Transcript PowerPoint

Electoral Law Enforcement and the Political
Economy of Long-Run Development: Evidence
from Latin America, 1800-2012
Rok Spruk
Research Fellow, University of Ljubljana – Laibach
E: [email protected], [email protected]
2017 Annual Meeting, American Economic Association, Chicago (IL)
January 6, 2017
Agenda
• Introduction
• Institutional Change and Long-Run Growth in Latin America
• Data
• Empirical Setup
• Results and Robustness Checks
• Counterfactual Scenario
• Concluding Remarks
6 april 2017
2
Introduction
• Why Latin America ended up with inferior growth and development outcomes and ultimately
fell behind the United States in spite of being initially rich?
• In 1500, PPP-adjusted per capita GDP of the U.S. was no different from that of Brazil
whereas Mexico's GDP per capita was 37% above the U.S level (Coatsworth 2008)
• In 1800, U.S GDP per capita overtook Argentina and Mexico by considerable margin. By
1900, U.S. enjoyed three-fold higher GDP per capita than core 8 Latin American countries
from Maddison's dataset (Bolt and Van Zanden 2013). By 2010, the difference in PPPadjusted terms amounted to four-fold
• Various theories have been proposed to explain long-run development gaps such as physical
geography (Diamond 1997, Bloom and Sachs 1998, Gallup et. al. 1999, Pomeranz 2000,
Presbitero 2005), international trade (Ben-David 1996, Frankel and Romer 1999, Dollar and
Kraay 2003, Földvari 2006), social, cultural and religious factors (Weber 1930, Landes 1998,
Guiso et. al. 2008, Becker and Wößmann 2009, Tabellini 2010, Gorodnichenko and Roland
2011) and human capital (Becker et. al. 1989, Galor and Weil 1996, Glaeser et. al. 2004,
Van Leeuwen and Földvari 2008, 2014, Becker and Wößmann 2012)
6 april 2017
3
Introduction
•
In spite of the initial economic advantage compared to North
America, the economic performance of Latin America continuously
fell behind the U.S. from 1500 onwards (Acemoglu et. al. 2002)
little empirical evidence is given to examine how (rather than why)
Latin America fell behind the United States?
•
Since institutions matter for long-run growth and development, the
natural starting question is which set of institutions has led to the
long-run comparative economic stagnation of Latin America in
comparison with the United States?
6 april 2017
4
Institutional Change and Long-Run Development:
Latin American Equilibrium
•
Existing empirical evidence suggest that relatively high per capita income of Spanish and
Portuguese colonies in early 16th century is largely a result of native population decline
(Mexico), silver boom (Bolivia) and low population density (Argentina) rather than
structural change and productivity improvements (Arroyo-Abad et. al. 2012)
•
Three tentative explanations for Latin America's failure to achieve comparable income
level to the U.S.: (i) institutional differences (Acemoglu et. al. 2001, 2002, 2005), (ii)
factor endowments (Engerman and Sokoloff 2000) and (iii) pattern of colonization
(Lange et. al. 2006, Mahoney 2010)
•
The inheritance of formal and informal Iberian political, legal and economic institutions
from Castillian social and legal norms (creation of the caste system, encomienda, mita,
haciendas) has affirmed the persistence of inefficient rent-seeking institutions in the
post-independence period despite the break from the Spanish empire
•
Extractive institutions transmitted from Iberian legal tradition led to high wealth, land
and income inequality (Milanovic et. al. 2011) which failed to extend the franchise and
secure property rights outside the elite groups
6 april 2017
5
Institutional Change and Long-Run Development:
Latin American Equilibrium
• Why Iberian colonialism failed to create dynamic societies capable of generating an independent technological
breakthrough and innovation (Coatsworth 2008)?
• Absence of real revolutionary changes and inexperience with self-government inherited from Spain and Portugal
established few institutional constraints on the powerholders asa check on the abuse of executive power
• Coupled with traditional disrespect for law and order, the failure to create a common market and the difficulty of
developing effective contracting institutions established an institutional environment uncapable of securing
property rights and encouraging sustained growth
• Independence from Spain (Castille) further consolidated extractive political and economic institutions
• Mexican independence in 1810 came through a virtual military coup by the colonial Creole elite which carried out
the coup to isolate Mexico from the wave of political and economic liberalization in Spain and to regain industrial
and land-owning privileges lost under the Bourbon era
• Rent-seeking coalitions comprised major asset holders, bankers and creditors seeking to consolidate massive
concentration of land ownership in the hands of elites, make the government too weak to establish a strong
state, and receive monopoly rents in exchange for enforcing the contract between asset holders and the
government
• Elsewhere in Latin America, inefficient and extractive economic institutions from 16th century Iberian colonialism
(encomienda and mita in Peru, cronyism of channeling railroad investment subsidies in Brazil) persisted into postindependence period in various forms (Dell 2010; Acemoglu and Robinson 2012; Summerhill 2000)
6
Electoral Laws and the Origins of Latin America‘s
Institutional (Under)Development
•
The independent Latin American countries did not break away from the
political institutions of the colonial period and periodically excluded nonproperty owners from the legal standing to vote
•
Participation in general election restricted to substantial landowners and elitepropriated class in a much more restrictive fashion compared to the United
States
•
Exclusive electoral laws allowed the post-independence elites to subordinate
the political institutions to their own benefit and exclude the non-elites from
