Political Change, Stability and Democracy - Penn...

Political Change, Stability and Democracy

Daron Acemoglu (MIT)

MIT

February, 13, 2013.

Acemoglu (MIT) Political Change, Stability and Democracy February, 13, 2013. 1 / 50

Political Change, Stability and Democracy Introduction

Motivation

Almost all social, political, economic institutions have evolved overdecades and hundreds of years.

e.g., present British democracy with universal suffrage evolved fromMagna Carta (1215), Bill of Rights (1689), and Reform Act (1832)

The process is slow, and political decisions are made by politicalactors along the way.

these may include kings, landlords, peasants...

Institutional arrangements are often very persistent, but suchpersistence also coexists with change

Mexican institutions are still shaped by its colonial history, but Mexicogained its independence almost 200 years ago and many of theeconomic institutions of colonial era have disappeared.



Motivation (continued)

Social conflict often key in institutional change:

E.g., the British case, leading up to the First Reform Act:



Key Ingredients of a Framework

1 A game theoretic environment capturing social conflict.2 Many agents.3 Limited discounting (or foresight): not too myopic, but also notextremely forward-looking agents.

4 Stochastic events: inability to predict the future perfectly results isneeded to model realistic evolutionary paths.

5 State variables creating potential persistence (both because ofequilibrium and because of costs of change), but also allow forchanges in political equilibria along transition path or in response toshocks.



Example I

Monarch may wish to grant constitution and some limited rights to asmall segment of society, but is opposed to full-scale democratization.

Constitution may be desirable because it encourages investment or ormay simply be a “necessary evil” from the viewpoint of the monarchwishing to stave off other actions.

But emerging middle class may demand (and stochastically succeededin obtaining) further democratization. This may even lead toinstability or revolutions empowering radical groups.

Would the monarch wish to grant such a constitution?

This will depend on:

Whether further democratization will occur (which in turn depends onwhether one more round of democratization will ultimately dislocatethe middle class from power also)Whether there is a natural coalition between other groups in societyHow soon further democratization might occur



Example I

autocracy limitedfranchise

fulldemocracy



Example I (the simple case)

Three states: absolutism a, constitutional monarchy c , full democracyd

Two agents: elite E , middle class M

wE (d) < wE (a) < wE (c)

wM (a) < wM (c) < wM (d)

E rules in a, M rules in c and d .

Myopic elite: starting from a, move to c

Farsighted elite (high discount factor): stay in a– as moving to c willlead to M moving to d

Also richer insights when there are stochastic elements andintermediate discount factors:

e.g., fear of shift of power to radicals may limit reform, but also itsinsulating effects.



Example II

What determines tolerance/intolerance towards different groups?

Example: secular vs. religious groups.

Different countries moving in different directions (e.g., Iran vs.Turkey vs. France)

what drives such change?

preferences?future changes of preferences?fear of changes in future states?



Example III

Does greater social mobility strengthen democracy?

De Tocqueville on America:

“Nor can men living in this state of society derive their belieffrom the opinions of the class to which they belong, for, so tospeak, there are no longer any classes, or those which still existare composed of such mobile elements, that their body can neverexercise a real control over its members.”

We would need a model in which individuals anticipates that theirposition in society my change, i.e., “reshuffl ing”of individuals.



Past Work

Game-theoretic models of regime dynamics: Acemoglu and Robinson“‘Why Did the West Extend the Franchise?”QJE 2000, Acemogluand Robinson “A Theory of Political Transitions”AER 2001,Acemoglu and Robinson Economic Origins of Dictatorship andDemocracy, Cambridge University Press 2006: focus on models withtwo groups and limited links between periods. No change/persistence.Issues of path dependence and institutions and political equilibria:Acemoglu and Robinson Why Nations Fail: The Origins of Power,Prosperity and Poverty, Crown 2012: no formal models.Modeling coexistence of change and persistence: Acemoglu andRobinson “Persistence of Institutions, Elites and Power”AER 2008:two groups and limited links between periods.Towards a more general framework: Acemoglu, Egorov and Sonin“Dynamics and Stability of Constitutions, Coalitions and Clubs”AER2012: focus on non-stochastic environments and high discountfactors. Limited dynamics.Acemoglu (MIT) Political Change, Stability and Democracy February, 13, 2013. 10 / 50


Ongoing Work

The words a general but tractable framework of social (political,economic) evolution.

