EC302 Political Economics
Lecture 7: Electoral Competition
Prof. Ronny Razin
November 2013
Razin (LSE) EC302 11/13 1 / 24
Electoral Competition
Background
Until now: we studied the bounds of what we can expect a theory of
political economy to yield.
This was done by studying social choice theory.
We now move head-on to use game theory to construct models of
political economy that are closely related to real-life institutions.
This will allow us to construct a theory that yields predictions
and to gain some intuitions about the design of political institutions.
We focus on majority rule as a real-life feature at the heart of
democracies.
... But before we start we have to think clearly about what an
exhaustive model of political economy should include.
Razin (LSE) EC302 11/13 2 / 24
The stages of a political model
1 Entry
2 Campaign
3 Voting
4 Election results political positions
5 Policy determination
One would love to have a model that would encompass all these
stages, but...
Unfortunately this is not feasible.
Razin (LSE) EC302 11/13 3 / 24
The stages of a political model
What we strive for is then a bag of models/intuitions about the
di¤erent stages of the political process.
With this bag of intuitions we can proceed to discuss the design of
institutions and comparisons between di¤erent regimes.
Razin (LSE) EC302 11/13 4 / 24
Two-candidate winner-take-all election.
The simplest case to start with is that of a two-candidate
winner-take-all election.
Our focus is on the e¤ect of the campaign process and political
competition on political outcomes.
Razin (LSE) EC302 11/13 5 / 24
Two-candidate winner-take-all election.
1 Entry
2 Campaign
3 Voting
4 Election results political positions
5 Policy determination
We are analyzing stages 2 to 5, taking stage 1 as given.
Razin (LSE) EC302 11/13 6 / 24
Two-candidate winner-take-all election: Implicit
assumptions
We assume that there are two candidates/parties.
We are analyzing parties as unitary actors.
We assume that the political spectrum is one-dimensional and that
voters have single peaked preferences.
Razin (LSE) EC302 11/13 7 / 24
Two-candidate winner-take-all election: Questions
Convergence.
Hotelling model of spatial (economic) political competition.
How robust is the convergence result to changing the assumptions?
Razin (LSE) EC302 11/13 8 / 24
Two-candidate winner-take-all election: Questions
To this end we will now analyze several models related to the
Hotelling model, but relaxing one of the assumptions at a time.
Candidate motivation:
o¢ ce-motivated.
ideological.
Campaign credibility/commitment.
Full commitment.
No commitment.
Information.
Perfect information
Imperfect information.
Razin (LSE) EC302 11/13 9 / 24
Motivation\commitment Full commitment No commitment
Office motivated Convergence Unclear
(Hotelling)
Ideological ? Divergence
Razin (LSE) EC302 11/13 10 / 24
Two-candidate winner-take-all election: Imperfect
Information
Motivation\commitment Perfect information Imperfect information
Office motivated Convergence ?
(Hotelling)
Ideological ? ?
Razin (LSE) EC302 11/13 11 / 24
Two-candidate winner-take-all election: Perfect
Information
Full Commitment and Ideological candidates
Assume policy space is unidimentional- [ 1, 1].
Ideological candiates?
A candidate j is represented by an ideal point xj 2 [ 1, 1],
The utility of candidate j is just from the policy that is implemented,
y , NOT whether he wins or not.
uj (y , xj ) is single peaked.
Razin (LSE) EC302 11/13 12 / 24
Two-candidate winner-take-all election: Perfect
Information
Full Commitment and Ideological candidates
Assume two candidates, L and R with xL = 1 and xR = 1.
What are the equilibria?
Proposition Assume two candidates that are ideological with ideal
policies on the two sides of the median voter. The unique Nash
equilibrium in the campaign has both candidates announcing the
median’s ideal policy.
Razin (LSE) EC302 11/13 13 / 24
Two-candidate winner-take-all election: Imperfect
Information
Imperfect information?
So far we have assumed that both candidates know exactly where the
median is.
More realistically, candidates’knowledge of the whereabouts of voters
is determined by polls.
Polls produce a statistical estimation about where the voters are.
For our purposes, we will model the uncertainty of candidates about
where the median is by assuming that both candidates believe that the
median of the voter distribution is distributed according to some
distribution function F .
Razin (LSE) EC302 11/13 14 / 24
Two-candidate winner-take-all election: Imperfect
Information
Example
Example 1: assume that both candidates believe that the median
voter is either at - 21 , at zero or at 12 .
Moreover, we assume that both candidates believe that the probability
of each of these occurrences is a third.
Example 2: assume that both candidates believe that the median
voter is distributed uniformly on [ 1, 1].
