Experimental Examples - PowerPoint PPT Presentation

1 / 63
About This Presentation
Title:

Experimental Examples

Description:

Cardinal utilities: Teams assign numbers indicating strength of preference. Four Methods ... Sum of Cardinal Rank: Bentham. Maximize Social Welfare' Multiply ... – PowerPoint PPT presentation

Number of Views:34
Avg rating:3.0/5.0
Slides: 64
Provided by: indridiin
Learn more at: https://www.msu.edu
Category:

less

Transcript and Presenter's Notes

Title: Experimental Examples


1
Experimental Examples
  • Selection of a pair of trajectories for NASA
    spacecrafts.
  • Consider four methods of aggregating teams
    preferences.
  • Two types of utilities considered
  • Ordinal utilities Only rank counts
  • Cardinal utilities Teams assign numbers
    indicating strength of preference

2
Four Methods
31
  • Sum of Ordinal Rank The Borda Rule
  • Rank, not only first preference, matters
  • Sum of Cardinal Rank Bentham
  • Maximize Social Welfare
  • Multiply Cardinal Utility Nash
  • Protect the losers 0.1111 vs. .5.5.50.125
  • Pairwise Comparison Condorcet

26
31
26
3
Simple Majority Decision
  • Different Rules can lead to Different Outcomes
    Chapter 2
  • Is there a Superior Rule ?
  • If there is, it is Simple Majority Rule
  • When there are only two alternatives it works
    without a problem
  • Problem Choices normally not dichotomous ! ?
    Primaries, for example

4
What is so great about Simple Majority Rule ???
  • Definitions
  • A set Xx,y of alternatives
  • A set N1,2,3,,n citizens/voters
  • Pi Individual is strict preference over X
  • Not complete No strict preference possible
  • Asymmetric xPiy implies not yPix
  • Ii Individual is indifference over X
  • Not complete if xPiy then not xIiy or yIix
  • Symmetric if xIiy then yIix

5
What is so great about Simple Majority Rule ???
  • Definitions
  • Ri Individual is weak preference over X
  • If xPiy and xIiy then xRiy
  • Complete Either xRiy or yRix
  • A Social Choice Rule C(X)
  • If xPiy then say Di 1
  • Then preferences of voters can be summarized
    D(D1, D2, D3,Dn)

6
What is so great about Simple Majority Rule ???
  • Definitions
  • D?D. D is the set all profiles
  • Two person society D(1,1),(1,0),(0,1),(0,0),(0,
    -1),(-1,0),(-1,-1)
  • Define a function F(D) equivalent to C(X)
  • F(D)1 ? C(X)x
  • Indecisive Rule F(D)0 for all D?D
  • Imposed Rule F(D)1 (or 1) for all D?D
  • Simple Majority Function Plurality
  • Absolute Majority Function gt50
  • Special Majority Function gtz, zgt50

7
Outcomes under SMR
  • SMR with equal weights Find outcome sum over
    D gt 0 ? x is chosen.

8
Three qualities of SMR
  • Monotonicity, Undifferentiatedness,
    Neutrality.
  • Monotonicity If some voters switch from x to y
    then xs outcome cannot worsen
  • Formally D, D ? D. If Di ? Di for all i?N
    then F(D) ? F(D).
  • Strong monotonicity A change breaks ties

9
Monotonicity Unanimity
  • Unanimity If everyone favours x then x wins
  • Weak Unanimity If everyone favours y then y
    does not win
  • Strong Monotonicity ? Unanimity
  • Monotonicity ? Weak Unanimity
  • Does Unanimity imply monotonicity ?

10
Violations of Monotonicity
  • Musical Contest Decision Rule
  • Judges award contestants 1-25 points
  • Median determined, outliers eliminated
  • Points summed, highest score wins
  • Assume three judges, two contestants
  • Does this make sense?

