Voting

Elliot Hunt, Maya Dhillon, Rahul Kumar and Ben Knight

Voting Systems and Types of Ballots

Rahul Kumar
Common Types Of Ballot

• Plurality

• Run-off Plurality

• Sequential Run-off

© Rahul Kumar 3
Common Types Of Ballot

• Borda Count
• With candidates x, y, z we can write the voters preferences as a vector
vi = (0, 1, 2).

• The Borda count for a candidate j is

n
X
vij .
n=i
.

© Rahul Kumar 4
Social Choice Theorems

Elliot Hunt
Arrow’s Impossibility Theorem

"It is impossible to achieve a fair result in a ranked voting system"

• Named after economist Kenneth J. Arrow

• Illustrates the flaws of a ranked voting system
• Arrow won the Nobel prize in Economic sciences for his work

© Elliot Hunt 6
Conditions for Fair Voting

The conditions for fair voting in a ranked voting system are as follows:

• Nondictatorship

© Elliot Hunt 7
Conditions for Fair Voting

The conditions for fair voting in a ranked voting system are as follows:

• Nondictatorship
• Pareto Efficiency

© Elliot Hunt 7
Conditions for Fair Voting

The conditions for fair voting in a ranked voting system are as follows:

• Nondictatorship
• Pareto Efficiency
• Independence of Irrelevant Alternatives

© Elliot Hunt 7
Conditions for Fair Voting

The conditions for fair voting in a ranked voting system are as follows:

• Nondictatorship
• Pareto Efficiency
• Independence of Irrelevant Alternatives
• Unrestricted Domain

© Elliot Hunt 7
Conditions for Fair Voting

The conditions for fair voting in a ranked voting system are as follows:

• Nondictatorship
• Pareto Efficiency
• Independence of Irrelevant Alternatives
• Unrestricted Domain
• Social Ordering

© Elliot Hunt 7
Example

Consider a scenario where Arrow’s Theorem is highlighted. Let there be 3

options, A, B and C, and an electorate of 100 people. Consider the following
voting profile:

• 40 votes A > B > C

• 35 votes B > C > A
• 25 votes C > A > B

© Elliot Hunt 8
May’s Theorem

"The only voting method for 2 candidates that is anonymous, neutral, monotone
and nearly decisive is the simple majority method"

Mathematician Kenneth O. May

© Elliot Hunt 9
May’s Theorem

• The candidate with the most votes will win

• An equal number of votes for each candidate will result in a tie

© Elliot Hunt 10
Proportionality vs. Majoritarianism

Ben Knight
What is Majoritarianism?

• Majority wins

• FPTP (UK and US)

• Works on Plurality

• Always majority but many unrepresented

© Ben Knight 12
What is Proportionality?

• Each vote counts for equal weight

• Additional Members (New Zealand)

• Single Transferable Vote (Scotland)

• Doesn’t guarantee majority

© Ben Knight 13
Criteria for Assessing Voting Systems

• Monotonicity:
• If one party receives more votes than another, they should receive more
electoral merits
• Typically not the case in Majoritarian systems

• Sainte-Lague:
(%seats−%votes)2
• A measure of proportionality:
P
%votes
• The lower the better

• d’Hondt:
• Simpler measure of proportionality: Maximum( %seats
%votes
)
• Closer to 1 = Better

• Regional Parties:
• Electoral systems should also try to ensure regional interests are represented to
some degree
• Often a trade off with monotonicity

© Ben Knight 14
Voting System Case Studies

• UK general election (FPTP) – 2005:

• Failed on Monotonicity

• Sainte – Lague of 20.65 – very high

• d’Hondt of 1.485 – high (also bad values for other parties)

• Regional parties represented (SNP and NI parties)

• New Zealand general election (AM) – 2020:

• Better on Monotonicity but Maori party overperformed

• Sainte – Lague of 6.39 – High ( mostly accounted for by Maori

overperformance (87%))

• d’Hondt of 1.42 – High (only for Maori party, around 1 for others)

• Regional parties represented(Maori)

© Ben Knight 15
 The UK needs electoral reform!

© Ben Knight 16
Tactical Voting

Maya Dhillon
Tactical Voting and the Myerson-Weber Strategy

• We have k candidates forming the set K = {1, 2, ..., k}.

• Each ballot can be written as a vector v = (v1 , v2 , ..., vk ).

• Each voter’s preferences can be written as a utility vector u = (u1 , u2 , ..., uk ).

• Pivot Probability is pij .

• We have vector p where p = (pij )ij∈H .

© Maya Dhillon 18
The Myerson-Weber Strategy

• We let G(p, v, u) be the expected gain in utility.

X X
G(p, v, u) = vi pij (ui − uj ).
i∈K i6=j

• We let X
pij (ui − uj )
i6=j

be the prospective rating.

© Maya Dhillon 19