January 2011
The Making of Policy:
Institutionalized or Not?
Carlos Scartascini
Mariano Tommasi
ABSTRACT: This paper attempts to build bridges in the formal study of policymaking across polities
of different degrees of institutional development. It explores the reasons why policymaking is fairly
institutionalized in some polities but not in others.
It suggests extending standard models of
institutionalized policymaking to allow for a wider set of actions, including the threat of violence or of
damage to the economy. It engages the discussion of institutions as rules and institutions as equilibria,
delivering multiple equilibria with different degrees of institutionalization. The likelihood of
institutionalized policymaking increases as the cost of alternative political actions increases, as the
damage these alternatives cause decreases, and as the economy becomes wealthier. In cases in which
the distribution of de jure political power is more asymmetric, it is more likely to observe use of
alternative political technologies as well as low degrees of institutionalization.
We received valuable comments from Facundo Albornoz, Mariana Chudnovsky, Thad
Dunning, Jon Eguia, David Epstein, Alberto Fohrig, Macartan Humphreys, Marcelo Leiras,
Johannes Lindvall, Cesar Martinelli, Massimo Morelli, Vicky Murillo, Pablo Pinto, Susan
Rose-Ackerman, Frances Rosenbluth, Alejandro Saporiti, Ken Scheve, Ernesto Stein, David
Stasavage, Susan Stokes, Kaj Thommson, Razvan Vlaicu, Federico Weinschelbaum, and
seminar participants at Columbia, George Washington, the Inter-American Development
Bank, Maryland, New York University, Universidad de Salamanca, Universidad de San
Andrés, Yale, the Hertie School of Governance, the Latin American and Caribbean
Economic Association Meeting in Rio de Janeiro, the Latin American and Caribbean
Economic Association Political Economy Group Meeting in Cartagena, and the Yale
Conference on Fighting and Voting. We are particularly indebted to three anonymous
referees for their criticism and suggestions; to Adam Przeworski for generously sharing his
ideas on these issues; to Fabiana Machado for allowing us to draw from our joint ongoing
empirical work on protests in Latin America, as well as for valuable suggestions; and to
Laura Trucco for extraordinary research assistance, useful comments, and for allowing us to
draw freely from joint ongoing work.
1
THE MAKING OF POLICY: INSTITUTIONALIZED OR NOT?
Institutional systems differ significantly in their capacity to absorb and process conflict. In some countries, for example
Argentina, almost every contentious issue finds thousands of people on the Plaza de Mayo, tractors blocking roads, and
pickets cutting a bridge in Neuquén. Yet in other countries, say Costa Rica, almost all conflicts are disciplined by political
parties and processed through the Congress, the Presidency, or the Courts. Adam Przeworski 2009 (p. 2)
Formal analysis of policymaking within the realm of political institutions has deepened our
understanding of how various legislative and electoral institutions shape economic policies. Political
economists have built on the foundations laid out early on by political scientists – such as Duverger,
Sartori, Rae, Lijphart, Cox and others – on the impact of formal rules on the workings of political
institutions to study the impact of political rules on policy outcomes. In almost all of this literature
(Persson and Tabellini 2000 is a standard textbook treatment), most or all of the relevant action takes
place within the context of these formal institutional rules and relatively formalized institutional arenas
(the voting booth, the building of Congress, etc.) That is a very good approximation for policymaking
in various countries at some moments in time (many developed countries in the last several decades),
but it is a much rougher approximation in other cases.
Other than voting, forming political parties, bargaining in the legislature, and the like, there are a
number of alternative political technologies (such as threats of violence and of disruption of economic
activity) that individuals or groups could utilize in order to influence collective decisions. This paper is
an exploration on the implications of an enlarged political action space for the study of institutions and
policymaking. We believe this is an important step towards the integrated study of policymaking in
polities of different levels of economic and institutional development. There is a tendency to study
policymaking in advanced industrial societies and in less developed countries in dichotomous ways,
2
using widely different models by assumption. These differential treatments are imposed exogenously;
in this paper we take a step in the direction of studying such different polities within a unified
framework that delivers different modes of policymaking as an outcome of the analysis.
This paper provides one specific example, extending a standard legislative bargaining model by
allowing actors the choice of investing in playing in the legislative arena or in utilizing an alternative
political technology (APT) that allows them to threaten to impose costs on society if some demands
are not met. We analyze that game, and use the example as motivation to address some broader issues
on institutions, institutionalization, and policymaking.1 We investigate the way in which the presence
of these alternative political technologies, in interaction with formal political institutions and
underlying socioeconomic structures, influences the workings of institutions, policy outcomes, and the
use of such technologies in equilibrium.
One implication of the model and logic presented here is that different countries might have
different degrees of institutionalization in the way in which collective decisions are made. Countries
do present a great deal of variation in the way policymaking is conducted, including differences in the
degree to which formal institutions such as Congress and political parties are the central locus of
1
In the paper, “political institutions” refers to Constitutional and electoral “rules of the game” that
define what political actors can do, as well as to the actors and arenas in charge of official
policymaking (Congress, the Courts). The “institutionalization” of policymaking refers to the degree
to which arenas such as Congress are indeed the places where societal actors focus their energies when
trying to influence policy. One can (as we do in our model) assign a great deal of structure to actions
outside the formal institutions of government and in a strict sense call such actions highly
“institutionalized.”
That said, most people would agree to call bargaining in Congress more
institutional behavior than burning tires in the street (no matter how structured the latter activity might
be). The founding reference on political institutionalization is Huntington 1968; a seminal application
(to the U.S. Congress) is Polsby 1968. See also Blondel 2006 and the discussion below.
3
programmatic demands by socioeconomic actors, and (conversely) the degree to which socioeconomic
interests use, instead, alternative political technologies to influence policymaking.2
Actors will concentrate their political investments and actions in the place where they believe
crucial decisions are made. Institutions reflect past investments; summarize information, beliefs and
expectations; and incorporate self-reinforcement effects. The relevance of specific institutional arenas
for policymaking is a self-reinforcement phenomenon, potentially subject to multiplicity if equilibria. 3
When Congress and the political party system are effective conduits of preference aggregation and
political bargaining, various relevant actors place their bets (investments) in those institutions, most
citizens believe that those are the spaces where relevant decisions are made, and this whole logic
reinforces and become self-fulfilling. On the contrary, if such institutional arenas are not taken
seriously and everybody knows that the way of getting something out of the political system is to
blockade a road or to bribe the president, those investments in the institutionalization of Congress
and/or parties are not undertaken and the weakness of formal institutions is reinforced. Polities might
be stuck with higher or lower levels of institutionalization. While polities in more institutionalized
equilibria will behave as predicted in the “tidy” literature on political institutions and policies, polities
2
These differences are striking in comparing advanced industrial countries with poor developing
nations, but also when comparing across polities of similar level of economic development, as the
epigraph from Przeworski 2009 reflects. According to O´Donnell (1996), only in Costa Rica, Chile,
and Uruguay do the executive branch, congress, parties, and the judiciary function in a manner that is
reasonably close to their formal institutional rules, making them effective institutional knots in the
flow of political power. The collection in Stein et al 2008 reflects important variation across countries
(and some over time) in the institutionalization of policymaking in eight Latin American countries.
3
In pursuing this argument, the paper concurs with an analytical current that views institutions as
equilibrium phenomena. Our use of “institutionalization” is quite close to the game theoretic notion of
“institutions as equilibria” by authors such as Calvert (1995, 95b), Greif (2006), and Aoki (2001).
4
in less institutionalized equilibria might behave differently. (One of the factors increasing the
likelihood of less institutionalized policymaking is the asymmetry of power within formal institutions).
Our contribution is obviously not totally novel; it draws inspiration and insights from various
extant literatures. Previous contributions considering the impact of alternative political technologies on
some aspects of the workings of formal political institutions include Ellman and Wantchekon’s (2000)
study of electoral competition under the threat of political unrest, and Dal Bo and Di Tella (2003) and
Dal Bo et al’s (2006) studies of political agency models in which interest groups can harm politicians.
Przeworski 2009 is an insightful recent effort motivated by the same questions and Latin American
observations as this paper.4 The literature on democratic consolidation is concerned with similar issues
at a higher level, where the variable of concern is democracy itself;5 for instance, in the words of
Adam Przeworski, “if democracy is to be consolidated, distributive conflict must be institutionalized:
all major political forces must channel their economic demands through the democratic institutions
and abjure other tactics” (1992: 127). Humphreys 2001 analyzes allocations when the enforcement of
bargains cannot be taken for granted; Humphreys defines such environments as “weakly
institutionalized.”6 There are very rich literatures studying contentious political actions and political
violence, but only recently have these types of political actions started to be studied in an integrated
4
Bueno de Mesquita et al 2003 is an ambitious effort that also goes beyond standard assumptions in
order to encompass a (much) broader set of situations than usually studied in the institutional
literature. Our discussion is also related to studies of economic agents’ choice whether to produce or to
fight, for instance Hirshleifer (2001) and Garfinkel and Skaperdas (1996).
5
Even though this paper touches upon general discussions, our empirical referent is modern day Latin
America: countries with intermediate levels of institutional development where electoral democracy is
not in question, but where the degree to which “places” such as Congress are effective institutional
knots in the flow of political power and policy (O’Donnell 1996) is the variable in question.
6
Levitsky and Murillo (2009) explore related issues, attempting to conceptualize the notion of
institutional weakness, and also emphasize the notion of investments in institutions.
