Académique Documents
Professionnel Documents
Culture Documents
Bruno Amable
University of Paris I
2011
Probabilistic voting
2011
1 / 47
Probabilistic voting
2011
2 / 47
Probabilistic voting
2011
3 / 47
Assumptions
Income distribution F () is discrete (i.e. a step function) and not
continuous as in the Median Voter Model.
F is the cumulative distribution function. There are three distinct income
groups,
P: the Poorare P x 100% of population and all members have
income y P where P < 21
F (Y
yP ) = P
M: The Middle Classare M x100% of population and all members
have income y M where M < 12
F (Y
yM ) = P + M .
M
m
y = y the median income
R: The Richare R x100% of population and all members have
income y R where R < 21
F (Y
yR ) = P + M + R = 1 Average (expected) income is given by
J =P ,M ,R y J J
Bruno Amable (Paris I)
Probabilistic voting
2011
4 / 47
Probabilistic voting
2011
5 / 47
Probabilistic voting
2011
6 / 47
E [iJ ] =
1
2J
1
2J
iJ g J (iJ )d iJ =
1 iJ 2 J 21 J
j 1
2
2J
= 0
The average voter in group J has neither bias towards B nor A and
only evaluates policy
However, the distribution of ideology may dier across income group
Often-made assumption: middle income voters may have a higher
density than the other groups, M > P and M > R [Why?]
As the density, J , increases, the distribution function becomes more
concentrated around the average (assumed to be equal to zero), and
hence the group becomes less ideological, as fewer voters have a
strong ideology or sympathy towards a candidate, while more voters
are neutral, or almost neutral.
Bruno Amable (Paris I)
Probabilistic voting
2011
7 / 47
Probabilistic voting
2011
8 / 47
1
1
2 , 2
.Probability that
On average, the population has neither bias towards B nor A and only
evaluates policy
Candidates cannot control their average popularity before the election
they are aware of the fact that for instance scandals may occur with
some probability
Bruno Amable (Paris I)
Probabilistic voting
2011
9 / 47
ideologyiJ
average popularity
Probabilistic voting
2011
10 / 47
Voter i in group J will vote for party B, if the utility associated with
the platform indicated by this party plus his individual ideology and
the average popularity is larger than the utility associated with the
platform launched by party A
iJ can be positive or negative, iJ > 0 indicates an ideology
closer to party B
can be positive or negative, and > 0 indicates an average
popularity in favour of party B
W iJ (gB ) + iJ + > W iJ (gA )
Probabilistic voting
2011
11 / 47
Timing of events
Probabilistic voting
2011
12 / 47
Probabilistic voting
2011
13 / 47
The swing voter divides the income group into two subgroups
Probabilistic voting
2011
14 / 47
Probabilistic voting
2011
15 / 47
CandidatesDecisions
Probabilistic voting
2011
16 / 47
By using the denition of swing voter, one can easily obtain the mass of
the voters for candidate A in group J:
J
1
2J
=
=
1
+ J J
2
h
1
+ J W J (gA )
2
Probabilistic voting
W J (gB )
2011
17 / 47
One applies the same reasoning to each group and sum up votes
across groups. Total vote share for party A, A is
A =
J =P ,M ,R
h
J
+ J J W J (gA )
2
W J (gB )
1
Party
h A wins the
i
n Jelection when A > 2 .The probabilityois
Pr J =P ,M ,R 2 + J J W J (gA ) W J (gB ) J > 12
Probabilistic voting
2011
18 / 47
W J (gB )
io
>
1
2
i.e. when
<
J =P ,M ,R J J W J (gA )
W J (gB )
Party A wins the elections if the realisation of the shock that takes
place before the election, , is below a certain threshold
Pr A >
1
1
= Pr < = +
2
2
Probabilistic voting
2011
19 / 47
Probabilistic voting
2011
20 / 47
2 J =P ,M ,R
the policy platform chosen by party A, gA , will aim at pleasing voters
in the more numerous (J ) and less ideological (J ) groups
Policies targeting more numerous (income) groups (large J ) have the
advantage of providing benets to more potential voters. However,
some voters in these numerous groups may be highly ideological, and
hence di cult to convince, despite the benets provided by the
platform. The idea of targeting less ideological groups (large J )
captures this element
Groups with a large mass of non-ideological voters that is, with
many swing voters are easy targets
Bruno Amable (Paris I)
Probabilistic voting
2011
21 / 47
1
= Pr > = 1
2
Pr
Probabilistic voting
2011
22 / 47
Probabilistic voting
) y i
2011
23 / 47
y i f y i dy i = E y i = y = g
W i (g ) =
= (y
g
y
g)
) y i + H (g )
y i + H (g )
yi
+ H (g )
y
Probabilistic voting
2011
24 / 47
If W J (gA ) = (y
pA =
1
+
2 J =P
,M ,R
J J (y
gA )
yJ
+ H (gA )
y
W J (gB )
pA
yJ
= 0 () J J Hg (gA ) = J J
gA
y
J =P ,M ,R
J =P ,M ,R
gA = Hg
With ye =
J J J y J
J J J
1 J J J y J
y J J J
= Hg
ye
y
= gB
gS
Probabilistic voting
2011
25 / 47
ye
y
Probabilistic voting
< Hg 1 ( 1 ) < Hg
1
1
ym
y
ym
y
2011
.If
26 / 47
Probabilistic voting
2011
27 / 47
Example
A 3-class economy: the poor (P), middle class (M) and the rich (R)
respective incomes: Y P = 12 , Y M = 23 , Y R = 1
proportions: P = 45%, M = 30%, R = 25%
Agentsutility function: U i (C , G ) = C + i ln G ; i = 1
i = P, M, R
Yi ;
Probabilistic voting
2011
28 / 47
1+
Median-voter result: G m =
1 Yi
G
1
3
G+ 1
Y i ln G
= 0 () G i = 1
Probabilistic voting
Yi
2011
29 / 47
Probabilistic voting
Individual party preferences within group are distributed according to
a uniform distribution with zero mean and unitary density
Probabilistic voting
2011
30 / 47
Vi (GB )
1
1
i = + i i
2i
2
V i (GA )
Vi (GB ) i i
V i (GA )
Vi (GB ) i i
i i +
Probabilistic voting
i
2
1
+
2
2011
31 / 47
V i (GA )
1
2
Vi (GB ) i i
i.e.
