Endogenous Political Institutions (A J W Philippe Aghion, Francesco Trebbi) 2004

ENDOGENOUS POLITICAL INSTITUTIONS*
Philippe Aghion
Alberto Alesina
Francesco Trebbi
A fundamental aspect of institutional design is how much society chooses to

delegate unchecked power to its leaders. If, once elected, a leader cannot be
restrained, society runs the risk of a tyranny of the majority, if not the tyranny of
a dictator. If a leader faces too many ex post checks and balances, legislative
action is too often blocked. As our critical constitutional choice, we focus upon the
size of the minority needed to block legislation, or conversely the size of the
(super)majority needed to govern. We analyze both "optimal" constitutional de
sign and "positive" aspects of this process. We derive several empirical implica
tions which we then discuss.
I. Introduction
Classical political theorists were well aware of the impor
tance of the trade-off between delegation of power to leaders and
the need to control them to avoid tyranny. For instance, in De
mocracy In America, Alexis de Tocqueville stressed that "Our
contemporaries are incessantly racked by two inimical passions;
they feel the need to be led and the wish to remain free."1 The
Founding Fathers of the American Constitution were also quite
aware of this dilemma. For instance, in Federalist Paper No. 70
Hamilton writes that, "Taking for granted . . . that all men of
sense will agree in the necessity of an energetic executive, it will
only remain to inquire what are the ingredients which constitute
this energy? How far can they be combined with those other
ingredients which constitute safety in the Republican sense?" The
* We thank Daron Acemoglu, Marios Angeletos, Olivier Blanchard, Matilde

Bombardini, Stefano Delia Vigna, Jeffrey Frieden, Bryan Graham, Oliver Hart,
Elhanan Helpman, Matthew Jackson, Miklos Koren, Vardges Levonyan, Roberto
Perotti, Torsten Persson, Andrei Shleifer, Adam Szeidl, Motohiro Yogo, and three
anonymous referees for useful comments and suggestions. Edward Glaeser was
exceptionally helpful at various stages of this project. Seminar participants at the
Center for Basic Research in the Social Sciences, Canadian Institute for Advanced
Research, Columbia, Copenhagen, and Harvard Universities, Massachusetts In
stitute of Technology, New York University, Universit?t Pompeu Fabra, Stock
holm University, University of Pennsylvania, Yale University, and the McArthur
Foundation group on inequality provided many helpful comments. Federico Etro
provided excellent research assistance. Alesina and Aghion gratefully acknowl
edge financial support from the National Science Foundation through the Na
tional Bureau of Economic Research and from the Canadian Institute for Ad
vanced Research, respectively.
1. Volume 2, part 4, Chapter 6, page 664 from the translation by Mansfield
and Winthrop [2000].
? 2004 by the President and Fellows of Harvard College and the Massachusetts Institute of
Technology.
The Quarterly Journal of Economics, May 2004
565
This content downloaded from
85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
All use subject to https://about.jstor.org/terms
566 QUARTERLY JOURNAL OF ECONOMICS
theory of checks and balances, embodied in the work by Montes

quieu [1748] provided the answer adopted by the framers of the
American Constitution. More specifically, the question of super
majorities as a way to restrain the "tyranny of the majority"
features prominently in the Constitutional theory by Hayek
[1960], Buchanan and Tullock [1962], and Buchanan [1975]. The
former, for instance, argues that the simple majority rule does not
have any particular "superior" standing and under certain condi
tions may lead to excessive costs imposed on individual liberties
by collective action. He argues that a "constitution of liberty" has
to be based on supermajority rules and judicial control. Our
emphasis on the endogeneity of institutions and how they may
evolve as certain characteristics of society follows the footsteps of
North [1981], the discussion of seventeenth century England by
North and Weingast [1989], and the discussion of the American
Constitution by Beard [1941], just to name a few.2
Given that we follow such giants, what is our contribution?
We make three points. First, we provide a simple model of this
trade-off between delegation of power and ex post control of
politicians, that generates a number of comparative statics re
sults in unified setup. We model delegation of power or, in one
word, "insulation" of leaders as the share of votes that can block
the leader ex post when he tries to implement legislation. A
Constitution that establishes a high share of votes needed to
"block" implies that leaders are more insulated. We show that the
optimal amount of insulation depends on politico-economic fea
tures such as the size of the aggregate improvement from reform,
the aggregate and idiosyncratic uncertainties over the outcome
from reform, the degree of polarization of society, the individual
degree of risk aversion, the availability and efficiency of fiscal
transfers, and the degree of protection of property rights against
expropriation. In doing so, we build upon a related framework of
incomplete institutional contracts by Aghion and Bolton [2003].
Second, we contrast normative versus positive implications
(as in Buchanan [1975]) and investigate the political economy of
institutional design. That is, we discuss how the optimal choice of
insulation would or would not be adopted in a system where the
choice was not made completely behind a veil of ignorance or only
a fraction of the population had a voice in the choice of institu
2. For a vast survey of the literature on Constitutional design, see Voigt

[1997].

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
ENDOGENOUS POLITICAL INSTITUTIONS 567
tions. One especially interesting case is a situation in which those

who choose a Constitution are also those who know who will
control political office after the Constitution is ratified. In this
case, what is optimal for them may not be optimal for society as
a whole. In particular, we focus on the following: in a fragmented
society, while it would be optimal to choose less insulation to
guarantee a "voice" to all groups, in practice an especially pow
erful group may take a hold of the constitutional process and
impose its rule. Results by Acemoglu and Robinson [2001] in the
context of institutional design in unequal society are somewhat
related and go in the same direction.3
Finally, we use our model as a basis for comparative cross
country analysis of constitution design. Our empirical analysis
focuses on the relationship between polarization (measured by
two indexes of socio-ethnic fractionalization) and various mea
sures of insulation, e.g., whether the regime is autocratic rather
than democratic, or whether it is more presidential, or whether
the voting system involves majority rules instead of proportional
rules. We uncover a positive and significant correlation between
polarization and insulation, in accordance with the "positive"
theoretical results above. Our findings, obtained on a much larger
sample of countries than Lijphart [1994] contrast with this au
thor's results. The reason is that our data set includes many more
countries including developing ones (not in Lijphart's data set); in
less developed democracies or semi-democracies, it is more likely
that one particular group (identified by income or ethnicity) can
impose undemocratic or semi-democratic rule. We also show how
our model can explain why political systems in less developed
economies tend to be more insulated than those with lower levels
of GDP per capita, why older constitutions insulate political lead
ers less than more recent constitutions, and why insulation tends
to increase in times of crisis or war.4
In moving from a "normative" model in which insulation is
chosen behind a veil of ignorance to one in which some groups
have more power than others at the constitutional table, we
3. These authors show how rich elites would impose dictatorship; they dis
cuss in much more detail than we do the issue of insurrection of the poor.
4. Our contribution should thus be seen as a complement to the recent vast
literature on the effects of institutions on economic outcomes, as in Persson and
Tabellini [2003] and the references therein. This literature takes institutions such
as electoral laws, level of democracy, presidentialism, etc. as exogenous, or at least
predetermined.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
connect with a recent literature on "choosing how to choose"

institutional rules, or voting on voting rules.5
Finally, our paper is complementary to a recent lively litera
ture that takes political institutions as exogenous (or predeter
mined) and studies their effects on various economic outcomes.
Rodrik [1999] and Persson and Tabellini [2003] among others
make a strong case for the relevance of various electoral laws and
other institutions on economic outcomes. On the opposite side is
work by Mulligan, Sala-i-Martin, and Gil [2004] who argue that
what really matters is not the nature of political institutions but
the strength of lobbies confronting each other. In fact, under
standing the endogeneity of political institutions may help re
solve the empirical disputes in this literature that puts institu
tions on the right-hand side of various regressions.
The paper is organized as follows. In Section II we describe
and solve the basic model. Section III illustrates several exten
sions. Section IV offers interpretation and discussion of the
model. Section V discusses the "political economy" of writing
constitutions. Section VI highlights several empirical implica
tions of our model and brings about empirical support. The last
section concludes.
II. Political Insulation

We begin with the discussion of optimal institutional design.
First, we present and solve the basic model, and in the next
section we discuss extensions.
ILA. The Basic Model

Consider an economy populated by a continuum of individu
als, assumed, for the moment, to be risk-neutral with respect to
income. Members of this polity will differ ex post on how much
they benefit from policy actions (labeled "reforms") which may be
implemented. If no reform is implemented, all individuals obtain
5. In particular, see Koray [2000], Barbera and Jackson [2001], and Polborn
and Messner [2002]. Barbera and Jackson investigate the endogenous choice of a
majority voting rule, investigating conditions of existence of a self-stable voting
rule, an issue that we shall revisit below. Koray instead explores social choice
functions and whether such functions are self-selecting. In an overlapping genera
tions setting, Polborn and Messner identify a trade-off arising in the selection of
voting mechanisms over a reform when only part of the population (the old) incurs
the cost of the reform, but not the subsequent benefits.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
the same income, which we normalize to 1. Individual income

from the policy reform is given by6
, v IX7 if reform occurs . ,
(1) * = {l otherwise, with 7 > 0
where
Xi=Xi + a,
with X? uniformly distributed over the support [X,X],7 with X < X
and
Xm=(X + X)/2
is the marginal valuation for reform for the average (or median)
individual in the population. We label I = X - X. With a we
indicate a random variable with mean zero, uniformly distributed
over the support [-A,A], with A > 0. The preference shock a has
to be interpreted as a change of the distribution of preferences
occurring after the leader has taken office and while he is imple
menting his policy. This is meant to capture the idea that as a
reform materializes through the effort of a leader new voters
come in or the population at large "matures" definitive prefer
ences about the reform, for instance, as they learn more precisely
who will be a winner or loser from the reform.
We employ the uniform distribution to obtain simple closed
form solutions, but in the Appendix we show how our results
extend to more general distributions. If \my > 1, the policy
reform is ex ante efficient in the sense that it makes the average
(and median) voter better off. The parameter 7 allows changing
the value of the reform without changing the distribution of the X
and the median voter in particular.
This community of individuals selects a leader to implement
reforms.8 With exogenously given probability p the selected
6. The following specification builds upon Krusell and Rios-Rull [1996] and
Aghion and Howitt [1998, Ch. 9] on the political economy of vested interests.
7. The constitutional decision is taken behind a veil of ignorance, before the
realization of the parameter \? for all ?'s and with all individuals facing the same
status-quo outcome in case reforms do not occur, an assumption we shall relax
below.
8. The role of the political leader in the basic model is highly stylized. A
leader is needed to promote the reform and to pass it (unless it is blocked).
However, the leader can take advantage of his position to expropriate. Obviously,
if the citizens could produce reforms without a leader, expropriation would not
occur. We rely on the realistic idea that a centralized entity is needed to coordinate
the reform policy.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
leader is "good" and promotes the reform; with probability ( 1 - p)