5
manner with more institutionalized ones (such as voting, Dunning 2009 and references there). See also
Lindvall 2008 for a recent model of protests.
The empirical agenda that could move this project forward is a challenging one. One will need to
develop convincing measures of the degree of institutionalization of policymaking, and to provide
evidence of the superiority of policies generated by more institutionalized environments. One would
also need proxies for the parameters highlighted in the model. Some of these measures (such as formal
institutional rules) are easy to come by, others are less obvious. While far from offering the last word
on these issues, throughout the paper we refer to preliminary attempts at measuring institutionalization
variables and relating them to other variables in the theory. (One intriguing stylized fact that shows up
in the data is the negative correlation between measures of institutionalization and measures of the use
of alternative political technologies.) One of the objectives of the paper is to try to convince other
scholars that pursuing such empirical efforts is a worthwhile endeavor.
POLICYMAKING WITH FORMAL INSTITUTIONS AND ALTERNATIVE
POLITICAL ARENAS
Typical models of institutional policymaking such as those summarized in Persson and Tabellini 2000
study policy determination specifying political institutions as the rules of the game in which policies
are decided. Such models are quite explicit about those formal rules, but for their analysis to be valid
there is also an implicit assumption that actors in the model can only play within those rules and
institutions; this would require that alternative actions are either technologically impossible, or
somehow forbidden by (enforced) decisions at some other level. That is not a bad approximation to
study policymaking in a number of modern polities (at least in normal times), but it seems to miss a
very important part of the action in other cases.
We suggest extending those models to include alternative political actions undertaken in additional
arenas, as well as the interactions among players in different arenas. Borrowing the language in
6
Acemoglu and Robinson 2006, we suggest calling de jure political power the one assigned by formal
political institutions, and de facto political power the one assigned by the technologies of alternative
political action. For instance, if the alternative to institutionalized decision-making was the threat of
violence, such power would be conferred by the capacity to exercise violence.
The capacity to function within formal institutions as well as the capacity to be effective through
those alternative channels will be, in part, endogenous to ex ante decisions by each player. Actors will
have to make an ex ante decision on whether to invest resources in the more institutionalized arenas
such as Congress or political parties, or to invest in alternative technologies (such as violence, or
corrupting politicians). (In the specific model below the investment stage will be followed by a rather
short game, which captures in expected value what is more properly a potentially long sequence of
policymaking events given those initial investments.)
Classifying a set of various possible political actions (actions to influence collective decisionmaking) into “institutionalized” and “alternative” is to some extent an arbitrary choice. Which way
one wants to classify things in practice depends on the exact question at hand. Figure 1 presents some
examples of political actions. Some of them, towards the left of the diagram in white, are closer to the
more institutionalized end of the spectrum (forming a political party, writing to your congressman),
while others, towards the right of the diagram in dark grey, are clearly within the realm of the
alternative (road blockades, physical threats). A number of other political actions are harder to
classify, including different forms of business and union influence,7 as well as demonstrations to
7
Some forms of lobbying would fit more naturally in the “institutionalized” part of the game—as
when Grossman and Helpman (2001: 4) say, referring to the US, “according to the survey findings, the
activities undertaken by the greatest numbers of organized interest groups are those intended to
educate and persuade lawmakers of the wisdom of the groups’ position.” Other forms of business
influence are much closer to the darker part of Figure 1. See Harstad and Svensson 2009 for an
7
inspire sympathy in public opinion. Going into the darker side of Figure 1 and of reality, some
strategies we would have no doubt in classifying as “alternative,” such as the one exercised by
Colombian drug lords when offering new judges the choice between plata (bribes) or plomo (bullets) –
Dal Bó et al 2006.
<Insert Figure 1>
In the rest of the paper we work at some level of abstraction, focusing on a “black or white”
classification in which some actions take place within the institutional realm and others outside formal
institutional channels. For concreteness, we will speak as if the alternative political technology (APT)
utilized are road blockades,8 but the abstract formulation could also represent other things, such as
collective action by some economic sectors to damage the economy.
In the model it is assumed that taking the institutional road or the “street” road are indeed
alternatives.
This is a simplified way of capturing an important point of our argument, which
emphasizes the investment component of institutions. Political actions such as voting, campaigning for
your favorite candidate, writing letters to your Congressperson, participating in peaceful protests,
participating in violent protests, threatening to kill your opponents, and killing Supreme Court Justices
need not be substitutes. Under some conditions they are used jointly and under some conditions they
insightful distinction between lobbying and bribing. Strategies employed by unions vary from
extremely institutionalized actions in Parliament and the constitutionally sanctioned right to strike, to
the borderline-criminal physical intimidation utilized in some countries at some points in time.
8
We are thinking about situations as those described in Przeworski’s epigraph in which every
contentious issue finds thousands of people in the streets, as it is these days the case in places like
Argentina or Bolivia. The distinction between institutional and alternative tries to contrast a way of
aggregating preferences and processing conflict that is delegated, temporally and spatially coordinated
and relatively orderly, with one that is sporadic, haphazardous, unpredictable, it involves physical risks
and different actors doing different things at different moments and places. It is natural to imagine that
these two modes of participating in politics lead to different outcomes.
8
are used separately by different actors in different political equilibria.
Pending a more general
understanding of the conditions under which different actions are taken jointly or separately, this paper
emphasizes mechanisms (such as investment under budget or time constraints) that make these actions
alternatives.9 Figure 2 suggests that citizens in Latin America tend to view contacting their
representatives in Congress and protesting in the streets as alternative political options.
<Insert Figure 2>
The actors in our game, then, will have to make a prior decision of whether to enter the more
formal or the less formal arenas of political action. These decisions at the level of individual actors are
then aggregated in equilibrium. Political actors take different roads depending on what the
environment has to offer them, and their aggregate choices have an effect on that environment. This
bidirectional causality is at the heart of the issues of reinforcement and multiplicity of equilibria,
which we elaborate upon after presenting the model.
A MODEL: CONGRESS OR THE STREET?
Any model within the framework suggested will need to specify three components that will constitute
sets of exogenous parameters for comparative statics purposes: (1) policy problem (mapping from
feasible policy vectors to utilities of the players, which subsumes aspects of the socioeconomic
structure and available policy technologies); (2) political institutions; and (3) alternative political
technologies and protocols of interaction among different political arenas. In the example here, the
9
In the specific model presented below, even if one were to allow actors to allocate their unit of
investment in a continuous manner across the two political arenas, the choice would still be a corner
concentration on one of the two. A more substantive extension, discussed in the conclusion, would
incorporate a third “opportunity cost” dimension, allow for further actor heterogeneity, and generate
richer predictions at the individual level (who participates and how) and in terms of country equilibria.
9
policy problem (1) is a standard allocation problem of splitting a pie of size 𝑋, and political
institutions (2) are a simple variant of the well-known legislative bargaining model set forth by Baron
and Ferejohn (1989).10 Any assumption about (3) will embed considerations about mappings from
alternative actions to utilities, as well as about collective action, commitment technologies, and
allocation protocols in informal arenas. Modifying any assumption made at this level constitutes part
of the comparative statics agenda we want to promote. For brevity and concreteness in this paper the
APT is described as a road blockade (“going to the street”). Participating in a road blockade has an
individual cost 𝑐. If at least 𝑇 players protest they can (credibly) threaten to inflict a damage of size 𝐾
to the economy, with 2 ≤ 𝑇 and 0 < 𝐾 < 𝑋. If less than 𝑇 people engage in the blockade, they cannot
cause any damage. This simplified damage technology is represented in Figure 3, where 𝑚 stands for
the number of people who choose this alternative. This formulation captures in a stark manner some
properties of the framework, in particular it embeds a scale economy / strategic complementarity that
will be the seed of multiplicity of equilibria.11 It is also assumed that “The Street” acts as a unified
actor, ignoring collective action issues among those blocking the streets (after the threshold 𝑇)
(assumption A1), it splits whatever it receives equally among its members (assumption A2), and it is
endowed with a commitment technology: it can commit ex ante to cause damage or not in response to
specified actions by players in the formal (Congress) arena (assumption A3). The latter assumption,
10
It is more straightforward to think about this example as direct democracy, where the players who
could go to Congress are the citizens themselves. Alternatively, players can be understood as the
underlying constituencies, each of which could be represented in Congress by a perfect agent (if the
constituency chooses to invest in that representation).
11
Parameters as 𝑐 or 𝑇 will depend on various issues, including “repression technologies” in place.
10
common in the related literature (Ellman and Wanchtekon 2000, DalBo et al 2006),
12
is imposed to
avoid a lengthy analysis of the conditions under which such threat would be credible, which would
require adding more structure to the extensive form or going to a repeated game altogether to include
considerations of reputation. In the context of non-repeated games the threat of damage would be
credible when it constitutes a weakly dominant strategy in the final move, something that could be
easily adjusted in our model.13 Furthermore, there is an important literature on fairness, altruistic
punishment and spitefulness, which argues that punishments are indeed instrumented even when they
are costly for the punisher in narrow self-interest terms.14 Those results, often obtained in rather
aseptic experiments, are even more likely to obtain in the specific context of our focus: once the
piqueteros are out there in the street, it might be easier to enforce “damage in case of unfulfilled
demands” than “no damage in case of fulfilled demands”.15 There are also some technologies, such as
permanent squatting, where it is more costly to stop inflicting costs to others (leaving) than to continue
doing so (staying).
