Simple changes to election procedures to eliminate gerrymandering in legislative elections

Thomas Malloy

Introduction

Gerrymandering, the process of drawing legislative district boundaries to advantage your allies or disadvantage your opponents, is a frequent target of reform especially around the time every 10 years when these boundaries are redrawn. It is possible to eliminate gerrymandering with a few simple changes to our election process.

Voters participate as they do now, by selecting a single, favorite candidate. Local election officials conduct the election just as they do now. In particular, they need not make any changes to the ballots that they prepare, distribute and tabulate. Even those responsible for drawing district boundaries may continue their work as they do today, be it in a partisan or non-partisan manner. The proposed changes place a minor extra burden on candidates and on state election officials, which is described below.

For the remainder of this paper, the proposed changes are referred to as enhanced delegate transferable voting (EDTV).

Changes to the election process

1) Candidates designate like-minded alternative candidates. Each candidate provides a list of vote transfer preferences; an ordered list of like-minded alternative candidates who should receive the votes of this candidate if she is not herself elected. The lists may contain candidates from other districts within the domain of responsibility of the election authority; e.g. for California Assembly elections the domain is the entire state and the authority is the Secretary of State. The election authority receives these lists and uses them to determine the set of elected candidates through the candidate selection process described below. The election authority also publishes the lists before voting begins so that voters may include them in the criteria they use to select the candidate for whom they vote.

2) Votes of losing candidates go to other candidates. The election authority defines and implements a candidate selection process that incorporates the vote transfer preferences described in 1, above. The candidate selection process proceeds in rounds and ultimately converges on a set of elected candidates. During each round at least one active candidate is either elected or eliminated until the number of elected candidates equals the number of seats to fill. There are a number of ways to define such an election process. Appendix B provides a detailed description of one such process.

Benefits of EDTV

Adoption of EDTV is straightforward, only requiring the candidates to supply their list of alternative candidates and the election authority to implement the candidate selection process.

EDTV eliminates gerrymandering by allowing candidates to include candidates from other districts on their transfer lists. In the most typical scenario, a candidate would prefer candidates from her own party in other districts to any candidates from other parties in her own district.

EDTV preserves the simplicity of the current election process because the responsibility for specifying vote transfer preferences is assumed by the candidates. Another proposed reform that shares aspects of this proposal, rank choice voting, requires each voter to specify their vote transfer preferences which in turn complicates the design of ballots by the local election authorities.

Simulated results

Assuming that candidates prefer only other candidates in their own party, it is possible to simulate the results EDTV would have had on past elections. Simulated results for 4 elections (2012, 2014, 2016, 2018) for the U.S. House of Representatives were obtained from Election Explorer, a free iPad application available for download from Apple’s AppStore.

North Carolina and Maryland are often mentioned as states that gerrymander districts to advantage Republicans and Democrats, respectively. For instance, in 2016 Republicans won 53% of the votes but those votes yielded 77% of the seats (10). The simulated results for EDTV yielded the Republicans 54% of the seats (7). In Maryland, Democrats won 88% of the seats (7) with only 60% of the votes. The simulation predicted that Democrats would win 62% (5) of the seats using EDTV.

Even in states that use independent commissions to avoid gerrymandering, the simulation of EDTV produces results where the number of seats for each party more closely matches the number of votes that party received. For instance, in California Democrats received 74% of the seats (39) from 64% of the votes. The simulation predicted that Democrats would receive 64% of the seats (34).

Appendix A: Comparison with other electoral systems

As mentioned above, ranked choice voting (RCV), also known as single transferable votes (STV), uses the same two mechanisms described here; vote transfer preferences and a candidate selection process based on those preferences. However, with RCV, each voter supplies a list of vote transfer preferences.

In practice, EDTV could produce results very similar to another system known as party-list proportional representation, if you assume that each candidate for a given party submitted identical vote transfer preference lists. If this were the case the likely result is that votes would transfer to same-party candidates in the order of their seniority or level of party influence/power. To counter this outcome, the Election Explorer simulation and the detailed algorithm of Appendix B assume that voters would prefer to have their votes transferred to a candidate from the same party in a nearby district. Neither approach is obviously right or wrong, but it is a detail that would need to be resolved when putting this proposal into practice.