the leader is "bad" and only expropriates resources from the
citizens.9 For the moment we assume that all individuals are ex
ante identical in terms of their wealth, so the cost of expropriation
for each individual is the same, and we label it bw where b is the
exogenously given rate of expropriation and w represents indi
vidual wealth. Since we assume that w is for the moment iden
tical for everyone, we normalized it to 1.
A (super)majority M of individuals can block the action of the
leader (expropriation or reform) once the aggregate shock on
preferences a is realized. We define M as the degree of insulation:
if M is high, only a large majority of voters can block the reform.
On the contrary, a low M means that when in office the leader is
kept checked by small fractions of the electorate. Thus, a leader
passes a reform only if a fraction (1 - M) of the population favors
it or can expropriate only if ( 1 - M) of the population does not
object to this policy.10 Note that when M < V2 then supermajori
ties are needed to pass legislation. Also, in order to expropriate,
the leader has to "buy off" a fraction (1 - M) of the population in
order not to be blocked. Ex ante each individual in the polity faces
probability M of being subject to the expropriation, if the latter is
not blocked.
In our terminology an insulated leader can be less easily
blocked, so the probability that a good reform passes is higher,
but individuals are also more likely to suffer losses from expro
priation. The choice of M occurs ex ante, before the realization of
a, and we make the incomplete contracting assumption that the
size of M cannot be made contingent upon the realization of a.
Thus, we assume that the corresponding events cannot be de
scribed ex ante, and we rule out social contracts contingent upon
9. The Constitution could prohibit expropriation, but not reforms that would
be a Pareto improvement. We have two comments. First, in reality it is difficult to
fully restrain the authority of the government in this respect to expropriation
without restricting its ability to govern in other areas. In the Federalist paper No.
73 Hamilton, for instance, elaborates on the fact that "the power of preventing bad
laws includes that of preventing good ones." Second, with risk aversion, even
without expropriation, we still have a well-defined trade-off between insulation
and ex post control.
10. The parameter M we view as a "summary statistic" for a wide variety of
institutional rules that limit the power of appointed leaders. The most direct
interpretation of M is the majority that an executive has to command in order to
pass legislation. In general terms the issue of the "optimal supermajority" rule is
a widely debated question by Constitutional theorists. Those who favor superma
jority rules (low insulation) worry about limiting the power of appointed leaders
and about the tyranny of the majority. Those who oppose them view the (simple)
majority rule as the essence of democracy.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
messages that voters would exchange ex post about the realiza

tion of these random variables. For the moment we assume that
the X? are not observed by the politician and that the politician
cannot compensate the losers, i.e., those who ex post do not want
the policy reform.
The timing of "events" can be summarized as follows.
i) M is chosen at the constitutional stage, by individuals
behind a veil of ignorance, that is before the realization of
the X? in the interval [X,X];
ii) X? is realized;
iii) the politician proposes the reform or the expropriation;
iv) the uncertainty on the distribution of ex post preferences
is realized;
v) blocking of the reform may occur, the reform is imple
mented if and only if it is not blocked by the voters, if the
leader is "bad" he expropriates, up to the point that
avoids blocking.
Stage i) represents the "constitutional level" in which deci
sions are made behind a veil of ignorance. Stage iii) is rather
trivial. The only role of the politician is to promote a reform, that
passes if not blocked, or to expropriate the citizens. Steps iv) and
v) capture the postelectoral "dynamics" between leader and vot
ers. The latter implies that after the realization of the shock a the
voters still retain a choice to block ex post undesirable reforms. If
the threshold for blocking M is set low, then the voters insure
themselves that they will have a "voice" ex post. However, this
makes reforms easily blocked. On the other hand, if M is high,
reforms pass more easily, but a larger fraction of the population
may be expropriated; thus, ex ante, the probability that each
person is taxed is higher.
ILB. Interpretation
The real world example closest to the letter of the model
would be a popular referendum on policy, an institution that is,
however, seldom used. In this case the most extreme form of
noninsulation would be a referendum that requires a majority of
100 percent to pass legislation, so that any individual voter can
block policy. This institutional arrangement would set expropria
tion to zero, but would make it impossible to pass any legislation
which is not a Pareto improvement ex post. Given that referen
dum is rarely used, in the majority of institutional settings block
ing takes place indirectly, within the institutional structure of

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
delegation. In the case of Presidential regimes like the United

States, one can view the Presidential-Congressional relationship
as a key element of the system of checks and balances.11 In
parliamentary democracies the question of insulation refers to
the control over the power of the Prime minister and the rela
tionship between majority and minority in parliament. For given
size of the parliamentary majority the power of the executive, the
agenda setter, is also determined by the voting rules within the
parliament, an issue that has received much discussion in the
literature.12 Various voting rules governing procedures within
legislatures can be interpreted as giving more or less insulation to
the executive, i.e., in most cases the "leader" who holds a major
ity.13 For example, an important distinction is one between "open
rules" and "closed rules" in parliamentary voting. With open rules
the legislature has a vast latitude in amending policy proposals of
the agenda setter (the government); with closed rules the govern
ment can prevent amendments to its proposals, and as a result, it
has a larger strategic power. One may a priori associate open
rules with low insulation (low M) and closed rules with high
insulation (high M), since they imply different degrees of strate
gic power for the executive. Similar arguments apply to "fast
track" legislation in trade. This procedure is viewed in the United
States as critical for the implementation of free trade agree
ments, which otherwise might be blocked by various special
interests.14
With regard to the role of legislative institutions, a broad
interpretation of M could include a comparison of different elec
toral rules. Proportional rules tend to produce political systems in
which "governing by coalition" is the norm. In majoritarian sys
tems, the majority party can govern with fewer constraints.15
Even more broadly, one could also use M to compare dictatorship
or oligarchy versus fuller democracy. In a sense, one can think of
a dictatorship as a system in which a ruler, when in office (no
11. This is a point already made by Hayek [I960]. See Alesina and Rosenthal
[1995] for an extensive formal discussion of this issue.
12. See, for instance, Baron and Ferejohn [1989] and Baron [1991]. On
bicameralism see Diermeier and Myerson [1995].
13. In some cases we can have minority governments, in which the executive
does not command a simple majority in the legislature. See Persson and Tabellini
[2000].
14. See Grossman and Helpman [2001].
15. Persson and Tabellini [2003] and Milesi-Ferretti, Perotti, and Rostagno
[2002] present recent studies which compare proportional versus majoritarian
systems concerning fiscal policy choices.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
matter how he gets there), is uncontrolled, while an essential

element of democracy is some sort of checks and balances on the
politicians, above and beyond the fact that the latter are elected.
U.C. Solution of the Model

We proceed by backward induction. In stage v) the voters
with low X? oppose the reform; those with high X? favor it. A cutoff
point divides these voters:
X = I/7.
The realization of a, for given M, determines whether or not the
reform passes or not. The reform will pass if and only if
(2) (X - (X + a))/l < M
or a > X - X - IM. Therefore, ex ante the expected utility of the
generic voter, who is behind a veil of ignorance, is given by
(3)
1
max
M
p 2A"
^-rda+\ (\m
v m "'r 2A + a)y wj d
Ji-\-lM
The first two terms in (3) represent the expected benefits of the
socially efficient reform (multiplied by the exogenously given
probability of such an event, p), the third term represents expro
priation. Note that ex ante, behind a veil of ignorance and with
risk neutrality, the generic voter acts as the "average" individual.
Looking first for an interior solution to maximizing (3) relative to
M, and remembering that X = I/7, we obtain, after straightfor
ward maximization,
1 2Ab(l-p)
(4) M* =-^-?
whenever the right-hand sid
side of (4) is negative, then t
M* = 0.
Proposition 1. The preferences of voters are single peaked on M*,
and either the optimal degree of insulation is zero or it is
interior to the interval (0,1) and given by (4) if positive. In the
latter case, the optimal degree of insulation is decreasing
with the potential loss from expropriation (6), rising with the

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
probability that the leader is a good (p), rising with the value
of reform (7), rising with the dispersion of idiosyncratic pref
erences over the reform (I), and decreasing with the extent of
aggregate uncertainty over the reform (A).
First, note that in the absence of expropriation (b = 0), or
with no bad leaders (p = 1), we have M* = V2. This result
follows from the utilitarian nature of the maximization problem
in (3) and symmetry in the distribution of X?, but it extends to
more general distributions of idiosyncratic and aggregate shocks
on preferences, as we show in the Appendix. The basic intuition is
that the representative voter ex ante does not want to prevent an
ex post majority to stop an ex post efficient policy.16 This result
also holds only with risk neutrality; as we show below under risk
aversion even with no expropriation we obtain a well-defined
interior solution for M*.
That insulation should decrease in the probability (1 ? p ) of
expropriation and in the loss b from it, that is in the expected loss
from a bad reform, but that it should increase in 7, the average
benefit from a good reform, is self-explanatory.
Finally, to understand why insulation is increasing in I (re
spectively, decreasing in A), first note that in the absence of
expropriation, optimal insulation should not depend on I or A.
This follows from the fact that increases in I and A do not affect
the marginal effect of M on the outcome from good politician.
However, an increase in I (respectively, in A) increases (respec
tively, reduces) the average outcome under a good politician by
increasing (respectively, reducing) the likelihood of reform. Since
an increase in I (respectively, in A) does not affect the average
(expropriation) outcome under a bad politician, the higher / (re
spectively, A) the higher (respectively, lower) the benefit of insu
lation relative to its expropriation cost, and therefore the higher
(respectively, lower) the optimal level of insulation.
These parameters can all be connected to the real world. In
particular, a high 6 or a low p captures economic environments
with poor legal protection of individuals' property rights and
wealth. A high 7 captures the case of economies with a high
aggregate value of reform, e.g., emerging market economies that
require stabilization or transition economies that require struc
tural market reforms. Also a country that precipitates in a "crisis"
16. See the Appendix and Rae [1969].

