Archive for the ‘Election Auditing’ Category

Were 2014 United States Senatorial and Gubernatorial Elections Manipulated?

Friday, April 17th, 2015

Is it possible electronic vote-count manipulation determined who won any 2014 United States’ elections? The probability that disparities between predicted and reported 2014 election gubernatorial and senatorial vote margins were caused by random sampling error is virtually zero. Larger than average magnitude disparities are exhibited primarily in states conducting no or low-quality post-election audits. An analysis of Maryland’s election data by ballot type statistically confirms vote miscount as an explanation for its unexpected outcome.

Maryland, Illinois, Florida, and Kansas gubernatorial contests exhibited sufficient disparities between polls and election results (PED) to alter election outcomes; all used inauditable voting systems or failed to conduct post-election audits (PEA)s. Vermont’s PED was within one percent of sufficient to alter its outcome. In Nevada, Tennessee, New York, Ohio, and South Dakota PED were large but smaller than winning margins.

Kansas and North Carolina senatorial contests exhibited sufficient PED to alter election outcomes and no audits were conducted. Virginia’s PED was within one percent of sufficient to alter its outcome. In Arkansas, Wyoming, Tennessee, Kentucky, and Nebraska PED magnitude were large but smaller than winning margins.

A case study of Maryland’s unexpected 2014 gubernatorial outcome affirms there is, as yet, only an explanation of vote manipulation consistent with the statistical disparity patterns in Maryland’s pre-election poll predictions, and its partisan voter registration, turnout and vote data by ballot type.

The full study is available at Social Science Research Network http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2524093
A shorter, less technical version that could be passed on to politicians or officials is available via email request.

When should a candidate concede a close election contest?

Tuesday, April 19th, 2011

No candidate in a close contest should concede (ever!) without first taking the following steps:

  1. Obtain detailed vote tallies broken out by ballot type (mail-in votes, polling place, early, provisional,…) and by precinct for each county if possible. New Mexico is the only state currently publicly posting the detailed vote tallies necessary to reveal patterns consistent with ballot box stuffing, vote switching, and subtraction differing by ballot types that are hidden by aggregating tallies.

    In counties delivering mail-in ballots to the precincts to be fed into precinct-based machines rather than counted at the election office, there may be no convenient way to separate the mail-in ballot tallies.

  2. Reconcile the number of voters and ballots for each jurisdiction (county, parish, or township) by counting all blank and spoiled ballots and voters to account for all printed ballots (including examination of blank ballot orders and receipts from the printer, perhaps verifying with the printer directly) in at least any counties having anomalous patterns (patterns showing margin differences in the mail-in versus polling place counts or in counties showing unusually high precinct turnout rates or unusual patterns compared to prior elections or other contests), by examining all poll books and all absentee ballot envelopes mailed in by voters with their signatures to verify the number of eligible voters who cast votes.

    If an election administration official cannot produce the unused blank ballots for reconciliation, examine the absentee ballot envelopes in a forensic manner to look for possible forgeries and compare the number of absentee ballots cast with the number of absentee ballot voters and compare the names of absentee ballot voters with the candidate’s version of the voter registration rolls.

    Note: Ballot-on-demand systems make the type of reconciliation described here virtually impossible. Thus, a different set of safeguards and procedures must be developed to try to reconcile ballot-on-demand systems.

  3. Confer with a computer voting system expert and request copies of all the voting and operating system electronic log files and the reports available for the particular voting system(s), and have the expert examine those files for evidence of tampering with electronic records of vote tallies.
  4. Analyze the results of the above steps and do forensic analysis of the ballots and other records if there are unaccounted for blank ballots or vote tally patterns consistent with ballot box stuffing or ballot switching, or if ballot tampering is suspected.
  5. Manually count sufficient voter-marked paper ballots to verify the publicly posted machine tally accuracy because machine retabulations do not detect voter-marking errors or all types of machine miscount.

There is a federal right to examine voter registration records (poll books and mail-in ballot envelopes) under the National Voter Registration Act of 1993, so the candidate’s attorney should have no problems obtaining access to these voter registration records necessary to evaluate the accuracy and currency of voter registration records, as well as the integrity of ballots.

By obtaining the above information (such as the number of unaccounted for blank ballots ordered from the printer that could have been used for ballot box stuffing or substitution for voters’ ballots) the candidate might be able to justify a forensic examination of ballots and a manual recount rather than mere machine retabulation that does not detect ballot box stuffing, ballot substitution, voter-marking errors, or many types of machine miscount.