the right to vote
•
In spite of the swift abolition of slavery, post-independence Latin American
countries quickly introduced a myriad of wealth- and literacy-based voting
restrictions, effectively excluding the non-elites from the legal standing to
vote
6 april 2017
7
Electoral Laws and the Origins of Latin America‘s
Institutional (Under)Development
•
Early institutional development after independence from Spain and Portugal
characterized by the near absence of suffrage outside the elite class
•
The 1853 Constitution of Argentina enshrined universal suffrage which had
been periodically violated and suppressed. The 1912 Sáenz Peña Law
introduced compulsory and secret vote, women enfranchised under Péron in
1952. In the aftermath, universal suffrage continuously violated by electoral
fraud and political oppression until the transition to democracy in 1983
•
In Mexico, general suffrage established in 1917 but periodically violated by
PRI-sponsored electoral fraud until 1983
•
The weakness and failure of electoral law enforcement in establishing broadbased de jure and de facto political institutions an unexploited source of
variation in institutional development independent of long-run economic
development
6 april 2017
8
Hypotheses:
H1: De jure and de facto political institutions matter for long-run
development
H2: Enforcement of general suffrage, abolition of slavery, the removal
of wealth- and literacy-based voting qualifications, and the end of
electoral fraud made the de jure and de facto political institutions more
inclusive and open-access
6 april 2017
9
Data: GDP
•
Data on GDP per capita (1990 $G-K) is from Bolt and Van Zanden
(2013), building on the earlier Maddison dataset (Maddison 2007)
and recent work by economic historians for Argentina (Newland &
Poulson 1998; Della Paolera et. al. 2003; Newland & Ortiz 2001),
Brazil (Leff 1982; Goldsmith 1986), Chile (Diaz et. al., 2007),
Colombia (Kalmanovitz Krauter & Lopez Rivera 2009), Mexico
(Coatsworth 1989), Uruguay (Bertola et. al. 1998) and Venezuela
(Baptista 1997)
•
Inter-benchmark year estimates produced by a linear interpolation
based on the temporal trend per capita GDP for the core 9 Latin
American sub-sample
6 april 2017
10
Data: Political Institutions
• Data on the structure of political institutions is used to explicitly consider the distinction between de jure and
de facto political power
• De jure political institutions are delegated by laws, electoral systems and constitutions whereas de facto
political institutions reflect the factual enforcement of de jure political institutions and denote the ability of
the non-elites to engage in various forms of collective action (Acemoglu and Robinson 2006b, Robinson
2013, Voigt 2013)
• Polity IV index from Marshall et. al. (2013) is used as a measure of de jure political institutions comprising
six main indicators: (i) regulation of chief executive recruitment, (ii) competitiveness of the chief executive
recruitment, (iii) openness of executive recruitment, (iv) executive constraints on decision rules, (v)
regulation of political participation, and (vi) competitiveness of the participation
• Vanhanen Index of Democracy from Vanhanen (2000, 2003, 2012) is used as a measure of de facto political
institutions comprising (i) the index of political competition (the share of smaller parties vote share in the
composition of the national parliaments), and (ii) the index of political participation (voter turnout in
parliamentary and/or presidential elections)
6 april 2017
11
Data: Electoral Law Enforcement
•
Episodes of electoral law enforcement in seven different domains of broadbased political institutions constructed on yet unexploited historical
bibliography of Latin America compared to less reliable estimates by Mariscal
and Sokoloff (2000)
•
Exogenous changes to the existing de jure and de facto political institutions
captured by the abolition of slavery, constitutional suffrage guarantee (de jure
and de facto), income- and literacy-related voting restrictions, electoral fraud,
female suffrage and the transition to democracy under free, fair and regular
elections
•
Episodes of electoral law enforcement constructed on the basis of large Latin
American histriography
•
Different dimensions of broad-based political institutions and specific
distinction between the de jure and de facto institutional component
6 april 2017
12
Data: Electoral Law Enforcement
Figure 1: Latin America Institutional (Dis)Equilibrium
6 april 2017
13
Data: Electoral Law Enforcement
Table 1: Dates of Electoral Law Enforcement Across Latin America
Year of
Independence
Abolition
of Slavery
Constitutional
Suffrage Guarantee
De Jure
De Facto
Restricted Suffrage
Income,
Literacy,
Property
1816-1912
Electoral
Fraud and
Oppression
1930-1943,
1946-1955,
1958-1961,
1966-1973,
1975-1983
1964-1965,
1969-1981
1889-1946,
1964-1987
1974-1989
Argentina
1816
1853
From
1853
From
1912
Bolivia
1825
1826
1822
1888
Chile
1810
1823
Colombia
1810
1851
From
1952
From
1933
From
1970
From
1959
1825-1982
Brazil
1810-1936
1881-1914,
1933-1958
Mexico
1810
1829
1917-1992
1825
1854
1860-1979
1893-1895,
1929-1979
Uruguay
1825
1830
1825-1918
Venezuela
1811
1821
From
1917
19311932,
19391947,
19501961,
19631967
19681979
From
1918
From
1959
1810-1953
Peru
From
1939
From
1889
From
1833
18531881,
from
1914
From
1917
From
1858
1931-1932,
1973-1983
1948-1957
From
1918
From
1947
1822-1988
1810-1989
1811-1946
Female
Suffrage
Full
Suffrage
with Free
and Fair
Elections
From
1951
From 1983
From
1956
From
1932
From
1949
From
1932
From 1982
From
1958
From
1955
From 1993
From
1934
From
1947
From 1983
From 1988
From 1989
From 1959
From 1980
From 1958
14
Identification Strategy: Model Setup
•
The goal of the empirical strategy is to examine the contribution of de jure
and de facto political institutions to long-run development consistenty
•
Consider a simple model of long-run development:
ln yi ,t 
Venezuela

