Allen County, KS
Finding possible voter/election problems by comparing voter registration data to election results is quite difficult in Kansas because different precinct codes are used by different computer systems. I made that claim at a recent Secretary of State’s study group meeting about the future of elections in Kansas, and gave Allen County as an example of where comparing voters and ballots by precinct was quite difficult.
Retrospective studies of voter/election data may be a way to identify problems statistically that should be addressed, or to confirm results and improve the public’s perception of the integrity of the process. For example, in the never ending Franken-Coleman race in Minnesota, CNN reports “25 precincts in Minnesota now have more ballots than voters.” But can we verify that in all Kansas precincts the number of ballots cast in Nov. 2008 was not more than then number voters? Surprisingly, this comparison cannot be made using public sources of information.
As part of the Congressional mandates of the Helping America Vote Act, a new statewide system was introduced in Kansas in 2006. The ELection Voter Information System (ELVIS) replaced the 105 systems used by Kansas counties with a single statewide system. This new system introduced some additional data to analyze (e.g., additional voter history in most counties), but mostly brought uniformity on how voter data was handled. All I know about the new system is from the number of changes that can be observed in the voter file sold to the public.
Prior to ELVIS, precinct information associated with each voter consisted of a township, ward, and precinct. After ELVIS was introduced in 2006 I wasn’t sure what field(s) uniquely identified precincts, especially since the last known published list of precincts was from 2004. Questions to the Secretary of State and various county clerk offices failed to provide clarity about precincts.
A recent Jan. 22, 2009 voter file shows four fields have precinct data, but no definition or explanation of these fields is available:
- precinct_text_name (3277 unique values),
- precinct_text_designation (2559 unique values),
- precinct_part_text_designation (5682 unique values),
- precinct_part_text_name (81 unique values).
Results from the 2008 Kansas General Election from the Kansas Secretary of State said there were 3298 reporting precincts. 3298 doesn’t match any of the above numbers.
The number closest to 3298 in the above list of precinct fields is the precinct_text_name value of 3277. But, some of the precinct_text_name values are duplicated among various counties. The number of unique combinations, county-precinct_text_name, is 3467. Since not all precincts have voters, it’s unclear how the reported number of election precincts of 3298 should be compared with the observed number of voter precincts of 3467.
The problem is the voter system (ELVIS) uses one set of precinct assignments, and the election reporting system (called GEMS in many counties) appears to use another set of precinct names. Often, the two lists can be matched by a human for a given county, but an exact computer match rarely works.
After discovering this problem in 2006 in several counties, I tried to establish how big the problem was among all 105 counties. I started a detailed study of the first county alphabetically, Allen County — so I’m not picking on Allen County, that county is just the first alphabetically. But, based on the problems in Allen County I gave up any hope of making a statewide comparison.
Let’s review the Allen County voters from Jan. 2009 with the number of ballots cast in the Nov. 2008 general election to understand the comparison problem. Let’s assume relatively few changes in the number of registered voters in Allen County occurred from Nov 2008 till Jan 2009.
A tally of the voters by precinct_text_name for all Allen County voters was created using data from Jan 22, 2009 obtained from the Secretary of State. The tallies are shown in Table 1.
Table 1. Allen County Voters by Precinct in Jan 2009.
Allen County Jan 2009 Voters
The precinct-by-precinct results for the 2008 general election are published online by the Secretary of State in a large Excel file.
We are not interested in the individual vote tallies by candidates, but rather only the totals. The totals for each precinct for various statewide election contests are shown in Table 2.
One assumes the largest number of votes was for the U.S. President, but it’s possible some did not cast ballots for the president. We assume that the number of ballots cast is less than or equal to the number of voters in any contest, but that’s what we’re trying to check with data from the voter registration list.
We’re only interested in total ballots cast here, but as an example, in the table below the 156 ballots cast for U.S. President in Carlyle Township had this further breakdown:
- Ralph Nader, Independent, 0
- Jonathan E. Allen, write-in, 2
- Frank Moore, write-in, 2
- Barack Obama, Democrat, 38,
- Bob Barr, Libertarian, 0
- Chuck Baldwin, Reform, 0
- John McCain, Republican, 114
The undervote in the various contests shown below is somewhat expected. Most people voted for president, and almost, but slightly fewer, voted for U.S. Congress and U.S. Senate. Slightly fewer yet again voted for Kansas State Representative, and quite a few less voted for Kansas Senate.
Table 2. Allen County Ballots Cast in Certain Election Contests in Nov. 2008.
Allen County Nov 2008 Ballots
Our goal here is to compare the number of voters from Jan 2009 in Table 1 to the number of ballots cast in Nov 2008 in Table 2. Let’s just use the total of presidential ballots for this comparison, shown in Table 3.
Table 3. Comparison by Precinct of Voters in Jan 2009 to Presidential Ballots Cast in Nov 2008.
Allen County Voters09 to Ballots08
The above table shows a number of problems (in red) in comparing voters to ballots cast:
- Automatic exact computer mathches of precinct names would fail in all cases. While “Carlyle Twp” and “Carlyle Township” represent the same precinct, one is abbreviated and the other is spelled out. A person can make the connection, but without additional programming, such matches would be difficult by computer.
- Some precincts used to describe voters (e.g., Bassett City, Elsmore City, Gas City, LaHarpe City, Mildred City, Moran City, Savonburg City) do not exist in the list of election result precincts. It’s impossible to know what comparison to make in these cases, or where the election results for these voters are reported.
- The voter system before ELIVS always showed two Humboldt City precincts and four Iola City precincts. But since 2006 with ELVIS, voter information shows one Humboldt City precinct and one Iola City precinct, even though election results continue to show two Humboldt and four Iola City precincts.
- One word was spelled two different ways, Elsmore vs Elmsmore, and is included in both North and South precincts: North Elsmore or North Elmsmore, and South Elsmore and South Elmsmore.
- Four precincts which seem to match (but the match is inexact) show more ballots cast for president in 2008 than voters in those precincts:
- East Elm Twp has 125 voters, but reported 289 ballots cast.
- Marmaton Twp has 207 voters, but reported 391 ballots cast.
- South Elsmore Twp has 53 voters, but reported 78 ballots cast.
- West Elm Twp has 365 voters, but reported 512 ballots cast.
- Almost unbelievable: North Elsmore Twp reported 106 voters, but 105 ballots cast — a near 100% voter participation.
We need to use consistent names for precincts in Kansas between the computer system with the list of voters, and the computer system reporting election results.
Retrospective studies like this can identify possible voter/election problems statistically, so improvements can be made. But, we need better precinct data before retrospective studies are possible.
Editor’s Note: This article is an experiment in using Scribd for reporting tabular data in WordPress. Let me know if you have any problems with the Scribd data displays.
Related:
- Study Group for the Future of Elections in Kansas, Kansas Meadowlark, Jan 23, 2009.