Some simplifying, but not too restrictive assumptions, so ingredients1—5 can be included.

In particular single crossing type assumption; states are ordered from1 to m, and players ordered in such a way as to satisfy single crossing(increasing differences).

Build a dynamic non-cooperative game, with fully rational agentsmaking strategic decisions.

Study the impact of discount factor and of stochastic events on thedecisions made and on social / political transitions

Provide general characterization results and comparative statics.

Also a variety of rich applications with “tight” results.



Outline

Model

Equilibrium

Special case: non-stochastic environment with high discount factor(Acemoglu, Egorov and Sonin AER 2011)General non-stochastic characterizationGeneral stochastic characterization

Comparative statics

Applications


Model Model

Model: Basics

Finite set of individuals N = {1, 2, . . . , n}Finite set of states S = {1, 2, . . . ,m}Discrete time t ≥ 1Individuals decide on possible transition

Key assumption: individuals and states are ordered

e.g., from less to more democratic,or from less tolerance to secular values towards less tolerance toreligiosity, etc.


Model Model

States and utilities

More formally, “society” starts period in st−1 and decides on(feasible) stLet us start with a non-stochastic case first.

Individual i in period t gets instantaneous utility

wi (st )

Strict increasing differences: For any agents i , j ∈ N such thati > j ,

wi (s)− wj (s)is increasing in s

This could be weakened to weak increasing differences for some results.But imposed for simplicity throughout this presentation.


Model Model

Transitions

Special case: only one step

S1 S2 Sm–1 Sm

General case: all transitions are allowed but potentially with transitioncosts:

S1 S2 Sm–1 Sm


Model Model

Stochastic shocks: example

Cost of transitioning to state sm becomes less prohibitive:

S1 S2 Sm–1 Sm


Model Model

Transition costs

General transition costs ci (x , y) ≥ 0 cost of transitioning from statex to state y for player i

These could all be zero

Or such that only one step transitions are possible

Generally only weak restrictions on transition costs


Model Model

Transition costs: assumptions

Assumptions on ci (x , y) ≥ 0Reverse triangle inequality:

Higher marginal cost in longer transitions:

These assumptions to ensure single crossing of value functions.Acemoglu (MIT) Political Change, Stability and Democracy February, 13, 2013. 18 / 50

Model Model

Winning Coalitions

A transition occurs from state s to feasible state s ′ if suffi cientlymany players agree on that

set of winning coalitions Ws ⊂ 2S defined for each s

Three assumptions need to hold:

if X ⊂ Y ⊂ N and X ∈ Ws , then Y ⊂ Wsif X ∈ Ws , then N \ X /∈ WsN ∈ Ws


Model Model

Quasi-median voter

Player i ∈ N is a quasi-median voter in state s if for any winningcoalition X ∈ Ws , minX ≤ i ≤ maxX

a player that belongs to any connected winning coalitiongeneralization of standard median voterset of quasi-median voters in state s denoted by Ms

Monotonic Quasi-Median Voter property: Sequences {minMs}s∈Sand {maxMs}s∈S are nondecreasing


Model Model

Illustration

Quasi-median voters:

simple majority 5/6 supermajority

Monotonic Quasi-Median Voter property:

1234

Robert’s model; ok

also oknot ok


Model Model

Stochastic environments

We introduce stochastic elements as follows.

An environment is

E =({Ws}s∈S , {wi (s)}s∈S ,i∈N ,

{ci(s, s ′

)}s ,s ′∈S ,i∈N , β

).

Set of environments E .Finite set of environments, and finite number of shocks captured byenvironment transition probabilities

q(E ,E ′

).

Non-stochastic model special case in which |E | = 1.When stochastic elements need to be emphasized, we write wE ,i (s)to denote payoffs in environment E , etc.


Model Model

Dynamic game

1 Period t begins with state st−1 and environment Et−1 inherited fromthe previous period (where s0 is exogenously given).

2 Shocks are realized.3 Players become agenda-setters, one at a time, according to theprotocol πst−1 . Agenda-setter i proposes an alternative state at ,i

4 Players vote sequentially over the proposal at ,i . If the set of playersthat support the transition is a winning coalition, then st = at ,i .Otherwise, the next person makes the proposal, and if the last agentin the protocol has already done so, then st = at ,i .

5 Each player i gets instantaneous utility

wEt ,i (st )− cEt ,i (st−1, st ) .

