Académique Documents
Professionnel Documents
Culture Documents
Jorge A. Gallego
Forthcoming, Journal of Theoretical Politics. I would like to thank Rebecca Morton, Adam
Przeworski, Alastair Smith, Joshua Tucker, Leonard Wantchekon, two anonymous referees and the
editor of the journal for their helpful comments.
1
(p wins) =
2
(p wins) =
3
(p wins) = +
with > , so that the third group still experiences disutility if the candidate
is elected. Hence, group 1 totally dislikes the candidate, group 2 remains neutral,
while group 3 likes him. Also, can be interpreted as a measure of heterogeneity
9
between the three groups, with a higher implying a higher degree of diversity in
their political preferences. For simplicity, assume that if this candidate is not elected
and his contender wins the election, the ideological payo for each group is zero. This
means that
i
(p loses) = 0, for i = 1, 2, 3.
To answer the questions posed at the beginning of this section, it is important to
recognize that the candidate has scarce resources he must use in order to obtain at
least two votes from the three groups. Assume that g > 0 represents an indivisible
benet that the politician can only give to one of the groups. We can think of g as
a public job, a local public good, or any other indivisible commodity that cannot
be shared simultaneously by two distinct groups. Consequently, the total payo for
group i, when the clientelistic candidate is elected, is given by
u
i
(p wins) =
i
+ g if i receives a gift
i
otherwise
Finally, assume that ps benet for winning the election is given by v > 0, and that
g < v, so that the oce is suciently valued regardless of the cost of buying the
clientele. The timing of the interaction, which will be called the clientelism-game,
is as follows:
1. The candidate p oers g to group 1, group 2, or group 3, or does not oer it
to anyone.
2. After observing ps action, the election takes place and each group votes simul-
taneously for or against p.
3. Payos are realized and p wins if he gets two or three votes. Otherwise, his
contender is elected.
Solving this extensive-form game yields ve subgame perfect equilibria (SPE): in one
of them the candidate is elected with three votes, and in the other four he receives
less than two votes and loses the election. In every case, the politician chooses not to
give the gift to any of the groups. Clientelism does not exist in equilibrium because
the candidate never has incentives to favor any of the groups. Consequently, if the
game is played once, the underlying structure makes it unsuitable for explaining the
emergence of clientelism within electoral systems.
Now assume that the clientelism-game described above repeats during t = 1, 2, . . .
periods, so that the players ignore the time horizon. Every period the candidate
10
decides who receives g and each group votes or not for the candidate. The feasibility
of this interaction naturally results from the fact that elections are held periodically
and politicians run for reelection, with the possibility of interacting with the same
(group of) citizens every period. Given this new structure, the candidate has no
incentives to give the gift to the same group during each period. But it might be
the case that making uncertain this allocation is in his interest, because this could
discipline at least two groups and make them vote for him. Consequently, in what
follows this paper assumes that the politician might follow, in each period, a mixed
strategy of the form (0, 0, , 1 ), where 0 < < 1 is the probability that group 2
receives the gift, while 1 is the probability that the benet goes to the third group.
The rst two zeroes in the vector indicate that the candidate assigns probability zero
to the event of not giving a gift to any group, and to give it to group 1, respectively.
In this context, the candidate randomizes between the neutral and the sympathetic
group, and excludes the group that is less likely to vote for him.
3
Being more precise, it assumed that players follow grim trigger strategies with
the following characteristics: the candidate plays (0, 0, , 1 ) in t = 1; For any
t > 1, he plays (0, 0, , 1 ) if he won the election in t 1; Otherwise, he plays
(1, 0, 0, 0) forever. Group 1 votes against the candidate in every t = 1, 2, . . .. Groups
2 and 3 vote yes in t = 1, and in t > 1 if the candidate won and played (0, 0, , 1)
in t 1. Otherwise, they vote against this candidate forever. Dene (0, 1) as
the common discount factor for the players. Under what conditions these strategies
form part of a SPE in the innitely repeated version of the game? Proposition 1
establishes sucient conditions for the above strategies conforming a SPE.
