An Introduction to Voting Theory: History and Procedures Arnold B. Urken Professor of Political Science Division of Humanities and Social Science Stevens Institute of Technology aurken@stevens.edu DIMACS Workshop, May 10, 2004 © 2004. Arnold B. Urken All Rights Reserved 1 Outline Top Six Voting Systems Pre-18th Century Voting Theory 18th Century France: The Golden Age? The Rediscovery of Voting Theory Preference Aggregation Issues Competence in Voting Theory © 2004. Arnold B. Urken All Rights Reserved 2 Top Six Voting Systems Plurality voting Borda voting Condorcet scoring Copeland scoring Approval voting STV (IRV) © 2004. Arnold B. Urken All Rights Reserved 3 Top Six Voting Systems [continued] Voting systems include rules for Vote Endowment: number of votes used to express preferences Vote Allocation: Saving or trading? Vote Aggregation: Standard for producing a collective outcome. Allocation => “fungible voting,” which allows votes to be saved and traded © 2004. Arnold B. Urken All Rights Reserved 4 Hypothetical Data Set Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush Nine voters rank 1. Bush 2. Kerry 3. Nader © 2004. Arnold B. Urken All Rights Reserved 5 Top Six Voting Procedures [continued] Plurality Voting Endowment: One vote for the most preferred choice Allocation: Trading/saving not explicitly allowed Aggregation: Choice with the most votes wins (plurality) © 2004. Arnold B. Urken All Rights Reserved 6 Plurality Voting Results Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Bush 4 votes Kerry 3 votes Nader 2 votes © 2004. Arnold B. Urken All Rights Reserved Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush 7 Plurality vs. Majority What’s the Difference? Absolute vs. relative majority Historically Sanior et major pars Right/healthy and greater part Used to overturn outcomes Sanior difficult to measure, so major used © 2004. Arnold B. Urken All Rights Reserved 8 Top Six Voting Procedures [continued] Borda Voting Endowment: Assign ranks to choices Allocation: Trading/saving not explicitly allowed Aggregation: Choice with the most votes wins (plurality) © 2004. Arnold B. Urken All Rights Reserved 9 Borda Voting Results Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Bush Nader Kerry Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush 18 points 18 points 18 points Plurality aggregation not satisfied. © 2004. Arnold B. Urken All Rights Reserved 10 Borda and Rankings First Place Bush Kerry Nader Second Place Third Place x x Illegal in some elections © 2004. Arnold B. Urken All Rights Reserved 11 Borda and Rankings [continued] First Place Bush Kerry Nader x Second Place Third Place x x Not used this way © 2004. Arnold B. Urken All Rights Reserved 12 Top Six Voting Procedures [continued] Condorcet Scoring Endowment: Ordinal rankings assigned to choices Allocation: Trading/saving not explicitly allowed Aggregation: Winner is the choice with the most victories in binary comparisons © 2004. Arnold B. Urken All Rights Reserved 13 Condorcet Scoring Results Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Joe Nader Kerry Bush Bush Kerry Nader Tom Ruth Kerry Nader Nader Bush Bush Kerry Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush 9 points 9 points 9 points Plurality aggregation not satisfied. © 2004. Arnold B. Urken All Rights Reserved 14 Top Seven Voting Procedures [continued] Copeland Scoring Endowment: Ordinal rankings assigned to choices Allocation: Trading/saving not explicitly allowed Aggregation: Winner is the choice with greatest net score in binary comparisons © 2004. Arnold B. Urken All Rights Reserved 15 Copeland Scoring Results Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Bush Nader Kerry Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush 0 points 0 points 0 points Plurality aggregation not satisfied. © 2004. Arnold B. Urken All Rights Reserved 16 Top Seven Voting Procedures [continued] Approval Voting Endowment: N votes where N = number of choices Allocation: One vote cast for each approved choice; no trading/saving Aggregation: Plurality, majority, or unanimity © 2004. Arnold B. Urken