Equilibrium concept: Markov Perfect equilibrium.But for most of the analysis, we will focus on simpler Markov votingequilibrium.Acemoglu (MIT) Political Change, Stability and Democracy February, 13, 2013. 23 / 50

Equilibria Equilibria

Transition mappings

Think of any MPE in pure strategies as represented by a set oftransition mappings {φE } such that

if st−1 = s, and Et = E , then st = φE (s) along the equilibrium pathwe write

φ : S → S .

Transition mapping φ = {φE : S → S} is monotone if for anys1, s2 ∈ S with s1 ≤ s2, φE (s1) ≤ φE (s2).

natural, given monotonic median voter property

Key steps in analysis

fix Echaracterize φE and expected payoffs when there is no stochasticitythen backward induction and dynamic programming.



Value functions

More explicitly for given fixed state E (i.e., non-stochastic case),continuation value not including transition costs is

V φi (s) = wi (s) +

∞

∑k=1

βk[wi(

φk (s))− ci

(φk−1 (s) , φk (s)

)].

Recursively

V φi (s) = wi (s) + β

[V φi (φ(s))− ci (s, φ (s))

], or

V φi (s) = wi (s)− βci (s, φ (s)) + βV φ

i (φ(s))

Also define continuation value inclusive of transition costs:

V φi (s | x) = V

φi (s)− ci (x , s)



Value functions (continued)

In the stochastic case:

V φE ,i (s) = wE ,i (s)+ βE ∑

E ′q(E ,E ′

) [V φE ′,i (φE ′(s))− cE ′,i (s, φE ′ (s))

].

And also:V φE ,i (s | x) = V

φE ,i (s))− cE ,i (x , s) .


Equilibria Markov voting equilibrium

Markov voting equilibrium

φ = {φE : S → S} is a Markov Voting Equilibrium if for any x , y ∈ S ,{i ∈ N : V φ

E ,i (y | x) > VφE ,i (φE (x) | x)

}/∈ WE ,x{

i ∈ N : V φE ,i (φE (x) | x) ≥ V

φE ,i (x)

}∈ WE ,x

The first is ensures that there isn’t another state transition to whichwould gather suffi cient support.

Analogy to “core”.

The second one ensures that there is a winning coalition supportingthe transition.


Equilibria Markov voting equilibrium

General result

Theorem

(existence) There exists a Markov voting equilibrium with monotonetransition mapping φ.

Theorem

(uniqueness) “Generically” there exists no other Markov votingequilibrium with monotone transition mapping if either every set ofquasi-median voters is a singleton or preferences are single-peaked (plusadditional conditions on transition costs; e.g., only one step transitions).

Thus monotone transition mappings arise naturally.

though equilibria without such monotonicity may exist.


Equilibria General characterization

Limiting states and effi ciency

Theorem

(limit behavior) In any Markov voting equilibrium, there is convergenceto a limiting state with probability 1.The limiting state depends on the timing of shocks.

Theorem

(effi ciency) If each βE is suffi ciently small, then the limiting state isPareto effi cient. Otherwise the limiting state may be Pareto ineffi cient.

Example of Pareto ineffi ciency: elite E , middle class M

wE (d) < wE (a) < wE (c)

wM (a) < wM (c) < wM (d)

E rules in a, M rules in c and d .



Comparative statics

Theorem

(“monotone” comparative statics) Suppose that environments E 1 andE 2 coincide on S ′ = [1, s ] ⊂ S and βE 1 = βE 2 , φ1 and φ2 are MVE inthese environments. Suppose x ∈ S ′ is such that φ1 (x) = x. Thenφ2 (x) ≥ x.

Implication, suppose that φ1 (x) = x is reached before there is aswitch to E2. Then for all subsequent t, st ≥ x .Intuition: if some part of the state space is unaffected by shocks, it iseither reached without shocks or not reached at all.



MPE vs. Markov voting equilibria

Theorem

(MPE ≈MVE) For any MVE φ (monotone or not) there exists a set ofprotocols such that there exists a Markov Perfect equilibrium of the gameabove which implements φ.Conversely, if for some set of protocols and some MPE σ, thecorresponding transition mapping φ = {φE }E∈E is monotone, then it isMVE.In addition, if the set of quasi-median voters in two different states haveeither none or one individual in common, and only one-step transitions arepossible, every MPE corresponds to a monotone MVE (under anyprotocol).