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
may have a very high value for a reform that eliminates the
source of distress. A high A corresponds typically to reforms such
as international treaties or constitutional change, that involve a
high degree of aggregate risk and irreversibility. Proposition 1
thus suggests that insulation should be higher in economies with
better legal protection, or facing an emergency or transition situa
tion, or for decisions that involve a limited amount of aggregate
risk or are easily reversible. In the next section we complete this
comparative statics by introducing polarization, risk aversion,
and compensation costs as additional determinants of insulation.
III. Extensions
HLA. Risk Aversion and Polarization of Preferences

Let us now add a very simple form of risk aversion, with ex
post individual utilities being given by
y i if y i ^ ?,
(4) u(yt) = y i ? u otherwise,
where y ? is ex post income and where u > 0. Thus, only if income

is above a threshold 0, do we have the same utility as in the basic
model; otherwise, the individual incurs a loss. We assume that
the status quo outcome is always above such a threshold (i.e., 6 <
1) linking more tidily risk aversion and reform. Moreover, for
simplicity we take the probability of a bad reform to be zero; i.e.,
p = 1. This also shows an important result: with risk aversion
one obtains a well-defined interior solution for M* even without
expropriation; that is, even when property rights are fully
protected.
In this case the optimal choice of M becomes
(5) *,*
M* ^/(X- - k) + u{1 " e)/^ " 2A6(1 " pVp
=-/ i m
(7/ + u)l
This implies the following.
Proposition 2. The optimal degree of insulation (M*) declines
with the risk aversion parameter (u).
Proof. See Appendix.
Thus, more risk aversion leads to lower insulation: in choos
ing insulation, ex ante the voter takes into account the risk of

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
falling below 0 ex post. Thus, more risk aversion leads to choosing

a system where ex post policy reforms can be more easily
blocked.17
Let us now examine the effect of polarization of preferences.
For simplicity, we introduce a very stylized form of nonlinearity
in the preference distribution, namely, we assume that a point
mass A(<1) is now added to the lower extreme of the distribu
tion's support [X,X]. The mean point of the distribution becomes
AX + (1 ? A)Xm. With risk aversion and polarization we obtain
the following key result.
Proposition 3. For a sufficiently large degree of risk aversion as

measured by u, more polarization (A) reduces the optimal
degree of insulation, whereas for low risk aversion more
polarization increases the optimal degree of insulation.
Proof. See Appendix.
The intuition for this result is straightforward: on the one

hand, more polarization increases the risk of ending up at the
bottom of the preference distribution, which in turn leads to a
utility loss when the reform is implemented; reducing insulation
will limit that risk; on the other hand, more polarization in
creases the probability of a reform being blocked by the lower tail
of the preference distribution and increasing insulation helps
neutralize the lower tail. This result rationalizes the fact that in
more polarized societies with highly risk-averse individuals, the
Constitution "should" imply less insulation, to avoid the risk of
being, ex post, unsatisfied with the insulated ruler. We stress the
word "should" because this result follows from a Constitutional
decision process in which everybody is behind a veil of ignorance.
As we will discuss below, this result may not apply in a situation
in which not everybody is behind a veil of ignorance at the
Constitution table.
17. Note that the critical assumption that drives this result is that there is
"risk" about the outcome of the reform for an individual voter, but no risk about
the status quo. This is a realistic assumption to the extent that one knows its own
status quo, but not the outcome of a possibly complex sequence of policy changes.
Yet one might think about the reverse situation. Think, for instance, about the
introduction of unemployment insurance, in which case the reform is meant to
remove uncertainty of outcomes. In this case the effect of risk aversion would be
reversed.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
III.B. Compensation
In general, those who are net losers from a policy reform can
be compensated by transfers, even though the latter will gener
ally induce welfare costs, such as the costs of distortionary taxa
tion. Suppose that after M is chosen, a fixed amount of resources
?a can be raised from all individuals through taxes, and assume
for simplicity that taxes are raised before the idiosyncratic and
aggregate shocks on preferences are realized. Note that this im
plicitly assumes that any increase in income obtained through the
reform cannot be used to compensate losers, since the amount
available for compensation is fixed ex ante. The maximum
amount of co is 1 which represents initial individual wealth. The
only purpose of taxation is to compensate losers for the reform,
which means that if the available resources for compensation
exceed the needs the resources in excess are returned lump sum
at no cost and the leader does not retain any revenue for his own
consumption.
Such a transfer scheme is assumed to involve a positive
deadweight cost k per unit of taxed funds, and the net tax reve
nues are used by the politician to compensate the required num
ber of losers in order to avoid blocking. Ex post, for given realiza
tion of the aggregate shock a, either more than (1 - M) individ
uals are willing to support the reform even without any
compensation (this will be the case whenever (X - a ? X)/(X -
X) < M), in which case no compensation will take place; or
passing the reform requires compensation to be made (this will be
the case when (X - a - X)/(X - X) > M: then the politician needs
to compensate the fraction (X ? a - X)/(X - X) - M of individ
uals for potential loss of utility due to the reform). Compensation
is paid to enough individuals who would, ex post, vote against the
policy reform, in order to keep them in. Obviously, the "cheaper"
individuals are compensated, i.e., those closer to the cut-off point
of indifference between having or not having the reform.
However, this requires that individual preferences be ex post
observable, since compensation is made dependent on those
preferences.18
Thus, the total amount of compensation needed to pass a
reform is given by
18. The case where individual preferences are not publicly observable can be
analyzed along the same lines as in Aghion and Bolton [2003].

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
c(a) = *7(X
X+IM
- a - X?) j d\t.
Two cases must be considered. The first case is when there

are enough funds in the compensation scheme so that the reform
will always take place no matter what the realization of a is. In
this case reform will always take place (as it will always be
affordable) and, at the constitutional stage behind a veil of igno
rance, the generic individual will choose M* in order to maximize
max {p(Xm7 - kEac(a)) - (1 - p)bM}9

where Eac(a) is the expected compensation cost. The second case
is when not enough resources can be raised to fully compensate
losers. In the Appendix we show that in either case,
Proposition 4. When the compensation scheme is available and k

is not too large, the optimal degree of insulation M * increases
with the taxation cost k; otherwise it satisfies the same
comparative statics properties as in Proposition 1.
The basic message of this proposition is that a less efficient

system of transfers should lead to higher insulation (higher M) in
order to reduce the need for compensation ex post.
IV. The Political Economy of Constitutions

Thus far, we have examined the case of a "perfect veil of
ignorance," behind which everybody is identical. This, in a sense,
is equivalent to a normative model of constitutional writing. In
reality, Constitutions are not written by social planners, and veils
of ignorance have large holes in them. In fact, in virtually every
instance of Constitutional reform, a large amount of bargaining
and conflict occurs at the Constitutional table. One simple way of
capturing the complexity of the political economy of writing Con
stitutions is to generalize our model by assuming that not every
body derives the same (known) utility from the reform. The ana
lytical structure that we have developed above allows us to ex
tend the analysis in this direction fairly easily. From the point of
view of empirical applications, a discussion of the political econ
omy of Constitutional design, that is, an analysis of deviations
from optimality criteria, is critical.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
IV.A. Wealth Distribution, Voting Rights, and Constitutions

For example, assume that individuals differ ex ante with
regard to their taxable wealth, namely a "^?-individual" expects
to be expropriated by an amount equal to (1 - p)bw?M on
average. We have already seen in Proposition 1 that the optimal
degree of insulation is negatively correlated with the scope for
expropriation. Then it immediately follows that an individual i
with higher w? at the Constitutional stage will optimally choose a
lower level of insulation, with
(6) M*(wl) = l~2Abwi:^j^,

and that individual ?'s preference for insulation is single peaked
around this maximum. This in turn has interesting implications
for the political economy of Constitutional writing. Consider a
symmetric distribution of expropriation losses w? between w and
w and suppose that the Constitution is decided by majority rule.
In this case, the median voter, wm will prevail and impose her
most preferred level of insulation; namely,
(7) M*(wJ = \-2Abwm1^.

Alternatively, if M had to be chosen by unanimity, any M >
M*(wm) would be vetoed by wealthy individuals. Allowing bar
gaining at the Constitutional table will make Constitution design
ultimately depend on bargaining rules and the distribution of
wealth.
Another example of interest is the writing of Constitutions by
a minority of wealthy individuals who might worry about the
possibility that new redistributive policies might be introduced by
future majorities as the extension of voting rights progresses over
time. In his economic interpretation of the American Constitu
tion, Beard [1941], and many followers after him, argue that the
critical preoccupation of the Founding Fathers was exactly this
one. An important aspect of insulation concerns the protection of
property rights against expropriation. James Madison, in particu
lar (see Federalist Paper No. 51), advocated separation of power
as a way of preventing an "overbearing majority" to become
tyrannical vis-?-vis the minority. He clearly identifies the minor
ity with that of wealthy men threatened by the majority of the
poor. This implies that older Constitutions, chosen when voting

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
rights were restricted to a fraction of wealthy men, should pre

scribe larger majorities (i.e., be less insulated) when it comes to
taxation and protection of property against redistributive re
forms. To put it differently, older Constitutions would make it
easier to block legislation that threaten property, a point empha
sized by Alesina and Glaeser [2004] as an explanation of the
difference between the American and European Constitutions
and the associated differences in welfare states. Acemoglu and
Robinson [2001] also investigate the redistributive motive as a
pivotal element of constitutional design and transition between
democracy and autocracy.
In the language of our model this can be interpreted as
follows. Suppose that individual wealth is heterogeneous across
individuals and distributed between w and w with density f(w).
Suppose also that the expropriation rate is the same b for all
individuals, and that those who decide on the Constitution lie in
the upper part of the wealth distribution, say between some wh
and w, with wh > w. Assuming that all individuals with wealth
w E [w,w] vote ex post, clearly the Constitution designers will
choose a lower degree of insulation than if the franchise was not
to be extended. In other words, Constitutions written with an eye
on defending property rights against future redistribution of ex
propriation will include a number of checks and balances and
require supermajorities, i.e., low insulation to pass legislation.
These checks and balances will be targeted especially toward
making it easy to block legislation against redistribution and
taxation of wealth.19
TV.B. Constitutional Choice by the Upper Tail

Consider a situation in which a minority who chooses the
Constitution also knows that it will always be able to appoint
leaders. For example, suppose that a minority of high X individ
uals, with X G [X ? 8,X], and X ? 8 > X, choose the constitution.
First, the minority will choose a higher M than the rule that
emerges under the veil of ignorance; that is, it will chose to have
more insulated leaders. In the most extreme example where the
minority is a singleton, one absolute dictator will "choose" M = 1,
which in turn will allow him to pass any legislation he will like ex
post. Second, the lower 8, the more likely it is that the (upper tail)
19. Acemoglu and Robinson [2000] and Lizzeri and P?rsico [2004] investigate
what leads wealthy minorities to extend the franchise.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
minority who chooses the constitution, will react to increased

polarization of preferences by increasing insulation, as this mi
nority is less subject to the risk of ex post utility losses than the
average individual in the population; thus, from the point of view
of constitution writers in the minority the main concern is that
polarization increases the probability of a reform being blocked by
the lower tail of the preference distribution, and as we stressed in
Proposition 3 above increasing insulation helps reduce this block
ing power by the lower tail.
More formally, we can show the following.
Proposition 5. Suppose that the insulation rule M is decided at
the constitutional stage by a minority of individuals who
know that their idiosyncratic preference for reform is distrib
uted on the interval [X ? e,X]. Suppose in addition that a
mass (1 ? A) of individuals are uniformly distributed over the
interval [X,X], with a discrete mass A of them being located at
the lower extreme ofthat preference interval. Then, we have
(1) the lower e, the higher the optimal insulation rule Me
chosen by this minority; (2) for 8 sufficiently small, Me in
creases with polarization A, no matter the degree of risk
aversion as measured by u.
Proof. See the Appendix.
The implication of this proposition is clear: when a minority
chooses the Constitution and knows that it will rule under this
constitution, it will choose more insulation, and in this case more
polarization leads to more insulation.
TV. C. Choosing How to Choose

If voters are not identical ex ante at the constitutional table,
then the question is which rules "should" and will be used to
choose a Constitution. That is, we have both a normative and a
positive question of voting rules at the Constitutional table.
One might think of some sort of "fixed point" argument in
voting rules; that is, one may want to argue that a choice of M*
has to be approved itself with a blocking rule M*. That is, a
Constitutional choice of M* can be vetoed only by a M* (super)
majority. This is exactly the approach taken in models by Barbera
and Jackson [2001] and Polborn and Messner [2002]. While this
self-stable solution is very elegant, its realism may be called into
question. In fact, voting rules and procedures to select or change