Caution: These procedures may not detect ballot tampering, such as deliberately spoiling votes cast for opposing candidates by over-voting cast ballots, and will not detect all types of vote switching by electronic ballot (digital recording electronic) voting machines. Digital recording electronic machines should not be used to cast anonymous ballots because at least four types of DRE vote switching schemes are undetectable by any post-election auditing method.

Unfortunately, today virtually all US jurisdictions use the discredited principle of security by obscurity giving inside administrators and their assignees opportunity to tamper with, pad, or substitute ballots or miscount votes. To be able to detect all types of post-election insider ballot or vote tampering would require publicly verifiable oversight over ballot security, which is virtually nonexistent, to my knowledge, in most states today. Post-election forensic analysis of ballots could be a substitute for publicly verifiable ballot security.

What state or federal laws would help to ensure honest elections?

State or federal laws would help to ensure honest, accurate election results by requiring jurisdictions to:

  1. use voter-marked paper ballots for all able-bodied voters;
  2. prior to the election, allow candidates and political parties to compare the printed ballots for all precincts with electronic ballot definition files to be used to electronically assign votes to candidates based on ballot positions;
  3. publicly post all polling place results at the polls during poll closing and publicly post detailed vote tallies (election results) broken out by precinct and by ballot type (polling place, absentee, early, provisional) on the state election web site (very small precincts can be combined if desire to protect voter anonymity);
  4. publicly post and allow public comment on, oversight of, and participation in ballot security procedures before, during, and after elections, including during transportation;
  5. preserve and secure all cast, unused and spoiled ballots, computer and voting system electronic log files, voting machine memory devices, mail-in ballot envelopes, poll books, and so forth that are necessary to reconcile and audit elections;
  6. allow public access to all electoral records necessary to evaluate the accuracy of election outcomes;
  7. conduct routine post-election independent manual counts to check the accuracy of the publicly reported electronic vote tallies of at least sufficient tallies to detect, with 99% probability, the minimum amount of miscount that could cause an incorrect election outcome, assuming any miscount is well-hidden in the fewest number of tallies necessary to cause an incorrect outcome, given the initially publicly reported tallies; and
  8. preserve a copy of the voter registration rolls as they existed when polls closed on Election Day, prior to continuing to do updates.

Please pass along this information to any candidate in a close contest who is declared the loser according to initially reported results.

Nate Silver, in NY Times Blog, Claims Vote-Counting Error In Wisconsin Points to Incompetence, Not Conspiracy

Friday, April 8th, 2011

The National Voter Registration Act of 1993 (NVRA) provides a public right to access electoral records necessary to evaluate the integrity of voter registration rolls, but U.S. citizens have no federal right to examine electoral records necessary to evaluate the integrity of vote tallies.

Does Wisconsin Count Votes Correctly?

http://www.fairelectionswi.com/Counting.htm

Despite the lack of any supporting evidence for vote count accuracy, American Politics scholars, whose published opinion poll work often assumes the accuracy of election results, often claim US election results are accurate, even in close contests where small percentages of manipulation could alter the outcome.

Vote-Counting Error In Wisconsin Points to Incompetence, Not Conspiracy
By NATE SILVER
http://fivethirtyeight.blogs.nytimes.com/2011/04/08/vote-counting-error-in-wisconsin-points-to-incompetence-not-conspiracy/?nl=us&emc=politicsemailemb2#preview

—————

According to my cursory analysis of the Wisconsin election results data by county found here: http://www.jsonline.com/news/statepolitics/115529044.html Nate Silver’s analysis’ assumes, without any supporting evidence, the following about elections in Wisconsin counties:

1. November 2008 elections were fair and accurate

2. November 2010 elections were fair and accurate

3. February 2011 elections were fair and accurate

In other words, the election officials and technicians who count U.S. votes with trade secret software and without public oversight, are infallible and honest in both prior elections and in other Wisconsin counties

4. there tends to be higher average voter turnout per precinct in counties tending to vote Republican (more ballots cast per precinct)

5. there tends to be higher Republican vote share in counties which use trade secret software to count ballots than in counties manually counting ballots (My cursory analysis shows the Democratic candidate won in the manually-counted counties and Republican won in the machine counted counties. However, I did not have time to phone counties that I was unsure of which category they fall into based upon the current data at Verified Voting. I hope someone will compile a validated list of the vote tallies counted by each method in each county – the candidates could do this.)