i  Argentina
 i  Di  1  ˆ iDe,t Jure  2  ˆ iDe,t Facto  X'i ,t  ui ,t
where y is per capita income in country i at time t, Z is the measure of
political institutions, X is the vector of control variables, D is the set of
country-fixed effects, and u is the stochastic disturbance
•
One possible setback of the empirical model of long-run development is the
omitted variable bias contaminating the true effect of political institutions
on long-run development, denoted by μ
6 april 2017
15
Identification Strategy: Assumptions
•
Our key identification assumption is that the changes in electoral law from Table 1 are
independent of the long-run development path and influence it only through the de jure and de
facto institutional channel
•
First, the electoral law enforcement changed the structure of existing de jure and de facto political
institutions and does not represent the independent source of variation from the setup of political
institutions
•
Second, the timing of electoral law enforcement allows to decompose the variation in institutional
development into before-law and after-law component which implies that post-treatment effects
of electoral law enforcement should matter only for the changes in the setup of political
institutions rather than directly for the changes in long-run development
•
And third, electoral law enforcement is essentially a flow variable changing the existing stock of
political institutions which implies that the failure to control for common trends in de jure/de facto
institutional path would make it impossible to identify the effects on long-run development
•
Econometrically, the key identification assumption is postulated as follows:
16
Identification Strategy: First-Stage DiD
• Seven binary treatment and post-treatment measures of electoral law enforcement is constructed and used
as an IV to address the endogeneity of de jure and de facto political institutions to long-run development
• Difference-in-difference approach to model institutional change in the spirit of DiD analysis from Card and
Kruger (1994), Autor (2003)
• An exogenous shock (treatment) independently affects the outcome of interest in countries which
underwent institutional changes (treatment group) compared to the countries which did not undergo such
changes (control group). The compact form of difference-in-differences (DiD) model setup for de jure and
de facto institutional development that takes place is:
j
j