11
Musical Contest
  • Performer A Performer B
  • Judge 1 15 10
  • Judge 2 16 10
  • Judge 3 24 15
  • Total 55 35

12
Musical Contest
  • Performer A Performer B
  • Judge 1 15 10
  • Judge 2 16 10
  • Judge 3 25 15

Total 31 35
Median 16
Median 8 24
13
Musical Contest
  • Performer A Performer B
  • Judge 1 15 10
  • Judge 2 16 10
  • Judge 3 25 15

Total 31 35
What else is bad about this system ?
14
Monotonicity with more than Two Alternatives
  • STV V voters electing S candidates
  • Calculate quota q(V/S1)1
  • If candidate gets q votes he is elected and
    surplus votes are distributed to second ranked
    candidate of his voters
  • Repeat until no one has q votes
  • Start dropping candidates with fewest votes and
    redistribute votes.
  • Repeat until someone has q votes.

15
The Single Transferable Vote
  • Example Xx,y,z,w, n26, S2
  • q(26/21)1 ? 9
  • 9 voters w ? z ? x ? y
  • 6 voters x ? y ? z ? w
  • 2 voters y ? x ? z ? w
  • 4 voters y ? z ? x ? w
  • 5 voters z ? x ? y ? w

i) w wins first seat
ii) No one has q
iii) z eliminated
iv) x wins second seat
16
The Single Transferable Vote
  • Example Xx,y,z,w, n26, S2
  • q(26/21)1 ? 9
  • 9 voters w ? z ? x ? y
  • 6 voters x ? y ? z ? w
  • 2 voters x ? y ? z ? w
  • 4 voters y ? z ? x ? w
  • 5 voters z ? x ? y ? w

i) w wins first seat
ii) No one has q
iii) y eliminated
iv) z wins second seat
17
The Majority Run-off
  • Xx,y,z, N1,2,3,,100

31 voters x ? y ? z 30 voters z ? x ? y 29
voters y ? z ? x 10 voters y ? z ? x
X 31 votes Y 39 votes Z 30 votes
18
The Majority Run-off
  • Xx,y,z, N1,2,3,,100

31 voters x ? y ? z 30 voters z ? x ? y 29
voters y ? z ? x 10 voters y ? z ? x
X 61 votes Y 39 votes
19
The Majority Run-off
  • Xx,y,z, N1,2,3,,100

31 voters x ? y ? z 30 voters z ? x ? y 29
voters y ? z ? x 10 voters y ? z ? x
20
The Majority Run-off
  • Xx,y,z, N1,2,3,,100

31 voters x ? y ? z 30 voters z ? x ? y 29
voters y ? z ? x 10 voters x ? y ? z
X 41 votes Y 29 votes Z 30 votes
21
The Majority Run-off
  • Xx,y,z, N1,2,3,,100

31 voters x ? y ? z 30 voters z ? x ? y 29
voters y ? z ? x 10 voters x ? y ? z
X 41 votes Z 59 votes
22
Undifferentiatedness
  • Not equality but sufficient.
  • Undifferentiatedness allows anonymity and
    anonymity allows equality.
  • Violations of anonymity Weighted voting some
    votes count more than others, e.g., U.K. in the
    good old days.
  • U.N. Security Council

23
Undifferentiatedness
  • Formally, any permutation of the vote vector
    results in the same outcome
  • 3 voters. The vectors (1,1,0), (1,0,1) and
    (0,1,1) all result in the same choice.
  • Undifferentiatedness vs. Anonymity
  • Possibility of coercion
  • Legislative Voting

24
Demand Revealing Processes
  • Necessarily violate undifferentiatedness
  • Argument for
  • Strength of Preferences should be taken into
    account
  • Argument against
  • Wasteful
  • Unfair Wealth not evenly distributed
  • Possibility of Coercion

25
Neutrality
  • A Neutral Method favours no alternative
  • Formally, if preferences are switched around
    then outcome changes
  • If (1,1,-1) 1 then (-1,-1,1) -1

26
Neutrality
  • Many rules violate
  • Super-majority rules
  • Simple Majority Rule with a Tie-breaker
  • Minority Decisions Rules (Agenda Setting)
  • Some that satisfy
  • The Jury Rule Tie mistrial
  • SMR without a tiebreaker (or a coin-flip)
  • Rules for gt2 alternatives Condorcet, Borda,
    Bentham, Plurality

27
Summary
  • Three qualities of SMR with Xx,y
  • Undifferentiatedness votes equal
  • Neutrality alternatives treated equal
  • Monotonicity Non-arbitrary
  • What is fair ?
  • Demand revealing processes strength of
    preferences.