I do not believe that Nate Silver has provided any evidence in support any of these claims which are logically required for him to claim the findings he makes from his analysis.

Unfortunately, publicly verifiable evidence of election outcome accuracy is unavailable in most US states today because election officials and their assignees have successfully asserted a special legal right to secretly count votes, denying the public the right to verify or participate in ballot security or tally procedures under the discredited concept of “security by obscurity”. Most U.S. counties or townships do not make their records for reconciling the number of printed ballots with the total number of voters casting ballots available publicly for all types of ballots, including mail-in and provisional ballots.

Although Wisconsin requires post-election manual audits, its audits are conducted by the same persons who initially count and process ballots; and the manual audit sample size is grossly insufficient to verify the accuracy of close contests like this one. In Wisconsin, a recount does not involve a 100% manual count of all ballots, but rather allows the same persons who conducted the initial count to re-count the ballots using the same machines again. The public is usually not able to verify that ballots are not tampered with, substituted, stuffed, and so forth because election officials adhere to the principle of keeping ballots secure from public by keeping security procedures secret and unverifiable by the public. The security-by-obscurity principle makes it easy for insiders within any system to undetectably produce errors, deliberately or by mistake.

In addition, only one state in the U.S. New Mexico, publicly reports its vote tallies broken out by ballot type. In all other U.S. states, vote padding for one candidate in one type of ballot and undercounting votes in another type of ballot for an opposing candidate are hidden by adding the two results together. As long as the amount of vote stuffing for one candidate is less than the number of undervotes for another, the evidence of both problems is hidden in the aggregated tallies.

Other countries would do well not to follow the lead of the U.S. in using trade secret electronic vote-counting devices allowing fraudulent vote manipulation to become virtually undetectable.

This election is high stakes and close enough to merit a full manual recount of ballots in all Wisconsin counties that are immediately secured by a court, regardless of which party is reported to win.

The losing candidate should ask for a detailed list of tallies from each county, broken out by ballot type and counting method and examine all precinct poll book records and absentee ballot envelopes voter signatures, counting to try to evaluate whether the number of voters voting equals the number of ballots cast, and to detect any anomalies in the poll book records. The candidate should also manually count and examine ballots forensically in order to try to detect possible ballot box stuffing, tampering, or substitution which may have occurred before, during, or after the election. Ideally, someone familiar with post-election auditing and forensic analysis of election records should assist the candidate.

Statement by Ramona Kitzinger, member of Waukesha Board of Canvassers since 2004
Monday, April 11, 2011
http://www.orchidforchange.com/parties/waukeshadems.com/ht/display/ArticleDetails/i/1343504

Waukesha County reported highly unusual 97% voter turnout in the 2004 election. The number of voters casting ballots in the current election could be verified against the number of votes in a forensic analysis:

http://www.waukeshacounty.gov/uploadedFiles/Media/List_Documents/County_Clerk/2004_Official_Election_Results/Summary_Report_Nov2_2004.lst

Comments by Bev Harris are apt:
Bev Harris 9:29pm Apr 15

Kathy Nickolaus is ruled out in terms of tampering with ballot counts from Brookfield for last week’s election. The City of Brookfield posted its results on Election Night, and the belated results sent by Kathy Nickolaus are identical to those posted on Election Night. Regardless of whatever illegal, uncertified, murky little data system Nickolaus uses, she was not in a position to alter the Brookfield count. Now the Brookfield City Clerk — that’s another matter. Like all other town and city clerks in Wisc. she and insiders who work for her WERE in a position to have their way with the election. But that swings both ways. Milwaukee is notorious for election tampering and it swings Dem. So for those wanting recounts, be careful what you ask for.

New York State’s post-election audits provide little confidence in close contests.

Thursday, December 30th, 2010

This post is in response to a December 29, 2010 editorial in the Times-Union, “Time to fix this election law” by Jay Jochnowitz, Editorial page editor.

