Home  >  Kansas  >  Dead voters in Kansas?

Dead voters in Kansas?

By   /   October 28, 2010  /   6 Comments

Current Kansas voter registration records from the Kansas Secretary of State show 93-year-old Mildred Crouch lives at the Wichita Presbyterian Manor in Wichita.  Did Crouch vote in 2006 three years after she died, or was there some sort of clerical error?

Older voter registration records as far back as 2001 show the same address for Crouch. No one else with that name has been registered to vote during this time anywhere in Kansas.

Kansas VoterView, an online service provided by the Kansas Secretary of State, shows Crouch’s current voting history:

Voting History for Mildred Crouch

But an online obituary in the Lincoln Journal Star said Mildred Crouch died on Jan. 9, 2003 in Nebraska and could not have voted in Wichita in 2006.

In a phone interview on Thursday daughter-in-law Jane Crouch from Lincoln, NE, said  Mildred moved to Lincoln in 2000.  Jane Crouch said “that’s so wrong” when she learned records say her mother-in-law voted in 2006.  She added, “and people say there is no fraud in the election….. It’s ludicrous to say that fraud does not impact the election process here in America.”

Bill Gale from the Sedgwick County Election Office investigated the matter on Thursday:

She did not sign the book or actually vote in that election after all.  The notation on the page (made here in the office) shows that we had mailed her an advance ballot but that it was rejected.  I’m assuming it came back undeliverable by the post office.

In further looking at her record I see that she was on our permanently disabled list to receive a ballot each election at the time.  She was removed from the list later that year.

While removed from the permanently disabled list, she remains on the list of registered voters now but has been flagged as “inactive.”

It’s unclear why Mildred was never removed from the voting rolls since her mail should have been forwarded to Lincoln in 2000.  Jane Crouch believes a Nebraska death certificate had been issued.  But Kansas officials would not have received that notification.

Smith County Clerk Sharon Wolters on Monday in a phone interview described a similar problem with notifications from part-year residents.  “A lot of them will go somewhere else and die … we don’t get official data on that.”

In a phone interview on Monday Kansas elections official Brad Bryant said that Kansas does a cross-state check for duplicate registrations with about 11 other states, but because of limited resources Kansas had not checked against the Death Master File from the Social Security Administration for dead registered voters.

Death Master File

Kansas Watchdog extracted 1.2 million records from the Death Master File and compared them with the 1.7 million registered voters in Kansas.  A scoring system was used to track different kinds of matches and how “good” a match was considered.  The results show a potential list of 1,966 dead registered voters that could be removed once official verification has been established.  This list will have false matches that need to be discarded.  Both systems have different ways of treating data, and errors can occur in any data field.  See the Methodology section below for more information.

In roughly one out of three records (especially in more recent years), we have observed online information can be used to connect the voter registration information with the Death Master File information.  But this varies a bit based on exactly what matched between the two files.  Matching immediate family members from an obituary with voter registration address information can sometimes establish connections.

The Death Master File gave a “LastZIP” for Mildred Crouch of 68521, which is Lincoln, Nebraska.  Finding her obituary in a Lincoln paper instead of a Wichita paper was not surprising.

Graham County Clerk Jana Irby looked at the list of potential dead registered voters for her Western Kansas county:

I personally know two or three of the individuals are deceased. We are a small town and we know most of the people. We try to follow guidelines and use all of the tools provided to us. The National Change of Address, mailings, duplicate checks, cancellations from other jurisdictions, obituaries and death notices and felony conviction notices.

After the election we plan to look at all inactive voters.

When matches are close but not perfect between voter registration and the Death Master File,  it’s often not clear if there is an error in the data, or what other explanation might exist.

There was a close match between a “Donald V. Hitsman” and a “Ronald V. Hitsman” in Clay Center.  A relative on the phone knew there was a mismatch because Donald and Ronald were twins, which is why they had a common birth date and similar names.  Donald died in 2000.

About 39% of the 1,966 potential dead registered voters have been flagged as “inactive” and possibly could be deleted after the Nov. 2010 election in keeping with federal law.

Dead Voters?

From the 1,966 potential dead registered voters a check was made between the death date and the date the last ballot was cast.  This resulted in 75 potential “dead voters.”

Dead registered voters or dead voters?

Additional analysis of the 75 potential dead voters suggest 27 were likely false matches for various reasons.  Common names, like Smith, Johnson and Jones, accounted for 15 of the potential dead voters, which are almost impossible to verify because of the number of false hits.  There was inconclusive information on another 17.  Two were likely errors caused by confusion with a father and son with the same name.