All Rights Reserved 17 Approval Voting Results Assuming one approval vote is cast for 1st and 2nd place choices Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Bush Nader Kerry Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush 5 points 6 points 5 points Nader is the plurality winner! Based on the number of voters who approve him © 2004. Arnold B. Urken All Rights Reserved 18 Top Seven Voting Procedures [continued] Observations about Approval Voting Empirical observation: Voters cast an approval vote for each choice ≥ average utility Ties possible under plurality, majority, and unanimous aggregation rules Definitions of base for aggregation All allocated votes The number of voters casting votes © 2004. Arnold B. Urken All Rights Reserved 19 Top Seven Voting Procedures [continued] STV (IRV--Proportional Representation) Endowment: Assign ranks to choices Allocation: One choice for each rank, trading/saving: not explicitly allowed Aggregation: Majority of first place votes, but if no choice wins, eliminate the most choice most frequently ranked last and count first place preferences again until a majority winner is produced © 2004. Arnold B. Urken All Rights Reserved 20 STV (IRV—Proportional) Scoring Results Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Bush Kerry Nader Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush 4 points 3 points 2 points Majority aggregation not satisfied. © 2004. Arnold B. Urken All Rights Reserved 21 STV (IRV—Proportional) Scoring Results One Round of Elimination Ranks John Mary Joe 1st Kerry Kerry Nader 2nd Nader Nader Kerry 3rd Bush Kerry Nader Tom Ruth Steve Debby Ed Jack Kerry Nader Nader Kerry Nader Kerry Nader Kerry Kerry Nader Kerry Nader Deleted Information Eliminated 5 votes 4 votes Kerry is the majority winner! © 2004. Arnold B. Urken All Rights Reserved 22 PR with Strategic Voting Ranks 1st 2nd 3rd John Bush Kerry Nader Mary Kerry Nader Bush Joe Nader Kerry Bush Tom Ruth Kerry Nader Nader Bush Bush Kerry Steve Bush Nader Kerry Debby Bush Kerry Nader Ed Bush Nader Kerry Jack Kerry Nader Bush Strategic Voting Rank John Mary Joe Tom Ruth Steve Debby Ed Jack 1st Bush Kerry Nader Kerry Nader Bush Bush Bush Kerry 2nd Nader Kerry Nader Kerry Nader 3rd First Round Elimination Rank John Mary Joe Tom Ruth Steve Debby Ed Jack 1st Bush Kerry Kerry Kerry Kerry Bush Bush Bush Kerry 2nd 3rd © 2004. Arnold B. Urken All Rights Reserved 23 Summary of Results Method Plurality Majority Borda Condorcet Copeland Approval (Plurality) Approval (Majority) PR (IRV) Sincere PR (IRV) Strategic © 2004. Arnold B. Urken All Rights Reserved Winner Bush None Tie Tie Tie Nader None Kerry Kerry 24 Pre-18th Century Voting Theory General Observations Theoretical insights were derived from practical problem solving Knowledge was not cumulative The communication of votes was an issue “Science” was • “pre-normal” Kuhnian framework • early stage Popperian “metaphysical” research program © 2004. Arnold B. Urken All Rights Reserved 25 Pre-18th Century Voting Theory [continued] Pliny the Younger Ramon Lull Nicolaus Cusanus The Venetian Mehod © 2004. Arnold B. Urken All Rights Reserved 26 Pre-18th Century Voting Theory [continued] Pliny the Younger Letter to Titius Aristo, A.D. 105 Agenda manipulation in the trial of Afranius Dexter’s slaves Slaves accused of murdering his master Options Acquittal Banishment Death © 2004. Arnold B. Urken All Rights Reserved 27 Pre-18th Century Voting Theory [continued] Execution faction leader leads switch from death to banishment Banishment is the majority choice Pliny’s faction favored leniency, but included less than one-half of all votes © 2004. Arnold B. Urken All Rights Reserved 28 Pre-18th Century Voting Theory [continued] Pliny calls for ternary vote (with division of the whole) Pliny knew that the opposition had the following preference orders: Death > Banishment > Acquittal Banishment > Acquittal > Death © 2004. Arnold B. Urken All Rights Reserved 29 Pre-18th Century Voting Theory [continued] Why? Neither Acquittal nor Death would get a majority in the first round of voting—in binary