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
the Constitution are generally different from the rules regulating

the passage of "normal" legislation. In general, the blocking coa
litions needed to prevent changes in the Constitution are lower
than those required to block "normal" legislation. In fact, our
model, and more specifically subsection IV.B. on risk aversion,
suggest one possible reason why changing the Constitution would
require smaller blocking coalition (larger majorities): Constitu
tional change may bring about a more uncertain distribution of
winners and losers, and voters may be especially risk averse
concerning radical changes of the rules of the game. Another
commonly discussed reason is the need to prevent an elected
leader from "easily" changing the rules of the game restricting
democratic rules in favor of the leader himself.
V. Empirical Implications and Discussion
In what follows, we discuss several empirical implications of our

theoretical analysis. Rather than formally "testing" our model, we
highlight several of its implications that seem to shed light on some
aspects of institutional choices and economic development. In the
previous sections we have discussed both normative and positive
aspects of constitutional design. As it is often the case in policy
analysis, one can take the normative case as a benchmark to analyze
and interpret the actual evidence on Constitutional design.
VA. Economic Development and Institutions

A well-known feature of developing countries is that they do
not have well-functioning fiscal systems. The share of transfers of
GDP is larger in OECD countries than in developing countries,
and more generally, the role of government in transferring re
sources across individuals, the welfare state, is far more wide
spread in richer countries. In the 1990s the average level of
subsidies and other current transfers as a percentage of current
expenditure in the high-income countries sample [World Devel
opment Report 2000-2001, World Bank] was about 60 percent. In
lower middle income countries it was 18 percent in 1990 and 26
percent in 1997. Part of the reason is that it is easier to collect
taxes in more advanced industrial countries and also targeting
compensation toward the truly deserving is particularly difficult.
These considerations suggest that developing countries should
adopt more insulated systems of government, since, in the lan
guage of our model they have a higher k.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
On the other hand, property rights tend to be less well

protected in developing countries, and insulated leaders may
have more latitude to pursue policies which favor the leader
himself and its close allies. The potential for "expropriation,"
broadly defined, is larger in developing countries. This is cap
tured by a higher b in our model. One may argue that the
technology for expropriation and that of taxing for compensation
go hand in hand, to the extent that they both involve collecting
fiscal revenue. However, a compensation scheme involves a fairly
sophisticated system of targeting, while expropriation, especially
in its more brutal form, can be rather easy to accomplish to the
extent that the government has the monopoly of coercion, a mo
nopoly that will itself increase with more insulation.
These considerations put developing countries between an
institutional rock and a hard place. High insulation leads to high
expropriation. Low insulation means that policy reforms are not
implemented. In fact, we believe that this "steep" trade-off be
tween the possibility of implementing reforms with winners and
losers and the likelihood that insulated leaders turn into dicta
tors, may be one of the key reasons for institutional failures in
developing countries.
A related point concerns the well-known fact that richer coun
tries tend to be more democratic. Note that richer countries have
better functioning fiscal systems, which allows for transfers and
social insurance, features that can be interpreted as a lower k in our
model, that is, lower costs of compensation. This reduces the need for
insulation and allows for better protection against socially ineffi
cient reforms. A more stringent implication is that countries with
lower costs of taxation or better functioning welfare states should
have lower insulation. Interestingly Milesi-Ferretti, Perotti, and
Rostagno [2002] and Persson and Tabellini [2003] find that more
proportional electoral systems are associated with a larger share of
transfer payments. Proportional electoral systems (as opposed to
majoritarian) can be thought of as systems with low insulation since
they often require large coalition governments to govern. These
authors interpret causality from the electoral systems to the welfare
state; in this paper we suggest that the alternative direction of
causation may also be present.
V.B. Polarization, Fragmentation, and Institutional Choice

An especially relevant question is what institutions are more
appropriate for more or less polarized or fragmented societies, a

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
preoccupation that goes back at least to the Founding Fathers. From

a "normative" point of view the above Proposition 3 suggests that
more polarization, that is, a higher scope for excluding a subset of
the population from the benefits of reform, should lead to the adop
tion of a lower level of insulation if constitution writers are suffi
ciently risk averse and choose under the veil of ignorance, as lower
insulation provides more checks and balances and avoids permitting
any group to impose its insulated authority on all the others. In his
analysis of advanced democracies Lijphart [1994] argues that this is
indeed the case. However, we have shown in Proposition 5 that
when constitution writers belong to a privileged minority far from
the 'Veil of ignorance" ideal of normative models, the opposite im
plication should hold: in more polarized, fragmented societies one
should observe more rather than less insulation.20
We now test the relationship between polarization and insula
tion in a vast sample of countries. Measurement issues are complex.
They affect both our independent and dependent variables. It is
almost impossible to construct an undisputable measure of "insula
tion" for a sample of more than a hundred countries in which
institutional arrangements vary on many dimensions. In our model
insulation is an ex post control of the voters over policy-makers. In
practice, this type of control, as we discussed above, can take many
different forms. For this reason, we shall consider different empiri
cal proxies for insulation.
To measure polarization, we consider several indices of
ethno-linguistic fractionalization, which we take as reasonably
exogenous, and have been widely used in the empirical litera
ture.21 Ideally, we would like to have a direct measure of polar
ization of the distribution of gains or losses from reform across
voters. Constructing such a measure would require that we iden
tify the size of the groups that are ex ante more likely to be
damaged by the reform process. However, our cross-country data
only provide information on the relative size of ethnic groups.
Thus, to link our theory to our empirical analysis, we need to
make the assumption that the extent of ethnic divisions is posi
tively correlated with the likelihood of a polarized distribution of
gains from the reform process.
The formula for the fractionalization index in country i is
20. In Aghion, Alesina, and Trebbi [2004] we explore these issues in the
context of the choice of forms of government of American cities.
21. See, for instance, Easterly and Levine [1997], La Porta et al. [1998],
Alesina et al. [2004], and Montalvo and Reynal-Querol [2002].

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
FRACTt = i - 2 (^) ,
where nÎNi is the relative size of group j in country ?, with j =
1,..., J. As for the data used, the first index is the one used
originally by Easterly and Levine [1997] ,22 It is an ethno-linguistic
fragmentation index based on a historical Russian classification of
languages in 1960. The second and third fractionalization indices
have been constructed by Alesina et al. [2004] by disentangling
linguistic measures from other ethnic variables, like racial origin.
We will employ a measure for ethnic fragmentation and another
measure for linguistic fragmentation.
We also check our results by using the polarization index pro
posed in Esteban and Ray [1994] and applied also by Montalvo and
Reynal-Querol [2002]. This index is computed as follows:
w&y
j j
POL, = KjJJJ
where K and a are consta

ethnic groups d = 1 if j'^ k
estimate the main paramet
show that a E (0, 8/5]. We
also indicated by the autho
to 1). Intuitively, the diff
fractionalization and polari
creases monotonically with
distance from a bimodal dis
ing to the polarization inde
polarization) is reached wh
other. In the fractionalization
reached when many small
group face each other.23
In Table I we describe al
tables including their sour
scriptive sample statistics.
22. A second "traditional" index (labeled AVELF) is an index proposed by

Easterly and Levine [1997] which averages over five related linguistic indices. All
the following results are robust to the use of this alternative index, and we do not
report them.
23. See Montalvo and Reynal-Querol [2002], for a detailed discussion.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
simple dichotomy: democracy versus autocracy. A dictator is the

most insulated leader of all. The second measure refers to demo
cratic forms of government, and we assign to Presidential sys
tems the role of "most insulated" form of government, Semi
Presidential (or Hybrid) the middle level?not reported?and
Parliamentary systems the least insulated. This choice is justified
by the substantial reduction in the number of veto players within
Presidential systems and their intrinsic winner-take-all nature,
which distinguishes them from Parliamentary systems.24 The
unconditional correlation between the form of government from
Presidential to Semi-presidential to Parliamentary and the Free
dom House index of political freedom is -.51, significant at the 1
percent level, for the sample of countries used below. That is,
more Presidential regimes are associated with lower political
rights, which we interpret as a proxy for more insulation.25 No
tably, one can identify an analogue winner-take-all nature in
electoral rules. For example, first-past-the-post formulae in Plu
rality systems may act as instruments of insulation of the elected
politicians from the minority of the electorate. This is the third
proxy we present. Table II shows that for all cases more fraction
alized regimes are more insulated. For instance, more fractional
ized systems are less democratic, more presidential, and tend to
have more Plurality rules.
In Table III we present Ordered Probit regressions where the
dependent variable is the Freedom House autocracy index in
1990. This table shows that the correlations indicated in Table II
survive after controlling for several other institutional and eco
nomic variables. Without controlling for real GDP per capita, the
fractionalization variables are statistically significant at the 1
percent level in explaining the probability of ending up in a more
autocratic regime. GDP per capita may be endogenous; however,
in three out of four cases the fractionalization variable remains
significant at standard levels, even after including together with
GDP a large set of controls (not reported, but available from the
24. For example, there is a relevant, negative (-0.501), and significant (at
the 1 percent level) correlation between presidentialism and a measure of political
constraints within the political arena (as presented in Henisz [2002]). For more
discussion on insulation and presidentialism, see Shugart and Carey [1992].
25. Further analysis shows that this correlation holds up even when control
ling for a battery of other variables, including log of per capita GDP, school
enrollment ratios, regional dummies, and openness.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE I
Variables Definitions:
Ethnic Fractionalization: Ethnic Fractionalization Index from Alesina et al.

[2004]; construction is described in the text.
Ethno-Linguistic Fractionalization Index in 1960: ELF from Easterly and
Levine [1997]. The original source is the 1964 Atlas Narodov Mira for the
year 1960.
Polarization: Measure constructed applying Esteban and Ray [1994] to the
Alesina et al. [2004] ethnic measure, a = Vs .
Linguistic Fractionalization: Linguistic Fractionalization Index from Alesina et
al. [2004].
Autocracy: Ordinal variable based on Freedom House [2002] Index in 1990.
Free = 1, Partly Free = 2, Not Free = 3. The ranking increases in the degree
of insulation of the Executive.
Separation of Powers: Ordinal variable from the Database of Political
Institutions [2001], available from DATAVINE/Harvard CID and the World
Bank [Beck et al. 2000] for year 1990. We consider the variable SYSTEM,
complemented to 2 to facilitate exposition. Direct Presidential is assigned 2,
Hybrid-Presidential = 1, and Parliamentary = 0. The ranking increases in
the degree of insulation of the Executive.
Executive Constraints: Measure of operational (de facto) independence of the
Chief Executive. From Polity IV data set. Ordinal, from 1 = minimum degree
of constraint to 7 = maximum degree of constraint. Average over 1990-1994
period for variable XCONST. The ranking decreases in the degree of
insulation of the Executive.
Powers of the Presidency: For columns (l)-(5) is Powers of the Elected
President from Shugart and Carey [1992, Ch. 8], dependent variable for (6)
(10) includes only nonlegislative powers (as oppose to legislative). High
nonlegislative powers imply high insulation of the Executive. The NCSEER
data expand the sample to post-Soviet Hybrid-Presidential countries and
conform to Shugart and Carey's criteria. The ranking increases in the degree
of insulation of the Executive.
Plurality: MAJ Dummy for plurality rule and majority systems (1 = Plurality,
0 = Proportional representation, mixed system or other). From Persson and
Tabellini [2003] variable MAJ. The ranking increases in the degree of
insulation of the Executive.
Control Sets:
Legal origin: French, Socialist, and other non-Common Law legal origin from
La Porta et al. [1999]; the Social legal origin dummy is dropped when ELF is
used, as non-Socialist countries were included in the original study.
Regional dummies: Sub-Saharan Africa, East Asia, and Latin America.
Colonial origin dummies: British, French, Spanish/Portuguese, other colonies;
from CIA World Factbook [2001].
Religion: Continuous variable indicating fraction of the total population that is
Protestant, Catholic, or Muslim, from La Porta et al. [1999].
Log(Real GDP) in 1960: Natural logarithm of real GDP per capita in 1960 from
Easterly and Levine [1997].