<
i V i (GA )
Vi (GB ) i i
Pr A >
1
1
>
2
2
1
1
= +
2
2
1
max +
GA 2
max
V i (GA )
Yi
Vi (GB ) i i
GA + 1
Y i ln GA i i
Probabilistic voting
Vi (GB ) i i
i
2011
32 / 47
1+
YP
GA
+ M
1+
YM
GA
+ R
1+
YR
GA
=0
GAPV = 0.325
Even with identical s, the equilibrium is dierent from the median
voter result
GAPV < G m
A high R would lead to a low GAPV
Probabilistic voting
2011
33 / 47
Lobbying models
In Probabilistic Voting Models:
Voting strategy of voter i in group J is aected by (1) the economic
policy that is implemented (2) his individual ideological bias iJ
toward party B, and (3) the popularity of B.
If parties apply probabilistic voting, they maximise a particular social
welfare function, weighing all votersutility. But weights are biased
towards groups with a narrow distribution of ideology (groups with
many swing voters, groups with high concerns of policy evaluation,
high J
Equilibrium outcome of public good provision (e.g. infrastructure,
social spending, health care, theaters ) does not coincide with the
median voter model outcome in 2-party competition
If voting is on multidimensional issues (q contains at least two
policies), an equilibrium exists with probabilistic voting, if the
probability of winning function p( ) is well behaved. The equilibrium
is unique if p( ) is concave
Bruno Amable (Paris I)
Probabilistic voting
2011
34 / 47
Lobbying models
Probabilistic voting
2011
35 / 47
Probabilistic voting
2011
36 / 47
Probabilistic voting
2011
37 / 47
CA )
Probabilistic voting
2011
38 / 47
Timing of events
2
3
4
5
Probabilistic voting
2011
39 / 47
Swing voter
J
= W J (gA )
= W J (gA )
W J (gB )
J
W (gB )
= Pr A
=
=
h (CB
CA )
1
j
2
n
h
io
1
J
J
J
+
W
(
g
)
W
(
g
)
h
C
C
(
)
A
B
B
A
2 J =P
,M ,R
(
)
h
i
1
J
J
J
+
W (gA ) W (gB ) h (CB CA )
2
J =P ,M ,R
Probabilistic voting
2011
40 / 47
Expected gain is utility its members derive from the election, given by
pAhW J (gA ) + (1 ipA ) W J (gB ) minus the cost of contributions
2
2
1
CAJ + CBJ
2
Probabilistic voting
2011
41 / 47
F.O.C.
Since
p A
C AJ
pA h J
W (gA)
CAJ
W J (gB )
h
hJ W J (gA )
CAJ
W J (gB )
Probabilistic voting
2011
42 / 47
=
=
=
1
+
2
1
+
2
1
+
2
J =P ,M ,R J W J (gA ) W J (gB )
h CB J =P ,M ,R J CAJ O J
J =P ,M ,R J W J (gA ) W J (gB )
h CB J =P ,M ,R J hJ W J (gA ) W J (gB ) O J
)
h
i
J W J (gA ) 1 + J hJ + ...
J =P ,M ,R
Probabilistic voting
2011
43 / 47
1
pA = +
2
J =P ,M ,R
gA ) yy + H (gA ), pA becomes
J (y
h
i
yJ
gA )
+ H (gA ) 1 + J hJ + ...
y
Probabilistic voting
2011
44 / 47
F.O.C.
pA
= 0 () J
gA
J =P ,M ,R
J =P ,M ,R
h
i
yJ
+ Hg (gA ) 1 + J hJ = 0 ()
y
h
i
J 1 + J hJ Hg (gA ) =
Probabilistic voting
J =P ,M ,R
i
yJ h
1 + J hJ
y
2011
45 / 47
gA
= Hg
= Hg
"
1 J =P ,M ,R J y J 1 + J hJ
y J =P ,M ,R J [1 + J hJ ]
yb
= gL
y
yb = y1 J =P ,M ,R J [1 +J hJ ] is lobbying-weightedincome
J =P ,M ,R
Probabilistic voting
2011
46 / 47
weights reect group size and whether or not the group is organised.
Organised groups receive greater weights and equilibrium is tilted
towards their groups bliss point which is determined by their income.
This only holds if not all groups are organised.
In comparison with the result of probabilistic voting, gS converges to
the bliss point of groups with many swing voters (high J ), these
have more political power for the election outcome
Lobbying equilibrium: more political power of group J through
campaign contributions in cash and kind
As candidates seek only election victory, and groups organised in
lobbies can help them achieve this goal by nancing their campaigns,
both parties bias their policy platforms towards the direction desired
by the lobbies i.e. the organised groups.
Probabilistic voting
2011
47 / 47