Proposition 1:
Consider the innitely repeated clientelism-game dened above, in which the candi-
date and groups 2 and 3 follow grim trigger strategies while group 1 always votes
against the candidate. Then, any mixed strategy of the form (0, 0, , 1 ) can be
part of a subgame perfect equilibrium as long as
g
1 +
+
g
(1)
for suciently close to one.
3
In the context of repeated games, we might interpret as the proportion of periods in which
group 2 is honored with the gift. Nonetheless, being strict with the denition of mixed strategy,
is just a probability. Assuming that g in fact is divisible, it is also possible to interpret as the
fraction of the gift given to group 2 during each period.
11
Proof:
It is straight forward to show that the politician has no incentives to unilaterally
deviate from his strategy. Suppose that groups 2 and 3 vote for the candidate. If he
follows his strategy, his discounted payo is
u
p
(coop) =
t=1
t1
(v g)
=
v g
1
If he cheats and deviates from giving a gift in every period, gets u
p
(cheats) = 0.
Then, because v > g, the candidate always honors his strategy. In subgames that
come from a defection of either group 2 or 3 (or both), the present value of the
candidates payo after following the grim trigger strategy is 0, while a one-stage
deviation (which implies giving g to either group after losing the election) yields a
payo of g. Then, the candidate has no incentives to deviate from his strategy.
Consider a subgame in which in the rst period of the stage game, g is given to
group 2. In this case, group 2s payo for cooperating is
u
2
(coop) = g + +
t=1
t
(g + )
= g + +
1
(g + )
After receiving the gift, if group 2 cheats, its discounted payo is
u
2
(cheats) = g + (0) +
2
(0) +
= g
Then, group 2 has no incentives to deviate from the grim trigger strategy as long as
g + +
1
(g + ) g
which is equivalent to
g
(2)
It is easy to show that (2) is also the necessary condition so that group 2 has no
incentives to deviate in subgames in which the gift was given to group 3. In these
12
type of subgames, group 3s payo for following the grim trigger strategy is
u
3
(coop) = g + + +
t=1
t
[(1 )g + + ]
= g + + +
1
[(1 )g + + ]
If this group cheats, receives u
3
(cheats) = g. Therefore, group 3 has no incentives
to deviate from the grim trigger strategy as long as
g + + +
1
[(1 )g + + ] g
or, equivalently, if
1 +
+
g
(3)
Again, (3) is also the condition for group 3s compliance in subgames in which the
gift is given to group 2. Finally, if the candidate and groups 2 and 3 follow their
strategies, no matter what group 1 does, it always receives an average payo of
u
1
() =
1
. For the result of the election and the materialization of payos, it does
not matter if this group votes against or in favor of the candidate. Hence, summing
up, a grim trigger strategy that supports a candidates mixed strategy of the form
(0, 0, , 1 ) constitute a subgame perfect equilibrium of the clientelism-game as
long as
g
1 +
+
g
which completes the proof.
Proposition 1 shows that there an innite number of mixed strategies that can
be implemented by the politician, in a SPE of the clientelism-game. Within this
paper, the set of mixed strategies that satisfy (1) is referred to as the feasible set.
What are the basic properties of this set of equilibria? How does the feasible set
respond to changes in dislike towards the party, heterogeneity among the groups, the
magnitude of the gift, or players patience? Proposition 2 shows that some interesting
comparative statics of the innitely repeated clientelism-game can be described.
Proposition 2:
Consider the clientelism-game in which any mixed strategy
which implies: a)
g
min
()
=
2
=
1
< 0
for any (0, 1); and c)
g
min
=
2 +
2
< 0 for any (0, 1) because we
assumed > .