For each Markov voting equilibrium, there exists a protocol π suchthat the resulting (pure-strategy) MPE induces transitions thatcoincide with the Markov voting equilibrium.


Equilibria Applications

Simple example

Suppose three groups: elite, middle class and workers.

The elite rule under absolutist monarchy, a.

Suppose that with limited franchise, c , the middle class rules withprobability p and workers rule with probability 1− p.Workers rule in full democracy, d .

The middle-class prefer limited franchise, workers prefer fulldemocracy.

Payoffs

wE (d) < wE (a) < wE (c)

wM (a) < wM (d) < wM (c)

wW (a) < wW (c) < wW (d)



Simple example (continued)


fulldemocracy

?

What happens if β large and p small?





fulldemocracy

!




What happens if p = 1 or close to 1?


fulldemocracy

!

What happens if β small or intermediate?




Now suppose that p takes different values in different environments.We start in E1 and then stochastically transition to either E2 or E3,both of which are absorbing, and pE2 = 1 and pE3 < 1. Is an earlyresolution of uncertainty good for transitioning to democracy?


fulldemocracy

?? ?



Application: Radicals in Politics

Most regime transitions take place in the midst of uncertainty andturmoil, which sometimes brings to power the most radical factionssuch as the militant Jacobins during the Reign of Terror in the FrenchRevolution or the Nazis during the crisis of the Weimar Republic.

The possibility of “extreme”outcomes also of interest because, inmany episodes, the fear of such radical extremist regimes has beenone of the drivers of repression against a whole gamut of oppositiongroups.

Leading example: the Bolshevik Revolution in Russia.

A fringe group that was repressed from the early 1900s came to powerafter the February Revolution, first entering into an alliance with theliberal left, then with the Social Revolutionaries, then with the leftSocial Revolutionaries, and at each stage, tilting power towards itself.



Application (continued)

Consider a society consisting of n+ 1 groups.

The stage payoff of each group depends on the current ‘political state’which encapsulates the distribution of political and economic rights.

Each group maximizes the discounted stage payoffs and may alsoincur additional costs from transitions across (political) states.

Stochastic shocks affect both stage payoffs and the likelihood of shiftsin political power in a given political state (e.g., in the Russiancontext, the possibility of a group inside or outside the Dumagrabbing power or sidelining some of the rest of the groups).

Suppose that a shock to the environment starting from the stabledictatorship of the tzar changes stage payoffs and makes it desirableto include share power with ‘moderate’groups.




Now several considerations are potentially important.

First, the tzar may not go all the way to including all left groups ormay maintain a veto power if feasible, because he may be worried of a‘slippery slope’– once power shifts to these groups, they may laterinclude additional groups further to the left, which is costly for thetzar.

Second, the probability that radical extremist groups may gain powermight be higher in states in which additional groups to the left areincluded in the decision-making process (again potentially as in theRussian case), further discouraging limited power-sharing.

Thus, in these first two scenarios, the tzar might be afraid of ourstylized description of the Russian path where power gradually (andstochastically) shifts from left liberal groups to the coalition ofsocialist/communist groups, and then ultimately to the most extremeelements among them, the Bolsheviks.




Third, and counteracting the first two, the most moderate left groupsmay be unwilling to enter into alliances with other groups to their leftbecause they are themselves afraid of a yet another switch of powerto groups to their left.

But if so, the tzar may be more willing to allow power-sharing in thiscase, calculating that further slide down the slope will be limited.

Can we model these dynamics and get more insights?



Application: Model

There is a fixed set of n players (groups) N = {−l , . . . , r} (son = l + r + 1).We interpret the order of groups as representing some economicinterests (poor vs. rich) or political views.The set of states is S = {−l − r , . . . , l + r} (so the total number ofstates is m = 2l + 2r + 1 = 2n− 1), and they correspond to differentcombinations of political rights.Repression: a way of reducing the political rights of certain groups

The set of players who are not repressed in state s is Hs , whereHs = {−l , . . . , r + s} for s ≤ 0 and Hs = {−l + s, . . . , r} for s > 0.Thus, the states below 0 correspond to repressing the rich (in theleftmost state s = −l − r only the group −l has vote), the statesabove 0 correspond to repressing the poor (again, the rightmost states = l + r on the group r has vote), and the middle state s = 0 involvesno repression and corresponds to full democracy (with the medianvoter, normalized to be from group 0, ruling in state 0).