<Insert Figure 3>
12
Even in Acemoglu and Robinson 2006, in a different context, exogenous assumptions are imposed
in order for the threat of revolution to be credible.
13
That would be more likely if the damage protesters can inflict does not affect them. In the specific
formulation in this paper that would be the case if the damage only affects the payoff of players in
Congress. Given that we have imposed A3, for notational brevity we will assume that Congress offers
protesters a share of the final pie, but it would be trivial to modify that and let them received a fixed
amount, in which case causing damage would indeed be weakly dominant.
14
See for instance Bahry and Wilson 2006, Henrich et al 2004, and Levine 1998.
15
As Maurice Thorez, longtime leader of the French Communist Party remarked in 1936, “One has to
know how to end a strike” – Przeworski 2009. These issues have been studied in a literature on
collective behavior, for instance Marx and McAdams 1994.
11
Let 𝒩 = {1,2, … 𝑛} be the set of players. The economy is subject to the aggregate constraint
∑𝑛𝑖=1 𝑥𝑖 ≤ 𝑋. Each player maximizes his or her piece 𝑥𝑖 ≥ 0, net of a possible cost 𝑐 that they pay in
some cases. Each agent is endowed with 𝑝𝑖 units of (ex ante) de jure political power, with 𝑝𝑖 < 𝑝𝑖+1 ,
and ∑𝑛𝑖=1 𝑝𝑖 = 1. The timing of the game, represented in Figure 4, is as follows.
<Insert Figure 4>
Investment stage:
Stage 1: Each player (simultaneously) chooses an action 𝑎𝑖 ∈ {0,1}, where 𝑎𝑖 = 1 means going to
Congress and 𝑎𝑖 = 0 going to the street, which carries a cost 𝑐. This action represents an investment
that enables the player to become effective at one of the two arenas. ∑ 𝑎𝑖 will be the aggregate
investment in formal institutions, capturing the degree of institutionalization, which we discuss after
solving the model. Let 𝑚 be the number of players who go to the street, and 𝑛 − 𝑚 the number of
players who go to Congress. Let ℳ denote the set of players in the street, and 𝒩\ℳ the set of players
in Congress.
Policymaking stages:
Stage 2: Among those that went to Congress, Nature chooses an agenda setter 𝜇. Each player 𝑖 in
Congress has an ex post probability ∑
𝑝𝑖
𝑗𝜖𝒩\ℳ 𝑝𝑗
of being recognized as the agenda setter.
Stage 3: The players in the street can (credibly) threaten to cause damage 𝑑 as a function of the share
𝑧 of the total pie received by each of them (𝑍 = 𝑚𝑧 denotes the total amount received by those in the
street). Let 𝑇 ≥ 2 be a threshold of critical mass for action in the street. If 𝑚 < 𝑇, the amount of
damage they can cause is zero, otherwise it is 𝐾.16
16
Let 𝑑(𝑧) be the possible damage function, or feasible damage contract. If 𝑚 < 𝑇, then 𝑑(𝑧) = 0 for
all 𝑧. If 𝑚 ≥ 𝑇, 𝑑(𝑧): [0,1] → {0, 𝐾}. We assume for brevity that the range of the damage function can
only take discrete values, but nothing of substance will change if we let a continuous range. If 𝑚 = 0
12
Stage 4: After listening to the threat of The Street, the agenda setter in Congress (𝑖 = 𝜇) proposes an
𝝁
allocation of the pie, a vector of shares 𝑆 𝝁 (𝑑) = {𝑠𝑖 }𝑖𝜖𝒩 , subject to ∑𝑛𝑖=1 𝑠𝑖 ≤ 1 and 𝑠𝑖 ∈ [0,1]. For
brevity we introduce the ex post distributional assumption A2 as an ex ante constraint on the agenda
setter’s proposal, by requiring at this stage that the share for each person in the street is the same, that
is 𝑠𝑖 = 𝑧 𝝁 ∈ [0, 1⁄𝑚] for all 𝑖 ∈ ℳ.
𝝁
Stage 5: The 𝑛 − 𝑚 players in Congress vote on 𝑆 𝝁 . Let 𝜙𝑖 ∈ {0,1} denote the choice of
congressperson 𝑖, where 0 and 1 respectively represent voting against and for 𝜇’s proposal. We assume
that players in Congress vote only on the basis of their individual share, so that 𝜙𝑖 (𝑠𝑖 ): [0,1] → {0,1}.
To avoid some sources of multiplicity not germane to the objectives of this paper, we assume that in
case of indifference, 𝜙𝑖 = 1 (assumption A4).17 The outcome of the legislative process, 𝑆, will be:18
𝑆𝝁
1
𝑖𝑓 ∑ 𝜙𝑖 ≥ (𝑛 − 𝑚)
2
𝑖𝜖𝒩\ℳ
𝑆=
{𝑠𝑖 = 0}𝑖𝜖𝒩
{
1
𝑖𝑓 ∑ 𝜙𝑖 < (𝑛 − 𝑚) .
2
𝑖𝜖𝒩\ℳ
this stage becomes irrelevant, and the rest of the game reduces to a traditional legislative game with
only stages 2, 4, and 5.
17
In this way two sources of multiplicity are eliminated. On the one hand, we are considering only
equilibria in which voters do not choose weakly dominated strategies, having them always act as if
they were pivotal. On the other, we make the payoff given to members of the winning legislative
coalition converge to their status quo payoff.
18
This very simple formulation of the legislative bargaining problem is used for brevity. The main
logic pursued in the study does not depend on any specific extensive form of the institutionalized
game; the crucial individual decisions and equilibrium features depend on comparing the expected
values of participating in the institutionalized decision-making process with that of using alternatives.
(We are indebted to Massimo Morelli and an anonymous referee for this point.) Nothing of substance
will change if the legislative s.q. vector of shares is allowed to be non-zero. This could be an
additional source of heterogeneity, which is captured here in an ex ante manner by the vector of
𝑖 ’s.
13
Stage 6: Those in the street implement the damage announced in 3.
Stage 7: The allocation is implemented. If no damage was caused, each player receives 𝑥𝑖 = 𝑠𝑖 𝑋. If
damage was caused, each player receives 𝑥𝑖 = 𝑠𝑖 (𝑋 − 𝐾).
We are implicitly assuming complete information, a common first modeling step. Interesting
additional insights could be obtained by adding asymmetric information. If some aspects of the
collective action or damage capacity of protesters were stochastic and unobserved by institutionalized
actors, we might have equilibria with occasional outbreaks of violence, a realistic feature shared by
models of incomplete information in oligopoly, war, and strikes.
SOLVING THE MODEL
The model is solved by backward induction.19 Given the assumption of commitment by protesters,
stages 6 and 7 are mechanically implemented after decision 4-5 is made. The latter is a standard
legislative bargaining game with a small twist. It is easy to show that the agenda setter will give 0 to
all other players in Congress and allocate the whole pie (except anything given to the protesters) to
herself.20 The novel question is how much to give to the protesters. Clearly, if 𝑚 < 𝑇, the agenda
19
The solution is proven formally in the Appendix.
20
More generally, the allocation among the players in Congress could be defined as 𝑠𝑖 = 0 for any 𝑖
not part of the winning coalition, 𝑠𝑖 = 𝜀 for those 𝑖’s in the winning coalition other than the agenda
setter, and the rest for the agenda setter. We follow the standard convention of letting ε go to zero, and
of assuming that players who are indifferent between two actions at zero will chose the one they would
have chosen for 𝜀 > 0.
As stated, the fact that the other members of Congress get zero is just for
simplicity. What actually matters for the relevant decisions is their expected utility in Congress before
the agenda setter was selected by nature. This is a reduction of a richer intertemporal structure (where
those “included” will eventually get their share) into a one-period model.
14
setter will give nothing to protesters. If 𝑚 ≥ 𝑇 , then the amount she will allocate to the protesters will
depend on the amount 𝑍 that they request. If they request 𝑍 > 𝐾, the agenda setter will give them
nothing, since it is better to keep (𝑋 − 𝐾) than (𝑋 − 𝑍). If they request 𝑍 ≤ 𝐾, then the agenda setter
will grant them 𝑍. This response will lead the protesters (in stage 3) to request exactly 𝑍 = 𝐾, since
they will get nothing if they go above 𝐾, and 𝐾 is preferred to anything below it. Each protester will
receive 𝐾/𝑚.
Given that stage 2 is a move by nature, this brings us to the key choice of arenas in stage 1. Each
player has to decide between the two arenas, given his expectation of what he will receive if he goes to
Congress and if he goes to the street, taking as given the choices of all the other (𝑛 − 1) players.
Finding equilibria to the whole game, then, consists of finding the Nash equilibrium to these 𝑛
individual decisions over 𝑎𝑖 . We present the results in the following proposition, which is presented
more formally and proven in the Appendix.
PROPOSITION 1:
(a) There always exists a Full Institutionalization Equilibrium (FIE) in which everybody is in
Congress and no alternative arenas are used.
(b) For some parameter values, there is also a Low Institutionalization Equilibrium (LIE) in which
𝑚∗ actors go to the street.
The existence of the FIE is guaranteed by the fact that if only one player chooses to be in the street
in order to set a blockade, he will be unable to credibly threaten to cause any damage, and hence
receive zero from the agenda setter, getting a final payoff of – 𝑐, which is less than what he would
expect to receive in Congress. No one would want to deviate unilaterally from a FIE.