2. The magnitude of the maximum level of dislike towards the candidate, in equi-
librium, is given by
max
=
g +
2
which implies: a)
(
max
)
g
=
2
> 0 for any (0, 1); b)
(
max
)
= 1/2;
and c)
(
max
)
=
g
2
> 0 for any g > 0.
14
Figure 1: Set of feasible mixed strategies as a function of g
g
(2+)
1
3. From (1), the minimum level of heterogeneity admissible for equilibrium is
min
= 2 g
Therefore: a)
min
()
= 2; b)
min
g
= < 0 for any (0, 1); and c)
min
g+
2
the gift. The intersection of and determines the minimum gift necessary for
equilibrium: g
min
=
2
g 2
max
=
g +
2
. This level is increasing in g, , and . Richer candidates can buy
voters that dislike more their party. More diversity makes party dislike less costly
to the politician. As stated above, this is because represents the sympathy that
group 3 feels when the candidate is elected. This attenuates and makes clientelism
easier. Higher patience makes intolerance towards the party less important because
the groups value more their future ow of payos.
Finally, gure 3 plots the set of feasible as a function of diversity . The lower
bound is constant in this case, because it does not depend on . Group 2s utility
does not depend on this parameter, so that the necessary condition for making
them comply is not a function of heterogeneity. Nonetheless, the upper bound is
increasing in . The intersection of both lines yields the minimum admissible level
of heterogeneity:
min
= 2 g
As rises, more heterogeneity is needed to compensate the increased dislike
towards the party. Higher gifts or higher patience, make group 3s sympathy less
necessary for the establishment of clientelism. In this section a simplied model
of political clientelism and the comparative statics just described serve to illustrate
how the feasibility of patronage varies as a function of ideological and economic
conditions.
17
4 Clientelism-Game With Random Shocks
Even though the model presented in the previous section provides some interesting
predictions pertaining to how clientelism becomes more or less feasible as relevant
parameters change, the fact that an innite number of mixed strategies can be imple-
mented does not solve some of the most interesting questions. As such, the following
section presents an extended version of the clientelism-game, incorporating random
shocks into the groups ideology payments. It was argued that groups exhibit a
natural aversion towards the party ( < 0) and that group 1 dislikes the candidate,
group 3 feels sympathy towards him, and group 2 remains neutral. This paper now
assumes that during each time period t in which the clientelistic candidate is elected,
the three groups are exposed to a random shock
i
, for i = 1, 2, 3, which increases or
decreases group is ideological payo at time t. More specically, groups ideological
payos now are given by
1
(p wins) = +
1
2
(p wins) = +
2
3
(p wins) = + +
3
where E[
i
] = 0 for i = 1, 2, 3. For simplicity, assume that
i
is uniformly distributed
over the interval [k, k], for k R
++
. Naturally, this implies that E[
i
] = 0 for each
group.
As before, assume that the game repeats during t = 1, 2, . . . periods and that
the candidate and groups 2 and 3 follow the grim-trigger strategies specied in the
previous section. Consequently, during each period the politician plays a mixed
strategy (0, 0, , 1 ) in which he assigns the gift g to group 2 with probability
(0, 1). In the last section necessary conditions for equilibrium where established
on . But how is really determined? What motivates the candidate to give the
gift with a higher probability to the close or to the distant group? Does he prefer to
more frequently solicit the group that ideologically is more distant to him, or is the
group that sympathizes with him awarded for this?