De Jure
i ,t
2012
k
 Electoral 
 Electoral 
j
j
  i  Di   j  





  


t
t

  i , k   i ,t
i =Argentina
j 1
t 
 Provision i ,t t 1800
 Provision i ,t

De Facto
i ,t
2012
k
 Electoral 
 Electoral 
j
j
  i  Di   j  





  


t
t

 i , k  i ,t
i  Argentina
j 1
 1
 Provision i ,t t 1800
 Provision i ,t
Venezuela
Venezuela
7
7
j
j
6 april 2017
17
Identification Strategy: A Note on Standard Errors
• Conventional standard errors in DiD applications may grossly understate the estimated treatment
effects due to serially correlated error terms which leads to overestimated t-statistics and generally
unreliable inference from the estimated parameters (Bertrand et. al. 2004)
• When the number of individual panel units is low, arbitrary variance matrix estimator artificially
increases the rejection rates above 5% significance level and renders the inference on treatment effects
unreliable, and the control for unobserved effects does little to mitigate the inconsistency of standard
errors
• Data collapse into pre-treatment and post-treatment averaging by ignoring time-series transformation
is not feasible in our case due to the varying composition of the treatment group over time
• A valid inference in our case requires multiway-clustered standard errors between countries and within
countries over time as the failure to do so leads to massively underestimated standard errors and
overrejection of the null hypothesis on the treatment and post-treatment effects (Moulton 1986, 1990,
Davis 2002, Kezdi 2004)
• Multiway clustering method by Cameron, Gelbach and Miller (2011) is used to ensure robust and valid
inference on the parameters of interest by adjusting standard errors for arbitrary heteroskedasticity and
serially correlated stochastic disturbances by extending the standard cluster-robust variance matrix
estimator (White 1980, 1984, Liang and Zeger 1984, Arellano 1987) to avoid over-rejecting the null on
the treatment and post-treatment effects
6 april 2017
18
Results – Second Stage
•
One basis point improvement in de jure political institutions
associated with 25.4 percent increase in long-run per capita income
and holds across a battery of specification checks
•
One basis point improvement in de facto political institutions
associated with 11.2 percent increase in long-run per capita income
and also holds across a broad array of specifications
•
The estimated effects of de jure and de facto political institutions on
long-run development appear to be robust to the unobserved
effects
•
The within-country of de jure and de facto political institutions on
long-run development dominates the between-country effect (!)
6 april 2017
19
Results – First Stage
•
The changes in electoral law enforcement associated with marked
improvements in the set of de jure and de facto political institutions towards
greater level-playing field
•
The onset of electoral fraud led to the 2.8 basis point decline in the
inclusivity of de jure political institutions
•
The switch to income- and literacy-based suffrage restrictions associated
with the deterioration in de jure and de facto political institutions between
2.2 and 7.8 basis points
•
The overidentifying restrictions remain stable even at artificially high
significance level, confirming the validity of law enforcement instruments in
providing the exogenous source of variation in long-run development
6 april 2017
20
Results – Second Stage
0.81
0.56
.2
.4
.6
.8
Figure 2: Effects of De Jure and De Facto Political Institutions on Long-Run
Development Across and Within Countries
0.09
0
0.12
De Jure "Within-Country" Effect
De Facto "Within-Country" Effect
De Jure "Between-Country" Effect
De Facto "Between-Country" Effect
6 april 2017
21
Results – Second Stage
tio
le
c
rE
Fa
i
an
