Robert Keating
Professor Georges
Public Choice Economics
May 19, 2026
5.1
Corollary 5.2 says it's impossible for a social choice function with 3+ candidates to be Pareto, independent, and
nondictatorial all at once. We need to show that dropping any one of those three requirements makes the corollary
fail - meaning we can find a method that satisfies the remaining two.
Drop Pareto: The all-ties method (which always declares everyone a winner) is anonymous (so nondictatorial) and
independent (the outcome never changes regardless of voter preferences, so in particular it never switches a loser to
a winner). It is not Pareto since it elects candidates even when everyone prefers someone else to them. So we have a
nondictatorial, independent method - the corollary fails without Pareto.
Drop independence: The plurality method is anonymous (so nondictatorial) and Pareto (if everyone prefers A to B,
B gets no first-place votes and can't win). It is not independent, as we showed in Chapter 4. So we have a
nondictatorial, Pareto method - the corollary fails without independence.
Drop nondictatorial: Any dictatorship is Pareto (the dictator's top choice is always the unique winner, and if
everyone prefers A to B then the dictator does too so B won't win) and independent (the outcome only depends on
the dictator's preferences, so changing other candidates' relative rankings never affects the A vs B outcome). So we
have a Pareto, independent method that is a dictatorship - the corollary fails without nondictatorial.
5.3
The method: if one candidate is at the top of every voter's list, that candidate wins uniquely. Otherwise, everyone
ties.
(a) Is it anonymous?
Yes. The rule only asks whether all voters agree on a top candidate - it doesn't care which specific voter submitted
which ballot. Swapping any two voters' ballots doesn't change whether unanimous agreement exists or who the top
candidate is. All voters are treated the same.
(b) Does it satisfy independence?
No. Consider:
Before:
1: A > B > C
1: A > C > B
A is on top of both lists, so A wins uniquely. B loses.
After (second voter moves C up):
1: A > B > C
1: C > A > B
No candidate is at the top of every list now, so everyone ties. B goes from losing to winning. But no voter changed
their relative ranking of A vs B - independence is violated.
(c) Does it satisfy Pareto?
No. The after profile above shows it. Every voter prefers A to B (A is above B on both ballots), yet B is a winner in
the tied outcome. Pareto says B can't win when everyone prefers A to B.
5.5
Best-of-two: whoever the majority prefers between Prince A and Prince B wins, regardless of where other
candidates are ranked. If it's exactly 50-50, A and B tie.
(a) Show the best-of-two method is independent.
Suppose candidate X wins and candidate Y loses in profile P, and in profile Q every voter has X and Y in the same
relative order as in P. We want to show Y still loses in Q.
X must be one of the two princes - they're the only ones who can win this method. Say X is Prince A. Now there are
two cases for Y:
If Y is Prince B: the A-vs-B head-to-head is determined entirely by each voter's relative ranking of A and B. Since
every voter has A and B in the same relative order in both profiles, the head-to-head result is identical. Prince A
defeats Prince B in Q just as in P, so Y (Prince B) loses.
If Y is not a prince: Y loses in every possible profile under this method, since only the two princes can ever win. So
Y definitely loses in Q.
Either way, Y loses in Q. The method is independent.
(b) Does the best-of-two method satisfy Pareto?
No. If every voter prefers some commoner C to both princes, C should win by Pareto. But the best-of-two method
ignores commoners entirely - one of the princes always wins. Pareto is violated. (Another way to see this: if best-of-
two were Pareto, it would be anonymous, Pareto, and independent - but it's also nondictatorial, which would
contradict Arrow's theorem.)
5.7
A coalition is influential if, whenever every voter in it ranks some candidate A first, A is guaranteed to be the unique
winner.
(a) Influential coalitions for plurality?
Any coalition with more than half the voters. If a majority all rank A first, A has a majority of first-place votes and
wins plurality. No smaller coalition can guarantee this - a coalition of exactly half or fewer voters could have all
their votes go to A while the other half splits differently, and A might not come out on top.
(b) Influential coalitions for a dictatorship?
Any coalition that contains the dictator. If the dictator ranks A first, A wins - end of story. It doesn't matter what
anyone else does. A coalition without the dictator has zero influence over the outcome.
(c) Influential coalitions for Borda count with 4 candidates and 10 voters?
Any coalition of at least 8 voters. Here's why: if 8 voters all rank A first, A earns at least 8 x 3 = 24 Borda points.
The best any other candidate can do is get 3 points from each of the 2 non-coalition voters (6 points) plus 2 points
from each of the 8 coalition voters if placed second on every coalition ballot (16 points), for a total of 22. Since 24 >
22, A wins.
A coalition of 7 is not influential. Here's a profile where 7 voters all rank A first but A still loses:
7: A > B > D > C
3: B > C > D > A
Borda scores: A = 7(3)+3(0) = 21; B = 7(2)+3(3) = 23; C = 7(0)+3(2) = 6; D = 7(1)+3(1) = 10. B wins with 23
points despite only having 3 first-place votes.
5.9
Nearly Pareto: if 90% or more of voters prefer A to B, then B must lose.
No social choice function can be nearly Pareto. Here's the argument using the following 10-candidate, 10-voter
profile:
Each voter's ranking is a cyclic shift of the candidates A through J:
Voter 1: A > B > C > D > E > F > G > H > I > J
Voter 2: J > A > B > C > D > E > F > G > H > I
Voter 3: I > J > A > B > C > D > E > F > G > H
Voter 4: H > I > J > A > B > C > D > E > F > G
Voter 5: G > H > I > J > A > B > C > D > E > F
Voter 6: F > G > H > I > J > A > B > C > D > E
Voter 7: E > F > G > H > I > J > A > B > C > D
Voter 8: D > E > F > G > H > I > J > A > B > C
Voter 9: C > D > E > F > G > H > I > J > A > B
Voter 10: B > C > D > E > F > G > H > I > J > A
In this profile, 9 out of 10 voters prefer A to B (every voter except voter 10). Nearly Pareto says B must lose.
Similarly, 9 out of 10 voters prefer B to C, so C must lose. Continuing the chain: D must lose, E must lose, F must
lose, G must lose, H must lose, I must lose, and J must lose.
But 9 out of 10 voters also prefer J to A - so A must lose too.
Every single candidate must lose. But a social choice function has to name at least one winner for every profile -
that's the universal domain requirement. We've reached a contradiction, so no social choice function can be nearly
Pareto.