Razin (LSE) EC302 11/13 15 / 24
Two-candidate winner-take-all election: Imperfect
Information
Full Commitment and O¢ ce -motivated candidates
What are the equilibria?
Proposition Suppose two o¢ ce-motivated candoidates compete in
an election. Assume that both candidates believe that the median
voter is distributed according to the distribution F . The unique
equilibrium has both announcing the median of the distribution of
medians F .
Proof:
If you think of the distribution of medians, F , as a distribution of
voters, you will see that this model is similar to a model with full
information and where the median is at the median of the distribution
F . Therefore, our proof for the full information model (Hotteling) can
be adopted here as well.
Razin (LSE) EC302 11/13 16 / 24
Two-candidate winner-take-all election: Imperfect
Information
Full Commitment and ideological candidates
Assume that there are two candidates, L and R with ideal points at -1
and 1 respectively.
UL (y ) and UR (y ) single-peaked.
Note, UL0 (xL ) = 0 and UR0 (xR ) = 0.
Candidates are imperfectly informed about the whereabouts of the
median: F .
Equilibria?
Razin (LSE) EC302 11/13 17 / 24
Two-candidate winner-take-all election: Imperfect
Information
Full Commitment and ideological candidates
Lemma When candidates are ideological and there is imperfect
information, no equilibrium has both candidates converging.
Proof:
Suppose both candidates converge on some policy y .
Focus on candidate L. His utility is UL (y ).
If he deviates a little bit to the left, y ε
If he looses he still gets UL (y ),
but if he wins he will get UL (y ε ) > UL ( y ) ,
As there is a chance he will win, he will deviate.
Razin (LSE) EC302 11/13 18 / 24
Two-candidate winner-take-all election: Imperfect
Information
Full Commitment and ideological candidates
Lemma When candidates are ideological and there is imperfect
information, no equilibrium has a candidate choosing his ideal point
as a platform.
Suppose there is such an equilibrium
Consider a small deviation away form the candidate’s ideal point.
This increases his probability of winning
Since his platform is close to his optimal policy the loss due to the fact
that his platform moved is negligable.
Focus on candidate L, assume that R chooses y > xL ,
x +y x +y
EUL = F ( L )uL (xL ) + (1 F ( L ))uL (y )
2 2
∂EUL 1 0 xL + y x +y
= F ( )(uL (xL ) uL (y )) + uL0 (xL )F ( L )
∂xL 2 2 2
1 0 xL + y
= F ( )(uL (xL ) uL (y )) + 0 > 0
2 2
Razin (LSE) EC302 11/13 19 / 24
Two-candidate winner-take-all election: Imperfect
Information
Full Commitment and ideological candidates
Proposition If candidates are ideological and if they are imperfectly
informed about the whereabouts of the median then in any
equilibrium we will have divergence but less than full divergence.
Equilibrium will be yL and yR such that,
∂EUL
(y jy ) = 0 ,
∂yL L R
1 0 yL + yR yL + yR
F ( )(uL (yL ) uL (yR )) = uL0 (yL )F ( )
2 2 2
and similarly for candidate R.
Razin (LSE) EC302 11/13 20 / 24
Two-candidate winner-take-all election: Imperfect
Information
Full Commitment and ideological candidates
Intuition: moving platform towards ideal point )
Cost: you loose with a higher probability.
Bene…t: when you win, you win with a better platform.
Equilibrium: cost=bene…t.
Razin (LSE) EC302 11/13 21 / 24
Two-candidate winner-take-all election
Motivation\commitment Perfect information Imperfect information
Office motivated Convergence Convergence
(Median) (Median of median
distribution)
Ideological Convergence ?
(Median)
Razin (LSE) EC302 11/13 22 / 24
Two-candidate winner-take-all election
What have we learned?
We learn that commitment, while an important aspect of
convergence, is not su¢ cient.
An additional important aspect of convergence is that candidates be
o¢ ce motivated.
As in politics we expect information not to be perfect, we will tend to
see convergence if candidates are o¢ ce motivated and we will tend to
see divergence when candidates/parties are ideological.
Razin (LSE) EC302 11/13 23 / 24
Two-candidate winner-take-all election
What have we learned?
Are these results satisfactory?
For one thing there is mixed evidence about convergence in empirical
work.
If anything the evidence suggests that yes candidates do converge
some way towards each other but they most certainly do not converge
the whole way.
But on theoretical grounds, as mentioned above, our analysis is not
complete.
One reason is that we have not modeled entry. This is the topic of
our next few lectures.
Razin (LSE) EC302 11/13 24 / 24