comparisons In the second round of voting, the winner of the first round of voting (Acquittal or Death) would lose to Banishment Sincere and manipulated voting produce the same outcome! Pliny uncomfortable: inconsistent with Senate customs? © 2004. Arnold B. Urken All Rights Reserved 30 Pre-18th Century Voting Theory [continued] Issues Raised Sincere voting: honest communication of preferences Strategic voting: changing “sincere” votes to manipulate the collective outcome Pliny anticipates Robin Farquharson, Theory of Voting. Yale, 1969 © 2004. Arnold B. Urken All Rights Reserved 31 Pre-18th Century Voting Theory [continued] Ramon Lull A.D. 1232-1316 Explored methods for honest church elections Two methods based on selections of pairs of choices from a larger set of ranked choices Blanquera (1285) De Arte Eleccionis (1299) © 2004. Arnold B. Urken All Rights Reserved 32 Pre-18th Century Voting Theory [continued] Blanquerna (1285) Mixed method (“art”) Borda and Condorcet Electors choose Blanquerna as bishop without following the “art” they generate an indecisive outcome and the decision must be appealed to the Pope to produce a winner Work reflects ambivalence about preference aggregation and making the right choice. © 2004. Arnold B. Urken All Rights Reserved 33 Pre-18th Century Voting Theory [continued] De Arte Eleccionis (1299) Condorcet scoring Uses matrix notation (next used by Dodgson in the 19th century) Method does not address collective intransitivity (later discovered by Condorcet and Arrow) © 2004. Arnold B. Urken All Rights Reserved 34 Pre-18th Century Voting Theory [continued] Nicolaus Cusanus (1430) Goal: design an “honest” voting procedure to elect a Holy Roman Emperor to end a long schism in the papacy Proposes a Borda system Applies it to propositions with more than two choices Criticizes manipulation of electorsand criticizes attempts to control the collective outcome by manipulating electors. Implicitly suggests that voting by ballot is new © 2004. Arnold B. Urken All Rights Reserved 35 Pre-18th Century Voting Theory [continued] The Venetian Method (13th Century) Similar to approval voting Simplified the process of selecting 41 electors from an initial assembly of 1500 members. © 2004. Arnold B. Urken All Rights Reserved 36 18th Century France: The Golden Age? Voting in the French Academy of Sciences Borda, Condorcet, and others Condorcet and the French Revolution Daunou and after Proportional voting © 2004. Arnold B. Urken All Rights Reserved 37 18th Century France: The Golden Age? [continued] Voting in the French Academy of Sciences Scientists recommend top three candidates to the King of France Plurality voting used since 1699, ties rare. 1770 Borda talk about plurality voting Borda paper not published until 1784 © 2004. Arnold B. Urken All Rights Reserved 38 18th Century France: The Golden Age? [continued] Voting in the French Academy of Sciences Borda and Condorcet were political enemies Borda fought in the American Revolution Condorcet, a modernist, won a manipulated election as Secretary © 2004. Arnold B. Urken All Rights Reserved 39 18th Century France: The Golden Age? [continued] Voting in the French Academy of Sciences No evidence of actual voting debate Condorcet regards Borda’s work as physicaille (petty experiments) Condorcet’s 1785 Essai Essai sur l’application d’analyse à probabilité des décisions rendues à la pluralité des voix © 2004. Arnold B. Urken All Rights Reserved 40 18th Century France: The Golden Age? [continued] Voting in the French Academy of Sciences The 1785 Essai Goal: analyze the probability of making a correct collective choice Introduction: identifies collective intransitivity Body: 13 hypothetical situations © 2004. Arnold B. Urken All Rights Reserved 41 Condorcet “Jury Theorem” Question: How does majority rule affect the group probability of making a correct choice? Assumptions • 50 or more voters 1.0 • Binary choice 0.5 • One Person, One Vote • Preferences a random variable 0 0 0.5 1.0 Individual Voter Competence © 2004. Arnold B. Urken All Rights Reserved • Individual competence