d
ni
by
Su
ffr
ag
e
R
es
tri
ct
e
d
U
In
co
ve
rs
a
m
lS
e,
L
uf
ite
fra
ge
ra
cy
w
an
ith
d
Fr
e
e
Pr
op
e
d
Fr
au
by
d
es
tri
ct
e
R
e
ffr
ag
Su
ns
t
m
en
eq
R
rty
an
d
Fe
m
O
al
e
pp
Su
ui
re
ffr
ag
re
ss
io
n
e
ge
na
lS
on
s
e
D
D
e
Ju
r
e
Fa
c
C
to
Ab
Su
tit
u
tio
ffr
ag
ol
iti
on
e
of
Sl
a
Ex
te
n
uf
fra
si
o
ve
r
n
y
-.4 -.2
0
.2 .4 .6
Figure 3: Post-Treatment Effects of Electoral Law Enforcement on De Jure
and De Facto Institutional Development
De Jure "Within-Country" Effect
De Jure "Between-Country" Effect
De Facto "Within-Country" Effect
De Facto "Between-Country" Effect
6 april 2017
22
Parallel Trend Assumption
•
The validity of DiD model of institutional change critically hinges on
the parallel path assumption
•
Undertaking the institutional reforms of electoral law in the
treatment group should follow the parallel trend with the control
group before the reforms were enforced
•
The pre- and post-treatment trends of de jure and de facto political
institutions in the treatment and control sample are tested on
country-specific subsamples are confirmed a marked break in the
path of de jure and de facto institutional development after the
electoral law enforcement
6 april 2017
23
Robustness Checks
•
Temporal and Spatial Sample Splits
•
Extreme Bounds Analysis of De Jure and De Facto Institutional
Development
•
Standardized Effects of De Jure and De Facto Political Institutions
•
Causal Effects of Electoral Law
•
DiD Identification Check with Country-Specific Time Trends
6 april 2017
24
Robustness Checks – Temporal and Spatial Sample Splits
Figure 4: Temporal Sample Split
(1)
(2)
(3)
(4)
(5)
Excluding Pre-1820
Excluding Pre-1830
Period
Period
Panel C::Endogenous (Structural) Long-Run Growth Model Setup
Polity2 Score
.144***
.139***
(.013)
(.012)
Vanhanen Index of
.085***
.083***
Democracy
(.013)
(.012)
(6)
(7)
Excluding Pre-1850
Period
.125***
(.012)
.112***
(.012)
.078***
(.012)
.070***
(.011)
Panel D: First-Stage Difference-in-Differences (DiD) OLS Regression for Polity2 Score and Vanhanen Index of Democracy
Abolition of Slavery
.794
-.018
1.553
-.310
3.579**
-.037
4.339**
(Post-Treatment Effect)
(1.259)
(1.013)
(.668)
(1.416)
(1.404)
(1.901)
(1.672)
De Jure Constitutional
2.548*
1.715
2.098
1.929
2.055
2.070
2.219
Suffrage
(1.246)
(1.676)
(1.247)
(1.900)
(1.367)
(2.128)
(1.666)
(Post-Treatment Effect)
De Facto Suffrage Extension
.467
.141
.543
.003
.622
-.198
.529
(Post-Treatment Effect)
(1.147)
(1.391)
(1.147)
(1.419)
(1.254)
(1.423)
(1.314)
Suffrage Restricted by
-1.713
-2.538*
-2.267
-2.522*
-2.370
-2.220
-2.629
Income, Literacy and
(1.322)
(1.241)
(1.493)
(1.375)
(1.382)
(1.484)
(1.672)
Property Requirement
(Post-Treatment Effect)
Suffrage Restricted by Fraud
-2.803*
-1.157
-2.832*
-1.076
-2.836*
-.910
-2.994*
and Oppression
(1.416)
(2.021)
(1.369)
(2.028)
(1.477)
(2.032)
(1.459)
(Post-Treatment Effect)
Female Suffrage
-.419
3.669**
-.573
3.668**
-.984
3.645**
-1.235
(Post-Treatment Effect)
(1.016)
(1.372)
(1.042)
(1.398)
(.919)
(1.380)
(.967)
Universal Suffrage with Free 6.464*** 13.035*** 6.238*** 13.141*** 6.381*** 13.508*** 6.147***
and Fair Elections
(.905)
(.937)
(.913)
(.946)
(.987)
(.949)
(1.069)
(Post-Treatment Effect)
Observations
1,728
1,728
1,638
1,638
1,467
1,467
1,287
Country-Fixed Effects
YES
YES
YES
YES
YES
YES
YES
First-Stage F-Test on Joint
Significance of PostTreatment Effects (p-value)
Hansen J-Test of
Overidentifying Restrictions
(p-value)
Cragg-Donald Weak
Identitifcation Wald Test
(p-value)
(8)
Excluding Pre-1870
Period
.028
(2.156)
2.102
(2.149)
-.446
(1.369)
-1.851
(1.569)
-.696
(2.045)
3.613**
(1.340)
14.015***
(.971)
1,287
YES
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.27
0.32
0.26
0.27
0.26
0.44
0.28
0.41
