Negotiated Consensus

A more representative way to VOTE

If more than three candidates run in a “first past the post” (FPTP) single-seat election, the results can be highly unrepresentative.

REPRESENT!

Typically a PLURALITY rather than a MAJORITY winner is declared, and frankly a winner receiving only 35% of the vote feels illegitimate. More worrisomely, in many elections consensus choices are undermined through the tactic of spoiler or cloned candidates skewing the outcome by vote splitting.

To address these inequities, Ranked or Preference Voting (RPV) methods have been proposed and, in some cases, adopted. But RPV demands voters invest the time, knowledge and interest to pre-record their alternate choices for a subsequent automatic vote-preference reallocation. Dishonest, lazy or misguided voters can undermine even the most carefully planned voting schemes. While intellectually fascinating to design and debate, RPV systems can also be opaque or misfire, leading to sometimes embarrassing disconnects between election results and common sense. Despite thousands of years of experimentation, no universally acceptable RPV system has emerged that can challenge FPTP.

Here we propose an alternative to RPV automatic-voting called Negotiated Consensus (NC). In the absence of a simple majority by election day, the candidates themselves negotiate, for a price, to transfer all (or some) of their votes to other aspirants. With NC, you vote for your preferred candidate, and trust them to reach a consensus on your behalf. In real-time negotiations. Similar to forming a ruling coalition in parliamentary systems. Any candidate accumulating a majority of the votes, including pledged transfers, wins.

Unlike RPV (see next section) where your alternate-votes are given away, having received nothing tangible in return, in NC small parties gain influence for their constituents during negotiations. Perhaps swapping votes for a future cabinet seat. Or agreeing to soften their position on climate adaptation to pick-up support. Unlike conventional runoff elections where third parties can only signal voters and hope they will maintain discipline, in NC vote transfers are guaranteed.

What happens if an agreement cannot be reached? Certainly, public pressure will increase, but if after three weeks there is no winner, the top two candidates (after any vote transfers) hold a runoff election.

NC eliminates the spoiler effect, disingenuous clones, plurality winners, mathematical voting aberrations, complexity and wasted votes. Negotiations force compromise and, over time, the pragmatic candidate’s best tactic is to behave more like a virtuous representative.

Issues with Ranked or Preference Voting:

How does one compare voting systems in a democracy? There is no generally accepted legal or political answer. Analysts often defer to some kind of general “satisfaction” with the outcome as a metric. Maximizing election satisfaction (presumably with at least a majority of the voters) is the goal.

It is possible for 70% of the voters to be satisfied, even if the winner only garnered 55% of the votes, as the losers might still agree the election was fair. The converse is also true- 40% could be satisfied even though they voted for the majority winner, because they preferred a different candidate as their standard bearer and thought the primary system was corrupt.

In the spirit of maximizing representation and voter satisfaction, ranking systems are generally designed to improve the chances of independent or smaller parties participating and winning. Increased competition slightly encourages candidates to become more moderate or centrist, but unless moderation improves overall voting satisfaction, this is a side-effect and not a democratic goal.

Among the dozens of proposed voting systems two seem closest to adoption: some variation on the theme of Ranked Choice or Ranked Preferences.

There are numerous technical and practical issues with RCV (see link for RCV tutorial). Because the vote-aggregation steps are non-linear, you cannot simply sum the outcomes from different precincts, but must wait until all votes are certified and merged. This can delay or call into question the outcome. Standard voting machines are not suited for RCV, so a few percent of people regularly spoil ballots by inadvertently declaring two second choices (and a few percent may be critical in tight elections). Exhausted ballots, where the automated process could not match a voter to the final two contenders due to the order of their choices, may disenfranchise 10% to 15% of the electorate.

RCV has failed in the field, for example, in the 2009 Burlington, VT mayoral election neither the plurality or the head-to-head preferred candidate won, leading to low voter satisfaction with the result. RCV was repealed later that year.

RCV can also prematurely eliminate the consensus candidate, which, to the extent they are the candidate satisfying most voters, implies the least democratic choice won. For example, imagine Alex receives 38% of #1 votes, Blake 35% and Carey 27%. No one has a majority on the first round, so an automated reallocation occurs. Alex’s and Blake’s voters hate each other’s guts, so only a few will rank each other as #2, but they generally approve of Carey. However, in RCV, the bottom vote-getter is dropped, which means the consensus candidate Carey loses. True, Carey’s second-choice selections will tip the election in a direction they prefer, but fewer people will be happy with that outcome compared to Carey winning.

There is also an existential question. Is your #2 ranked candidate slightly worse than #1, or just over the bar? In either case they receive a fully-transferred vote on the second round, which may over or under-estimate voter intent. That is, if maximizing voter satisfaction is the goal, should two lukewarm #3 votes be worth twice as much as one strong #2?

STAR voting was invented to avoid many of these aberrations, and is often compared to submitting a Yelp review. In STAR you don’t rank the candidates against each other, but compare individuals to an abstract concept of “goodness” on a five-point scale. Similar to a restaurant review, where you rate your experience compared to an ideal meal. Except the rankings are weighted by the number of diners. So, a restaurant with 100 five star reviews gains 500 points, while one with 1000 four stars are assigned 4000 points. And thus wins the “Best Of” award. Note the more popular establishment may not actually serve the best food, but satisfies the most people most of the time.