Figure 5 provides an intuitive representation of the LIE. The figure plots the value of being in
Congress (the upward-sloping curve) and the value of being in the street (the downward-sloping
15
curve) from the point of view of an individual player as a function of his or her probability of being
the agenda setter (𝑝𝑖 ), under the assumption (which is true in the type of LIE we select, see Appendix)
that all players of lower 𝑝𝑗 will be in the street, and all players of higher 𝑝𝑗 will be in Congress. Let
𝑚∗ be the highest integer to the left of the point where these two curves intersect. Players 1 to 𝑚∗
(those with the lowest ex ante probabilities of being selected agenda setter in Congress) will go to the
street, and players 𝑚∗ + 1 to 𝑛 will go to Congress. The result if fairly intuitive: players with weaker
chances of getting what they want in formal political institutions are the ones more likely to utilize
alternative political technologies.
<Insert Figure 5>
The low institutionalization equilibrium takes us back to the world of the models described in the
introduction, where the institutionalized part of the game is the only relevant one – public policies are
decided in formal arenas (and actors invest in institutional capabilities, as we argue in the next
section). The low institutionalization equilibrium, on the other hand, will be characterized by the use
of alternative political technologies and by weakness of formal institutional arenas.
ON INSTITUTIONALIZATION (AND MULTIPLE EQUILIBRIA)
The simple static model above provides glimpses into more general issues we want to raise regarding
different degrees of institutionalization of policymaking across polities. In particular, the model
captures in a very compressed way (through the initial investment, and through the beliefs about other
people’s actions implicit in different equilibria) some richer dynamics characterizing institutions and
institutionalization. Institutions reflect past investments, summarize information, beliefs and
expectations, and incorporate self-reinforcement effects.
Recognizing that there is no agreed terminology with respect to institutions and
institutionalization, we would say that “the U.S. Congress is more institutionalized than the Argentine
16
Congress”. The U.S. Congress is a central institution in U.S. policymaking; its members exhibit
remarkable longevity and specialize in committees; it engages in considerable oversight of the public
bureaucracy, and it is the focus and main entry point of political action by any interest group
attempting to influence American policymaking.21 The Argentine Congress does not share any of these
features, in spite of the fact that Argentina has a constitutional structure similar to that of the US
(Jones et al, 2002). The comparison can also be taken among countries of similar levels of
development: the Chilean legislature plays a much more important role in the policymaking process of
that country than does its Argentine counterpart (Stein et al, 2008). These different roles in
policymaking are associated with various measures of legislators’ and legislatures’ capabilities, which
have a clear investment component or that reflect the beliefs of various actors: longevity of legislative
careers, legislators’ education, specialization in policy committees, resources available for policy
analysis, esteem in the eyes of the public, and appreciation of Congress as an important place in
politicians’ careers – Saiegh (2010) builds measures of such characteristics.
When Congress and the political party system are effective conduits of preference aggregation and
political bargaining, various relevant actors place their bets (investments) in those institutions, most
citizens believe that those are the spaces where relevant decisions are made, and this whole logic
reinforces and becomes self-fulfilling. On the contrary, if such institutional arenas are not taken too
seriously and everybody knows that the way of getting something out of the political system is to
blockade a road (or to bribe the president), those investments in the institutionalization of Congress or
21
Polsby 1968 is the authoritative reference on the institutionalization of the U.S. Congress. Our
notion of institutionalization shares Polsby’s suggested components of a definition of
institutionalization (focused on internal organization), but adds other elements that reflect its
importance in the broader equilibrium, including the beliefs about its relevance, along the lines of the
“external considerations” emphasized by Blondel (2006).
17
parties are not undertaken and the weakness of formal institutions is reinforced. The logic of selfreinforcement of institutions is central to the concerns of various strands of institutionalism in the
social sciences.
Scholars in the game theoretic tradition (as well as common usage) refer to institutions “as rules”,
and also “as equilibria”. The standard “Northian” definition has become “the rules of the game in a
society, the humanly devised constraints that shape human interaction”. A number of scholars have
turned to a conception of institutions as equilibrium phenomena. It is ultimately the behavior of others
rather than the rules themselves that induces each person to behave (or not) in the way prescribed by
the rules.22 The model above captures these two distinct usages of the notion of institutions in a simple
manner. On the one hand, it has political institutions as rules in the allocation of de jure political
power (the rules of the Baron-Ferejohn game). On the other, it generates different degrees of
institutionalization in equilibrium, as captured by ∑ 𝑎𝑖 , the aggregate investment in formal institutions
(which is a fixed point to the beliefs of all actors with respect to the actions of everybody else).
Even within the context of a simplified static model, the result of having multiple equilibria
(with different equilibria characterized by its degree of institutionalization) captures an important
aspect of institutions. It also opens the door for the very relevant considerations so keen to historical
institutionalists (Pierson and Skocpol 2002, Thelen 1999) about the way in which specific temporal
sequences, critical junctures, and historically grounded events matter to explain different trajectories.23
22
This view has a long pedigree, but in its modern formulation has been pioneered by Greif (see for
instance 2006), Calvert (1995, 1995b) and Aoki (2001).
23
If the world as seen with game theoretic eyes did not include multiplicity of equilibria, it would be
impossible for history to matter. Additionally, the self-reinforcement effects that in a static game
theoretic equilibrium appear in a stark and compressed manner, in reality take place in historical
sequences, as emphasized by historical institutionalists.
18
Multiplicity of equilibria arises in the model of this paper due to a stark strategic complementarity
introduced by the shape of the damage technology. Beyond the specificities of our formulation, an
extreme simplification of threshold models of collective action (Oliver 1993, Lohmann 1994,
Granovetter 1978), there are a number of reasons to expect strategic complementarities, reinforcement
effects, and hence the possibility of multiple equilibria in institutionalization. Unconventional forms of
political participation tend to be chosen where institutions are of lower quality, but those very
decisions feed back into the weakness of formal institutions. Strategic complementarities are likely to
be present in the action choices of players with regards to any given APT (as already present in the
model),24 but also with regards to the actions and investment choices within any specific formal
institution such as Congresses. Furthermore, there are reasons why alternative activities of one type
are likely to induce alternative activities of other types. For instance, Trucco 2009 shows a
complementarity between “bribes by the rich” and “street protests by the poor” as countervailing
forces that tend to happen together in polities with weaker political institutions. Similarly, there tend to
be important systemic effects across various institutionalized arenas. For instance, Stein and Tommasi
2007 and Scartascini et al 2009 show a positive correlation among measures of institutional strength in
various domains (Congress, party system, Judiciary, the bureaucracy) across countries.
Even though not quite modeled explicitly in this paper, having different degrees of
institutionalization of policy making has important welfare consequences. In our simple model the cost
of incomplete institutionalization is just 𝑐𝑚∗ . There are various reasons (some of which could be
incorporated in a richer dynamic formulation of this model) why more institutionalized collective
decision-making arenas tend to lead to better policies and better outcomes. Investing in technologies
that increase strength and capabilities in the institutionalized game (such as hiring PhDs in Public
24
Machado et al 2009 finds that the likelihood that any individual would participate in street protests
is increasing in the total number of people protesting.
19
Policy) or in technologies that increase strength and capabilities in alternative arenas (such as hiring
goons), are likely to have a different impact on the strength and productivity of formal institutions, as
well as on the quality of the output the polity generates. Scartascini et al 2009 shows that more
institutionalized policymaking environments are associated with better public policies, which in turn
induce better economic outcomes. One of the reasons behind that result is likely to be that more
institutionalized policymaking arenas provide a better structure for exchange of information and for
the agreement and enforcement of intertemporal cooperation. According to Pierson (2004: 107)
“political institutions can serve to coordinate the behavior and expectations of decentralized actors
(Carey, 2000) and to facilitate bargaining by creating monitoring bodies, issue linkages, and
mechanisms for making credible commitments.”
Multiplicity of equilibria has important implications in terms of theory, in terms of interpreting and
analyzing empirical evidence, and in terms of thinking about possible practical recommendations.
Before closing the section we want to say a word regarding the implications of this analysis for crosscountry econometrics on the effects of formal political rules. The effects of constitutional rules on
policymaking and performance might be conditional on the type of equilibrium achieved by each
polity. The standard predictions in the literature might apply for countries at the full
institutionalization equilibrium, but not necessarily for countries with low institutionalization. Caruso
et al 2010 explores this issue by clustering the countries into those with high and those with low levels
of institutionalization.25 Using those clusters, it reproduces within each cluster the empirical exercises
in Persson and Tabellini 2003. All of the results in Persson and Tabellini 2003 are replicated for the
25
The clustering of countries is performed using as proxies for the degree of institutionalization
available measures of congressional capabilities, party system institutionalization, bureaucratic
capabilities, and judicial independence.
20
high-institutionalization sample, but almost none of the results are significant for the lowinstitutionalization sample of countries.
COMPARATIVE STATICS / EMPIRICAL PREDICTIONS
One important question is what are the implications of equilibrium multiplicity for attempting to map
the predictions of the model into empirical evidence. Fortunately, the simple structure of the model
enables us to make some comparative static predictions in spite of having multiple equilibria. Under
the maintained assumptions of the model, the FIE always exists, while the LIE exists for some
parameter values and not for others. This means that for some parameter values the FIE will be the
only equilibrium possible (call that set of parameters F), while for other parameter values both types
of equilibrium are possible (call that set of parameters B). If the real world were fully described by this
model, then we would expect that in polities characterized by a vector of parameters belonging to F
we should observe the characteristics associated with the FIE, while in polities characterized by a
vector of parameters belonging to B we could observe either of the equilibria. Assigning in that latter
case a non-zero probability to each of the equilibria will be enough to generate empirical predictions.