“Today’s editorial: When an election is as close as the one in the 7th Senate District was, a recount should be automatic, not left to the judges’ discretion.” http://blog.timesunion.com/opinion/time-to-fix-this-election-law/8371/#comment-2671
——–

Great editorial by Jay Jochnowitz. The Board of Elections, via its own regulations on post-election auditing, gave away its authority to ensure accurate election outcomes by:

1. requiring every county separately audit a flat 3% rate of “machines” that each county selects;

2. neglecting to require the counties to publicly post, or to submit to the Board, election results broken out by precinct (or district) and by absentee and Election Day (thus allowing unaudited district tallies to be manipulated to match any fraudulent overall tally and the audits be not publicly verifiable. In auditing any field, the data needs to be committed prior to the audit);

3. exempting all absentee ballots from any manual checks;

4. not allowing the public to observe the audits, unless appointed by a candidate (unlike most other states);

5. not requiring the counties to reveal their ballot security procedures or allow the public to verify or assist with verification of the security of ballots; and

6. not requiring any jurisdiction-wide reconciliation of the printed ballots to detect any possible ballot substitutions or ballot box stuffing.

The Board of Elections could work with the State Legislature to use the same amount of resources more efficiently and effectively by

1. allowing the audit units to be smaller units than machines (such as districts) as long as the unit tallies are publicly reported prior to the random selections, and

2. allowing risk-limiting audits rather than requiring 3% flat rate audits,

3. requiring random selections from *all* the ballots (reported tallies) that sum to the overall total, and

4. allowing random selections weighted by the upper margin error bounds within each publicly reported tally, rather than requiring a uniform selection probability.

The state legislature needs to change a few laws to enable the Board to exercise more flexibility and use resources more efficiently and effectively, and the Board of Elections needs to work with a mathematician who understands how to conduct risk-limiting post-election audits so that the sample size automatically goes to 100% when needed to limit the risk of certifying incorrect election outcomes.

Jay correctly notes that a 3% sample of “machines”, excluding all absentee ballots from examination, allows a small number of precincts or districts tallies or absentee ballots to be manipulated to change the outcome of a very close contest like the Long Island NYS Senate race, or several other NYS contests in November 2010.

New York State’s current post-election audit procedures provide virtually no public confidence in any close contests.

Steven Colbert — Inside the E-Voting Machine

Wednesday, November 3rd, 2010
The Colbert Report Mon – Thurs 11:30pm / 10:30c
Indecision 2010 – GOP Takes House
www.colbertnation.com
Colbert Report Full Episodes 2010 Election March to Keep Fear Alive

2010 November Election: Questions the Press could ask about State Voting Systems

Tuesday, November 2nd, 2010

The press can help ensure the integrity of U.S. elections by asking election officials questions about the public verifiability of the accuracy of state election outcomes. For instance:

1. Does the state use auditable voting systems allowing the accuracy of publicly reported vote tallies to be checked? (I.e. voter marked paper ballots because digital recording electronic (DRE) voting machines can be fraudulently manipulated without possibility of detection and studies have shown 2/3rds of voters do not bother to check the accuracy of their paper roll ballot printouts and less than 9% of voters accurately check their paper roll ballot records .)

2. Is the public allowed to oversee the security of ballots from the time ballots are printed until after the post-election manual accuracy checks (audits)? (I.e. How do we know that the ballots counted are the same ballots cast by voters?) This is particularly troublesome for absentee mail-in ballots. Are mail-in ballots only opened and processed only in a public meeting? What are the security procedures for ballots?

3. What are the machine tallies and number of voters who cast ballots for each precinct for each ballot type (absentee, early, provisional, and polling place)? If detailed vote tallies are not publicly reported, then aggregated tallies hide any ballot box stuffing or vote padding for one candidate in one type of ballot and subtraction of votes for another candidate in another ballot type or by normal undervoting. When the reported tallies are aggregated, even if the number of voters voting in each precinct is publicly reported, absentee ballot padding is undetectable as long as less padding occurs than the number of polling place undervotes. Aggregating tallies also hides suspicious-looking counts occurring in any one ballot type.

4. Are post-election manual counts performed to check the accuracy of publicly reported machine tallies? If so, are the machine tallies publicly reported before randomly selecting which tallies to manually count showing that all the precincts and ballots were included in the random selection that add up to the overall results? Note: In some states the machine tallies selected for manual auditing are never publicly reported and shown to add up to the final result. Such audits are mere showmanship.

5. If manual post-election audits are performed, are they performed prior to certification; is the public allowed to observe the manual counting process; and are more vote tallies manually audited if the contest margin is close, than if the contest is less competitive? Close contests can be stolen by fraudulent manipulation of fewer vote tallies, requiring larger audit sample sizes than wide-margin contests.