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE II
Sample Means
Sample: DEMOCRACY
Variable Obs Mean Std. dev. Min Max
Ethnic Fractionalization Index 69 0.30999 0.22927 0.002 0.7872
Ethno-Linguistic Fractionalization
Index 1960 46 0.28522 0.24319 0 0.82
Polarization (alpha = 4/5) 69 0.12976
Linguistic Fractionalization Index 66 0.272
Sample: PARLIAMENTARY
Index 1960 38 0.30158 0.25149 0.01 0.89
Polarization (alpha = 4/5) 49 0.12939 0.
Sample: NONPLURALITY
Index 1960 59 0.32593 0.27512 0.01 0.89
Polarization (alpha - 4/5) 66 0.14849 0.
authors).26 In columns (l)-(8) we alternative

and colonial origin, geographical, and religio
each country; in column (9) we control, as m
levels (we try to reduce the endogeneity is
levels). Usually, the control sets we employ
cance well into the critical region, and we
26. The results in Table II differ slightly from those re

Using a different sample and a different set of controls
of ethnic fractionalization has the same sign as ours in
democracy index, but it is not significant. In fact, this
nothing except level of per capita GDP affects his dem
many of his controls could be endogenous.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE II
(continued)
Sample: AUTOCRACY
Index 1960 20 0.58 0.32216 0.01 0.93
Polarization (alpha = 4/5) 41 0.17759 0
Sample: DIRECT PRESIDENTIAL
Ethnic Fractionalization Index 68 0.54252 0.26042 0 0.9302
Index 1960 55 0.48455 0.30358 0.01 0.93
Polarization (alpha = 4/5) 68 0.17862 0
Sample: PLURALITY
Ethnic Fractionalization Index 42 0.48886 0.25831 0 0.8635
Index 1960 36 0.5125 0.271 0.01 0.93
Polarization (alpha = 4/5) 42 0.1755 0.07478 0 0.2774
Linguistic Fractionalization Index 40 0.45338 0.30719 0.0103 0.8983
Variables are defined in Table I. The table reports number of observations, the mean, the standard
deviation, and the minimum and maximum values within the subsamples reported for different indicators of
fractionalization. Democracy refers to the classification of "Free" by Freedom House, Autocracy of "Not Free."
report) that French legal origin and Socialist legal origin are
associated with less democracy relative to the Anglo-Saxon sys
tem (the omitted category). With few exceptions, fractionalization
seems to increase the probability of ending up in a more auto
cratic (and more insulated) regime.27 Finally, with reference to
the potential endogeneity of ethnic fractionalization with respect
27. It should be noted that many of these control sets are correlated with each
other. Adding all the possible controls in the same regression causes two main
problems: collinearity and a reduction in the sample size. We verified both prob
lems as being relevant. We tested simultaneously for all the possible controls and
observed a reduction in the precision of the estimate for fractionalization as well
as a loss of significance for a large majority of the controls.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE III
Autocracy and Polarization
ORDERED PROBIT
(1) (2) (3) (4)
Autocracy Autocracy
(1990) (1990) (1990) (1990)
Ethnic Fractionalization Index 1.926 2.021 1.700 1.434
(0.355)*** (0.390)*** (0.418)*** (0.448)***
[174] [163] [174] [163]
Ethno-Linguistic 1.678 1.625 0.526 0.383
Fractionalization Index (0.421)*** (0.450)*** (0.481) (0.499)
1960 [107] [107] [107] [107]
Polarization (alpha = 4/5)4.229 3.387 4.112 2.854
(1.148)*** (1.226)*** (1.173)*** (1.296)**
[174] [163] [174] [163]
1.395
Linguistic Fractionalization 1.576 0.681 0.562
Index (0.309)*** (0.333)*** (0.365)* (0.396)
[168] [157] [167] [156]
Control sets
Legal origin Included Included

[0.005] [0.000]
Regional dummies Included Included
[0.002] [0.000]
Colonial origin
Religion
Log(Real GDP) in 1960
Marginal effect at mean from

partly free to not free status 0.553 0.478 0.469 0.297
to insulation, note that the bias should go against finding these

correlations. In fact, more insulated and less democratic systems
should be those more likely to engage in active policies toward
reducing fractionalization, such as expulsion of minorities, geno
cide, etc. A more subtle problem of endogeneity concerns the case
in which individuals self-classify themselves in certain ethnic or
racial groups as a function of the feature of the institutional

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE III
(continued)
ORDERED PROBIT
(5) (6) (7) (8) (9)
Autocracy Autocracy Autocracy Autocracy Autocracy
(1990) (1990) (1990) (1990) (1990)
1.866 1.506 1.743 0.982 1.700
(0.383)*** (0.455)*** (0.386)*** (0.463)** (0.513)***
[146] [146] [145] [145] [109]
1.545 0.678 1.654 0.429 0.731
(0.422)*** (0.464) (0.443)*** (0.475) (0.457)
[107] [107] [107] [107] [103]
4.281 3.637 3.031 1.630 3.995
(1.226)*** (1.272)*** (1.251)** (1.377) (1.608)**
[146] [146] [145] [145] [109]
1.436 0.668 1.431 0.458 1.140
(0.316)*** (0.395)* (0.339)*** -0.417 (0.453)**
[141] [140] [140] [139] [106]
Included Included
[0.026] [0.001]
Included Included
[0.002] [0.014]
Included Included
[0.000] [0.000]
Included
[0.000]
0.453 0.344 0.402 0.199 0.231
Variables are defined in Table I. Each column in the table reports Ordered Probit coefficients, robust
standard errors (in parentheses, below coefficient estimates), and number of observations (in brackets, below)
for the measures of polarization in society described in Table I. The table includes/?-values for Chi-square test
of joint significance of the control sets described in Table I. The marginal effects and p -values reported refer
to the specification employing the Ethnic Fractionalization Index of Alesina et al. [2004]. * significant at 10
percent; ** significant at 5 percent; *** significant at 1 percent.
system, for instance, because certain groups or others are more or

less favored.
From the coefficients reported in Table III one can compute
the marginal effects on the probability of a political system of
being less democratic. The marginal effect of ethnic fractionaliza
tion on the probability of ending up in the less democratic group
for a country already partly lacking political freedom is about

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
0.38 on average. Including income, the size of the marginal effect

is almost halved, but remains quantitatively important. This
seems to suggest not only that fractionalization seems to explain
insulation, but also that it has a quantitatively large role.
In summary, there seems to be evidence that in more ethically
fragmented societies, political systems are less democratic. We find
this result interesting because two strands of cross-country empiri
cal literature have independently emphasized the effect of ethnic
fragmentation on economic outcomes and the effect of presidential
ism and democratic status on politico-economic outcomes. These
results seem to suggest that the two sets of variables, institutions,
and racial fragmentation are not independent from each other. Our
interpretation, consistent with our model, is that in more frag
mented systems, political systems are chosen to insulate certain
groups and prevent others from having a voice.
Table IV moves a step forward in focusing on institutional
details and mapping insulation into specific political systems' fea
tures. The table repeats the analysis in Table III for Presidential
regimes now, focusing on the issue of Separation of Powers (i.e.,
insulation of the Executive from the Legislative Power). In Table III
the effect of all measures of fragmentation is strong and highly
significant if we exclude income from the specification (columns
(1M8)), even after employing a wide range of controls. The individ
ual effects are not reported, but we note that French legal origin and
Socialist legal origin are associated with more Presidential regimes
relative to the Anglo-Saxon system, which is the omitted category.28
When we control for income, the effect of fractionalization becomes
weaker and in some specifications loses significance at standard
confidence levels. However, the size of the marginal effects never
falls below .38. Table V maintains the same focus as Table IV, i.e.,
investigates insulation in terms of the characteristics of the Execu
tive Power. We concentrate here on an index of intensity of the
constraints on the de facto independence of the Chief Executive in
different systems as another measure of insulation/delegation. Al
though the results are less robust to changes in the specification
than in Table IV, it is still the case that the expected correlations
between fractionalization and insulation hold. More fractionalized
countries appear associated with lower levels of constraints for their
28. In some specification the impact of polarization/fragmentation appears

weaker, but this is mostly due to the inclusion of an irrelevant set of controls, as
shown by the Chi-square tests reported. This seems to be the case, for example, in
columns (4) and (6) with respect to regional dummies. It may also, ex post, explain
the weakness of specification (3).

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
Chief Executive. Table VI provides an analysis of the role of polar

ization and fractionalization within the form of government. It re
ports results related to Presidential regimes, gaining focus on char
acteristics highlighted in Tables IV and V. Within Presidential re
gimes there are evident differences in the degree of control by the
elected president over both legislative and nonlegislative processes.
Shugart and Carey [1992] provide an insightful analysis of the issue
and a taxonomic framework as well. As we interpret a more power
ful president as a more insulated one, we expect to find that more
polarized countries are also characterized by a more powerful presi
dent. Moreover, we would expect it to be particularly so for those
powers that are more closely related to insulation of the Executive
(the nonlegislative ones, such as the power of dissolving the assem
bly). This seems to be the case in the simple linear relationships of
Table V. Not only does there appear to be the expected positive
correlation between insulation and polarization, but also its signifi
cance actually increases when we consider specifically those powers
that are more closely related to the insulation mechanisms we
described. Unfortunately, the scarcity of data points does not allow
us to get into such a detail for Parliamentary regimes. Similar
patterns are evident, however, in additional research, not reported
here. Control of the agenda timing, for example, appears to be more
frequently handed over to the Government, as opposed to the Par
liament, in more fractionalized countries. Table VII concludes our
empirical overview focusing on Electoral Rules. The unconditional
results of Table II are confirmed in the signs of the coefficients sign
(Plurality systems are chosen wherever the population is more frac
tionalized), but the results are less robust. This result is in direct
contrast with studies limited to a smaller number of countries (par
ticularly democracies) suggesting a positive association between
ethnic fragmentation and proportional representation as in Lijphart
[1994], who studies a small group of advanced democracies.29
Overall, we find significant evidence that various indices
of insulations are positively correlated with measure of frac
tionalization and polarization. Thus, more polarized socie
ties tend to have more "insulated" rulers. Also, forms of gov
ernment appear to be endogenous to ethnic fractionalization.
29. These results are robust to controlling for various other variables includ
ing the fraction of population above 65, fraction of urban population, fraction of
population in agriculture, and life expectancy. These are all variables highly
correlated with GDP per capita. All these sensitivity tests are available upon
request.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE IV
Separation of Powers and Polarization
ORDERED PROBIT
(1) (2) (3) (4)
Separation Separation Separation Separation
of powers of powers of powers of powers
(PRES, (PRES, (PRES, (PRES,
1990) 1990) 1990) 1990)
Ethnic 1.964 1.724 1.384 1.256
Fractionalization (0.390)*** (0.464)*** (0.500)*** (0.535)**
Index [141] [140] [141] [140]
Fractionalization (0.392)*** (0.428)*** (0.502) (0.485)**
Index 1960 [106] [106] [106] [106]
Polarization (alpha 4.731 3.5 3.301 2.603
= 4/5) (1.451)*** (1.633)** (1.499)** (1.629)
[140] [139] [140] [139]
Linguistic 0.938 0.932 0.5 0.55
Fractionalization (0.326)*** (0.356)*** (0.430) (0.428)
Index [135] [134] [135] [134]
Control sets