If conducting comparative statics on any given parameter increases the set of other parameters for
which LIE is an equilibrium, then we will say that increasing that parameter increases the possibility
of observing the Low Institutionalization Equilibrium (and its associated properties).
It turns out that in the model any comparative statics exercise that increases the possibility of LIE,
also increases the number 𝑚∗ of people going to the streets within the LIE (see Appendix), so that the
derivative of 𝑚∗ with respect to any parameter is a sufficient statistic for the way in which that
parameter affects the degree of institutionalization in either interpretation. Using that logic, it is easy to
show the comparative static results summarized in Table 1, which we state as conditions under which
non-institutionalized politics are more likely to be sustained.
21
<Insert Table 1>
Not surprisingly, the probability of being in the LIE increases as the use of APTs becomes less costly,
as the potential damage APTs can cause increases, and as the wealth of the economy decreases. More
interestingly, the probability of being in a Low Institutionalization Equilibrium increases as the
distribution of de jure political power becomes more asymmetric. As the vector of 𝑝𝑖 ’s becomes more
asymmetric, more actors at the lower end will find participation in formal political institutions less
appealing, inducing more street action and lower institutionalization. The model predicts that those
less favored by de jure political power are more likely to engage in non-institutionalized actions,26 and
furthermore, that the more asymmetric political power is, the greater the numbers taking the outside
option and the lower institutionalization in equilibrium.
We have undertaken some preliminary steps towards the empirical study of some of the
correlations predicted by the theory, using international measures of institutional quality which we
believe serve as proxies for the strength of policymaking institutions as understood in the paper. These
measures, taken from Scartascini et al 2009, include proxies for the strength and involvement of the
legislature in policymaking, some aspects of party system institutionalization, and bureaucratic
capabilities. These proxies for institutional strength have been combined with information on a
number of political activities from the Cross-National Time-Series Data Archive, often combined
under the heading of “Internal Conflict,” which we take as rough measure of the APTs emphasized in
this paper. The logic presented above implies that in equilibrium there should be a negative
26
This result has a flavor of “exclusion of the poor”. Yet, the result is more general than that, and it
applies to any case in which the de jure power of a relevant set of political actors tends to underrepresent them in comparison to their ability to put collective action together and threaten economic
disruption. The very visible 2008-2009 demonstrations of rural producers in Argentina against large
increases in export taxes by the government are a case in point. Those demonstrators were not the
dispossessed, but an economic sector underrepresented in the Argentine political system.
22
correlation between the strength of political institutions and the use of alternative political. Our
measures of institutional strength tend to be negatively correlated with the measures of conflict, with
the coefficients being statistically significant in univariate and multivariate regressions with standard
controls such as GDP per capita, regional dummies, and level of democracy. Figure 6 shows one of
these correlations between internal conflict and a combined measure of institutional strength.
<Insert Figure 6>
The argument presented in this study has also implications for correlations among institutional and
conflict variables with some characteristics of policy outcomes; for instance, a positive correlation
between weaker institutions, more use of alternative political technologies, and public policies that
favor some actors disproportionately. Political institutions that generate more biased policy outcomes
are likely to lead to increased use of alternative political technologies and this induces weaker
investments in the institutionalization of policymaking. We have also found a positive correlation
between the use of APTs and a measure of biases in government policy called favoritism compiled by
World Ec. Forum’s GCR, as well as a negative correlation of favoritism and institutionalization.27
CONCLUSIONS AND EXTENSIONS
The objective of this paper is to suggest a framework (a modeling strategy) to articulate insights from
the rich literature on the effects of political institutions on policy into a broader view of policymaking,
by endowing political actors with a larger action space. Such an effort could prove particularly useful
for the study of policymaking in developing democracies, as well as for integrating such analyses
across countries of different levels of economic and institutional development.
27
This is consistent with observations in O’Donnell (1996) who characterizes many Latin American
democracies as having both weak institutions and particularism (a combination that he defines as
“another institutionalization”).
23
The model presented in this paper has the following implications. 1. Different countries have
different degrees of institutionalization in their policymaking. 2. There is multiplicity of equilibria -this allows similar countries to be stuck at different levels of institutionalization, self-reinforcing
dynamics, and the possibility of equilibrium switches. 3.
The
likelihood
of
institutionalized
policymaking increases as the cost of alternative political actions increases, as the damage these
alternatives can cause decreases, and as the economy becomes wealthier. 4. If the distribution of de
jure political power is very asymmetric, it is more likely to observe use of alternative political
technologies as well as low degrees of institutionalization.28
There are various pending tasks in the agenda described in the paper, such as enriching various
aspects of the model to permit comparative static exercises on economic structure, APTs, and formal
political institutions. To begin with, one can give more structure to the economy and more specific
characteristics to the actors in the policy game. The APT in the model presents a fairly “flat”
topography. Costs and potential effects of various forms of collective action are distributed in much
more specific manners. The various 𝑐(𝑖)’s of different actors could be a function of who else is
participating in that activity, facilitating collective action across particular sets of agents, such as
workers in sectors with high “damage capacity”. More generally, there are various structural,
historical, and perceptual factors that affect specific forms of collective action. 𝑐(𝑖)’s as well as
potential rewards from such activities will be a function not only of who else protests, but also of the
history of organization of such movements, and of the beliefs about the legitimacy of the protest by
other actors who might matter for the degree of repression.
28
Ongoing work indicates two additional implications: (5) High costs (for instance due to agency
problems) in using formal institutions can lead to the use of APTs and to low institutionalization; (6)
there are strategic complementarities across the use of different APTs.
24
An interesting extension would consist of a richer set of APTs with different socioeconomic actors
having differential access to each of the technologies. Various technologies used extensively by the
well-off in Latin America give them privileged access to policymaking. Trucco (2009) provides a start
in that direction, and her results suggest the presence of a strategic complementarity between the uses
of bribes and of road blockades. “Bribes by the rich” and “protests by the poor” can be seen as
countervailing forces that tend to happen together in polities with weaker political institutions. (Such
result is related to the asymmetry result 4 above).29 Other extensions might allow actors to invest in
both institutions and alternative technologies and to use the outside threats in their institutionalized
negotiations. Including also a “third activity” or opportunity cost, would allow studying the form of
political participation by different actors across countries and relating that to institutionalization.
Natural extensions might come from moving away from our simple legislative bargaining
formulation and fishing in the abundant pond of formal models of political institutions for richer
institutional comparative statics. For instance, what would be the effect of alternative electoral rules
(proportional versus majoritarian) on the degree of institutionalization of policymaking? (One might
hypothesize that PR systems, allowing a better representation of minorities, might lead to more
symmetric distributions of de jure political power and hence to more institutionalized behavior.) What
would be the effect of regime type (parliamentary of presidential) on the degree of institutionalization
of policymaking? To deal with such questions, one will need to move in the direction of representative
democracy, with models that permit exploring the electoral connection and agency issues.
Another line of useful extensions might come from moving to more dynamic specifications. The
paper couches some of its interpretation in terms of the “dynamics of institutionalization”. This is just
a heuristic that captures the extent to which institutionalization is a function of shared expectations.
29
Such an extension’s empirical predictions – bribes positively correlated with other APTs as protests,
and negatively correlated with institutionalization -- seem consistent with preliminary evidence.
25
(Almost all static game theoretic equilibria are simplified depictions of processes that indeed take
place over time.) Another sense of dynamics would be the question of how a country arrives to (or
switches towards) a particular equilibrium. Addressing such transitions is an important pending task,
which we plan to undertake via theoretically informed narratives of some Latin American experiences.
Also, the model has treated constitutional rules as exogenous, a natural first modeling step (Diermeier
and Krehbiel, 2003). This can be embedded in a richer game in which such rules are chosen at an
earlier stage, having our model here as a continuation subgame. Such modeling strategy would be
quite complementary to the Acemoglu-Robinson (2006) logic, providing a bit more structure to their
(unmodeled) threat of collective action due to temporary de facto power.
There are various possible implications from the results here and from the eventual next steps. To
highlight one of them (in light with the results in Caruso et al 2010), the existence of different levels of
institutionalization may imply that some influential research in “political economics” is not necessarily
applicable to every type of country. Studies on the impact of political institutions on policies should
incorporate the possibility that political rules might have different effects depending on the arenas in
which policies are determined in each polity; as well as the possibility that the arena itself might
depend on political rules. This should also raise a flag for those in the business of recommending
institutional reforms for dealing with failing policies.
APPENDIX
Solution of the Game (and proof of Proposition 1)
The parameters of the game are 𝑛, 𝑋, 𝐾, 𝑇, 𝑐, {𝑝𝑖 }𝑛𝑖=1 . Let Γ = [ℳ, 𝒩\ℳ] denote a partition of the set of players.
Take any Γ and any realization of the agenda setter 𝜇 consistent with that Γ, and call 𝐺(Γ, 𝜇) the game from
stage 3 onwards with partition Γ and agenda setter 𝜇. The equilibrium play of that game will generate payoffs
𝑉𝑖 (Γ, 𝜇) for each player 𝑖. We solve the game in two steps: first finding the (subgame perfect) equilibrium for
26
any 𝐺(Γ, 𝜇), and then the Nash equilibria to the venue choices of all players in stage 1 with 𝐸𝜇 [𝑉𝑖 (Γ, 𝜇)] as
continuation payoffs. We focus on pure strategies throughout.
