Tutorial 2: Social Choice

ECNM10118 Political Economy∗

Question 1
In the lecture we discussed how voters can change the election results in their favour by voting
strategically, taking as given that others vote sincerely. Such approach shows that there may be
profitable deviations from sincere voting, but does not confirm whether strategic voting can be part
of an equilibrium (as well as whether others being sincere is optimal). This question aims at closing
this gap.

Consider a situation with three voters (i ∈ {1, 2, 3}) and the set of alternatives {a, b, c}. Voters
have the following preferences (as in slides 17-20 of Lecture 2):

1:a≻b≻c

2:b≻a≻c

3:c≻b≻a

To choose among the three alternatives, the two round system is adopted. In the first round,
each voter casts one vote. If one option has the majority (more than 50% of the votes), then it is
adopted. If not, then the two alternatives with the highest support pass to the second round, and
the option with majority support wins in the second round. Given the simple setup with just three
voters, if no option gets the majority (at least two votes) in the first round, it means that three
options tie and it is impossible to determine the top two. To resolve this, assume that option b is
eliminated (to match b having the minority support in the lecture) and there is the second round
between a and c.

∗
Questions? Email: [Link]@[Link]

1
(a) Solve the game assuming that voters vote sincerely according to their true preferences.

Answer: In the first round, each voter votes for her preferred alternative (i.e. 1 for a, 2 for b and
3 for c). There is a tie between the three alternatives and therefore there is a second round between
a and c, with a winning (since 1 and 2 prefer a over c).

(b) Confirm that sincere voting in the first round cannot be part of a Subgame Perfect Nash
Equilibrium (SPNE). [Hint: assume that SPNE where everyone votes sincerely does exist, and find
a voter who has a profitable deviation from sincere strategy in the first round. Note that sincere
voting is the only equilibrium behaviour in the second round given that there are just two options.]

Answer: Suppose that sincere voting is SPNE. Then, the equilibrium outcome is that a wins in the
second round, as found in (a). Consider voter 3 deviating to voting for b in the first round instead
of casting his sincere vote for c (while 1 and 2 still vote sincerely). Such deviation leads to option
b winning in the first round with majority support (from voters 2 and 3), and so voter 3 is strictly
better off from deviating, as she prefers b to a. Hence, sincere voting cannot be SPNE.

(c) Assume now that all voters behave strategically (i.e., play best responses) and exclude weakly
dominated strategies (i.e., a voter never casts a vote for some option x, if there is another alternative
y that leaves her at least indifferent between x and y in terms of election outcome and sometimes
makes her strictly better off with y, for any possible actions of other voters). We want to find a
pure strategy SPNE and show that it is unique (given elimination of weakly dominated strategies).
For this, follow the steps:
(i) Show that voter 1 has a weakly dominant strategy to vote for option a in the first round, i.e.
voting for b or c cannot make her strictly better off for any possible voting decisions of the other
voters. [Hint: it is convenient to consider if voter 1 can become better off by changing her vote
from a to another alternative for any election outcome rather than any strategies of other voters -
i.e. for when a wins in the first round, b wins in the first round, c wins in the first round, or there
is a tie in the first round.]

2
Answer: Suppose that voter 1 votes for a. To confirm weak dominance, we need to check that she
can never become strictly better off by deviating to b or c (for at least one strategy profile by voters
2 and 3), and that sometimes such deviation makes her strictly worse off. Consider deviations to
b or c in four possible election outcomes:
a wins in the first round: this outcome is already the best one for voter 1, so she trivially cannot
become strictly better off by any deviation. Yet, her deviation can make her strictly worse off (eg.
suppose voters 1 and 2 vote for a, while voter 3 votes for b (or c). Then, by deviating to b (or c)
voter 1 becomes strictly worse off as the winner shifts to b (or c)).
b wins in the first round: this is possible only if voters 2 and 3 both voted for b in the first round.
In this case voter 1’s vote cannot change the majority (she is never pivotal), and so any deviation
cannot make her strictly better off (as well as strictly worse off ).
c wins in the first round: same intuition as when b wins in the first round (voter 1 is not pivotal).
Tie in the first round, so a wins in the second round: similarly to a winning in the first round, the
outcome is already best for voter 1, so she cannot become strictly better off. In fact, any deviation
would make her strictly worse off, as it would create a majority for less favourable option in the
first round.