28
Summary
  • SMR with Xx,y the only rule that satisfies
    the conditions of monotonicity, neutrality and
    undifferentiatedness (Mays Theorem).
  • Is SMR with Xx,y superior ?
  • X is rarely x,y naturally.
  • Any process of narrowing choices will violate
    some of the conditions.

29
Lessons
  • Arguments favouring a Responsible Party System
    misguided
  • England Disciplined Parties offering real
    choices
  • Two counter-arguments
  • Dahl Observes that policy is made after
    elections building coalitions out of many
    minorities.

30
Lessons
  • Two counter-arguments (cont.)
  • Downs Proposes a model of two candidate/party
    competition. Party platforms should converge on
    median.

31
Three or More Alternatives
  • SMR Some great qualities but rarely applicable
  • Choice between two alternative
  • Modify definitions
  • X x,y,z,.
  • Di complete and transitive
  • xy xPiy, yx yPix, or (xy) xIiy
  • Indifferent voters dont vote

32
Majoritarian Methods
  • Condorcet Rule
  • Often no best alternative the Paradox of
    Voting
  • Policy Cycles can occur

33
Condorcet Rule
  • Xw,x,y,z, N1,2,3,4

D1 w ? y ? x ? z D2 x ? w ? z ? y D3 y ? z ?
w ? x D4 x ? w ? y ? z
w wIx, wPy, wPz x xIw, xIy, xPz
Choice?
34
Majoritarian Rules
  • The Amendment Procedure
  • The goal is to find a Condorcet Winner
  • X t,w,x,y,z,s
  • t Motion
  • w Amendment
  • x Amendment to Amendment
  • y Substitute Amendment
  • z Amendment to Substitute
  • s Status quo

35
The Amendment Procedure
w vs. t
D D1, D2 s t w D3, D4 t s w D5
w s t
w wins
t wins
w vs. s
t vs. s
w
s
t
s
36
The Amendment Procedure
w vs. t
D D1, D2 t w s D3 s t w
D4 s w t D5 w s t
w wins
t wins
w vs. s
t vs. s
w
s
t
s
37
The Amendment Procedure
  • Xw,x,y,t,s, N1,2,3

Step 1 x vs. w ? w Step 2 w vs. y ? y Step 3 y
vs. t ? t Step 4 t vs. s ? t
D1 w ? x ? t ? y ? s D2 y ? w ? x ? t ? s D3
s ? x ? t ? y ? w
Yet x is unanimously preferred to t !!!
38
The Successive Procedure
  • Pick a Candidate and pit against every other
    Candidate in X
  • A Voter votes for the Candidate (x) if he is
    preferred to all other candidates in X\x
  • Problems
  • Sometimes no winner (POV)
  • Condorcet winner may lose

39
The Successive Procedure
  • Xw,x,y,z, N1,2,3

D1 w ? x ? y ? z D2 y ? x ? w ? z D3 z ? x ?
y ? w
The Successive Procedure fails to select a winner
Yet x is a Condorcet winner !!!
40
The Majority Runoff
  • Again the goal is to select the Condorcet Winner
    or at least a large plurality winner.
  • Need not find Condorcet Winner

41
The Majority Runoff Condorcet Winners
  • Xx,y,z, N1,2,3,4,5

D1,D2 x ? z ? y D3,D4 y ? z ? x D5 z ? x
? y
X 2 votes Y 2 votes Z 1 votes
X 3 votes Y 2 votes
Yet z is a Condorcet winner !!!
42
Some other rules
  • Copeland
  • Schwartz Top cycles
  • Kemmeny

43
Positional Methods
  • Majoritarian Methods generally use binary
    comparisons
  • Positional Methods aim at using all the
    information we have about the voters preferences

44
Borda Rule
  • Xx,y,a,b,c, N1,2,3,4,5

4 3 2 1 0
D1 x ? y ? a ? b ? c D2 y ? a ? c ? b ? x D3
c ? x ? y ? a ? b D4 x ? y ? b ? c ? a D5 y ? b
? a ? x ? c
x 12 y 16 a 8 b 7 c 7
But x is the Condorcet winner !
45
The 1912 Presidential Election
  • XWilson, Roosevelt, Taft, ?, N?