Step 1. Solving 𝐺(Γ, 𝜇).
𝜇
{𝑑∗ (𝑧), 𝑆 𝝁∗ (𝑑), {𝜙𝑗∗ (𝑠𝑗 )}
𝑗𝜖𝒩\ℳ
} constitute a Subgame Perfect Equilibrium for 𝐺(Γ, 𝜇) if:
𝜇
𝜇
1. For all 𝑑(𝑧), 𝑆 𝝁: 𝜙𝑖∗ (𝑠𝑖 ) maximizes payoffs to 𝑖, given {𝜙𝑗∗ (𝑠𝑗 )}
𝑗𝜖𝒩\ℳ\{i}
𝜇
2. Given {𝜙𝑗∗ (𝑠𝑗 )}
𝑗𝜖𝒩\ℳ
𝜇
3. Given {𝜙𝑗∗ (𝑠𝑗 )}
𝑗𝜖𝒩\ℳ
, for all 𝑖𝜖𝒩\ℳ.
, for all 𝑑(𝑧): 𝑆 𝝁∗ (𝑑) maximizes payoff to the agenda setter µ.
and 𝑆 𝝁∗ (𝑑): 𝑑∗ (𝑧) maximizes payoffs to The Street.
We solve 𝐺(Γ, 𝜇) by backward induction from stage 5 to stage 3.
Stage 5: given the status quo payoffs of zero, assumption A4 assures that any proposal greater or equal to zero
will be accepted by all members of Congress. This implies 𝑆 = 𝑆𝜇 .
𝝁
Stage 4: given 𝑆 = 𝑆𝜇 , 𝑠𝑖 = 0 ∀𝑖 ∈ 𝒩\ℳ\{μ}, since the agenda setter will try to keep as much of the pie as
possible for himself. With regards to how much to give to The Street, the decision depends on the number of
players in the street, as well as on the amount they request. If 𝑚 < 𝑇, given 𝑑(𝑧) = 0 ∀z, 𝑧 𝜇 = 0. If 𝑚 ≥ 𝑇,
the amount given to The Street by the agenda setter will depend on 𝑑(𝑧). The agenda setter will pick the lowest
𝑧 such that 𝑑(𝑧) = 0, and compare giving at 𝑧 to all players in the street, to giving them 0 and suffering the
max{{1 − 𝑚 ×
damage 𝑑(0). Hence, his problem is
𝑧̂ }𝑋, 𝑋 − 𝑑(0)},
𝑧
(1)
where 𝑧̂ = 𝑚𝑖𝑛{𝑧 ∈ [0,1]|𝑑(𝑧) = 0}.
Stage 3: (The problem for The Street is trivial when 𝑚 < 𝑇; we focus on 𝑚 ≥ 𝑇.) The Street’s problem consists
of maximizing 𝑧 subject to (1). Given that in the second option they get zero, their problem reduces to
announcing a function 𝑑(𝑧) that: (i) minimizes 𝑋 − 𝑑(0): and (ii) minimizes {1 − 𝑚 × 𝑧̂ }𝑋, subject to {1 −
𝑚 × 𝑧̂ }𝑋 ≥ 𝑋 − 𝑑(0). (i) implies 𝑑(0) = 𝐾. (ii) implies
𝐾
𝑚𝑋
= 𝑧̂ . Any function 𝑑(𝑧) that satisfies those two
conditions constitutes a solution to The Street’s problem. For brevity we focus on
27
𝐾
𝑑(𝑧) = {
𝐾 𝑓𝑜𝑟 𝑧 < 𝑚𝑋
0 𝑓𝑜𝑟 𝑧 ≥
(2)
𝐾
𝑚𝑋
The above reasoning can be summarized in the following Lemma.30
Lemma 1: under the maintained assumptions of the model, subgame perfect equilibria to G(Γ,µ) will imply the
following equilibrium play:
𝝁
(a) If m<T: the Street announces anything; the agenda setter proposes 𝑆𝜇 = {𝑠𝑖 }𝑖𝜖𝒩 such that
1
𝜇
𝑠𝑖 = {
0
𝑖=𝜇
;
𝑖≠𝜇
all players in Congress accept the proposal; no damage is caused; and the allocation is
𝑋
𝑥𝑖 = {
0
𝑖=𝜇
.
𝑖≠𝜇
𝜇
(b) If m≥T: the Street announces (2); the agenda setter proposes 𝑆 𝝁 = {𝑠𝑖 }𝑖𝜖𝒩 such that
1 − 𝐾⁄𝑋
𝝁
𝑠𝑖 = { 0
𝐾⁄
𝑚𝑋
𝑖=𝜇
𝑖 ∈ 𝒩\ℳ\{μ} ;
𝑖∈ℳ
all players in Congress accept the proposal; no damage is caused, and the allocation is
𝑋−𝐾
𝑥𝑖 = {0
𝐾⁄
𝑚
𝑖=𝜇
𝑖 ∈ 𝒩\ℳ\{μ}.
𝑖∈ℳ
From Lemma 1 it is easy to compute the continuation values of the different configurations of venue choices in
stage 1, before 𝜇 is realized in stage 2. These expected values from venue choice 𝑎𝑖 , as a function of all other
players’ choices 𝒂−𝒊 ,will constitute the elements of the payoff matrix of the stage 1 game.
Step 2. Nash Equilibria of Stage 1 Game.
There are two types of equilibria in this game: (a) Full Institutionalization Equilibria (FIE) in which all
players choose to go to Congress and a standard Baron-Ferejohn game results from stage 2 onwards, and (b)
30
In the case in which , there will be no one in Congress; but it is easy to see that cannot happen in
equilibrium, since in that case any player switching venues would become the agenda setter and
increase his payoff.
28
Low Institutionalization Equilibria (LIE) in which 𝑚∗ > 0 players go to the street. We start by showing that a
FIE exists for all parameter values under our assumptions. It is easy to see that if everybody is in Congress,
nobody wants to deviate from that, for all the feasible parameters of the game. If everybody was in Congress,
and one player was to deviate and go to the street, the number of players in the street would be one, which is
smaller than 𝑇 and hence insufficient to produce any damage. The allocation resulting from that path will lead
to 𝑠𝑖 = 0 for the deviating player. His payoff in the street will then be equal to – 𝑐, while his expected payoff in
Congress would have been 𝑝𝑖 𝑋. This deviation is never profitable. (The continuation of the game in this type of
equilibrium is as specified in part (a) of Lemma 1).
We now proceed to characterize LIE. Let Γ ∗ denote a partition being considered as possible equilibrium,
with ℳ its associated set (with measure 𝑚∗) of players in the street. In any LIE it has to be the case that no
∗
player wants to switch venues. It is easy to see from Lemma 1(a) that it is not possible to have a LIE in which
m*<T; hence in any LIE the continuation game has to be as specified in part (b) of Lemma 1.
For a partition to be an equilibrium one, it has to be the case that no player in Congress would rather be in
the street. That is, ∀𝑖 ∈ 𝒩\ℳ∗, (𝑋 − 𝐾) [∑
𝑝𝑖
∗ 𝑝𝑗
𝑗∈𝒩\ℳ
𝑝𝑖 ≥
]≥
𝐾
− 𝑐,
𝑚∗ +1
∑𝑗∈𝒩\ℳ∗ 𝑝𝑗
(𝑋−𝐾)
which implies
𝐾
−
𝑚∗ +1
(
𝑐) ≡ 𝑝𝐿 (Γ ∗ );
(3)
so that 𝑝𝐿 (Γ ∗ ) constitutes the lowest possible 𝑝𝑖 that will rather stay in Congress for an arbitrary partition Γ ∗ .
Condition (3) is necessary for Γ ∗ to be an equilibrium partition.
A similar reasoning can be applied to players in the street, only that two situations have to be considered in
that case, depending on whether the number of players in the street m∗ induced by Γ ∗ is equal to or greater than
𝑇. For Γ ∗ such that 𝑚∗ > 𝑇 to be an equilibrium partition, it has to be the case that ∀𝑖 ∈ ℳ∗ ,
𝑝𝑖
],
𝑝𝑖 +∑𝑗∈𝒩\ℳ∗ 𝑝𝑗
(X − K) [
𝐾
𝑚∗
−𝑐 ≥
which implies
𝐾
𝑝𝑖 ≤
∑𝑗∈𝒩\ℳ∗ 𝑝𝑗 ( ∗ −𝑐)
𝑚
𝐾
(X−K)−( ∗−𝑐)
𝑚
≡ 𝑝𝑈 (Γ ∗ ).
(4)
For Γ ∗ such that 𝑚∗ = 𝑇 to be an equilibrium partition, it has to be the case that ∀𝑖 ∈ ℳ∗:
29
𝐾
𝑝𝑖
− 𝑐 ≥ X[
],
𝑇
𝑝𝑖 + ∑𝑗∈𝒩\ℳ∗ 𝑝𝑗
𝑝𝑖 ≤
which implies
𝐾
𝑇
∑𝑗∈𝒩\ℳ∗ 𝑝𝑗 ( −𝑐)
𝐾
𝑇
X−( −𝑐)
≡ 𝑝𝑈 (Γ ∗ ).
(4’)
The difference between (4’) and (4) lies in the absence of the term 𝐾 in the denominator, since when 𝑚∗ = 𝑇,
anyone moving from the street to Congress will make The Street’s threat of damage no longer feasible.