[0.000] [0.000]
[0.044] [0.190]
Colonial origin
Religion
Log(Real GDP) in
1960
Marginal effect at
mean from
partly free to not
free status 0.782 0.685 0.551 0.499
V. C. Insulation in Times of Crisis

A "crisis" can be defined as a situation in which a policy
action is especially desirable, even though not everybody may
benefit equally from such action. In our model, then, we can

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE IV
(continued)
ORDERED PROBIT
(5) (6) (7) (8) (9)
Separation Separation Separation Separation Separation
of powers of powers of powers of powers of powers
(PRES, (PRES, (PRES, (PRES, (PRES,
1990) 1990) 1990) 1990) 1990)
1.642 1.485 1.704 0.957 L265
(0.457)*** (0.539)*** (0.406)*** (0.525)* (0.468)***
[140] [140] [139] [139] [109]
1.236 0.929 1.704 1.043 0.391
(0.515)** (0.622) (0.432)*** (0.579)* (0.439)
[106] [106] [106] [106] [103]
3.787 3.437 3.475 1.684 3.391
(1.584)** (1.598)** (1.482)** (1.596) (1.583)**
[139] [139] [138] [138] [108]
0.983 0.735 1.104 0.642 0.269
(0.399)** (0.474) (0.342)*** (0.502) (0.412)
[134] [134] [133] [133] [105]
Included Included
[0.801] [0.060]
Included Included
[0.000] [0.000]
Included Included
[0.000] [0.000]
Included
[0.000]
0.653 0.591 0.678 0.381 0.503
Variables are defined in Table I. Each column in the table reports Ordered Probit coefficients, robust
standard errors (in parentheses, below coefficient estimates) and number of observations (in brackets, below)
for the measures of polarization in society described in Table I. The table includes p -values for Chi-square test
of joint significance of the control sets described in Table I. The marginal effects and p -values reported refer
to the specification employing the Ethnic Fractionalization Index of Alesina et al. [2004]. * significant at 10
interpret a "crisis" as a situation where 7 is especially high. A

straightforward implication of our comparative static analysis is
that one should observe a movement toward more insulation in
times of emergency. One extreme example is wartime, when,

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE V
Executive Constraints and Polarization
OLS
(1) (2) (3) (4)
Executive Executive Executive Executive
constraints constraints constraints constraints
Ethnic -3.768 -3.681 -2.633 -1.748

Fractionalization (0.619)*** (0.740)*** (0.843)*** (0.863)**
Index [121] [113] [121] [113]
0.2 0.25 0.28 0.4
Ethno-Linguistic -2.869 -2.737 -0.593 -0.437
Index 1960 [104] [104] [104] [104]
0.14 0.19 0.32 0.36
Polarization (alpha -7.182 -4.041 -6.142 -2.372
= 4/5) (2.368)*** (2.684) (2.324)*** (2.398)
[121] [113] [121] [113]
0.06 0.1 0.26 0.37
Linguistic -2.742 -2.688 -1.196 -0.605
Index [119] [111] [119] [111]
0.13 0.21 0.22 0.36
Control sets

[0.070] [0.022]
[0.001] [0.000]
Colonial origin
Religion
Log(Real GDP) in
1960
often, democratic rule is limited and even democratic countries

adopt a more hierarchical structure of power. In the terminology
of our model, one can think of "winning a war" as a policy with a
very high 7 requiring a very high degree of insulation of leaders.
Less extreme examples involve reforms of economic institu
tions. For instance, often Central Banks have been made more

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE V
(continued)
OLS
(5) (6) (7) (8) (9)
Executive Executive Executive Executive Executive
constraints constraints constraints constraints constraints
-2.988 -1.803 -3.135 -1.219 -1.99

(0.719)*** (0.895)** (0.612)*** (0.838) (0.669)***
[113] [113] [112] [112] [105]
0.32 0.38 0.33 0.44 0.42
-1.943 -0.472 -2.388 -0.362 -1.074
(0.745)** (0.845) (0.663)*** (0.658) (0.643)*
[104] [104] [104] [104] [100]
0.34 0.41 0.3 0.45 0.4
-3.842 -2.617 -3.75 -0.489 -1.635
(2.549) (2.386) (2.281) (2.286) (2.147)
[113] [113] [112] [112] [105]
0.23 0.36 0.21 0.42 0.37
-2.035 -0.695 -2.381 -0.714 -1.056
(0.629)*** -0.826 (0.589)*** -0.706 (0.630)*
[119] [111] [110] [110] [102]
0.29 0.35 0.31 0.43 0.38
Included Included
[0.011] [0.009]
Included Included
[0.003] [0.282]
Included Included
[0.000] [0.001]
Included
[0.000]
Variables are defined in Table I. Each column in the table reports OLS coefficients, robust standard
errors (in parentheses, below coefficient estimates), number of observations (in brackets, below), and R2s (in
italics, further below) for the measures of polarization in society described in Table I. The table includes
p-values for F-test of joint significance of the control sets described in Table I. Thep-values reported refer to
the specification employing the Ethnic Fractionalization Index of Alesina et al. [2004]. * significant at 10
independent, in order to "insulate" monetary policy, in periods of

very high inflation, that is in periods where anti-inflationary
policies have an especially high 7. The most famous example
involves the Bundesbank and the German hyperinflation. The
recent adoption of fiscal rules with the European Union that limit
the discretion of fiscal decisions on budget deficits can also be

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE VI
Powers of the Presidency and Polarization
OLS
(l) (2) (3) (4) (5)
Total powers Total powers Total powers Total powers Total powers
of the of the of the of the of the
presidency presidency presidency presidency p
Ethnic 7.164 7.723 5.523 6.887 5.269
Fractionalization (3.940)* (5.322) (6.326) (4.359) (3.341)
Index [31] [31] [31] [30] [56]
0.10 0.19 0.33 0.15 0.03
Ethno-Linguistic 4.321 5.092 3.928 3.753 4.321
Fractionalization (3.281) (4.305) (4.231) (3.648) (3.281)
Index 1960 [30] [30] [30] [30] [30]
0.04 0.15 0.32 0.09 0.04
Polarization (alpha 31.057 31.841 21.929 4.152 17.824
= 4/5) (10.556)*** (13.776)* (13.615) (3.145) (10.085)*
[31] [31] [31] [28] [56]
0.26 0.30 0.39 0.21 0.06
Linguistic 4.484 8.123 7.959 29.898 2.161
Fractionalization (2.918) (4.019)* (3.356)** (11.064)* (3.063)
Index [29] [29] [29] [30] [53]
0.06 0.30 0.56 0.29 0.01
Control sets
Legal origin Included

[0.240]
Regional dummies Included
[0.000]
Log(Real GDP) in Included
1960 [0.296]
Sample Shugart and Shugart and Shugart and Shugart and Shugart and
Carey Carey Carey Carey Carey [1992]
[1992] [1992] [1992] [1992] and NCSEER
seen as a response to the fiscal crises of the 1980s and

Latin America several institutional reforms leading to
sulation" of monetary and fiscal policy from the ebb an
changing political majorities have followed the "lost d
the 1980s.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE VI
(continued)
OLS
(6) (7) (8) (9) (10)

Nonlegislative Nonlegislative Nonlegislative Nonlegislative Nonlegislative
(checks and (checks and (checks and (checks and (checks and
balances) balances) balances) balances) balances)
powers of the powers of the powers of the powers of the powers of the
presidency presidency presidency presidency presidency
5.566 5.765 3.456 5.522 4.545

(2.665)** (3.993) (4.703) (2.981)* (2.286)*
[31] [31] [31] [30] [56]
0.13 0.19 0.29 0.14 0.05
3.928 4.395 3.809 3.686 3.928
(2.248)* (3.195) (2.969) (2.494) (2.248)*
[30] [30] [30] [30] [30]
0.07 0.17 0.31 0.09 0.07
20.137 19.559 12.819 3.736 12.74
(7.369)** (10.900)* (10.188) (2.176)* (6.740)*
[31] [31] [31] [28] [56]
0.22 0.25 0.33 0.18 0.06
3.843 6.16 6.537 19.677 1.733
(1.948)* (3.380)* (2.135)*** (7.936)** (2.138)
[29] [29] [29] [30] 153]
0.09 0.33 0.59 0.23 0.01
Included
[0.321]
Included
[0.023]
Included
[0.569]
Shugart and Shugart and Shugart and Shugart and Shugart and
Carey Carey Carey Carey Carey [1992]
[1992] [1992] [1992] [1992] and NCSEER
Variables are defined in Table I. Each column in the table reports OLS coefficients, robust standard
errors (in parentheses, below coefficient estimates), number of observations (in brackets, below), and R2s (in
italics, further below) for the measures of polarization in society described in Table I. The table includes
p -values for F-test of joint significance of the control sets described in Table I. The p -values reported refer to
An even more extreme case of a crisis is a war. We can

interpret winning a war as a "policy" which is considered essen
tial by most, so a war implies a large increase in y. It is generally
believed that during wars it is necessary to increase the executive
power of unrestricted action and even temporarily suspend cer

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE VII
Electoral Rule and Polarization
PROBIT
(1) (2) (3) (4)
Plurality Plurality Plurality Plurality
rule rule rule rule
Ethnic Fractionalization 1.071 0.997 0.351 0.186
Index (0.488)** (0.565)* (0.611) (0.65)
[109] [109] [109] [109]
Fractionalization Index (0.479)*** (0.541)* (0.588) (0.650)
1960 [95] [95] [95] [95]
Polarization (alpha = 4/5) 2.716 2.336 1.493 1.248
(1.588)* (1.844) (1.788) (2.008)
[108] [108] [108] [108]
Linguistic Fractionalization 0.94 0.726 -0.251 -0.594
Index (0.427)** (0.462) (0.576) (0.614)
[104] [104] [104] [104]
Control sets
[0.002] [0.002]
[0.048] [0.031]
Colonial origin
Religion
Log(Real GDP) in 1960
Marginal effect at mean

from partly free to not
free status 0.408 0.376 0.133 0.070
tain types of checks and balances. The army itself has a very
"insulated" constitution; that is, nobody can question directives
from superior officers. In fact, the case of wars may be one
example in which Constitutions can be made contingent on one
particular event, namely war. So the Constitution can prescribe
special and more insulated procedures in times of declared wars.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
TABLE VII
(continued)
PROBIT
(5) (6) (7) (8) (9)