Unlike Yelp, the “AR” in STAR stands for “Automatic Runoff”. After the preference scoring round, the top-two highest scorers are compared in an instant-runoff re-using earlier preference scores, winning points for every head-to-head ballot with a higher preference. The head-to-head winner is elected.

Since all preferences are combined before candidates are compared, every vote counts and there are no exhausted ballots in the scoring stage. However, if you did not indicate a relative head-to-head score for the eventual runoff candidates, or scored them identically, your ballot had no effect in the runoff. So effectively an exhausted ballot.

In STAR, voters assign preferences from 5 to 1 to each candidate, and thus each step in preference is downgraded by 20% when preferences are averaged. But why 20% and uniform steps, except for mathematical ease? Depending on the degrading factors between levels, different candidates can win.

Dishonest voting is an issue, as are clones and collusion. In addition, the “stars” have no fixed meaning. Similar to RCV, combining preferences is problematical where each voter had a different scale in mind.

Moreover, as with all automatic voting systems, ranks/prefs are preloaded into the system BEFORE voters are aware of the results of the first round. Voters might prefer to readjust their second-choice preferences in light of events. As they do in a runoff election.

Issues with Negotiated Consensus:

• Perhaps illegal in some jurisdictions, and certainly unfamiliar. Will definitely raise red flags.
• The cost of holding a potential runoff.
• Possibly lower turnout in the runoff (then again, possibly higher as is typical with primaries vs regular elections).
• Vote buying and collusion is possible, but if you don’t trust your candidate in NC, you won’t trust them in office.
• Small parties may be thrust into the role of “king makers”. But larger parties can moderate this tendency if they agree to cooperate. Or default to runoff.
• Candidates might renege on a promise, but assuming elections are held again in a few years, they will have no credibility in future negotiations. So, everyone has an enormous incentive to honor their commitments. Classic game theory.
• As in many voting systems, a minimum level of support may be required to qualify for the ballot- perhaps via a primary or enrollment stage to weed out niche parties.
• Americans expect a winner will be announced the day after the election, and initially may be uncomfortable with NC’s longer time horizon.

Random Variations in Negotiated Consensus:

Note any rigid forcing-function to break an NC deadlock, such as an instant runoff or preference vote instead of a two-person runoff, will be gamed by candidates in their favor. For example, if you knew you would win an instant runoff you have no incentive to negotiate vote transfers in good faith.

A compelling but unconventional alternative is Athenian randomness. If after three weeks there is no consensus winner, instead of facing each other in a runoff election, one of the top two (after transfers) candidates wins by random selection.

Athenian randomness is also realistic. Imagine two candidates have 48.5% and 48.6% of the vote. If you re-ran the election the numbers would undoubtably shift and possibly change the winner. Systemic voting suppression, including disinformation, gerrymandering, endless, pointless debates over hanging chads, partisan election officials, etc. puts its thumb on the scales. Randomness undermines these attempts at gaming the election, and thus sits on the side of greater democracy.

Sortition is hard to game, and is rough justice. But I fear injecting intentional randomness into an election will be a hard sell, no matter how efficacious. So NC defaults to a top-two runoff election.

Multi-member Districts and Negotiated Consensus:

A similar approach can be used in multi-member districts. Let’s say there are three open positions and a dozen candidates. Each voter has three votes they can apply entirely to one, or split among two or three candidates. Any candidate who can win or gain transfers of 25%+1 of the total cast votes (the Borda threshold) is provisionally elected. Any excess can be transferred to others.

Again, multi-member negotiations will determine the outcome. In the event one or more seats are unfilled after three weeks, a runoff election is held. The runoff must include the candidates who passed the 25% threshold, and the top two or three of the remainder (two, if two candidates achieved the threshold, three if only one). Why are the “winners” of the first round included in the runoff? To avoid the tyranny of the majority. If the winners of the first round are excluded, then the largest plurality basically double-votes, knowing their preferred candidate is already safe . In fact, a large plurality might discourage negotiation to throw the election into a runoff- instead of splitting the seats, as they could win every seat.

Alternatively, a random winner could be selected from among the top two (or three). Avoiding the cost and complexity of a runoff.

Useful References:

• Charles Dodgson, aka Lewis Carrol, was fascinated by voting system design. He invented some of the earliest proportional voting methods, and balanced mathematical optimization methods with real-world concerns over voting privacy and manipulation. There was even an “NC”-like feature involving the formation of coalitions:

May I, in conclusion, point out that the method advocated in my pamphlet
(where each elector names one candidate only, and the candidates themselves can, after the numbers are announced, club their votes, so as to bring in others besides those already announced as returned) would be at once perfectly
simple and perfectly equitable in its result?

C. L. Dodgson, The Principles of Parliamentary Representation: Supplement (Oxford 1885), p7.

• NC is a variation of “Asset Voting”, which like NC, has been independently reinvented a number of times. It is also loosely related to proxy voting. As always, the devil is in the details.
• Australian elections are predominately a form of RCV. Parties, after mutual negotiations and strategizing, parties issue “cheat sheets” (how-to’s) to guide their member’s ranked selections in the voting booth. But compliance only runs from 20% to 50%.
• Jobst Heitzig and Forest W. Simmons propose a nice mathematical stochastic voting method allowing minorities to gain leverage when reaching consensus.
• https://blog.yelp.com/news/new-study-compares-yelp-review-quality-with-competitors/
• The mathematical-political papers of C. L. Dodgson, 1982 Francine Abeles, in Lewis Carroll, a Celebration ISBN 13: 9780517545577