Expressions (3) and (4) are necessary and sufficient conditions for Γ ∗ such that 𝑚∗ > 𝑇 to be an equilibrium
partition. (3) and (4’) are necessary and sufficient conditions for Γ ∗ such that 𝑚∗ = 𝑇 to be an equilibrium
partition. These conditions allow for two types of LIE: one in which the partition is such that the greatest 𝑖 in
ℳ∗ is smaller than the least 𝑖 in 𝒩\ℳ∗ , that is all types in Congress are larger than all types in the street (call
those LIE1); and one in which that condition is not satisfied (LIE2). In the rest of this appendix as well as in the
text we focus on equilibria LIE1, and reserve the terminology LIE to refer to LIE1. Other than being more
intuitive and allowing for a briefer exposition of the subsequent analysis, this choice can be justified by two
results (available upon request): (i) for any vector of parameters for which a LIE2 exists, there exists also a
LIE1; and (ii) take two partitions that differ only in that one player from each venue is switched to the other: the
sum of utilities of those two players will be higher in the partition in which the higher type is in Congress and
the lower type is in the Street.
We analyze now the parameter values under which LIE exists. We can explain the results with the visual
aid of Figure A-1. Let the horizontal axis represent ℝ+ as well as the set of players 𝑖 ordered from 1 to 𝑛.
<Insert Figure A1>
Take any vector of all the parameters of the game except 𝑇. Let 𝑉𝐷𝑎 (𝑖) represent the value of action 𝑎 ∈ {0,1}
for player 𝑖 when potential damage is 𝐷 ∈ {0, 𝐾}, under the assumption (from LIE1) that all players to his left
are choosing the street (𝑎𝑗 = 0 ∀𝑗 < 𝑖) and all players to his right are choosing Congress (𝑎𝑗 = 1 ∀𝑗 > 𝑖). We
𝑝𝑖
have then 𝑉𝐾1 (𝑖) = (𝑋 − 𝐾) [∑𝑛
𝑗=𝑖 𝑝𝑗
𝑝𝑖
], 𝑉01 (𝑖) = 𝑋 [∑𝑛
𝑗=𝑖 𝑝𝑗
], 𝑉𝐾0 (𝑖) =
𝐾
𝑖
− 𝑐, and 𝑉00 (𝑖) = −𝑐. The first three
30
functions are plotted in figure A-1.31 Let 𝑚∗ be the largest integer such that 𝑉𝐾0 (𝑖) ≥ 𝑉𝐾1 (𝑖). The set of
parameter vectors for which there is a LIE can be visualized by juxtaposing 𝑇 with the rest of parameters
summarized in the functions plotted in Figure A-1. If 𝑇 < 𝑚∗, then the threat of damage 𝐾 is credible, the
relevant curve for the value of being in Congress is 𝑉𝐾1 (𝑖), and we have a LIE in which players 1 through 𝑚∗ go
to the street and players 𝑚∗ + 1 through 𝑛 go to Congress. If 𝑇 > 𝑚∗ , then the value of being in the street
would be 𝑉00 (𝑖) = −𝑐, and that cannot be an equilibrium, as already stated. For 𝑇 = 𝑚∗, the relevant
comparison for the (now pivotal) player 𝑖 = 𝑇 = 𝑚∗ is between 𝑉01 (𝑖) and 𝑉𝐾0 (𝑖). If 𝑉01 (𝑖) ≤ 𝑉𝐾0 (𝑖), then he
chooses the street and we have a LIE. If 𝑉01 (𝑖) > 𝑉𝐾0 (𝑖), then he would choose Congress and we do not have a
LIE.
We summarize the analysis in the following proposition, which is the more formal equivalent of (heuristic)
Proposition 1 in the text.
Appendix Proposition 1:Let 𝑚∗ ≡ max{i| 𝑖 ∈ 𝒩 𝑎𝑛𝑑 𝑉𝐾0 (𝑖) ≥ 𝑉𝐾1 (𝑖)}.
(a) A Full Institutionalization Equilibrium in which ℳ∗ = ∅ and 𝒩\ℳ∗ = 𝒩 exists for all values of
parameters.
(b) A Low Institutionalization Equilibrium in which ℳ = {1,2, … 𝑚∗ } and 𝒩\ℳ = {𝑚∗ + 1, 𝑚∗ + 2, … 𝑛}
∗
∗
exists if T<𝑚∗ or if T=𝑚∗ 𝑎𝑛𝑑 𝑉𝐾0 (𝑖) ≥ 𝑉01 (𝑖)
We have established conditions for the existence of LIE, in which players up to 𝑚∗ are in the street and
players from 𝑚∗ + 1 are in Congress. In such cases, the equilibrium conditions 𝑉𝐾0 (𝑚∗ ) ≥ 𝑉𝐾1 (𝑚∗ ) and
𝑉𝐾1 (𝑚∗ + 1) ≥ 𝑉𝐾0 (𝑚∗ + 1) are satisfied. For simplicity, in any case of indifference we assign 𝑚∗ to the street
and 𝑚∗ + 1 to Congress. Under those conditions 𝑚∗ is unique for any set of parameters for which LIE exists,
which facilitates the comparative statics.
Comparative Statics
31
The functions are written and plotted as continuous for expositional purposes, even though for a
given set of s they are discrete functions.
31
We can use Figure A-1 to intuitively visualize comparative statics results from the parameters of the model to
𝑚∗. These comparative statics could be interpreted as referring both to the number of people in the street within
a LIE, and to the possibility of existence of LIE (vis-à-vis FIE that always exists), where “possibility” is defined
in terms of the size of the set of parameters (other than the one being changed) for which the conditions for
existence of LIE are satisfied. First, it is clear that a larger 𝑇 makes LIE less likely. We can see the effect of
changing any parameter other than 𝑇 by shifting the curves 𝑉𝐾0 (𝑖) and 𝑉𝐾1 (𝑖) in Figure A-1, and by noticing that
the larger 𝑚∗, the weaker the requirements on all other parameters for the existence of LIE. It is easy to see that
an increase in 𝑐 shifts 𝑉𝐾0 (𝑖) down, tending to decrease 𝑚∗, hence decreasing the possibility of LIE. Increasing
𝑋 has the same effect through an upward shift of 𝑉𝐾1 (𝑖). Decreasing 𝐾 also decreases both 𝑚∗ and the
probability of LIE through leftward shifts in both curves.
Comparative statics on the distribution of 𝑝𝑖 s can be a bit more involved. Here we prove the effect of
increasing the asymmetry of 𝑝𝑖 s (the inequality in the distribution of de jure political power) in a particular way.
Let 𝑃(𝑖) = ∑𝑛𝑗=𝑖 𝑝𝑗 be the cumulative density function counting from 𝑛 to 𝑖. Notice that 𝑃(𝑖) decreases in 𝑖. Let
𝑞 be another distribution, with cumulative 𝑄(𝑖) = ∑𝑛𝑗=𝑖 𝑞𝑗 . For brevity we will speak as if the functions where
continuous, while the proper logic is the discrete equivalent of what we say. We say that 𝑄(∙) is more unequal
than 𝑃(∙) if 𝑄(∙) is equally or more elastic than 𝑃(∙) for all 𝑖, and it is strictly more elastic for at least one 𝑖.
This means that changes in 𝑖 lead to larger changes in the probability under 𝑞 than under 𝑝. Notice that this
implies that 𝑃(∙) first order stochastically dominates 𝑄(∙). We show that a more unequal distribution leads to
more people in the street. 𝑄(∙) more unequal than 𝑃(∙) implies
𝜕𝑄 𝑖
𝜕𝑃 𝑖
| 𝜕𝑖 𝑄| ≥ | 𝜕𝑖 𝑃| ⇒
𝜕𝑄 𝑖
𝜕𝑖 𝑄
≤
𝜕𝑃 𝑖
𝜕𝑖 𝑃
⇒
𝑞𝑖
𝑄(𝑖)
≤
𝑝𝑖
𝑃(𝑖)
⇒
𝑞𝑗
𝑛
∑𝑗=𝑖 𝑞𝑗
𝑝𝑖
≤ ∑𝑛
𝑗=𝑖 𝑝𝑗
.
As we can see in Figure A-2, this implies that in the case of more unequal distribution of power, there will be
more (or at least not fewer) people in the street.
<Insert Figure A2>
32
REFERENCES
Acemoglu, Daron, and James A. Robinson. 2006. Economic Origins of Dictatorship and Democracy.
Cambridge, United Kingdom: Cambridge University Press.
Aoki, Masahiro. 2001. Toward a Comparative Institutional Analysis. Cambridge, MA: MIT Press.
Bahry and Wilson. 2006. “Confusion or Fairness in the Field? Rejections in the Ultimatum Game
under the Strategy Method”, Journal of Economic Behavior and Organization 60: 37-54.
Baron, David, and John Ferejohn. 1989. “Bargaining in Legislatures.” American Political Science
Review 83(4): 1181-1206.
Blondel, John. 2006. “About Institutions, Mainly, but not Exclusively, Political” in Rhodes, Binder
and Rockman Oxford Handbook of Political Institutions. Oxford: Oxford U. Press.
Bueno de Mesquita, Bruce, Alastair Smith, Randolph Siverson, and James Morrow. 2003. The Logic
of Political Survival. Cambridge, MA: MIT Press.
Calvert, Randall. 1995. “Rational Actors, Equilibrium and Social Institutions.” In Knight and Sened.
Explaining Social Institutions. Ann Arbor, United States: University of Michigan Press.