Given these conditions, it is natural to consider cases in which group 1 votes for
the candidate, even though they strongly dislike him and never receive a gift. Dene
p
i
, for i = 1, 2, 3, as the probability that group i votes for the candidate. Therefore,
if [0, 1] represents the probability that the candidate gets elected in a given
period, = p
1
[1 (1 p
2
)(1 p
3
)] + (1 p
1
)p
2
p
3
. Assume that in equilibrium the
18
three groups play according to the following rule: If
i
x
i
, for certain threshold
value x
i
R, group i votes for the candidate. Hence, x
i
simply represents a cutpoint
which determines if a group supports or not the clientelistic candidate. Therefore,
the probability of having group i voting for the candidate is
p
i
= 1 F(x
i
)
where F(x
i
) = Pr(
i
< x
i
) is the cumulative distribution function (cdf) of
i
. Con-
sider the case in which the candidate chooses in order to maximize his probability
of being elected in any given period. Then, the clientelistic politicians problem for
each period can be stated as
max
= p
1
[1 (1 p
2
)(1 p
3
)] + (1 p
1
)p
2
p
3
s.t. [0, 1]
p
i
= 1 F(x
i
)
Naturally, we have to carefully establish the cumulative distribution functions F(x
1
),
F(x
2
) and F(x
3
) in order to solve this game. We know that each period group 1
votes for the clientelistic candidate as long as
+
1
0
This equation implies that group 1s threshold value is
x
1
= + (4)
Similarly, group 2 supports the candidate if
+
2
+ g 0
which implies
x
2
= g (5)
Finally, the candidate receives support from group 3 if
+ +
3
+ (1 )g 0
Consequently, group 3s threshold value is
x
3
= (1 )g (6)
19
Now that we explicitly know the threshold values for each group, it is time to present
the main result of the model: the optimal value
=
1
2
+
2g
Proof:
We know that
i
is uniformly distributed in the interval [k, k]. Then, if f(x) is the
probability density function for any x [k, k] it is true that f(x) =
1
2k
. Therefore,
for group i
p
i
= 1 F(x
i
)
=
k
x
i
f(x)dx
=
k
x
i
1
2k
dx
=
1
2
x
i
2k
Substituting (4) into the last equation, we nd that
p
1
=
1
2
+
2k
(7)
Similarly, using (5) we nd that for player 2
p
2
=
1
2
+
+ g
2k
(8)
Finally, from (6) for player 3
p
3
=
1
2
+
+ + (1 )g
2k
(9)
The candidate wants to maximize the probability of being elected, which can be
written as
= p
1
[1 (1 p
2
)(1 p
3
)] + (1 p
1
)p
2
p
3
= p
1
(p
2
+ p
3
) + p
2
p
3
(1 2p
1
)
20
From (7) it is clear the p
1
is not a function of , so we can treat it as a constant.
Also, from (8) and (9)
p
2
+ p
3
= 1 +
2 + + g
2k
Consequently, p
2
+p
3
also does not depend on and can be considered as a constant
in this analysis. Therefore, in an interior solution of the candidates maximization
problem, the rst order condition is
=
(p
2
p
3
)
[1 2p
1
] = 0
From (7), it is clear that p
1
< 1/2. Hence, the rst order condition implies that
(p
2
p
3
)
= 0
Also, from (8) and (9) we know that
p
2
p
3
=
k + + g
2k
k + + + (1 )g
2k
k + + + (1 )g
2k
g
2k
k + + g
2k
= 0
and solving, we nd the optimal mixed strategy for the candidate:
=
1
2
+
2g
(10)
The second order condition is
g
2
2k
2
< 0
for any . Therefore,
is a maximum.
It is important to note that since
2g
> 0, it is always the case that
> 1/2.
This means that group 2, the neutral group (which at the same time is more distant
from the candidate than group 3), receives the gift with a higher probability than
the sympathetic group 3. The politician gives the gift more frequently to the
group that is ideologically less identied with him, probably because a bigger eort
is necessary in order to make this group comply. In other words, 1
< 1/2 is a
consequence of group 3s sympathy to the candidate. This does not mean that the
politician only allocates resources to one group. He still has incentives to randomize
in order to make clientelism self-enforcing. Some interesting features characterize
this randomization, as established by proposition 4.