(ii) Similarly, show that voter 2 has a weakly dominant strategy to vote for option b in the
first round.

Answer: Similar intuition as for part (i). Assume voter 2 votes for b and consider if she can
benefit from deviating:
a wins in the first round: voter 2 is not pivotal, so cannot become strictly better (or worse) off.
b wins in the first round: this is first best outcome for voter 2, so she cannot become better off. But
can become strictly worse off if she shifts her vote (as then it could either shift majority to less
favourable option or cause a tie with a winning in the second round).
c wins in the first round: voter 2 is not pivotal, so cannot become strictly better (or worse) off.
Tie in the first round, so a wins in the second round: voter 2 deviating from b to a leads to a win-
ning in the first round (same outcome, so voter 2 is neither better off nor worse off ). Deviating
from b to c makes her strictly worse off, as this leads to c winning in the first round (and c is

3
worse than a for voter 2).

(iii) Find the best response for voter 3 (given others play weakly dominant strategies), and
conclude that this constitutes the unique SPNE.

Answer: Given voter 1 votes a and voter 2 votes b in the first round, voter 3 voting for a or c
leads to a winning (either directly with majority in the first round, or through a tie and winning
in the second round), while voting for b creates a majority for b in the first round and so b wins.
Since voter 3 prefers b to a, her best response is to vote for b in the first round.
This confirms the SPNE: {(a,a), (b,a), (b,c)} - equilibrium strategies for voters 1, 2 and 3 (re-
spectively) in the first and second round (respectively). It is unique given elimination of weakly
dominated strategies, since voters 1 and 2 have all the other votes weakly dominated, while voter
3 has a unique best response to weakly dominant strategies of voters 1 and 2.
Overall, we confirmed that acting strategically can form an equilibrium.

(d - Advanced) To make the two rounds system more realistic in terms of which options proceed
to the second round, we need to have at least 5 voters (i.e. at least three votes are required to win
in the first round, and the vote split of 2 - 2 - 1 determines which options proceed to the second
round). Assume that there are two voters with preference 1, one voter with preference 2, and
two voters with preference 3. Confirm that the unique SPNE (given elimination of weakly domi-
nated strategies) is such that in the first round type 1 voters vote for a, while type 2 and 3 voters
vote for b, and so b wins the election in the first round. [Hint: follow the same steps as in part (c).]

Answer: We can similarly show that any voter with preference 1 will find it weakly dominant to
vote for a, while voter with preference 2 still finds it weakly dominant to vote for b. Given that the
intuitions remain similar, let us only confirm weak dominance for voters with preference 1.
Now that there are 5 voters, there are more possible election outcomes. Specifically, there is a
majority in the first round if there are at least three votes for the same option. Alternatively, the
votes may split as 2 - 2 - 1, and so the contestants in the second round (where voters cannot do
better than vote sincerely) may be different. If it is a vs c in the second round, then a wins (as
before). If it is b vs c, then b wins (preferred to c by types 1 and 2 voters). Finally, if it is a vs

4
b, b wins (with support of types 2 and 3). Let’s go over the possibility to profitably deviate from a
for type 1 voter in each case (note that we only consider one of type 1 voters deviating at a time,
while the other sticks to some other strategy):
a wins in the first round: this is first best outcome for voter 1, so she cannot become better off (but
can become strictly worse off ).
b wins in the first round: voter 1 is not pivotal, so cannot become strictly better (or worse) off.
c wins in the first round: voter 1 is not pivotal, so cannot become strictly better (or worse) off.
a vs c in the second round, so a wins: first best outcome for voter 1, so she cannot become better
off (but can become strictly worse off ).
b vs c in the second round, so b wins: this means that the voter we are thinking about was the only
vote for a in the first round. Shifting her vote would create a majority for b or c in the first round,
so b or c would win in any case (where b is preferred and was already happening given her vote for
a). So she cannot become better off, but may become worse off with deviation.
a vs b in the second round, so b wins: this means that the voter we are thinking about was one of
the two votes for a in the first round. The only ways to become strictly better off is to create
majority for a in the first round (impossible in this case as there is only one other vote for a here,
while three are needed), or ensure that the second round is a vs c (also impossible given this vote
split, as if she does not vote for a, a will be left with just one vote and so will never pass to the
second round). As such there is no way to become strictly better off (while strictly worse off is still
possible).
The above confirms type 1 voters find it weakly dominant to vote for a. Similarly, it can be shown
that the type 2 voter weakly prefers to vote for b (proof omitted here). We need to determine the
best response of type 3 voters given the weakly dominant strategies of others (that guarantee two
votes for a and one vote for b) - where two type 3 voters can act differently potentially. The matrix
below determines the winner given all possible strategies of two type 3 voters:

Type 3 voter 1 / 2 vote a vote b vote c

vote a a a a
vote b a b b
vote c a b a

Option b is preferred to a by type 3 voters, and voting b weakly dominates voting for a or c for each
of these two voters individually (i.e. gives highest chance of b winning for any strategy of another

5
type 3 voter). Hence, their best response is to vote for b.
The above analysis leads to unique SPNE (with dominated strategies elimination) where in the
first round type 1 voters vote a, type 2 voter votes b and type 3 voters vote b, and so b wins with
majority in the first round. Notice that extending the setup from three to five voters did not change
the equilibrium outcomes, as well as strategies.

Question 2

Consider a society inhabited by a continuum of voters of size 1 (there is an infinity of voters

and each voter is negligible). The utility function of a voter i over private consumption ci and
public good y is:
yγ
ui (ci , y) = ci + αi , 0<γ<1
γ
The parameter αi represents a voter i’s preference for the public good. It is distributed over the
interval [0, 1] according to the continuous strictly increasing CDF F (·) with mean ᾱ ∈ (0, 1) (that
is, for all z ∈ [0, 1], P r(αi ≤ z) = F (z) - this is by definition of CDF and is the only knowledge
about the CDF that you may need in this question). Each voter is endowed with an income 1.
The income is taxed at rate τ and the proceeds are used to finance public good so y = τ . The
after tax income can be consumed by voter i so ci ≤ (1 − τ ) (obviously, holding with equality in
equilibrium). Given the set-up, the only relevant policy variable is the tax rate τ .

(a) Suppose voter i can choose the tax rate. What tax rate would she choose? Denote her pre-
ferred tax rate as a function of αi : τ (αi ).

Answer: The maximization problem of voter i after rewriting everything as a function of τ is:
1
γ
maxτ (1 − τ ) + αi τγ . The FOC yields: −1 + αi τ γ−1 = 0, or τ (αi ) = αi1−γ once rearranged.

(b) Are the preferences single-peaked? Justify your answer.

γ
Answer: The function û(γ) = (1 − τ ) + αi τγ is strictly concave in τ (since the second derivative
of the utility function is αi (γ − 1)τ γ−2 < 0 for any τ given γ < 1). Hence it is strictly increasing
for all τ < τ (αi ) and strictly decreasing for all τ > τ (αi ). Therefore, it satisfies single-peakedness,

6
with the bliss point at τ = τ (αi ).

(c) What is the tax rate chosen by majority rule? (Imagine all voters can propose a tax rate and
policy is chosen by pairwise majority rule)

Answer: Given single-peakedness, the tax rate preferred by the median (denoted αm ) is picked:
τ (αm ) - this is by the Median Voter Theorem. Note that the median is defined as F (αm ) = 1/2,
since this implies that exactly half of the population has αi < αm and exactly half has αi > αm .

(d) Suppose a social planner who cares about the average of the voters’ utilities were to choose
the tax rate. What is the socially optimum tax rate then? How does it compare to your answer
in (c)?

γ γ γ
Answer: The social planner’s problem is: maxτ E[(1−τ )+αi tγ ] = (1−τ )+E[αi ] tγ = (1−τ )+ ᾱ tγ .
Hence, the social planner chooses τ (ᾱ) (same maximization problem as in part a). It is greater
than the citizenry’s choice whenever ᾱ > αm (the median values public good less than the average
citizen), equals when ᾱ = αm , and strictly lower when ᾱ < αm . This question illustrates that social
outcomes in a citizenry are not necessarily socially optimal (eg., under ᾱ > αm more citizens care
less about public good than the average value, hence select lower taxation in elections, which leads
to the public good being under-provided).