Calvert, Randall. 1995b. “The Rational Choice Theory of Institutions: Cooperation, Coordination, and
Communication.” In Banks and Hanushek, Modern Political Economy. Cambridge University Press.
Carey, John M. 2000. “Parchment, Equilibria, and Institutions.” Comparative Political Studies 33(67): 735-761.
Caruso, Germán et al. 2010 “Are We All Playing the Same Game? The Economic Effects of
Constitutions Depend on the Degree of Institutionalization.” Mimeo.
Dal Bó, Ernesto, and Rafael Di Tella. 2003. “Capture by Threat.” Journal of Political Economy
111(5): 1123-54.
33
Dal Bó, Ernesto, Pedro Dal Bó and Rafael Di Tella. 2006. “Plata o Plomo?Bribe and Punishment in a
Theory of Political Influence.” American Political Science Review 100(1): 41-53.
Diermeier, Daniel and Keith Khrebiel. 2003. “Institutionalism as a Methodology” Journal of
Theoretical Politics, 15(2): 123-144.
Dunning, Thad. 2009. “Fighting and Voting: Violent Conflict and Electoral Politics” Mimeo, Yale.
Ellman, Matthew, and Leonard Wantchekon. 2000. “Electoral Competition under the Threat of
Political Unrest.” Quarterly Journal of Economics 115(2): 499-531.
Garfinkel, Michelle and Stergios Skaperdas. 1996. The Political Economy of Conflict and
Appropriation, New York: Cambridge University Press.
Granovetter, Mark. 1978. “Threshold Models of Collective Behavior.” American Journal of Sociology
83(6): 1420-1443.
Greif, Avner. 2006. Institutions and the Path to the Modern Economy: Lessons from Medieval Trade.
Cambridge, United Kingdom: Cambridge University Press.
Grossman, Gene, and Elhanan Helpman. 2001. Special Interest Politics. Cambridge: MIT Press.
Harstad, Bard and Jakob Svensson.2009. “Bribe or Lobby? Rule Bending, Rule Changing, and
Economic Growth” Mimeo, Northwestern University.
Henrich et al. 2004. Foundations of Human Sociality. Oxford: Oxford University Press.
Hirshleifer, Jack. 2001. The Dark Side of the Force: Economic Foundations of Conflict Theory.
Cambridge, United States: Cambridge University Press.
Humphreys, Macartan. 2001. “To Bargain or to Brawl.” Mimeo.
Huntington, Samuel P. 1968. Political Order in Changing Societies. Yale University Press.
Jones, Mark et al. 2002. “Amateur Legislators, Professional Politicians: The Consequences of PartyCentered Electoral Rules in Federal Systems.” American Journal of Political Science 46(3): 656-669.
34
Levine, David. 1998. "Modeling Altruism and Spitefulness in Experiments," Review of Economic
Dynamics, 1: 593-622, 1998.
Levitsky, Steven and M. Victoria Murillo. 2009. “Variation in Institutional Strength” Annual Review
of Political Science 12: 115-33.
Lindvall, Johannes. 2008. “A Model of Protests” Mimeo. Dept. of Political Science, Oxford U.
Lohmann, Susanne. 1994. “The Dynamics of Informational Cascades: The Monday Demonstration in
Leipzig, East Germany 1989-91.” World Politics 47(1): 42-101.
Machado, Fabiana et al. 2009. “Political Institutions and Street Protests in Latin America.”
Marx, Gary and Douglass McAdams. 1994. Collective Behavior and Social Movements: Process and
Structure. Prentice Hall: New Jersey.
O’Donnell, Guillermo. 1996. “Another Instititutionalization: Latin America and Elsewhere” Working
Paper 222: Kellogg Institute for International Studies, University of Notre Dame.
Oliver, Pamela. 1993. “Formal Models of Collective Action” Annual R. of Sociology 19: 271-300.
Persson, Torsten, and Guido Tabellini. 2000. Political Economics: Explaining Economic Policy. MIT
Press.
Persson, Torsten, and Guido Tabellini. 2003. The Economic Effects of Constitutions. Cambridge,
United States: MIT Press.
Pierson, Paul. 2004. Politics in Time: History, Institutions, and Social Analysis. Princeton, United
States: Princeton University Press.
Pierson, Paul. and Theda Skocpol.
2002. “Historical Institutionalism in Contemporary Political
Science”, in Katznelson and Milner (eds.) Political Science. State of the Discipline, Centennial
Edition. New York: W.W.Norton & Co.
35
Polsby, Nelson W. 1968. “The Institutionalization of the House of Representatives. American Political
Science Review, 62(1): 144-68.
Przeworski, Adam. 1992. “The Games of Transition” in Mainwaring et al Issues in Democratic
Consolidation. University of Notre Dame Press.
Przeworski, Adam. 2009. “Representative Institutions, Political Conflicts, and Public Policies” Mimeo
Saiegh, Sebastian. 2010. “Active Players or Rubber-Stamps? An Evaluation of the Policy-Making
Role of Latin American Legislatures.” In Scartascini, Stein and Tommasi. How Democracy Works:
Political Institutions, Actors and Arenas in Latin American Policymaking. Cambridge, U.S.: Harvard
University, David Rockefeller Center for Latin American Studies.
Scartascini, Carlos, Ernesto Stein and Mariano Tommasi. 2009. “Political Institutions, Intertemporal
Cooperation, and the Quality of Policies.” WP 676. Washington, DC: IADB.
Stein, Ernesto, and Mariano Tommasi. 2007. “The Institutional Determinants of State Capabilities in
Latin America.” In Bourguignon and Pleskovic, Annual World Bank Conference on Development
Economics Regional: Beyond Transition. Washington, DC: World Bank.
Stein, Ernesto et al. 2008. Policymaking in Latin America: How Politics Shapes Policies. Cambridge,
U.S.: Harvard University, David Rockefeller Center for Latin American Studies.
Thelen, Kathleen. 1999. “Historical Institutionalism in Comparative Politics”, Annual Review of
Political Science, 2: 369-404.
Trucco, Laura 2009. “Instituciones y Tecnologías Políticas Alternativas en los Procesos de
Policymaking.” Buenos Aires: Universidad de San Andrés, Dept. of Economics. MA thesis.
36
Figure 1. Alternative Political Actions
More Institutionalized
Less Institutionalized
forming a political party
demonstrations
to inspire sympathy
writing to your congressman
legal strikes
road blockades
campaign contributions
subtle coordination of capital
flight
coordinated economic disruption
clean lobbying
bribes (“plata”)
violent threats (“plomo”)
37
.3
Figure 2. Congress or the Street? (17 Latin American Countries, 2008)
Bolivia
.2
.25
Argentina
.15
Ec uador
Peru
Venezuela
Honduras
Paraguay
.1
Nic aragua
Panama
Uruguay
Guatemala
Colombia
Costa Ric a
Mexic o
.05
Rep. Dominic ana
Brasil
El Salvador
.02
.04
.06
Percentage who contacted Mem ber of Congress
.08
38
Figure 3. Damage Technology
39
Figure 4. Timing of the Game
40
Figure 5. Low Institutionalization Equilibrium
(𝑋 − 𝐾) [
𝑝𝑖
]
𝑛
∑𝑗=𝑖 𝑝𝑗
𝐾
−𝑐
𝑖
𝒎∗
max 𝑝𝑖
iϵℳ
Go to street
𝒎∗ + 𝟏
𝑖
𝑛
min 𝑝𝑖
iϵ𝒩∖ℳ
Go to Congress
41
Table 1. Comparative Statics
Parameter
Probability of being in the noninstitutionalized equilibrium
(or
inverse degree of institutionalization )
 Threshold of collective action in the street (𝑇)
(-)
 Cost of using APT (c)
(-)
 Damage potential (K)
 Wealth (X)
 Asymmetry distribution of de jure political
power (asymmetry distribution of 𝑝𝑖 ’s)
(+)
(-)
(+)
42
2
0
-2
Internal Conflict
4
Figure 6. Correlation (Institutional Strength; Use of one particular APT)
-2
-1
0
Institutional Strength Index
1
2
* controlling for LnGDPpc and regional dummies
43
Figure A1. Low Institutionalization Equilibria
𝑉01 (𝑖) = 𝑋 [
𝑝𝑖
]
𝑛
∑𝑗=𝑖 𝑝𝑗
𝑉𝐾1 (𝑖) = (𝑋 − 𝐾) [
𝑉𝐾0 (𝑖) =
𝑝𝑖
]
𝑛
∑𝑗=𝑖 𝑝𝑗
𝐾
−𝑐
𝑖
𝑖
𝑇1
𝑇2
𝑇3
𝑇1
𝑇2
𝑇3
𝑇4
𝑇4
𝑛
< 𝑚∗ → ∃ 𝐿𝐼𝐸
= 𝑚2∗ → ∃ 𝐿𝐼𝐸
= 𝑚3∗ → ∄ 𝐿𝐼𝐸
> 𝑚∗ → ∄ 𝐿𝐼𝐸
44
Figure A2. More Unequal Distribution 𝒒 Leads to Larger 𝒎∗
(𝑋 − 𝐾)
𝑝𝑖
∑𝑛𝑗=𝑖 𝑝𝑗
(𝑋 − 𝐾)
𝑞𝑖
𝑛
∑𝑗=𝑖 𝑞𝑗
𝐾
−𝑐
𝑖
𝑛
𝑚∗ (approx.) for
distribution 𝑝
𝑖
𝑚∗ (approx.) for
distribution 𝑞
45