21
Proposition 4:
In the SPE in which the candidate and groups 2 and 3 use the grim trigger strategies
established in the clientelism-game with random shocks, the optimal mixed strategy
implemented by the politician has the following properties:
1. Higher heterogeneity between groups implies higher odds of giving the gift to
the neutral group.
2. Higher gifts available to the candidate imply lower probability of giving them
to the neutral group.
3. In equilibrium, groups 2 and 3 vote for the clientelistic candidate with the same
probability
Proof:
Partial derivatives (comparative statics) of the optimal mixed strategy reveal that:
1.
is increasing in :
=
1
2g
> 0
for any pair g > 0.
2.
is decreasing in g:
g
=
2g
2
< 0
for any pair (g, ) R
++
R
++
.
3. Substituting (10) into (8) yields
p
2
=
2k + 2 + g +
4k
(11)
While (10) into (9) yields
p
3
=
2k + 2 + g +
4k
Therefore, in equilibrium, p
2
= p
3
.
Proposition 3 describes the unique mixed strategy that a clientelistic politician fol-
lows in equilibrium. Proposition 4 shows how this mixed strategy varies as a function
of the relevant parameters of the model. First, as diversity between groups rises, the
probability that group 2 receives the gift also increases. This should be intuitive.
22
More diversity, in the context of this model means that group 3 feels more sympathy
towards the candidate, while group 2 remains neutral. Consequently, the candidate
makes a higher eort trying to convince the more distant group (2). Second, as the
gift is higher,
I
=
16k
2
16k
2
(2k + 2 + g + )
2
if the candidate is the incumbent at the beginning of the game, or
C
=
(2k + 2 + g + )
2
16k
2
(2k + 2 + g + )
2
if he is the challenger.
Proof:
If group 1 never votes for the candidate, p
1
= 0 and = p
2
p
3
= p
2
2
. Therefore, from
(11) we nd out that
=
(2k + 2 + g + )
2
16k
2
(12)
23
Also, if the candidate is the incumbent at the beginning of the game, the expected
time in oce is given by
I
= 1(1 ) + 2(1 ) + 3
2
(1 ) +
= (1 )
i=1
i
i1
which simplies to
I
=
1
1
(13)
Substituting (12) into (13), yields
I
=
16k
2
16k
2
(2k + 2 + g + )
2
If the candidate is the challenger at the beginning of the game, his expected time in
oce is
C
= 0(1 ) + 1(1 ) + 2
2
(1 ) +
= (1 )
i=1
i
i
which is the same as
C
=
1
(14)
Using (12) and (14) we nd that
C
=
(2k + 2 + g + )
2
16k
2
(2k + 2 + g + )
2
Which completes the proof.
Finally, proposition 6 presents the comparative statics of both the candidates
likelihood of being elected in each period, and his expected time in oce.
Proposition 6:
Suppose that groups dislike towards the party is suciently small, so that 2 <
2k +g +. Then, the probability that the candidate gets elected in each period, the
expected time in oce when he is the incumbent at the beginning of the game, and
this expected time if he is the challenger, are increasing in group heterogeneity and
the size of the gift, and decreasing in dislike towards the party.
24
Proof:
From (12) we know that
=
2k + 2 + g +
8k
2
;
g
=
2k + 2 + g +
8k
2
;
=
2k + 2 + g +
4k
2
These three partial derivatives are all positive as long as 2 < 2k+g+. Also, from
the chain rule we know that
j
n
=
j
n
, for j = I, C and n = , g, . Additionally,
=
C
=
1
(1 )
2
> 0
Therefore, it is also the case that
j
,
j
g
, and
j
=
1
2
+
2g
+
k
1
(k
3
k
2
)
2g( + )
Proof:
Following the same procedure of proposition 3, we know that groups support the
clientelistic candidate with probabilities:
p
1
=
1
2
+
2k
1
; p
2
=
1
2
+
+ g
2k
2
; p
3
=
1
2
+
+ + (1 )g
2k
3
Once more, assuming that in order to win the candidate needs the support of at
least two groups, he is elected with probability
=
k
1
+
2k
1
1 +
+ g
2k
2
+
+ + (1 )g
2k
3
+
k
1
k
2
+ + g
2k
2
k
3
+ + + (1 )g
2k
3
+
k
1
g(k
3
k
2
+ + g 2g)
4k
2
k
3
= 0
and solving, we nd the optimal probability of rewarding group 2:
=
1
2
+
2g
+
k
1
(k
3
k
2
)
2g( + )
(15)
26
The second order condition is
+
k
1
2g
2
4k
2
k
3
< 0
Therefore,
k
1
=
k
3
k
2
2g( + )
Therefore
k
1
> 0 if k
3
> k
2
.