Stage payoff with policy p (and repression of groups j /∈ Hs ):

ui (p) = − (p − bi )2 −∑j /∈Hs γjCj .

The weight of each group i ∈ N is denoted by γi and represents thenumber of people within the group, and thus the group’s politicalpower.

in state s, coalition X is winning if and only if

∑i∈Hs∩X γi >12 ∑i∈Hs γi .



Application: Model (continued)

It is possible that a radical will come to power without havingmajority because of shocks and crises.

Let us model this by assuming that there is a set of k environmentsR−l−r , . . . ,R−l−r+k−1, and probabilities λj ∈ [0, 1], j = 1, . . . ,m, totransition to each of these environments; the environment Rj is thesame as E , except that in states −l − r , . . . , j , the decision-makingrule comes into the hands of the most radical group −l .In other words, the probability of a radical coming to power if thecurrent state is s is µs = ∑s

j=−l−r λj , and it is (weakly) increasing ins.



Application: Results

Proposition

(Equilibria without radicals) In the absence of shocks (i.e., ifenvironment E never changes), there exists a unique MVE given by afunction φ : S → S. In this equilibrium:

1 Democracy is stable: φ (0) = 0.2 For any costs of repression {Cj}j∈N , the equilibrium involvesnon-increasing repression: if s < 0 then φ (s) ∈ [−s, 0], and if s > 0,then φ (s) ∈ [0, s ].

3 Consider repression costs parametrized by parameter k: Cj = kC ∗j ,

where{C ∗j}are positive constants. Then there is k∗ > 0 such that:

if k > k∗, then φ (s) = 0 for all s, and if k < k∗, then φ (s) 6= 0 forsome s.



Application: Results (continued)

But if there are radicals, then radicals themselves will use repression,and this may increase repression.

Proposition

(Radicals) There exists a unique MVE. In this equilibrium:

1 After radicals came to power, they are more likely to move to theirpreferred state −l − r (repress everyone else) if (a) repression is lesscostly (in the sense that k is lower, as in parametrization above), and(b) they are more radical (meaning their ideal point b−l is lower).

2 Before radicals come to power, if s ≤ 0 then φ (s) ≥ s, but φ (s) > sis possible if s > 0.




In fact, the fear of radicals can push moderates to repression. Define:

Wi (s) =1− µs

1− β (1− µs )ui (s)+

µs(1− β) (1− β (1− µs ))

ui (−l − r) .

Proposition

(Repression by moderates anticipating radicals) Suppose that theradicals, when in power, move to their preferred state.

1 If W0 (0) < W0 (x) for some x > 0, then there is a state s ≥ 0 suchthat φ (s) > s. In other words, in some state there is an increase inrepression in order to decrease the chance of radical coming to power.

2 If for all states y > x ≥ 0, WMx (y) < WMx (x), then for all s ≥ 0,φ (s) ≤ s. (This will happen if costs C are high enough.)




Comparative statics of repression:

Proposition

(More repression) Suppose that there is a state s ≥ 0 (i.e., fulldemocracy or some state favoring the right), which is stable in E for some

set of probabilities{

µj

}. Let us consider an anticipated or unanticipated

change from{

µj

}to{

µ′j

}such that µ′j = µj for j ≥ s. After this change,

there will never be less repression of the left, i.e., φ′E (s) ≥ φE (s) = s.

Both greater and lesser power of radicals in “left” states leads togreater repression.




Path dependence in history:

Proposition

(Role of radicals in history) Suppose the society was in a stable statex ≥ 0 before the radical came to power. Then the ultimate state, after theradical comes and possibly goes, will be some y ≤ x.




Strategic complementarities and repression:

Proposition

(Strategic Complementarity) Suppose the costs of repressing othergroups declines for the radicals. Then it becomes more likely thatφ (s) > s for at least one s ≥ 0.

The history of repression in places such as Russia may not be due tothe “culture of repression”but to small differences in costs ofrepression (resulting from political institutions and economicstructure).


Roadmap Roadmap

Roadmap

Work in progress for a general framework for the analysis of politicalchange, stability and regime dynamics.

Plan:

Complete development of the framework.Systematic general comparative static results.Sharper results for specific applications.

Many general insights in the context of economic, social and politicalchange.

Many new applications (as well as new areas).

Empirical and historical applications.


Political Change, Stability and Democracy - Penn...

Documents

Transcript of Political Change, Stability and Democracy - Penn...