Professional Documents
Culture Documents
Democratization
Camilo Garcı́a-Jimeno
2/ 55
Acemoglu, Daron y James A. Robinson (2006). Economic Origins
of Dictatorship and Democracy, Caps. 5-7, Cambridge University
Press.
3/ 55
Democratization, Repression and Institutional Change
4/ 55
Some patterns of political change: Britain
I This may allow a better understanding of historical events
such as the democratization of the west in the 19th Century.
I The recurrence of coups in Latin American countries in the
20th Century
I Britain in the 19th Century democratized gradually:
I First Reform Act in 1832 eliminates “rotten boroughs” and
increased the electorate to 14% of adult males.
I Second Reform Act in 1867 doubles the electorate.
I Third Reform Act in 1884 expanded the electorate to 60% of
adult males.
I Representation of the People Act in 1918 established universal
suffrage for males 21+ and for ratepayer women 30+.
I Finally universal suffrage for women in 1928.
5/ 55
Some patterns of political change: Argentina
6/ 55
Some patterns of political change: Argentina
7/ 55
Some patterns of political change: Singapore
8/ 55
Some patterns of political change: South Africa
I Achieved independence from Britain in 1910.
I From the beginning whites in power began a process of
exclusion of blacks that culminated in full Apartheid in 1948.
I Political organization by blacks also arose right after
independence around the ANC.
I During the 50s and 60s the government attempted to fully
segregate all blacks into 10 ‘homelands’ restricting their
mobility and occupational choices.
I Tight control over the media, and lots of repression well into
the 90s.
I Only after lots of riots, strikes and international sanctions did
whites have to give up Apartheid in 1994.
9/ 55
Some patterns of political change: South Africa
Van
der
Berg,
Servaas
and
Megan
Louw
(2004).
“Changing
Patterns
of
South
African
Income
Distribution:
Towards
Time-‐Series
Estimates
of
Distribution
and
Poverty”,
in
South
African
Journal
of
Economics,
Vol.
72,
No.
3,
pp.
546-‐572.
10/ 55
Some patterns of political change: South Africa
Van
der
Berg,
Servaas
(2010).
“Current
Poverty
and
Income
Distribution
in
the
Context
of
South
African
History”,
Stellenbosch
Economics
Working
paper
22/10.
11/ 55
Some patterns of political change
12/ 55
Acemoglu and Robinson (2000), (2001)
13/ 55
Acemoglu and Robinson (2000), (2001)
I Average income is y .
I We parametrize income inequality with θ:
(1 − θ)y θy
yp = yr =
1−δ δ
14/ 55
Acemoglu and Robinson (2000), (2001)
Taxation:
I We restrict the space of fiscal instruments to linear income
taxes τ ≥ 0 and lump-sum transfers T .
I Taxation is distortionary, and we capture the costs of taxation
(incentive effects, ‘leaky bucket’, etc.) with a convex function
C (τ )y .
I Government budget constraint is
15/ 55
Acemoglu and Robinson (2000), (2001)
Preferences:
I Discrete time, infinite horizon.
I Utility is defined over discounted sum of post-tax income,
with discount factor β ∈ (0, 1):
∞
X
Ui = E β t (1 − τt )y i
t=0
I Indirect utility:
V (y i |τ ) = (1 − τ )y i + (τ − C (τ ))y
16/ 55
Acemoglu and Robinson (2000), (2001)
I Thus, single-peaked preferences.
I ⇒ Median voter theorem holds. Under democracy, fiscal policy
chosen by the median, which is a citizen since δ < 1/2.
I Under non-democracy, fiscal policy is just chosen by the elite.
I Under non-democracy citizens may influence the equilibrium
choice, but only indirectly through the exercise of their de
facto power (threat of revolution).
I The ideal tax rate of a citizen is implicitly given by
θ−δ
C 0 (τ p ) =
1−δ
τr = 0
17/ 55
Acemoglu and Robinson (2000), (2001)
Revolution:
I Very simplified environment.
I If revolution is attempted, it always succeeds but fraction µS
of all income is destroyed forever.
I After revolution, citizens receive all remaining income, forever.
I Revolution is an absorbing state.
I Therefore, if a revolution happens, citizens get every period
income
(1 − µS )y
1−δ
I µS is indexed by S ∈ {L, H}, and S will denote the state of
effectiveness of collective action, low or high.