2. Relative density between neutrals and supporters:
(k
3
k
2
)
=
k
1
2g() +
> 0
3. Intergroup heterogeneity:
=
1
2g
k
1
(k
3
k
2
)
2g( + )
2
Hence,
> 0 if k
3
k
2
<
(+)
2
k
1
.
4. Candidates budget:
g
=
k
1
(k
3
k
2
)
2g
2
( + )
2g
2
Consequently,
g
> 0 if k
3
k
2
<
(+)
k
1
.
5. Party identication
()
=
k
1
(k
3
k
2
)
2( + )
2
Therefore,
()
> 0 if k
3
< k
2
.
The rst item of proposition 8 illustrates that having less swing voters in group 1
(higher k
1
) favors group 2 only if this group is comprised of more swing voters than
group 3. In other words, as group 1 becomes less swing, the candidate allocates
more resources to the group with the higher population of swing voters between
groups 2 and 3. Item 2 depicts that, as the number of swing voters in group 2 in-
creases relative to group 3 (higher k
3
k
2
), the probability of group 2 being rewarded
also increases. Thus, with all else being equal, possessing a greater population of
undecided or easy-to-buy voters results in a guarantee of more resources. The
third item of this proposition concludes that when there are suciently more swing
voters in group 3 compared to group 2 (k
3
k
2
suciently small), more intergroup
heterogeneity increases the amount given to group 2.
4
The rationale for this outcome
4
It makes sense to talk about amounts if we interpret as a fraction, instead of a probability.
28
is as follows: When group 3 has more swing voters, it is natural to infer that more
resources are given to that group. But if grows, group 3, as a whole, becomes more
loyal so more resources can be given to the neutral group.
The fourth item of proposition 8 suggests that if there are suciently more swing
voters in group 3 compared to group 2, more resources are given to group 2 as the
candidates budget increases. The rationale is similar to before. When group 3 has
many more swing voters than its counterpart, it is natural to infer that it receives
more resources. However, if the budget increases, those extra resources can then be
allocated to persuade members of group 2. Finally, item 5 reveals that as dislike
towards the clientelistic party increases (higher ), the probability of rewarding
the neutral group grows if there are more swing voters in group 3 and vice-versa.
Hence, the model predicts that as party identication grows, the group comprised
of more extreme (inexible) voters will benet, while the group comprised of more
swing voters will be punished.
5 Discussion: What is Clientelism?
The model presented in this paper describes a particular type of clientelism: the
transfer of material benets (cash) in exchange for political support. This particular
choice does not mean that any type of transfer between a candidate and the elec-
torate will equate to clientelism. For example, if an incumbent politician running
for reelection favors certain group of voters through a subsidy, it is not necessarily
representative of clientelism. Instead, such strategy could be another form of redis-
tribution as it could simply be the result of the ideological and programmatic agenda
of the candidate. Nonetheless, the way clientelism is modeled in this paper distin-
guishes it from other forms of redistribution for one basic reason: the allocation of
gifts is conditioned in accordance with previous voting behavior.
