# Bibliography on IRV / STV

###### tags: `bibliography`

# Cryptography

## [A toolbox for verifiable tally-hiding e-voting systems](https://eprint.iacr.org/2021/491.pdf)
* Builds tally-hiding e-voting systems using MPC.
* Provides a solution for STV but covers IRV as the single-winner case.
* FH: 1) voters computation O(n^2), authorities computation O(n^3); 2) implementation used 3 trusttees (no threshold)l 3) voters must perform cryptography and shuffling of ballots; trustees must use mixnets to do additional shuffling 4 "Homomorphic tally can only be applied to simple vote counting functions." (not correct)

## [Extending the tally-hiding Ordinos system: implementations for Borda, Hare-Niemeyer, Condorcet, and Instant-Runoff Voting](https://eprint.iacr.org/2021/1420.pdf)
* Extends Ordinos to permit other voting methods including IRV.
* IRV implementation runs in $O(n!)$ time for $n$ candidates.
* Doesn't consider indifference

## [Universally Verifiable MPC and IRV ballot counting](https://dl.acm.org/doi/10.1007/978-3-030-32101-7_19) / [Universally Verifiable MPC with Applications to IRV Ballot Counting](https://eprint.iacr.org/2018/246)
* Provides a universally verifiable implementation of IRV using MPC.
* Uses a very similar matrix representation to ours, but instead of shifting rows, they keep the structure of the matrix the same but strike out the column for any eliminated candidate. Then, $\sum\mathbf{v}_{i} = 1$ iff $\mathbf{v}_{i}$ is the encoding for an uneliminated candidate, and otherwise $\sum\mathbf{v}_{i} = 0$. (FH: does the difference matter? Which one is better?)
* Allows for only a subset of candidates to be ranked (rather than all of them).
* FH: 1) pairing based; 2) implemented the single-TA version; 3) used IRV data from Australian elections; 4) 40K ballots, 6 candidates, 4 rounds -> 15 hours (not the actual time; for the actual time, need multiplying a constant factor, but the value for the constant factor is unclear); 5) distributed settings (voters perform encryption); 6) "The implementation did not include the construction or checking of proofs, and ran as a single party."
* FH: what's the complexity of TA?

## [Kryvos: Publicly Tally-Hiding Verifiable E-Voting](https://eprint.iacr.org/2022/1132.pdf)
* Proposes a tally-hiding and verifiable e-voting system called Kryvos.
* Has an instantiation for IRV.
* FH: 1) TA-based; 2) used pairing-based SNARKS; 3) n! encoding for n candidates (120 for n=5 candidates) - this encoding method explains why they need a fully tally-hiding scheme; 4) used same Australia NSW data as used in (Ramchen etc, FC'19) with candidates n=5 and 6 in Table 3. 5) used a single trustee in the implementation

## [Shuffle-sum: coercion-resistant verifiable tallying for STV voting](https://www.talmoran.net/papers/BMNRT09-shuffle-sum.pdf)
* Proposes *shuffle-sum* - a verifiable STV protocol.
* Main idea is to use different ballot representations for the different operations needed in STV.
* Elimination is handled by an encrypted *indicator* ($1$ for eliminated, $0$ otherwise).
* FH: 1) used mix-nets; 2) TA-based; trustees cooperate to decrypt the first-preference tallies; 3) implementation used a single TA, and tested using Victorian State Election STV data.

## [Event driven private counters](https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=9143126ce2ac5140234af4b5ed9ff271eb37e293)
* Proposes a new primitive called *private counters* that can be used to update preferences privately and non-interactively.
* Authors propose a protocol for running a preferential election that has an elimination procedure (appears to be IRV but not explicitly stated).
* Computational cost of $O(nt^{4})$ for $n$ voters and $t$ candidates.
* Requires election authorities to collect votes, verify them and compute the result of the election.

## [Implementing STV securely in Prêt à Voter](https://openresearch.surrey.ac.uk/esploro/fulltext/journalArticle/Implementing-STV-Securely-in-Pr%C3%AAt-%C3%A0/99515194202346?repId=12139654060002346&mId=13140654570002346&institution=44SUR_INST)
* Extends Pret a Voter to work with STV elections.
* Elimination is similar to ours but moves the eliminated candidate to the end of the list.
* Main idea is to encode a vote as a sequence of cryptograms, each decrypting to a candidate identifier. The head of votes are decrypted, sorted into piles, then the smallest pile is redistributed. Piles are anonymised and mixed so transferred votes cannot be distinguished from original votes.
* Paper largely focuses on single-winner STV (IRV) but briefly explores the multi-winner case.

# Other

## [Instant Runoff Voting: what Mexico (and others) could learn](http://archive.fairvote.org/rcv/brochures/FairVote_Article_On_IRV.pdf)
* Motivates the need for IRV in comparison to plurality.

## [The Case for Instant Runoff Voting](https://link.springer.com/article/10.1007/s10602-022-09379-5)
* Discusses the benefits of using IRV for elections compared to other methods including plurality and Condorcet.

## [Four Condorcet-Hare hybrid methods for single-winner elections](https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=49dba225741582cae5aabec6f1b5ff722f6fedf1)
* Looks into the resistance of different voting methods to tactical voting.
* IRV is found to be one of the most resistant to tactical voting.

## [Single transferable vote resists strategic voting](https://dspace.mit.edu/bitstream/handle/1721.1/49116/singletransferab00bart.pdf)
* Proves strategic voting is NP complete for STV.

# Secondary

## Links
* https://labour.org.uk/people/leadership-elections-hub-2020/leadership-elections-2020-results/
* Results for the 2020 Labour leadership election.
* It states using STV, but it is single-winner (so IRV).
* https://www.theguardian.com/politics/2022/jul/12/how-does-the-tory-leadership-contest-work
* Article on the Tory leadership contest.
* Voting procedure has multiple steps, but a subset of those is runoff rounds with elimination (IRV).