statistically independent 42 18th Century France: The Golden Age? [continued] Condorcet and the French Revolution Creates practical voting plan for the Republican Constitution with binary agendas Recommends jury design for the trial of the King of France Robespierre’s hit list drives him underground Dies in prison? © 2004. Arnold B. Urken All Rights Reserved 43 18th Century France: The Golden Age? [continued] Daunou and after FAS becomes the Institute of France New election method needed Napoléon interested Borda and Daunou on commission Daunou writes critique of Borda voting © 2004. Arnold B. Urken All Rights Reserved 44 18th Century France: The Golden Age? [continued] Daunou and after (continued) Voting theory is lost in French probability theory (Cf. Daston) Ideas rediscovered by Dodgson (Lewis Carroll) Nanson (Australia) refers to Condorcet’s ideas in designing elections for scientists © 2004. Arnold B. Urken All Rights Reserved 45 18th Century France: The Golden Age? [continued] Daunou and after (continued) Proportional voting developed for allocating seats in legislatures Ideas are not integrated with voting theorists © 2004. Arnold B. Urken All Rights Reserved 46 The Rediscovery of Voting Theory Black Does archival research on Condorcet Coins “jury theorem” to explain Condorcet’s interest in competence Develops “single-peakedness” concept to explain collective intransitivity © 2004. Arnold B. Urken All Rights Reserved 47 The Rediscovery of Voting Theory [continued] Arrow Relies on Black to understand Condorcet Invents the term “social choice” Axiomatizes collective intransitivity problem in impossibility theorem © 2004. Arnold B. Urken All Rights Reserved 48 The Rediscovery of Voting Theory [continued] Arrow Unrestricted domain or universality Non-imposition or citizen sovereignty Non-dictatorship Monotonicity Independence of irrelevant alternatives Impossible to satisfy all conditions simultaneously © 2004. Arnold B. Urken All Rights Reserved 49 The Rediscovery of Voting Theory [continued] Brams and Fishburn Develop formal proposal for approval voting Scientific societies adopt approval voting Articulate theoretical and empirical arguments © 2004. Arnold B. Urken All Rights Reserved 50 The Rediscovery of Voting Theory [continued] Saari Develops a geometric framework for comparing voting methods for three choices Does not address Ties Truncated preferences Competence © 2004. Arnold B. Urken All Rights Reserved 51 The Rediscovery of Voting Theory [continued] Preference Aggregation Issues Vote trading and fungible voting Manipulation: potential vs. actual Voter use of voting methods Ranking choices (STV) Identifying approved set of choices © 2004. Arnold B. Urken All Rights Reserved 52 The Rediscovery of Voting Theory [continued] Competence in Social Choice Young: Maximum likelihood interpretation of Condorcet’s rule Grofman (Owen, Feld) Explore models of competence Show that Condorcet solved Rousseau’s problem of reconciling “general will” and the “will of all” © 2004. Arnold B. Urken All Rights Reserved 53 The Rediscovery of Voting Theory [continued] Grofman-Shapley Theorem How to weight votes in interdependent collective decisions Don’t weight votes by using the ratio of p/1-p (ratio of competence to incompetence) Instead use ln p/1-p Experimental Confirmation © 2004. Arnold B. Urken All Rights Reserved 54 Average individual competence equals group competence. Average individual does better than the group. 1.0 Group does better than the average individual 0.5 00 0.5 1.0 Individual Voter Competence Non-monotonic pattern in approval voting © 2004. Arnold B. Urken All Rights Reserved 55 Reconciling Competence and Preferences 1.0 Minimum group competence Optimal group competence 0.5 Better than minimum performance 0 Suboptimal performance 0.5 1.0 Average Voter Competence Low © 2004. Arnold B. Urken All Rights Reserved High 56 Perspective History not of purely antiquarian interest Draws our attention to models and problems of integrating ideas Unresolved dualism Preference aggregation Competence © 2004. Arnold B. Urken All Rights Reserved 57