In the context of the model, if g were to represent a subsidy, for instance, it
would then not be necessary to employ a grim trigger strategy, involving allocation
by the candidate only if the groups formerly supported him. Under programmatic
redistribution, allocation of resources would be modeled based on the maximization
of an ideological utility function and its outcome. Time would only matter if the
structural economic conditions that determine such allocation change. It would not
be because citizens vote in a particular manner. Therefore, in order to appropriately
model clientelism, this paper includes the repeated interaction between the candidate
and the groups of voters even when this is not mathematically required: the one-shot
29
version of the probabilistic voting model presented in section 4 would be sucient.
But the repeated framework developed in section 3 is kept precisely to distinguish
between clientelism and other forms of redistribution.
Naturally, the transfer of material benets or cash based on past or present
voting behavior is not the only plausible form of clientelism. In many cases, politi-
cians discipline voters by using mechanisms that determine future ows of income
based on electoral results. Typical examples of this method include employment
in the public sector or public contracts. Robinson and Verdier (2003) discuss this
other form of clientelism, illustrating that, if incumbents and challengers compete
for votes through policies and transfers, then under certain circumstances the incum-
bent has incentives to reduce investment in order to attract supporters. A reduction
in investment makes the private sector less productive, which makes employment
in the public sector appear more attractive for certain groups. Hence, those voters
will support the incumbent, because his victory will promote public sector employ-
ment. Consequently, under a clientelistic regime, a lower level of investment results
in negative consequences to productivity and eciency, thereby increasing poverty
and inequality.
There is no doubt that employment in the public sector is an important compo-
nent of clientelism. But it is not the only one. In fact, most of the empirical litera-
ture on vote-buying and clientelism shows that cash and other immediate material
goods are widely used all across the globe [see Vicente (2010), Gonzalez-Ocantos
et al. (2012), Finan & Schechter (forthcoming), and Gallego & Wantchekon (2012)].
Robinson and Verdier (2003) do not have a satisfactory theory for this form of pa-
tronage. In fact, in their model the optimal amount of transfers that candidates give
to voters is zero. On the contrary, the model presented in this paper provides a the-
oretical description of why this form of clientelism takes place and is self-enforcing.
6 Conclusion
One of the major puzzles of political clientelism is explaining why agents comply with
these types of agreements, even when they have incentives to cheat and even provided
that in most democracies the vote is condential. In this paper, a simple mechanism
that explains why clientelism is self-enforcing is presented. The result of the election
provides politicians with a signal of voting behavior. However, more interestingly, the
election results provide a powerful mechanism for disciplining voters. Therefore, in
a repeated context in which votes are unobservable but results are, clientelism might
30
emerge as a result of equilibrium behavior. Citizens have incentives to honor their
agreements in order to maximize their future ow of payos, which in many cases
is higher when the clientelistic candidate wins the election and gives future benets.
In addition, the candidate has incentives to cultivate his clientele and provides them
with goods and benets (even at a personal expense), in order to promote voter
compliance and secure their vote. It is interesting to note that in this model the
promise of future gifts attenuates the ideological disutility that a citizen experiences
when a disliked party comes into power.
Some contexts are more favorable than others for patron-client relationships. In
the model presented here, the clientelistic politician is favored when dislike towards
his party is not so high, when diversity among neutral and sympathetic groups is
higher, when agents are more patient and place higher value on the future, as well
as when his budget allows him to oer better and more valuable goods or benets.
Nonetheless, its results are extremely important to understanding what determines a
politicians allocation of scarce resources. Under certain circumstances, swing voters
are more valued and, consequently, they receive higher benets more frequently. But,
the case also exists that other circumstances make core voters more important in
relative terms, so that allocations are more egalitarian in equilibrium.