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Notes: the table presents IV-2SLS and DiD estimates of institutional changes and the effects on long-run economic growth
and development of Latin America using temporal exclusion criteria to test the robustness of the underlying estimates from
Table 5. Treatment effects of institutional changes are estimated but not reported. Standard errors in the parentheses are
clustered across countries and within countries over time using Cameron-Gelbach-Miller non-nested multi-way clustering
scheme for finite-sample adjustment of the empirical distribution function and cluster-robust coefficient inference to remove
the inconsistencies arising from arbitrary heteroskedasticity and serially correlated residuals. Asterisks denote statistically
significant coefficients at 10% (*), 5% (**), and 1% (***) respectively.
• Excluding the temporal
sub-periods from the core
sample does not appear to
shape the effects of de jure
and de facto political
institutions on long-run
development
• In the first stage, the
transition to universal
suffrage with free and fair
elections associated with
large-scale improvements
in de jure and de facto
political institutions
6 april 2017
25
Robustness Checks – Spatial Sample Split
•
Excluding individual countries from the core sample does not appear to render the estimated
effects of de jure and de facto political institutions on long-run development unstable or
insignificant
•
The spatial sample split clearly suggests the effects of de jure and de facto political
institutions are not driven by idiosyncratic country-specific effects given a relatively low
sample size (N = 9)
•
Overidentifying restrictions remain stable, indicating the validity of instruments and the
structural effects of political institutions do not appear to be weakly identified
•
One basis point improvement in the level-playing field of de jure political institutions
associated with 14% permanent increase in long-run per capita income
•
One basis point improvement in the level-playing field of de facto political institutions
associated with 8% permanent increase in long-run per capita income
6 april 2017
26
Robustness Checks - Extreme Bounds of De Jure and De
Facto Political Institutions
•
Does the electoral law enforcement robustly explain the paths of de jure and de
facto political institutions on long-run development?
•
An alternative check on the robustness and identification of the structural effects
•
The post-treatment effects of abolition of slavery, de jure/de facto suffrage
extension, ending of electoral fraud, removal of wealth- and literacy-based literacy
qualifications robustly explain the paths of de jure and de facto institutional
development
•
At extreme bounds, the re-estimated post-treatment effects are within the
expected range and correctly predict the paths of institutional development in more
than 99 percent of CDF(0)
•
Post-treatment effects replicated in more than 1.3 million regressions (not much)
6 april 2017
27
Robustness Checks - Standardized Effects of De Jure and
De Facto Political Institutions on Long-Run Development
•
Are de facto political institutions more important for long-run development
than de jure political institutions?
•
Which type of electoral law enforcement is the most important one for the
trajectory of long-run de jure and de facto institutional development?
•
One should distinguish between within-country and between-country
effects in the first (DiD) and second (IV) stages:
•
Taking the weighted mean coefficient from extreme bounds analysis,
compute the following set of beta coefficients:
 within 
ˆb within  ˆ    De Jure 
De Jure
1 
within 
  ln y 
 between 
ˆbbetween  ˆ    De Jure 
De Jure
1  between 
  ln y 
within 
  De
within
Facto
ˆ
bˆDe




