CVD homepage
What's new?
Online library
Order materials
Get involved!
Links
About CVD

Recommended Implementation Policy for Instant Runoff Voting

November 2002

[NOTE:  This document has been prepared by the Center for Voting and Democracy to assist the Department of Elections in implementing Proposition A in a way that ensures that voting is as easy as possible for all voters and that all votes are counted to the maximum extent possible.  This document was not commissioned by or paid for by the City and County of San Francisco.]

General guidelines

  1. The charter requires Proposition A to be implemented for the November election 2003 for mayor, district attorney and sheriff.
  2. Proposition A will be implemented using existing voting equipment (Optech Eagle in precincts, Optech IV-C for absentee and provisional) with ballot format and polling procedures that are as similar as possible to current practices.
  3. Community education will be broad and last for a minimum of six months.
  4. To ensure an efficient implementation of Proposition A, acceptance of the system by voters and low rates of spoiled ballots, the Elections Commission will contract as necessary with consultants who have expertise in ranked ballot systems and in community education about ranked ballot systems.
  5. The Elections Commission will require timelines from the vendor chosen to implement Proposition A, including an approximate date on which the vendor will apply for certification of hardware, software, and procedures from the Secretary of State's office.
  6. The Elections Commission or its consultant will carefully monitor progress of implementation, including vendors, community education, and departmental staff and resource requirements.
  7. The Department shall apply this policy in the spirit of ensuring that every vote is counted in a way that honors the intent of the voter to the maximum extent possible under the charter and state law.
  8. The Department shall work with the school district and the vendor to conduct a youth vote at as many high schools as possible in a way that simulates a live vote as closely as possible.  Target data for the vote is early October.

Some technical guidelines

  1. Voters will have the option of indicating first, second and third choices.
  2. The following types of ballots will be accepted as valid votes by the Eagle:
    1. 1st choice only;
    2. 1st and 2nd choices only;
    3. 1st, 2nd and 3rd choices.
  3. The following types of ballots will be ejected by the Eagle and the voter will be given the option of correcting the ballot, filling out a new ballot or overriding the error warning:
    1. Ballots that do not list any choices
    2. Ballots that skip a rank before or between choices

                                                               i.      2nd choice only

                                                             ii.      3rd choice only

                                                            iii.      2nd and 3rd choices only

                                                           iv.      1st and 3rd choices

    1. Ballots with a duplicate ranking

                                                               i.      Two candidates listed as 1st choices

                                                             ii.      Two candidates listed as 2nd choices

                                                            iii.      Two candidates listed as 3rd choices

    1. Ballots that list the same candidate for two choices

                                                               i.      Same candidate listed as 1st and 2nd choices

                                                             ii.      Same candidate listed as 1st and 3rd choices

                                                            iii.      Same candidate listed as 2nd and 3rd choices

  1. For ballots ejected, the message printed on the Eagles tape shall describe the reasons or reasons and the race:
    1. ���Undervote for mayor���
    2. ���Skipped (1st and/or 2nd) choice(s) for mayor���
    3. ���Duplicate (1st, 2nd or 3rd) choice for mayor���
    4. ���Same candidate listed as (1st, 2nd) and (2nd, 3rd) choices for mayor

Interpreting ballots with errors on them

In instant runoff voting, each ballot counts for the voter���s 1st choice until that candidate is eliminated. Then the ballot counts for the voter���s next choice candidate who is still in the race.  The process of eliminating candidates and counting every ballot for the voter���s highest choice who is still in the race continues until one candidate receives a majority of the votes and is declared a winner.

By complying with language of the charter amendment and by striving to honor the intent of the voter to the maximum extent possible, we can specify how to interpret each type of voter error.

 

Error type

Interpretation

1.      Duplicate ranking (e.g., two 1st choice candidates)

The charter states that the ballot becomes exhausted when the duplicate ranking is reached.

2.      Same candidate given two rankings (e.g., Jones as a 1st choice and a 2nd choice)

Each ballot counts for the voter���s highest-ranked choice who has not been eliminated.  In the example, the ballot would count for Jones until Jones is eliminated.  Then it would count for the voter���s 3rd choice.

3.      Skipped ranking

The ballot counts for the voter���s next listed choice once the skipped ranking is reached.

Here are examples of each type of error

 

 

Voter's ranking Error Interpreted and stored ballot image
Ballot 1st 2nd 3rd type (1st choice, 2nd choice, 3rd choice)
1. Ann/Cat Bob Don 1 (exhausted)
2. Bob Ann/Don Cat 1 (Bob)
3. Don Bob Ann/Cat 1 (Don, Bob)
4. Ann Ann Don 2 (Ann, Don)
5. Bob Ann Bob 2 (Bob, Ann)
6. Ann Bob Bob 2 (Ann, Bob)
7. (none) Don Cat 3 (Don, Cat)
8. Cat (none) Bob 3 (Cat, Bob)
9. (none) (none) Cat 3 (Cat)

Note these errors could be combined.  For example, a ballot might list 3 (or more) candidates as a 1st choice.  A ballot could skip the 1st choice and then list 2 candidates.  By applying the above rules in the order they are encountered, multiple errors can be interpreted.

 

Cleaning of data (revised Dec 12)

 

The preferable method of storing and cleaning data is for the voting equipment to store a complete image of each ballot scanned, including duplicate rankings, skipped rankings and so forth.  Cleaning of the data should occur in a later step after all votes are aggregated. For example, if a ballot lists two 1st choices and a 3rd choice, the stored ballot image would contain two 1st choices and a 3rd choice.  When the ballot is cleaned, this ballot would become exhausted due to the duplicate 1st choice.

 

If the preferable method is not possible, the voting equipment could both scan the ballot and clean the data.  In this case, there would be no errors stored in the complete set of rankings.

 

Release of results and application of IRV tabulation

 

  1. The official application of the IRV tabulation method to the entire set of voting data will not occur until all ballots (precinct, absentee and provisionals) have been scanned by voting equipment.
  2. Unofficial election results will be reported for instant runoff races by two methods.
    1. 1st choice totals shall be reported in all instant runoff races in each regularly-scheduled election report, starting with report #1 consisting of early absentee ballots and continuing through each subsequent election report.
    2. Unofficial data sets consisting of all votes (each voter���s 1st, 2nd and 3rd choice candidates) will be released on paper and on the Internet on the following schedule:

                                                               i.      The first data set consisting of early absentee ballots will be released as soon as possible after 8pm.

                                                             ii.      The next data set will be released after all or nearly all precincts report on Election Night.

                                                            iii.      A data set will be released at the end of each subsequent day of counting absentee and provisional ballots.

  1. To protect the secrecy of ballots, these unofficial data sets will not indicate votes by precinct, and they will be delayed if the number of ballots in any data set is small enough to potentially compromise ballot secrecy.  For example, the Director could decide not to release reports with fewer than 500 ballots.  In this case, once there remained fewer than 500 absentee and provisional ballots to count, the process of result reporting would stop until all ballots were counted.
  2. When finally released, the official statement of vote will include all rankings separated by precinct number and by absentee versus precinct votes, in a way that is as similar to current practices as possible.

top of page


______________________________________________________________________
Copyright 2002     The Center for Voting and Democracy
6930 Carroll Ave. Suite 610, Takoma Park, MD 20912
(301) 270-4616        [email protected]