We need to know if trade secret machine counts performed by private companies are accurate and if the ballots counted were the same ballots voters cast. In states where inauditable (DRE) voting systems are used or in which no post-election manual counts are performed to verify machine count accuracy of paper ballots, close contests are susceptible to undetectable fraudulently manipulation.

Pre-election voting machine testing cannot detect all innocent programming errors and almost never detects deliberate election tampering.

The right to cast an anonymous ballot does not confer a special right to election officials or to private voting vendors to secretly count U.S. votes without public oversight.

For information on what type of voting systems and manual audit provisions each state uses, see http://www.verifiedvoting.org/

For a computer scientist, voting system expert to interview:
Rice electronic voting expert available through Election Day
http://www.newdesignworld.com/press/story/207442

For detailed information about the hackability of any DRE voting machines (with or without paper trails), see:
Evaluating the Security of Electronic Voting Systems
http://www.cs.ucsb.edu/~seclab/projects/voting/index.html

Posted by kathy dot dopp at gmail dot com

Are New York’s Voting Machines Secure?

Tuesday, September 14th, 2010

Today, Tues. Sept. 15, 2010, I arrived at my polling place at 10:30 a.m. (the polls did not open until noon). I vote at a fire station in Albany County. I found the room open and empty and the voting machines and materials sitting unguarded. According to a fireman in another part of the building, the voting machines were delivered and left unsecured in the upstairs voting room, accessible by an unlocked backdoor, since last Thursday. I took 8 photos. Here are three:

Here is what voting security expert, Professor Douglas Jones, U of IA computer science dept., has to say about it:
——————
Doesn’t look very secure to me. Tamper evident seals are nice, but I have no reason to believe that someone couldn’t have lots of fun with them given the time they were accessible.

The mere fact that nobody noticed you as you lifted the covers and took pictures suggests inadequate security.

I looked up the security seals from A Rifkin Co. on the web:
– http://www.secureelection.com/catalog/index.php?function=viewitem&idn=265

It’s a classic tamper-evident seal that says “void void void void” if you peel it off. It’s easy to peel off, so an attacker intent on having the results of your precinct thrown out could have, in the time you were there with the machine, peeled off a few seals so they say “void”, then after the polls close, demand that the results from that precinct be thrown out on the grounds that the ballot box had been tampered with (assuming that the election officials notice the voided seal).

My advice has long been that exposed seals are themselves a problem unless they are protected from casual tampering — I first saw this problem in Miami many years ago, when several seals were broken during pre-election testing.

Doug Jones
————————–

I have posted a recent research paper I wrote on NY State that provides a list of eight other weaknesses in New York State electoral laws, regulations, or procedures, along with a list of eleven electoral reform suggestions at http://electionmathematics.org

Kathy Dopp
kathy dot dopp at gmail dot com

Fundamentals of Publicly Verifying Election Outcome Accuracy

Thursday, September 9th, 2010

Kathy Dopp, MS Mathematics, Ph.D. Student
kathy dot dopp at gmail dot com
August 28, 2010

I developed this with the United States in mind, but have tried to make it general to verifying the accuracy of electoral outcomes obtained by electronic tallies of paper ballots. All of the following conditions are necessary to avoid certifying incorrect election outcomes:

  1. Independently-auditable voting systems, i.e. voter-marked paper ballots, not electronic ballots
  2. Public observation of, and participation in (a) manual counts or audits and (b) secure handling, storage and transportation procedures for electoral records, including ballots.
  3. Initially manually counting all ballots or, at a minimum, counting of adequate vote tallies (audit units) to limit the risk of certifying any incorrect initial election outcome by detecting miscount even when the fewest audit units that could cause an incorrect outcome are miscounted. (If calculations assume maximum margin error within each audit unit is less than its upper margin error bound, then candidates may select discretionary units for manual auditing in addition to the random sample.) If not initially manually counting all ballots:
    • Public reporting of all vote tallies (audit units) used to tally overall election results prior to randomly selecting a sample for auditing
    • Publicly-verifiably fair random probability selection methods (Calandrino, Halderman, & Felten, In defense of pseudorandom sample selection, Woodrow Wilson School of Public and International Affairs. 2007), preferably weighted by within audit unit upper margin error bounds rather than by a uniform sampling distribution
    • Prohibiting ballot and electoral record access between the time of random selections and manual audits
    • Expansion of the sample size, perhaps to a full recount, or certification of the election using an algorithm based on the premises of the sampling method, treating any missing paper ballots as discrepancies when deciding whether to expand or certify
  4. Polling place and jurisdiction-wide reconciliation of printed, used, unused, and spoiled ballots with absentee ballot, polling place and other electoral records
  5. Timely public access to electoral records necessary to evaluate the accuracy and integrity of reconciliation and manual counting or auditing processes
  6. Voter intent as the standard for manually counting ballots
  7. Public reporting of any discrepancies found during the manual count or audit, using manual counts to correct any initial reported results
  8. Completion of the manual count or audit prior to certifying election results