Nevertheless, several puzzles remain unanswered. What is the relation between
clientelism and poverty? And with income inequality? In the model presented,
groups diered in their ideological preference towards the candidate. However, it
would be interesting to understand how a clientelistic politician allocates resources
among groups that dier in income. Are poor voters more frequently bribed? Is the
medium class the main target of politicians? Some studies suggest (Stokes, 2005)
that clientelism is more frequent in poor and unequal societies. This seems reasonable
given that for a politician it is cheaper to buy poor voters. In addition, as inequality
increases, the gap between the rich elites that support certain candidates and the
poor voters increases. As such, it would be certainly intriguing to incorporate these
dynamics into the framework described in this paper.
References
Chandra, K. (2007). Counting Heads: a Theory of Voter and Elite Behavior in
Patronage Democracies. in Kitschelt and Wilkinson eds. (2007).
Dixit, A., & Londregan, J. (1996). The Determinants of Success of Special Interests
in Redistributive Politics. Journal of Politics, 58.
31
Fearon, J. (1999). Why Ethnic Politics and Pork Tend to go Together. Working
Paper, Stanford University.
Finan, F., & Schechter, L. (forthcoming). Vote-Buying and Reciprocity. Economet-
rica.
Gallego, J., & Wantchekon, L. (2012). Experiments on Clientelism and Vote-
Buying. in New Advances in Experimental Research on Corruption, D. Serra and
L. Wantchekon eds., Research In Experimental Economics Volume 15, Bingly:
Emerald Group Publishing.
Golden, M., & Picci, L. (2008). Pork Barrel in Postwar Italy, 1953-1992. American
Journal of Political Science, 52(2).
Gonzalez-Ocantos, E., Kiewiet, C., Melendez, C., Osorio, J., & Nickerson, D. (2012).
Vote-Buying and Social Desirability Bias: Experimental Evidence from Nicaragua.
American Journal of Political Science, 56(1).
Gosnell, H. (1937). Machine Politics: Chicago Model . The University of Chicago
Press.
Greene, K. (2001). Against the Machine: Party Organization and Clientelist Politics
in Mexico. Working Paper, University of Texas.
Kitschelt, H., & Wilkinson, S. (2007). Patron, Clients, and Policies. Cambridge
University Press.
Kobayashi, M. (2006). Political Clientelism and Corruption: Neo-structuralism and
Republicanism. In: Kawata, J. (ed.), Comparing Political Corruption and Clien-
telism, Cambridge University Press.
Lemarchand, R. (1972). Political Clientelism and Ethnicity in Tropical Africa: Com-
peting Solidarities in Nation Building. American Political Science Review, 66(1).
Medina, L., & Stokes, S. (2007). Monopoly and Monitoring: an Approach to Political
Clientelism. in Kitschelt and Wilkinson eds. (2007).
Myerson, R. (1993). Incentives to Cultivate Favored Minorities under Alternative
Electoral Systems. American Political Science Review, 87(4).
Robinson, J., & Verdier, T. (2003). The Political Economy of Clientelism. Working
Paper, University of California.
32
Scott, J. (1972). Patron-Clients Politics and Political Change in Southeast Asia.
American Political Science Review, 66(1).
Stokes, S. (2005). Perverse Accountability: A Formal Model of Machine Politics
with Evidence from Argentina. American Political Science Review, 99(3).
Stokes, S. (2007). Political Clientelism. in: The Oxford Handbook of Comparative
Politics. Oxford University Press.
Vicente, P. (2010). Is Vote-Buying Eective? Evidence from a Field Experiment in
West Africa. Working Paper, Department of Economics, Trinity College Dublin.
Vicente, P., & Wantchekon, L. (2009). Clientelism and Vote Buying: Lessons from
Field Experiments in African Elections. Oxford Review of Economic Policy, 25(2).
Wantchekon, L. (2003). Clientelism and Voting Behavior: Evidence from a Field
Experiment in Benin. World Politics, 55(2).
Weitz-Shapiro, R. (2007). Political Competition, Poverty, and Incentives for Good
Government and Clientelism in Argentina. Document presented for the Midwest
Political Science Association.
33