Plurality Plurality Plurality Plurality Plurality
rule rule rule rule rule
0.749 0.200 0.877 0.133 1.051
(0.572) (0.66) (0.518)* (0.638) (0.565)*
[109] [109] [108] [108] [96]
0.966 0.392 1.205 0.566 1.357
(0.553)* (0.696) (0.513)** (0.621) (0.540)**
[92] [92] [95] [95] [93]
2.426 1.294 2.690 1.224 2.809
(1.815) (1.948) (1.698) (1.861) (1.748)
[108] [108] [107] [107] [95]
0.496 -0.296 0.649 -0.218 0.901
(0.464) (0.630) (0.461) (0.604) (0.492)*
[104] [104] [103] [103] [93]
Included Included
[0.312] [0.242]
Included Included
[0.000] [0.001]
Included Included
[0.003] [0.026]
Included
[0.115]
0.275 0.073 0.331 0.050 0.402
Variables are defined in Table I. Each column in the table reports Probit coefficients, robust standard
errors (in parentheses, below coefficient estimates) and number of observations (in brackets, below) for the
measures of polarization in society described in Table I. The table includes p-values for Chi-square test of
joint significance of the control sets described in Table I. The marginal effects and p -values reported refer to
VI. Conclusions
This paper has moved some steps forward in analyzing the

"endogenous choice of political institutions." We have focused
on one important, and general question of institution design,
namely how insulated political leaders are expected to be ex

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
post, or to put in reverse, how large the (super)majority should

or will be to pass legislation. Rather than reviewing our re
sults, we close by highlighting a few directions of further
research.
One is to investigate other aspects of institutional design,
for instance, term limits as another form of insulation. A leader
facing short term limits may not feel particularly interested in
putting effort into policy reforms, especially if they require
"time to build." On the other hand, without term limits incum
bents may achieve an entrenched power structure and restrict
political competition. Once again, the Founding Fathers had
captured the essence of the problem of term duration. In The
Federalist Paper No. 71, in fact, Hamilton puts it beautifully by
writing that "It may be asked also whether a duration of four
years would answer the end proposed; and if it would not,
whether a less period, which would at least be recommended by
greater security against ambitious design would be too short
for the purpose of inspiring the desired fairness and indepen
dence of the magistrate."
An important question related to insulation is the choice of
electoral rules, majoritarian, proportional, etc. In order to
properly address this issue, one needs a richer model of voting,
rather than the simple referendum type structure that we used
here.
A third extension could explore the issue of entry of politi
cians and their quality.30 First, voters may adjust whom they
select as their leader as a function of how much the leader himself
is insulated. Different politicians may be chosen in different types
of systems. Also the nature of insulation might affect the entry
decision of potential leaders. High quality candidates may opt out
if they feel that they have no real power when in office.
Appendix 1: Optimality of the Simple Majority Rule under Risk

Neutrality and No Expropriation: The Basic Intuition31
Suppose that there were only two alternatives, x andy, to be

chosen between ex post. Ex ante the individuals in the constitu
ency (of size normalized to 1) are under the veil of ignorance and
30. For a recent discussion of the quality of politicians, see Caselli and
Morelli [2002].
31. This example was kindly provided to us by Matthew Jackson, and it
builds on Rae [1969].

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
do not know whether they will prefer x ory. Suppose that k is the
number of individuals who prefer x to y. If an individual has ex
post income a if her preferred alternative is selected and income
( ??) if the other alternative is chosen, then under risk neutrality
the ex ante total utility of choosing alternative x, is equal to
U(x) = ak- ?(l - k);
similarly,
U(y) = -?? + a(l-?)
if alternative y is chosen. Choosing M = V2 will then guarantee
that the alternative that maximizes total ex ante utility is always
chosen, namely x whenever k > V2 , and y otherwise. This rea
soning extends to collective decision problems like ours that boils
down to a utilitarian maximization problem: if M > V2 (respec
tively, M < V2), then reforms would take place too (respectively,
not sufficiently) often from the standpoint where voters expect
their preferences for reform to lie ex post.
Appendix 2: Proofs of Propositions
Generalization of Proposition 1
Let us consider a ~ g(a), a E [-A,A], where g(a) =
dG(a)/da and G is the cumulative distribution function for the
shock a, and X? - f(X?), X? G [X,X] and A > 0, X > X, 7 > 0. For
simplicity let us fix p = 1 in this analysis. The maximization
problem for the voter under a veil of ignorance is
(?i-\-lM Ci
f(Xi)g(a) dX? da
+ l(Xl + a)f(Xl)g(a)dXlda
H-k-lM ' \
= max {G(X -X-IM)~ G(-A) + (G(A)

-G(X-X- lM))EXn + y(AG(A)
- (X - X - IM)G(X -X- IM)) - 7 G(a) da,

J\-\~IM

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
where we make straightforward use of integration by parts and E

is the expectation operator. Now, by imposing the first-order
conditions, we obtain
M* = (EX?- X)/(X- X),
which corresponds to V2 for f(\?) uniform.
Proof of Proposition 2
Suppose that whenever (X? + a)y < 6, where 6 < 0, individual /
incurs a utility loss of ?u. Recall that the indifference voter between
reform and no reform, satisfies X? + a = X = I/7. The ex ante
maximization problem at the constitutional stage now becomes
\-\-lM
Cd/y-a 1 \ 1
max
M
Ip 2Ada ? up jdxA^da
X-X-IM
+ Jk-k~lM
p (\m-r-?)y^da-(l-p)Mb\
Rewriting the middle term in u as
CQ/y-a -|_
up
X-X-IM ldXi \2?da
-up 9 \ 1
~2M --\)(A-(X-X-ZM))-2(A2-(X
we can reexpress the original problem as
max
M {?- (X - X - IM - A) + ^ Kmy(A - (X - X - IM))
e
X (A-(X-X-ZM))
+ ?^(A2-(^-^-/M)2)-S
- 2 (A2 - (X - X - IM)2) (l-p)bM\
Taking first-order conditions with respect to M, we obtain
e
-K + IM
1-p 2Ab = 0.
G(u, M) = 7/(Xm - X - ZM) - u

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
The optimal degree of insulation is

yl(Xm - X) + u(l - 6)/7 - 2Afe(l - p)/p
(Al) M*
(yl + u)l
We can easily verify now the comparative statics dM*/du < 0.
To see this, we use the fact that u only enters the maximization
program if and only if 0/7 - X > X. - X - IM. We then have
dM* _ dG IdG _ e/7 - X + IM
du ~ " ~du~/dM " (yl + u)l K ?'
which establishes the proposition.
With polarization at the lower end of the preference inter
val and risk aversion as in the previous proposition, the ex ante
maximization problem at the constitutional stage, now
becomes
-\-(M-A)Z/(l-A) "L
UT\PJA 2?da
A+l ??-^dXi
(A / & \ 1
+ p \ Um - ~2 + a rt 2? da ~ ^
>\-\-(M-A)l/(l-A) \ j J
Taking first-order conditions with respect to M,
M (M - A)A I
y[im--7r-k
2 ^ 1-A/l-A
- M -- \ + ^?-irv)-2Ab =
It follows that under risk aversion and polarization t
degree of insulation is given by
yl(Xm - AZ/2 -\ + AZ/(1 - A))
(8) M* =+ u(\ - 9Ay)(l - A) - 2A6(1 - A)(l - p)/
l(-yl/(l - A) + u) '

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
For large u the optimal level of insulation M* is approximately

equal to
X - 8/7 1-6
M* = (1 - A)-j-1 = (1
7/ - A)
so that
dM* 1-6
<0.
dA yl
For u = 0, the optimal degree of insulation become
2
1= ,.(1 .?
M* =2 . 1-P?? (1-A)
v" -"' A)2' " +p A-
"~"2Ab
7Z2 '
so that
dM* 1-p 2(1-A)
dA p 7Z2
The proposition then follows by cont
spect to u.
A policy reform will pass with com
if
(1 4- k)c(a) < a)
or equivalently
2W \1/2
/ 2?co
X-X-IM - a
,(i + *h
given that
p~? (X - a - X,)7 7 t< , .-_2

C(a)= -ldKl-=2z(^-?-^-ZM)2.
Two cases can be considered:
(1) X - X - IM* + A < (2Z(o/(l + *h)1/2, which will be true
whenever w is sufficiently large (i.e., enough funds have been
raised); in this case reform will always take place (as it will
always be affordable) and, at the constitutional stage behind a
veil of ignorance, the generic individual will choose M* in order to
maximize

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
max {p(Xm7 - kEac(a)) - (1 - p)bM},
where Eac(a) is the expected compensation cost, and given by
l-?~lMy(X-X-lM-a)2 1
Eac(a)=\ -2/-2?da
da
7(X-X-ZM + A)3
12AI
The solution of this problem is
\V) M (1) - 7
_ i-\+A-2y?(Ab(l-p)/P1k)
A sufficient condition for individuals to opt for compen

at the constitutional stage instead of the no-compensa
tion analyzed in the previous section, is obtained by co
utility levels achieved under the two systems. One can s
utility under compensation is not defined for k = 0, b
continuous, monotonie, strictly decreasing in k for any
we consider the limit as k -? 0, it is possible to show
diverges to +?? while utility without compensation i
Hence, there exists a cutoff level k such that compensation
a viable alternative whenever k E (0, k],
(2) X - X - IM* + A > (2Z<?/(l + k)y)1/2 identifi
second case, when not enough resources can be raised t
compensate losers. Hence, the problem for the voter be
?I fi-h-lM-(2lu>/(l + k)y)1/2 "L

M
' -A
P\ V?â
1
+ <Xn + ah 2A da
X-\-ZM-(2/W(l + ?h)1/2
3/2\
7 / 2Zco , ,
where the third term in the parentheses multiplied by p is the

expected deadweight loss from compensation, calculated as

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
Eac(a) =J X-X-lM-(2lo>/(l+k)y)1/2
2?\ (X - X - IM - a)2^da
12AI \y(l + k)J '

while the first two terms represent the expected income level. The
maximization problem implies that
..? 1 2Ab(l-p)/p ( 2(0

(11) M*2> = 2-?S<-\WTTk)
A sufficient condition of the type derived above for the first case
can be derived for this second case as well.
Finally, it is immediate to see that both M*x) and M*2> satisfy
the conditions stated in the proposition.
We want to show that in a population symmetrically polar
ized and risk averse, if the minority sitting at the Constitutional
table, Ce, know that their type X is distributed over the interval
[X - e,X] with X - 8 > X, then
1. for all ? the level of insulation Me chosen by this minority
is greater than the optimal level M* chosen under the veil
of ignorance;
2. for 8 sufficiently small, constitution writers in Ce react to
an increase in polarization A by always increasing insu
lation Me.
First, note that the ex ante maximization problem faced by
the constitutional minority, can be written as
?fx-X-(M-A)?/(1-A) -[_
P 9~?da
B/y-a (]_ _ A)
dXi 2A
I "H -?rj da
A 1
+
X-\-(M-A)Z/
(EM p |
+
where