Does the state you live in follow all these procedures?

Why are each of these conditions necessary to verify election outcome accuracy?

An example of how elections can be stolen if each of the 12 conditions alone is missing. (Dozens of additional ways to rig elections exist when more than one condition is missing.)

1. There are at least four ways for a single insider to rig digital recording electronic (DRE) voting machines in an entire county with a few seconds access using an inexpensive memory stick, in a way that pre-election testing could not detect. (Are Your Votes Really Counted? Testing the Security of Real-world Electronic Voting Systems, D. Balzarotti, G. Banks, M. Cova, V. Felmetsger, R. Kemmerer, W. Robertson, F. Valeur, and G. Vigna, in Proceedings of the International Symposium on Software Testing and Analysis, Seattle, WA July 2008.) http://www.cs.ucsb.edu/~seclab/projects/voting/

2. (a) Miscount, misreport or under-report discrepancies found during manual counting (b) Substitute, lose, tamper with, or spoil ballots or stuff ballot boxes

3. Rig close elections by hiding miscount in the smallest number of precincts possible to avoid detection by a small fixed-rate audit, or use sham audit procedures that never check the accuracy of reported vote tallies

a. Misreport all audit unit tallies that were not previously randomly selected for auditing

b. Sampling can be fixed to audit only tallies that will match reported tallies (as in OH 2004)

c. Tamper with ballots selected for auditing between the time of the random selection and the audit

d. Failure to properly analyze the amount of errors or missing ballots can cause election certification of incorrect election outcomes despite evidence in the sample

4. Lack of polling-place reconciliation covers up polling place ballot box stuffing or ballot substitution and lack of jurisdiction-wide reconciliation covers up ballot box stuffing or ballot substitution when mail-in voting or early voting are allowed

5. Lack of public access to records permits fictitious numbers to be reported during reconciliation, covering up certain types of ballot substitution, fraudulent vote reporting and ballot box stuffing

6. If systematic errors in ballot design, or voter understanding are not identified in the audit process, it could result in certification of an incorrect result. Voter ballot marking errors or poll workers misinforming voters on how to fill out ballots may cause an outcome different than desired by voters

7. If all discrepancies are not reported and recognized as such, many may be explained away as happening as the result of a mistake or error in one place even though error in one place can be an indicator that similar mistakes or errors may exist elsewhere.

8. If the election is certified prior to completing the audit, there may be no legal recourse to remove a person from office that was not elected by voters.

Are Election Outcomes Verifiably Accurate? 12 Fundamentals of Auditing Vote Tallies

Saturday, August 28th, 2010

All of the following conditions are necessary to avoid certifying incorrect election outcomes:

1. Independently-auditable voting systems, i.e. voter-marked paper ballots, not electronic ballots

2. Public reporting of all vote tallies (audit units) used to tally overall election results prior to randomly selecting a sample for auditing

3. Meaningful public observation of, and participation in (a) manual audits and (b) secure handling, storage and transportation procedures for electoral records, including ballots.

4. Adequate sample size and methods to limit the risk of certifying any incorrect initial election outcome by detecting the minimum amount of miscount, in the fewest audit units that could change the outcome (If calculations assume each audit unit’s maximum margin error is less than its upper margin error bound, then candidates must be allowed to select discretionary units for manual auditing in addition to the random sample.)

