes in the future.
A note about the numbers. The initial dupe and death check turns up raw numbers. There are a lot of Robert Smith's or Mary Smith's born on the same day -- even living in the same city. An elections official must go though the dupes and compare the signatures. Many raw dupes turn out to NOT be dupes. Notations are added to the records so they do not turn up in future checks once we are certain they are two different people.
The dupes and deaths being turned up are from the initial load of the old databases that resided in the 39 counties. Now, with the VRDB centralized at the state-level, with systematic integrity checks, we expect smaller numbers in subsequent months.
The felon check will occur in March. The dupe and death checks will occur monthly. We will produce a statistical report of what we find each month in the subsequent month. A new version CD-ROM of the VRDB will be available after each check.
The VRDB is dynamic and in real-time so that means the NUMBERS ARE CONSTANTLY CHANGING as election officials go through reports and clear the records that turn up. By issuing statistical report AFTER all of the checking is done, everyone should have access to reliable numbers.
Yes, we will try to see if any dupes or the dead voted. We expect very small numbers, if any. The Governor's election contest ferreted out much of these and those persons' records have been cleaned-up since then.
The dupe check goes after First_Name, Last_Name and Birth_Date…whether in one county or cross-county-lines. These three fields exist for all voters. The reason most dupes turned out to be “cross-county” is due to most of the counties cleaning their own lists after the Governor’s Race Election Contest, but having no way to check with other counties. So, the results were heavily cross-county, but we did find some dupes within counties also.Sam Reed's press secretary, Trova Heffernan, posted this in the comments in response to my question wondering why the numbers in a KOMO news story differed from the numbers in a SoS press release:
Sam and I (and our Elections Division VRDB staff) were at a 2-day (Thursday and Friday, Jan. 26-27) Western Interstate Elections Official Summit in Portland --- exploring how to check for dupes across state lines. Two Election Assistance Commissioners were also present (Donnetta Davidson and Ray Martinez). California, Montana, Idaho, Oregon and Washington were all represented. Washington was one of the few states that was 100% on time for the HAVA January 1st deadline, and Washington State was the only state using the Social Security Master Death index to search for deaths nationwide. Other states were interested in doing the SSA Death Check and we’ve already sent info on how to buy the CD-ROMs. Oregon has 1 or 2 counties still to load and they will soon be 100% complete. I’ll predict that Washington and Oregon may end being the first two states to compare for interstate dupes. (Sam and Secretary Bradbury coordinated on this Summit and were the only two Secretaries at the Summit.) www.ElectionLine.org attended and covered the summit so you may find a news account there.
FYI: The SSA Master Death Index check did turn-up a large number of deaths out of state not reflected in our Dept. of Health’s vital records database. Example: a Clark County voter who dies in a Portland, OR, hospital would not show up on our DOH’s database – but does show up on the SSA Master Death Index check.
We are indeed cleaning up the voter registration rolls and we looking at ways beyond the minimum effort that Federal or state laws require.
The numbers in the KOMO story and the SOS press release are different b/c they were pulled from the database on different dates.Posted by Stefan Sharkansky at January 30, 2006 05:07 PM | Email ThisThe database flags computerized matches but human beings must then examine each match to determine if it is in fact a duplicate registration, a case involving a convicted felon or someone who is deceased. As election administrators examine these computerized matches each day, the numbers will continue to change. Once they finish, we will issue a complete list of registered voters and hope to do so next month. Even that list will not be perfect.
Most duplicate registrations exist b/c people move and forget to cancel their old registrations. They marry and forget to notify the County Auditor. Felony convictions will continue and the database will likely always hold names of the deceased.
Thank you for the opportunity to respond. This is an important issue.
After the 2004 election, we found numerous duplicates where the voter registered using their full name on one - then used initials or nicknames on others...
Examples include:
Patricia Marie smith
P M Smith
Patty M Smith
How does the database identify these as duplicates and deal with them? Can there be a cross check against the Drivers license database?
Or using Social security numbers?
It was a shot across the bow for WA Dems, but they were too busy partying down on the lower decks to notice. I'm sure a direct hit in November and in 2008 will sober them up a bit.
Posted by: Jeff B. on January 30, 2006 08:57 PMMy personal opinion is that the number and rough apportionment would more than make up for the election margin Queen Gregoire enjoyed.
Posted by: curious observer on January 30, 2006 09:04 PMI would like to give credit where due, and note that at least in my experience (a half-dozen or so visits to file initial and then final initiative paperwork over the last 3-4 years), the working-level people at the SOS - Elections Division office have been unfailingly courteous and helpful.
Another little-known and underappreciated island of non-partisan professionalism: The Washington State Office of the Code Reviser: They are the people who have to slog thru and put in proper legislateze every last one of the thousands of bills and hundreds of initiatives that are filed every biennium; and their helpful assistance goes a LONG way towards making initial amateur initiative language meet the requirements of the Code Reviser's style guide and RCW conventions.