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
g[e] = E[Xi\i E C] = X - e/2

is the average idiosyncratic benefit from reform among individu
als in the minority.
Taking the first-order condition for the above program, we
obtain
/X-8 (M-A)Z\ I ( /e ,
7^--T^^j(r^^-^m^l0'l7~x~z + ?
Z(M-A)\\ 1-p
+ -T^))-VÂfe = o,
and therefore,
8 AZ
(13) Me = Z(7Z/(1 - A) + u) yl\l-?2+ (1 - A)
+ u max ( 0, ( X - - + Z - 8 )(1 - A) + ZA
1-p
?^2A6(1-A)
That Me is greater than the corresponding rule M* under the veil
of ignorance, follows immediately from the comparison between
(8) and (13) together with the fact that
^[8] = X - 8/2 > Xm .

and
u(i - ?)(1 - A) < u^X - ? + Z - e)(l - A) + Za),

which implies that the minority loses less and gains more from
reform than the representative individual under the veil of igno
rance. This establishes the first part of the proposition.
Finally, that dMe/cZA increases when 8 is sufficiently small,
follows immediately from the fact that in that case for all a E
[X - X - (M - A)Z/(1 - A), A], we have
X - 8 > e/7 - a
when 8 is small. Hence in that case the equilibrium level of
insulation becomes

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
M
MP =
yP yl[l 2 + (1 - A)
2A6(1 - A)
Then:
dM? = 1 Z-e/2 1-p 1-A
--4A6
~d~K I yP >0,
so that for sufficiently low e the more constitution writers in the
minority react to an increase in polarization by increasing insu
lation. This establishes the proposition.
Hahvard University and Canadian Institute for Advanced Research

Harvard University, National Bureau of Economic Research, and Centre
for Economic Policy Research
Harvard University
References
Acemoglu, Daron, and James Robinson, "Why Did the West Extend the Franchise:
Democracy, Inequality and Growth in Historical Perspective," Quarterly
Journal of Economics, CXV (2000), 1167-1199.
Acemoglu, Daron, and James Robinson, "A Theory of Political Transition," Ameri
can Economic Review, XCI (2001), 938-963.
Aghion, Philippe, and Patrick Bolton, "Incomplete Social Contracts," Journal of
the European Economic Association, I (2003), 38-67.
Aghion, Philippe, and Peter Howitt, Endogenous Growth Theory (Cambridge, MA:
MIT Press, 1998).
Aghion, Philippe, Alberto Alesina, and Francesco Trebbi, Choosing Forms of
Government: Evidence from US Cities," mimeo, Harvard University, 2004.
Alesina, Alberto, Arnaud Devleeschauer, William Easterly, Sergio Kurlat, and
Romain Wacziarg, "Fragmentation," Journal of Economic Growth, VIII
(2003), 155-194.
Alesina, Alberto, and Edward Glaeser, Fighting Poverty in the US and Europe: A
World of Difference (Oxford, UK: Oxford University Press, 2004, forthcoming).
Alesina, Alberto, and Howard Rosenthal, Partisan Politics Divided Government
and the Economy (New York, NY: Cambridge University Press, 1995).
Barbera, Salvador, and Matthew Jackson, "Choosing How to Choose: Self-Stable
Majority Rules," mimeo, California Institute of Technology, 2001.
Baron, David, "Majoritarian Incentives, Pork Barrel Programs and Procedural
Control," American Journal of Political Science, XXXIII (1991), 1048-1086.
Baron, David, and John Ferejohn, "Bargaining in Legislatures," American Politi
cal Science Review, LXXXIII (1989), 1181-1206.
Barro, Robert, "Democracy and Growth," Journal of Economic Growth, I (1996),
1-27.
Beard, Charles, An Economic Interpretation of the American Constitution (Lon
don, UK: Macmillan, 1941).
Beck, Thorsten, George Clark, Alberto Groff, Philip Keefer, and Patrick Walsh,
"New Tools and New Results in Comparative Political Economy: The Data
base of Political Institutions," World Bank Policy Research WP No. 2283,
2000.
Buchanan, James, The Limits of Liberty; Between Anarchy and the Leviathan
(Chicago, IL: University of Chicago Press, 1975).
Buchanan, James, and Gordon Tullock, The Calculus of Consent (Ann Arbor, MI:
University of Michigan Press, 1962).

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00
Caselli, Francesco, and Massimo Morelli, "The Quality of Politicians," mimeo,

Harvard University, 2002.
CIA (2001), World Factbook.
Diermeier, Daniel, and Robert Myerson, "Bicameralism and Its Consequences for
the Internal Organization of Legislatures," American Economic Review,
LXXXV (1995), 1182-1196.
Easterly, William, and Ross Levine, "Africa's Growth Tragedy: Policies and Ethnic
Divisions," Quarterly Journal of Economics, CXII (1997), 1203-1250.
Esteban, Joan Maria, and Debraj Ray, "On the Measurement of Polarization,"
Econometrica, LXII (1994), 819-852.
Freedom House, "Freedom in the World," 2001.
Grossman, Gene, and Elhanan Helpman, Special Interest Politics (Cambridge,
MA: MIT Press, 2001).
Hayek, Friedrich, The Constitution of Liberty (Chicago, IL: University of Chicago
Press, 1960).
Henisz, Witold, "The Institutional Environment for Infrastructure Investment,"
Industrial and Corporate Change, XI (2002), 355-389.
Koray, Semih, "Self-Selective Social Choice Functions Verify Arrow and Gibbard
Satterthwaite Theorems," Econometrica, LXVIII (2000), 981-996.
Krusell, Per, and Victor Rios-Rull, "Vested Interests in a Positive Theory of
Stagnation and Growth," Review of Economic Studies, LXIII (1996), 301-329.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny,
"The Quality of Government," Journal of Law Economics and Organizations,
XV (1999), 222-279.
Lijphart, Arend, Electoral Systems and Party Systems A Study of 27 Democracies,
1945-1990 (New York, NY: Oxford University Press, 1994).
Lizzeri, Alessandro, and Nicola P?rsico, "Why Did the Elites Extend the Suffrage?
Democracy and the Scope of Government, with an Application to Britain's Age
of Reform," Quarterly Journal of Economics, CXIX (2004), 707-765.
Madison, James, Alexander Hamilton, and John Jay, The Federalist Papers (New
York, NY: American Library, 1788, 1961).
Mansfield, Harvey, and Delba Winthrop, Alexis de Tocqueville Democracy in
America (translation) (Chicago, IL: University of Chicago Press, 2000).
Milesi-Ferretti, Gian Maria, Roberto Perotti, and Massimo Rostagno, "Electoral
Systems and Public Spending," Quarterly Journal of Economics, CXVII
(2002), 609-657.
Montalvo, Jos? G., and Marta Reynal-Querol, "Why Ethnic Fractionalization?
Polarization, Ethnic Conflict, and Growth," mimeo, 2002.
Mulligan, Casey, Xavier Sala-i-Martin, and Ricard Gil, "Do Democracies Have
Different Public Policies than Nondemocracies?" Journal of Economic Per
spectives (2004), forthcoming.
North, Douglas, Structure and Change in Economic History (New York, NY: W. W.
Norton and Company, 1981).
North, Douglas, and Barry Weingast, "Constitution and Commitment: The Evo
lution of Institutions Governing Public Choice in 17th Century England,"
Journal of Economic History, XLIX (1989), 803-832.
Persson, Torsten, and Guido Tabellini, Political Economics. Explaining Economic
Policy (Cambridge, MA: MIT Press, 2000).
Persson, Torsten, and Guido Tabellini, The Economic Effects of Constitutions
(Cambridge, MA: MIT Press, 2003).
Polborn, Mattias, and Matthias Messner, "Voting on Majority Rules," mimeo,
2002.
Rae, Douglas, "Decision-Rules and Individual Values in Constitutional Choice,"
The American Political Science Review, LXIII (1969), 40-56.
Rodrik, Dani, "Democracies Pay Higher Wages," Quarterly Journal of Economics,
CXIV (1999), 707-739.
Shugart, Matthew, and John Carey, Presidents and Assemblies: Constitutional
Design and Electoral Dynamics (New York, NY: Cambridge University Press,
1992).
Voigt, Stefan, "Positive Constitutional Economics: A Survey," Public Choice, XC
(1997), 11-53.

85.240.80.124 on Thu, 14 Mar 2024 22:05:34 +00:00

Endogenous Political Institutions (A J W Philippe Aghion, Francesco Trebbi) 2004

Uploaded by

Copyright:

Available Formats

Endogenous Political Institutions (A J W Philippe Aghion, Francesco Trebbi) 2004

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Endogenous Political Institutions (A J W Philippe Aghion, Francesco Trebbi) 2004

Uploaded by

Copyright:

Available Formats

ENDOGENOUS POLITICAL INSTITUTIONS*

A fundamental aspect of institutional design is how much society chooses to

* We thank Daron Acemoglu, Marios Angeletos, Olivier Blanchard, Matilde

theory of checks and balances, embodied in the work by Montes

2. For a vast survey of the literature on Constitutional design, see Voigt

This content downloaded from

tions. One especially interesting case is a situation in which those

This content downloaded from

connect with a recent literature on "choosing how to choose"

II. Political Insulation

ILA. The Basic Model

This content downloaded from

the same income, which we normalize to 1. Individual income

This content downloaded from

leader is "good" and promotes the reform; with probability ( 1 - p)

This content downloaded from

messages that voters would exchange ex post about the realiza

This content downloaded from

delegation. In the case of Presidential regimes like the United

This content downloaded from

matter how he gets there), is uncontrolled, while an essential

U.C. Solution of the Model

This content downloaded from

16. See the Appendix and Rae [1969].

This content downloaded from

HLA. Risk Aversion and Polarization of Preferences

where y ? is ex post income and where u > 0. Thus, only if income

This content downloaded from

falling below 0 ex post. Thus, more risk aversion leads to choosing

Proposition 3. For a sufficiently large degree of risk aversion as

Proof. See Appendix.

The intuition for this result is straightforward: on the one

This content downloaded from

This content downloaded from

Two cases must be considered. The first case is when there

max {p(Xm7 - kEac(a)) - (1 - p)bM}9

Proposition 4. When the compensation scheme is available and k

The basic message of this proposition is that a less efficient

IV. The Political Economy of Constitutions

This content downloaded from

IV.A. Wealth Distribution, Voting Rights, and Constitutions

(6) M*(wl) = l~2Abwi:^j^,

(7) M*(wJ = \-2Abwm1^.

This content downloaded from

rights were restricted to a fraction of wealthy men, should pre

TV.B. Constitutional Choice by the Upper Tail

This content downloaded from

minority who chooses the constitution, will react to increased

TV. C. Choosing How to Choose

This content downloaded from

the Constitution are generally different from the rules regulating

V. Empirical Implications and Discussion

In what follows, we discuss several empirical implications of our

VA. Economic Development and Institutions

This content downloaded from

On the other hand, property rights tend to be less well

V.B. Polarization, Fragmentation, and Institutional Choice

This content downloaded from

preoccupation that goes back at least to the Founding Fathers. From