5. Publicly-verifiable random selection methods, preferably weighted by within audit unit upper margin error bounds rather than by using a uniform sampling distribution

6. Prohibiting ballot and electoral record access between the time of random selections and manual auditing

7. Polling place and jurisdiction-wide reconciliation of printed, used, unused, and spoiled ballots with absentee ballot, polling place and other electoral records

8. Timely public access to electoral records necessary to evaluate the accuracy and integrity of reconciliation and manual auditing processes

9. Voter intent as the standard for manually counting ballots during audits

10. Public reporting of discrepancies found during the audit, using manual counts to correct initial reported results

11. Expansion of the sample size, perhaps to a full recount, or certification of the election using an algorithm based on the premises of the sampling method, treating any missing paper ballots as discrepancies when deciding whether to expand or certify

12. Completion of the audit prior to certifying election results

Does the state you live in do all these things?

Kathy Dopp, MS Mathematics
Ph.D. Student
Rockefeller College of Public Affairs and Policy
State University of New York at Albany
kathy dot dopp at gmail dot com
August 28, 2010

South Carolina 2010 Democratic United States Senate Primary Election

Wednesday, June 16th, 2010

Abstract

For Election Day voting, South Carolina uses an electronic ballot system with no auditable, voter-verified record of votes. However, South Carolina’s absentee votes are cast with voter-marked paper ballots that can be manually counted after the election to check the accuracy of the optical scanner’s tallies. Without the ability to conduct risk-based post-election manual audits of Election Day tallies, the only available way to investigate for electronic vote count error in South Carolina is an indirect method — compare the tallies for absentee and Election Day tallies to look for suspicious-looking tallies consistent with a pattern produced by outcome-altering vote miscount. Such a suspicious pattern exists in the 2010 South Carolina Democratic US Senate primary contest.

Comparing South Carolina’s Three 2010 Democratic Primary Elections

Of the three Democratic primary elections, the greatest difference between the winning margins in the two different types of voting (absentee paper ballots and Election Day electronic ballots) occurs in the US Senate contest between candidates Greene and Rawl. Table 1 below provides the winning margins for the three Democratic primary contests for absentee and Election Day ballots. The margin for Election Day tallies is almost twelve times greater than for absentee tallies in the US Senate contest, the largest difference of any of the three contests. In fact, if the county-level absentee ballot margins are the more accurate reflection of voter intent, then the winner of the US Senate primary election could be Rawl rather than Greene.



Table 1. Comparison of 2010 South Carolina Democratic Primaries


US Senate SC Gov SC Sup. of Ed.
Absentee Ballot Overall Winning Margin Share 1.64% 37.50% 4.23%
Election Day Overall Winning Margin Share 19.38% 35.81% 12.93%
Ratio of Margins (Election Day)/ (Absentee) 11.81 0.96 3.06
Overall Winning Margin 17.9% 36% 12.3%
The Overall Margin if Election Day County-level Margins were the Same as the County Absentee Margins -0.7% The Winner Changes! 36.5% 9%

Even though the overall margin of absentee ballot tallies is 1.6%, favoring candidate Greene, this overall margin does not indicate who would be the winner if the county-level absentee ballot margins held for all votes because the relative number of votes cast by each method (absentee versus electronic ballot) in each county and the total number of votes cast in each county vary widely. Hence, to determine the above result, simply multiply the reported absentee ballot margin in each county times the number of Election Day votes in each county to determine the margin (in number of votes) in each county for the case that the absentee ballot tally margins truly represented how voters cast their ballots. Then add this margin to the reported absentee ballot margin (in votes) to determine the margin that would have occurred if the absentee ballots reflect true voter preferences.

In other words, as shown in Table 2 below, if the inauditable county-level Election Day voting margins were in truth equal to the margins in the county absentee-ballot tallies, then Rawl rather than Greene would be the declared winner of the primary election for the Democratic candidate for the United States Senate.



Table 2: If the County-level Absentee Ballot Margins are the Correct Margins for Election Day Voting


Apparent Winner would Lose Apparent Runner-up would Win
US Senate Yes Yes
SC Governor No No
SC Sup. of Ed. No No

Conclusion

While it is physically impossible to verify who won any election in South Carolina due to its use of electronic-ballot voting, analysis of the available data shows that the reported Democratic US Senate primary election results are consistent with sufficient vote miscount to cause an incorrect election outcome. The outcome of the South Carolina Democratic primary election for the United States Senate candidate is in doubt.

A less costly voter-marked paper ballot system would improve South Carolina’s capacity to check the accuracy of its election outcomes.

A PDF version of this is available here.