But 14 registered voters deserved additional scrutiny from votes from 1996 to 2010 when records suggest a match to the Death Master File, including:

  • Mathew B. Andrews from Coffeyville died 9/24/2008 but voted in the 2010 August Primary according to voting history.
  • Is the William J Kelly who died on 2/7/2007 with a connection to ZIP 66013 the same as the voter with the same name and birth date from ZIP 66010 and voted in the 2008 general election?
  • Sr. Virginia Minton died in Atchison in 2001.  Someone with the name “Helen Virginia Minton” is still registered to vote in Atchison with a Mt. St. Scholastica address and last voted in 2002.  While it is an apparent match, we are trying to contact someone there for verification.
  • Voter J. E. Young from Lawrence matched the birth and death date (2/19/1994) of Jasper E Young who had a Lawrence zip code in the Death Master File. It’s unclear why Young is still on the list of voters since he last voted in the Aug. 1996 primary, possibly two years after he died.
  • Jeroldine Raredon from Overland Park died in 2005 but appeared to vote in 2008.  However, on Thursday, Brian Newby from the Johnson County Election Office confirmed the voting history was wrong because someone signed on the wrong line.
  • Dorothy Carlson, who lived at a nursing home in Overland Park, appeared to have voted in 2004 several years after the Death Master File said she died.  Newby investigated the matter and late on Thursday said, “I think, simply, history was wrongly applied to her.”  Newby said this was likely a scanning error.

The ultimate check on any of these suspicious votes is to compare the signature from the poll book with the reference signature card on file with an election authority.

Will dead registered voters or some of the 138,000 inactive voters with unknown mailing addresses affect the November election?  Kansas Watchdog will be watching and will provide updates after analyzing the data following the elections.

A list of prisoners in Kansas was not available for analysis until Nov. 10.


Methodology

The Death Master File was compared with Kansas voter registration data to find dead registered voters, including ones that show voting history after their death.

The Death Master File from the Social Security Administration contains about 87 million records listing all deaths since 1937 for those receiving Social Security payments.

Investigative Reporters and Editors through the National Institute for Computer-Assisted Reporting (NICAR) at the Missouri School of Journalism provide this data for use by journalists.

These Death Master File Notes provide descriptive statistics of the various fields in the database to explain limitations in their use.  Appendix C of that document explains the four fields used to create a state subset of the Death Master File.

The Kansas Death Master File subset currently has 1,260,464 records.

Kansas voter registration data is an open record under the Kansas Open Records Act that can be purchased from the Kansas Secretary of State.  The file obtained on Sept. 7, 2010 contained information on 1,712,726 registered voters.

The 1.7 million registered voters were matched with the 1.2 million Death Master File subset if the last name and the birth date (mm/dd/yyyy) matched.  This “minimal match” resulted in 15,405 potential matches.  This match ignored 526 voters with a birth date of Jan 1, 1900 and 249 with a birth date of Jan 1, 1801, which were dates likely used in the voter registration system as an indication of a missing date.

A scoring system was used to score all the records in the minimal match.  In theory, a higher score was a “better” match.  The scoring system was designed to know what matched and what did not.  This scoring system is described in a document, Matching Death Master File with Kansas Voter Registration Data.

Each of the approximate 30 different scores represented different combinations of matches.  Each set of matches was scanned for obvious mismatches that should be ignored.  This system needs to be tuned to result in fewer false matches, but a “fuzzy match” with data that can have errors will always have some false matches.

The initial match and review resulted in 1,966 records and were ordered from “best” match score to “worst” match score.  This was the potential dead registered voter list to use for further verification.

“Dead voters”, those apparently casting ballots after their death, were determined by studying voting history and the death date.  The easiest comparison with voting history was to use the 4-digit year from an election code with the 4-digit death year.  This process needs to be improved since the exact election date should be used, and many ballot codes must be ignored since they do not contain a 4-digit year.

A list of 75 potential “dead voters” was derived from the 1,966 potential “dead registered voters.”   Each of these voters was scrutinized to confirm a connection between the death and voting information.

Obituaries were obtained when possible from “America’s Obituaries & Death Notices” from NewsBank.  Newsbank news searches, prior voter registration information, and information found from Google was used to connect the voter information with the death records for verification.


Raw Data Results (Excel files):

Technical information about data fields in files above.

Social Security Numbers, voter registration IDs, street house numbers were cleared in the files above.  date_of_birth was replaced with year_of_birth.  These modified columns are shown with a gray background in the Excel file.


Related:


Paul Soutar in Wichita contributed to this story.

Contact: Earl F Glynn, earl@kansaswatchdog.org, KansasWatchdog.org

Click here to LEARN HOW TO STEAL OUR STUFF!