I We assume Pr (S = H) = q, and for simplicity, µL = 1 so in
18/ 55
periods S = L there is no point in revolting.
Acemoglu and Robinson (2000), (2001)
Revolution and Collective Action:
I The idea is that collective action is transient, and may only be
effective or feasible in some periods (e.g. during economic
crisis, or war).
I Fluctuations in µ capture the uncertainty about when
collective action problem can be solved.
I This will be a source of commitment problems:
I For example, if S = H today, there is a threat of revolution.
Elites may want to appease citizens with concessions.
I But if tomorrow S = L, elite will have an incentive to renege
on that promise.
I Inherent problem in politics because there is no outside party
to enforce any promises.
19/ 55
Acemoglu and Robinson (2000), (2001)
Timing of Events:
I At each t the timing is:
I µS is revealed.
I Elite set the tax rate τ N (upper-script denotes
‘non-democracy’).
I Citizens decide whether to revolt or not (ρ ∈ {0, 1}, ρ = 1
means revolution).
I If ρ = 1, citizens consume remaining 1 − µS of output forever.
I If ρ = 0 elite remain in power onto next period.
20/ 55
Acemoglu and Robinson (2000), (2001)
Equilibrium Concept:
I We will focus on Markov Perfect Equilibria.
I Strategies only a function of the current state. Here state
space is {N, R} × {L, H}.
I Elites’ strategy is the tax rate:
τ N : {L, H} → [0, 1]
21/ 55
Acemoglu and Robinson (2000), (2001)
Payoffs after Revolution
I Define V i (R, µS ) the value of individual in group i ∈ {p, r } if
revolution takes place in period where state is S.
I ⇒
(1 − µS )y
V p (R, S) =
(1 − δ)(1 − β)
I For the elite the value is
V r (R, S) = 0
23/ 55
Acemoglu and Robinson (2000), (2001)
yr
V r (N) =
1−β
yp
V p (N) =
1−β
24/ 55
Acemoglu and Robinson (2000), (2001)
Revolution Constraint:
I The revolution constraint will be binding if in state S = H
citizens prefer to revolt than to live under non-democracy
without any redistribution:
(1 − µ)y yp
> ⇐⇒ θ > µ
(1 − δ)(1 − β) 1−β
I Inequality must be high enough for the revolution threat to be
credible
I =⇒ For θ ≤ µ, there is never a revolution and there is no
redistribution on the equilibrium path
25/ 55
Acemoglu and Robinson (2000), (2001)
I When θ > µ revolution constraint binds:
I Citizens would revolt if elite provides no redistribution when
the state is S = H.
I Elite will have to make concessions to avert the revolution:
τ N = τ̂ > 0.
I In this case, the value for citizens is
V r (N, H, τ̂ ) = y r + (τ̂ (y − y r ) − C (τ̂ )y )
and
V p (N, H, τ̂ ) = y p + (τ̂ (y − y p ) − C (τ̂ )y )
y p + (1 − β(1 − q)) [τ p (y − y p ) − C (τ p )y ]
V p (N, H, τ p ) =
1−β
28/ 55
Acemoglu and Robinson (2000), (2001)
I If this condition does not hold, then averting revolution with
redistribution is not an equilibrium.
I This inequality also tells us what is the lowest possible µ
under which redistribution can be sustained:
I When µ ≥ µ∗ the elite make the smallest possible concession
τ̂ that satisfies the no-revolution condition:
30/ 55
Acemoglu and Robinson (2000), (2001)
31/ 55
Acemoglu and Robinson (2000), (2001)
32/ 55
Acemoglu and Robinson (2000), (2001)
33/ 55
Acemoglu and Robinson (2000), (2001)
Democratization
I In the model we have been discussing the elite only has one
possible strategy to prevent revolution: redistribution.
I The elite is unable to avert revolution for some range of
parameters. This is because of a commitment problem.
I But in a more general model elites have another possible
strategy: to change political institutions (i.e., democratize).
34/ 55
Acemoglu and Robinson (2000), (2001)
Democratization
I This works because elite’s commitment problem arises
because citizens have de facto power today but not tomorrow.
I But de jure power, arising from formal institutions, is more
persistent.
I By giving citizens de jure power by changing political
institutions, elites are able to commit to future redistribution
that averts a revolution.