42 Wilson ? Roosevelt ? Taft 27 Roosevelt ?
Taft ? Wilson 24 Taft ? Roosevelt ? Wilson 7
Other
46
Approval Voting
  • Xx,y,z, N1,2,3,,101

D1-D61 x ? y ? z D62- D81 y ? x ? z
D82-D101 z ? y ? x
Assume all voters cast two votes.
x 81 y 101 z 20
Condorcet winner ?
47
Borda Count
  • Ranking may reverse if an alternative is removed
  • Alternative y may win in X, but lose in all
    proper subsets of X.

48
Utilitarian Methods of Voting
  • Majoritarian Methods rely on pairwise
    comparisons of alternatives
  • Positional Methods rely on comparisons of rank
    of all alternatives
  • Neither takes strength of preference into
    account.
  • Utilitarian Methods incorporate intensity of
    preference

49
Utility
  • A utility is a measure of preference
  • It is not obvious how to assign utilities
  • Von Neumann-Morgenstern utilities
  • Obtain a rank ordering (say x, y, z)
  • Assign 1 to top, 0 to bottom
  • Generate a lottery over x and z pu(x)
    (1-p)u(z)
  • Find a p such that i is indifferent between
    lottery and z.

50
Utilitarian Methods
  • Already seen
  • Bentham Social Welfare
  • Demand Revealing Truthful
  • Nash Consistent and Protective

51
Criteria for Judging Methods
  • Undifferentiatedness
  • Monotonicity
  • Neutrality
  • Condorcet Criterion A majority winner?
  • Consistency Divide X if y wins in both
    subsets it should also win in X
  • Independence of Irrelevant Alternatives Whether
    x or y wins should not depend on other
    alternatives

52
Violations
  • Certain types of violations can be associated
    with different methods
  • Majoritarian Consistency
  • Positional Condorcet
  • Borda also IIA
  • Approval Undifferentiatedness
  • Utilitarian Methods IIA

53
Copeland Consistency
N1
N2
17 x ? y ? z 8 y ? z ? x 5 z ? x ?
y
14 x ? z ? y 16 y ? x ? z 15 z ? y ? x
x 2 points y 0 points z -2 points
x 0 points y 0 points z 0 points
54
Copeland Consistency
N
N1
N2
?
17 x ? y ? z 8 y ? z ? x 5 z ? x ?
y
14 x ? z ? y 16 y ? x ? z 15 z ? y ? x
x 0 points y 2 points z -2 points
55
Borda IIA
D1 D2 D3 Total
D1 a ? b ? c D2 c ? a ? b D3 c ? a ? b
a 2 1 1 4 b 1 0 0 1 c 0
2 2 4
a and c tie
56
Borda IIA
D1 a ? b ? c D2 c ? a ? b D3 c ? a ? b
57
Borda IIA
D1 D2 D3 Total
D1 a ? b ? c D2 c ? b ? a D3 c ? a ? b
a 2 0 1 3 b 1 1 0 2 c 0
2 2 4
c wins !
58
Approval Vote Undifferentiatedness
  • Assumptions about how many votes a voter casts
    these depend preferences
  • Rikers example assumes that the rule determines
    the number of votes
  • Skip example but need to know Approval Rule

59
Bentham IIA
D1 D2 Total
D1 x ? y ? z D2 y ? x ? z
x 1.0 0.6 1.6 y 0.5 1.0 1.5 z 0
0 0
60
Bentham IIA
D1 D2 Total
D1 x ? y ? z D2 y ? x ? z
x 1.0 0.5 1.5 y 0.6 1.0 1.6 z 0
0 0
Why is this a violation of IIA ?
Because of the way we construct the utilities
they are in essence relative measures
61
What to do ?
  • No fair way of aggregating preferences
  • No fair voting methods
  • Moreover, no Philosopher Kings.
  • Even if we know everyones preferences fair
    choices can not always be made
  • Many additional criteria of fairness

62
Rikers Recommendations
  • Different voting rules for different situations
    because choices must be made
  • Legislatures Amendment, Borda, Kemeny
  • Elections of Officials Plurality
  • Primaries Approval
  • Economic Planning Demand Revealing

63
Protecting the Two-Party System
  • In a previous chapter Riker demonstrates the
    hopelessness of a Responsible Party System
  • Why is Riker so protective of the two-party
    system?
Write a Comment
User Comments (0)
About PowerShow.com