Delegated transferable voting (DTV), as described by Bishop and Beach, assigns the responsibility of transferring votes to the candidates, just as does EDTV. EDTV and DTV differ in the details. For instance, candidate’s transfer preferences are specified before the election in EDTV. In DTV, candidates negotiate vote transfers after the election. EDTV adds a specific anti-gerrymandering rule; the rule that allows a candidate to specify transfer preferences outside her own district.

Appendix B: A Detailed Candidate Selection Algorithm

//Initialization

Define 3 sets of candidates; active, elected and eliminated. Initialize the active set to all candidates. Initialize the elected and eliminated sets to empty.

Define a threshold vote count sufficient to elect a candidate according to the “Droop quota” formula:

Droop_quota = ((Total votes cast)/(Number of seats to fill + 1)) + 1

Initialize each candidate’s total vote count to the count received directly from voters.

//Loop

While there are candidates in the active set {

While there are candidates whose total vote count is greater than or equal to the Droop quota {

Select one of the candidates with the largest total vote count. If there is more than one, select one at random. They will all be processed before the while loop is done.

Move the selected candidate from the active set to the elected set.

// Transfer any of the candidate’s total votes that exceed the Droop quota, countP:

Create a list of possible like-minded candidates who can receive the candidates votes using the procedure, BuildTransferList, described below. The listT parameter to BuildTransferList should be initialized to an empty list.

listT now contains all active candidates directly or indirectly included on the transfer list of c.

The target candidate is the “nearest” candidate to c listT.

// See below for a discussion of “nearnesss”

If such a target candidate exists {

Increase the target candidate’s total vote count by the partial vote count, countP.

}

}

//At this point no active candidate has enough votes to be elected

If the number of active candidates is (less than or) equal to the remaining number of seats to fill {

Move all remaining active candidates from active set to the elected set.

Exit the main loop

}

Otherwise {

Select the active candidate with the smallest total vote count. If there is more than one such candidate, choose one of them using some tie breaking rule. Possible rules are, smallest direct vote count or some random mechanism such as drawing straws.

Move the selected candidate from the active set to the eliminated set. Create a list of possible like-minded candidates who can receive the candidates total vote count using the procedure, BuildTransferList, described below. The listT parameter to BuildTransferList should be initialized to an empty list.

listT now contains all active candidates directly or indirectly included on the transfer list of c.

The target candidate is the “nearest” candidate to c on listT.

// See below for a discussion of “nearnesss”

If such a target candidate exists {

Increase the target candidate’s total vote count by the total vote count of c.

}

}

}

// End of candidate selection algorithm

// BuildTransferList, given a candidate, c, and a transfer list, listT

// Add to the transfer candidate list, listT, for c, according to the following rules:

Add to the transfer candidate list, listT, all of the active candidates on the transfer list provided by the candidate.

For each candidate, cT, on the transfer candidate list provided by the candidate (active or not!) {

Add to the end of listT any candidates on the transfer list of cT that are not already on the candidate transfer list and who are still active.

Recursively call BuildTransferList passing cT and listT as parameters

}

// “Nearness“ is determined through the use of a distance metric between any two candidates. Any candidate, c, has a set of nearest candidates; those candidates whose distance metric is smaller than all candidates not in the set.

// One possible candidate distance metric is inferred from a distance metric on districts. If two candidates are in the same district the distance between them is zero. If they are in separate districts that share a border, the distance between them is one. If two districts each share a border with a common third district, the distance between them is two. Proceeding recursively, it is straightforward to assign a distance to any two districts connected by a path through other districts. For districts, that are not connected by such a path – e.g. the islands of Hawaii - The election authority must assign distances between the unconnected portions.

Reference

Farrell, David M (2011). Electoral Systems, A Comparative Introduction. Palgrave Macmillan

Bishop, J., & Beech, M. (2017). Exploring the Counting of Ballot Papers Using “Delegated Transferable Vote”: Implications for Local and National Elections in the United Kingdom

About the author

Thomas Malloy is a retired software executive. He can be reached at thomasjmalloy@kttk.org