Posted by: Methow Ken on January 30, 2006 11:13 PMON MONDAY FEBRUARY 27TH AT U.W. TACOMA AT 6:30 PM PAT MC CARTHY PIERCE COUNTY AUDITOR, APIRL BRINKMAN ELECTION INSPECTOR & LOLI ANDERSON FROM THE PDC ARE PUTTING ON A SEMINAR TITLED "AM I BLUE?; VOTING AND ELECTIONS IN WASHINGTON AND PIERCE COUNTY." THIS IS BEENING SPONCERED BY THE U.W. TACOMA AND THE "URBAN GOVERNMET LEADERSHIP
INSTITUTE"
LOVE TO SEE EVERYONE THEIR IT IS OPEN TO THE PUBLIC AND SHOULD BE GOOD FOR SOME LAUGHS!!!
It has been publicly identified that the past voter data base is flawed.
The Sec of States office and County Auditors appear to be taking a "minimalist" approch to cleaning things up.
Specifically, the minimalist approach is......
"...The dupe check goes after First_Name, Last_Name and Birth_Date…whether in one county or cross-county-lines..."
An agressive check would go after say same Birthdates and then let someone look at the First, Last Name to see if there were slight spelling differences or shortened first names that should be further investigated. There would also be a check on same addresses and a check on same last names.
The point is that those in charge of the data base only want to weed out the obvious. People who register as John James Doe or J. J. Doe or who live at the same address with the same name but have birthdays one day apart will continue to be able to vote more than once.
UNTIL THE SEC OF STATE GETS AGRESSIVE ON CLEANING THE VOTER ROLES, VIGILANTEEISM IS WHAT IS REQUIRED.
Stephan, Thank you for your monitoring this government largess and source of potential corruption.
Posted by: Bob on January 31, 2006 12:10 PMhttp://datamining.anu.edu.au/projects/linkage.html
Read the papers.
Matching on first name, last name, and ssn is not sufficient. Lotsa people pay big bucks for this kind of service. There's no easy fix.
For a more layperson introduction, search for "record linkage problem" (an extra credit self-study assignment).
In response to Mr. Cynical above: Everyone in all 50 states are already cross-referenced. And probably Canada and much of Latin America, too. The company is called Seisent. It was bought by Lexus/Nexis. You may recall the name from news about security problems last year. The government calls it "THE MATRIX". I've assumed that's the "data crunching" part of widespread illegal NSA wiretapping.
Is that really what you're asking for? You trust the government with that kind of comprehensive database? How about private corporations?
Just checking.
Posted by: Zappini on January 31, 2006 11:03 PMThe model Washington has (will have) will be "federated". Meaning the counties with their own pre-existing voter databases will keep them and the state will tie all the record together. Imagine the state using the counties has additional data feeds, in addition to the depts of health and licensing.
I'm luke warm on this approach. Note that King Co uses a DIMS (Data Information Management Systems) system (acquired by Diebold in 2003).
http://www.diebold.com/dieboldes/dims_vr.htm
I want to learn more about this system before I weigh in. For now, color me skeptical. It is Diebold, after all.
On the plus side, by going to a "federated" model, there's less disruption than if all the counties had to adopt a new system at the state level. So you get the benefits of state-wide duplicate detection and tie-in with health and licensing, without the switch over costs.
Posted by: Zappini on February 1, 2006 09:45 AM
It's not a matter of "giving" the government your data. Today, under our laws, your personal information is not yours. Anyone, including government, corporation, private investigator, can buy your data on the open market.
As I stated before, first name, last name, plus SSN number is not sufficient. I was surprised to learn that SSNs are not unique. That mistake has burned more than a few database designers who used the SSN as the "unique key". Further, SSNs are duplicated, used in identity theft, or just invented by illegal immigrants and others.
I don't know anything about ChoicePoint. They're probably using probablistic matching algorithms, please special sauce (a set of custom rules), just like everyone else.
I haven't researched the Seisent technology. It's on the to do list. What I do know is they can uniquely identify every individual.
Last I knew, Seisent sucks in 1,800 data sources (including public records), crunches the data, and trace everyone's path through life. And I mean everything. Everywhere you've ever worked, ever lived, every person you're affiliated with, every court document, every mortgage. Everything.
With Seisent's tech, it'd be trivial to incorporate voter registration information.
I should state an assumption I have: Maintaining a list of eligible voters isn't sufficient. To absolutely guarantee one citizen, one vote, you have to track everyone. Both voters and nonvoters. Including dead people.
That kind of tracking is occurring today, right now. I'm pretty uncomfortable with the situation.
Cheers, Jason