Professional Documents
Culture Documents
Political Economy
Tim Besley
March 2022
IIA: The ranking between two policies ga and gb should depend only
the individual preference ranking between a and b
What does it mean in practice?
Think of a jury of …ve people deciding to rank 5 candidates in a
competition
It decides to rank them 1,2,3,4,5
Then add up the score (this is called the Borda count)
It then announces the winner
This violates IIA as if one judge changes their ranking between two
candidates, it can change ranking between the winning two
J1 J2 J3 J4 J5
C1 1 1 1 5 5
C2 2 2 2 4 4
C3 3 3 3 1 2
C4 4 4 4 2 3
C5 5 5 5 3 1
The order is
J1 J2 J3 J4 J5
C1 1 1 1 5 5
C2 2 2 2 2 3
C3 3 3 3 1 2
C4 4 4 4 4 4
C5 5 5 5 3 1
The judges change their preferences only over the pairs [C2, C4], [C2,
C5] and [C4, C5].
BUT
The choice has changed the ranking of [C2, C1] and [C2, C3].
This violates IIA
C2 now wins instead of C3, even though no voter changed their
preference over [C2, C3]!
This rules out the Borda count as a way of deciding if we care about
IIA
but what about everything else?
The starkest version of this can be seen with three voters and three
alternatives
The society will base its decision on what the majority wants
BUT: 1 beats 2, 2 beats 3, and 3 beats 1.
So whichever pair of alternatives is voted over, there would not be a
clear winner.
(And if all three were on the table they would tie.)
After the 2016 referendum where citizens voted for Brexit, there was
a question of how to leave
There was no single alternative
Eventually Mrs May reached a deal with the EU
But she had to get it through Parliament
And there was an eye to what the public wanted too
sooner or late the government has to face the electorate
See https://blogs.lse.ac.uk/politicsandpolicy/brexit-condorcet/
What would happen with three alternatives
Deal
Remain
No Deal
In some polling data: Deal > Remain > No-deal > Deal
So there was no majority for anything over anything else just as in the
example above!
In the jargon of voting theory, there is no Condorcet winner, i.e.
something beats everything else in pair-wise comparisons.
The public good can be provided via taxation at a per capita cost of
c.
Each has the same income of m
θ i log (G ) + m cG
or
θi
Gi =
c
and
GA < GB < GC .
Not in general
Suppose that the social welfare function is Utilitarian:
h i h i h i
θ A log G + x A + θ B log G + x B + θ C log G + x C
.
3
What level of public goods maximizes this?
θ A +θ B +θ C
Let θ̄ = 3 then this maximizes
θ̄ log (G ) cG
or
θ̄
G = .
c
So welfare wants the outcome for the mean public goods preference
Politics delivers what the median voter, B, wants
They need not be the same.
The proposition is true in general for any social welfare function
unless by chance, the welfare function is perfectly aligned with the
median voter.
But this is a purely distributional di¤erence
it a¤ects who gets what.
This reinforces the idea politics will not generally maximize a social
welfare function.
And we could regard this as a “government” failure if there us a big
di¤erence between what a welfare function says what we want and
what politics delivers.
For the purposes of the analysis, we will suppose that the ruling elite
is somewhat self-interested.
B + G = T,
We will now suppose that citizens are identical and value public goods
equally
So we do not worry about preference aggregation across citizens
Preferences of citizens are
θ log (G ) + m T
θ log (G ) + log (B ) + m
where log (B ) is the utility from the transfers that they receive which
are funded out of taxation.
For simplicity, we are assuming that the elite do not pay taxes.
so
θ
G = T < T.
1+θ
She spends a proportion θ/ (1 + θ ) of tax revenue on public goods
Citizens would like G = T .
We need to see whether the utility of citizens is higher when they pay
more tax as long as that tax is used to …nance public goods
starting from
θ
G = T
1+θ
Their utility is
V (R ) = θ log (G + R ) + m R T
They get G + R in public goods now and pay a bit more tax which
reduces their income.
Is V (R ) increasing in R?
Is
V (R ) = θ log (G + R ) + m R T
increasing in R at R = 0.
Di¤erentiate with respect to R :
∂θ log (G + R ) R θ
= 1>0
∂R G +R
θ
Set G = 1 +θ T and R = 0 when this becomes
θ 1+θ
1= 1
θ
1 +θ T
T
The World Values asks: "Is it justi…able to cheat on your taxes if you have
a chance?"
answered on 10 point scale
asked in all six survey waves across 94 countries (1981-2014)
about 250,000 observations
around 63% of the population do not think that cheating is justi…able
Justi…able cheating
increasing in income
decreasing in education and age
more prevalent among men
correlated with lack of con…dence in government
https://www.ft.com/content/4b3e6db0-e57a-11e7-8b99-0191e45377ec
(R, B 0 , G 0 ) maximizes
θ log G 0 R
subject to
and
B 0 + G 0 = T + R.