I This argument suggests that democratization (more generally
institutional change), can be seen as a commitment device.
35/ 55
Acemoglu and Robinson (2000), (2001)
36/ 55
Acemoglu and Robinson (2000), (2001)
I So let’s augment the model by allowing a larger action space
for the elite:
I At the start of every period elite can democratize.
I For now assume democracy D is an absorbing state.
I The tax rate is decided by majority voting under D. Because
δ < 1/2, median voter is a citizen, who prefers τ p .
I Let’s also allow elite to use repression to avert the revolution.
I Assume for simplicity that repression is always successful in
preventing revolt.
I Repression is costly. It also destroys a fraction of output (e.g.
see current-day Syria).
I When will the elite choose redistribution, repression or
democratization?
37/ 55
Acemoglu and Robinson (2000), (2001)
38/ 55
Acemoglu and Robinson (2000), (2001)
39/ 55
Acemoglu and Robinson (2000), (2001)
40/ 55
Acemoglu and Robinson (2000), (2001)
Timing:
I State S ∈ {L, H} is realized.
I Elite decides to repress or not ω ∈ {0, 1}.
I If ω = 1 citizens cannot revolt, stage game ends.
I If ω = 0 elite decides to democratize or not φ ∈ {0, 1}
I If φ = 0 elite chooses the tax rate τ N .
I Citizens choose to revolt or not ρ ∈ {0, 1}.
I If ρ = 1 game moves to state R forever.
41/ 55
Acemoglu and Robinson (2000), (2001)
y p + τ p (y − y p ) − C (τ p )y
V p (D) =
1−β
y r + τ p (y − y r ) − C (τ r )y
V r (D) =
1−β
42/ 55
Acemoglu and Robinson (2000), (2001)
I Democratizing will prevent revolution if
V p (D) ≥ V p (R, H)
µ ≥ θ − [τ p (θ − δ) − (1 − δ)C (τ p )]
V p (N, µ∗ , τ p ) = V p (R, µ∗ )
43/ 55
Acemoglu and Robinson (2000), (2001)
I What will the elite choose?
I Payoffs from equilibrium with repression for every period when
S = H:
h i
V i (O, H|κ) = (1 − κ)y i + β qV i (O, H|κ) + (1 − q)V i (N, L)
h i
V i (N, L) = y i + β qV i (O, H|κ) + (1 − q)V i (N, L)
I =⇒
y r − (1 − β(1 − q))κy r
V r (O, H|κ) =
1−β
y p − (1 − β(1 − q))κy p
V p (O, H|κ) =
1−β
44/ 55
Acemoglu and Robinson (2000), (2001)
I When µ < µ∗ redistribution is not enough. The choice will be
between democracy and repression.
I When µ ≥ µ∗ redistribution can avert revolution, so choice is
between redistribution and repression.
I These two comparisons produce two cutoffs for κ.
I κ∗ : Cost of repression for which elite is indifferent between
redistribution at rate τ̂ and repression:
1
=⇒ κ∗ = [δC (τ̂ ) − τ̂ (δ − θ)]
θ
I For κ ≥ κ∗ elite prefers to redistribute than to repress.
45/ 55
Acemoglu and Robinson (2000), (2001)
I κ: Cost of repression for which elite is indifferent between
democratization and repression:
1 δC (τ p ) − τ p (δ − θ)
=⇒ κ =
θ 1 − β(1 − q)
I For κ ≥ κ elite prefers to democratize than to repress.
I Notice that κ > κ∗ .
46/ 55
Acemoglu and Robinson (2000), (2001)
47/ 55
Acemoglu and Robinson (2000), (2001)
48/ 55
Acemoglu and Robinson (2000), (2001)
dκ dκ∗
>0 >0
dθ dθ
49/ 55
Acemoglu and Robinson (2000), (2001)
50/ 55
Acemoglu and Robinson (2000), (2001)
dκ
>0
dq
51/ 55
Acemoglu and Robinson (2000), (2001)
52/ 55
Acemoglu and Robinson (2000), (2001)
53/ 55
Acemoglu and Robinson (2000), (2001)
Discussion:
I Key ideas in this model:
I Distributive conflict.
I De facto and de jure sources of power.
I Revolution threat.
I Commitment problems.
I Institutional change as a credible promise.
54/ 55
Acemoglu and Robinson (2000), (2001)
55/ 55