Facto
2 
within 

  ln y 
 between 
ˆbbetween  ˆ    De Facto 
De Facto
2 
between 
  ln y 
6 april 2017
28
Robustness Checks - Standardized Effects of De Jure and
De Facto Political Institutions on Long-Run Development
b) First Stage
a) Second Stage
Standardized Effects of Electoral Law Enforcement
Standardized Effects of De Jure/De Facto
Political Institutions on Long-Run Development on De Jure/De Facto Institutional Development
Full Sample
De Jure
De Facto
Polity2 Index
Vanhanen Index of Democracy
Within-Country
BetweenCountry
Within-Country
Between-Country
0.12
0.81
0.09
0.56
Panel A: Excluded Subset
Argentina
0.12
1.74
0.12
0.66
Bolivia
0.11
1.74
0.11
1.14
Brazil
0.11
0.81
0.07
0.39
Chile
0.13
1.64
0.10
0.99
Colombia
0.11
0.73
0.09
0.47
Mexico
0.12
0.66
0.09
0.47
Peru
0.12
0.85
0.11
0.58
Uruguay
0.12
0.83
0.10
0.34
Venezuela
0.12
1.71
0.11
1.18
Pre-1820
0.10
0.91
0.07
0.41
Pre-1830
0.10
0.96
0.07
0.43
Panel B: Excluded Period
Pre-1850
0.08
0.81
0.05
0.36
Pre-1870
0.03
0.31
0.03
0.28
Notes: the table presents the standardized effects of de jure and de facto political institutions on the paths of
long-run development within and across Latin American countries. The baseline standardized effects is
decomposed into “within-country” and “between-country” component. Panel A presents the standardized
effects on the subsamples with a single excluded country per subsample based on Table 7. Panel B presents
the standardized effects on time-specific subsamples by specifically excluding each period from the full
sample based on the underlying estimates in Table 5.
De Jure
Polity2 Index
De Facto
Vanhanen Index of Democracy
WithinCountry
BetweenCountry
WithinCountry
BetweenCountry
0.05
0.01
0.03
0.18
0.02
0.09
-0.01
0.02
0.01
0.00
0.00
0.00
0.00
0.00
-0.01
0.00
-0.02
-0.06
-0.05
-0.02
0.00
0.00
0.02
0.00
0.04
0.14
0.14
0.03
0.30
0.31
0.16
0.25
0.62
0.11
0.10
0.13
0.34
0.11
0.36
0.32
-0.04
-0.06
-0.19
-0.41
-0.26
-0.31
-0.20
-0.31
0.23
0.08
0.46
0.20
0.59
0.33
0.62
0.46
Panel A: Treatment Effects
Abolition of Slavery
De Jure Constitutional Suffrage)
De Facto Suffrage Extension
Suffrage Restricted by Income, Literacy and
Property Requirement
Suffrage Restricted by Fraud and Oppression
Female Suffrage
Universal Suffrage with Free and Fair
Elections
Panel B: Post-Treatment Effects
Abolition of Slavery
De Jure Constitutional Suffrage)
De Facto Suffrage Extension
Suffrage Restricted by Income, Literacy and
Property Requirement
Suffrage Restricted by Fraud and Oppression
Female Suffrage
Universal Suffrage with Free and Fair
Elections
Notes: the table presents the standardized effects of electoral laws on de jure and de facto
political institutions within and across countries. The set of standardized effects is decomposed
into “within-country” and “between-country” component. Panel A presents the standardized
treatment effects while Panel B displays the standardized post-treatment effects of electoral
laws.
6 april 2017
29
Robustness Checks – Causal Effects of Electoral Law
Enforcement
•
The full sample covers many years which immediately invokes whether the established effects
of de jure and de facto political institutions are causal
•
Causal relationship tested using Granger (1969), Angrist and Pischke (2009) model setup to
examine whether electoral laws happened before the changes in de jure and de facto political
institutions rather than vice versa
•
Using 10 lags and 10 leads similar to Autor (2003), the core DiD model of de jure and de facto
institutional change is rearranged into a flexible setup:
j
j
De Facto
i ,t
10
 Electoral 
 Electoral 
J
'
  i  Di    t Dt   j 1     

  j 1     
  Xi ,t   wi ,t
Provision
Provision
i  Argentina
t 1800
 0
 0

i ,t

i , t
Z
Venezuela
•
j
10
 Electoral 
 Electoral 
J
'
  i  Di   t  Dt   j 1     

  j 1     
  Xi ,t   vi ,t
Provision
Provision
i =Argentina
t 1800
 0
 1

i , t

i ,t
Venezuela
Z
j
De Jure
i ,t
2012
2012
J
J
10
10
The effect of constitutional suffrage extension and the introduction of female suffrage on de jure
institutional development appears to be Granger-causal (!)
6 april 2017
30
Robustness Checks – DiD Identification Checks with
Country-Specific Time Trends
•
Country-specific time trends are added to the list of control variables
•
Electoral law enforcement varies greatly throughout the sample and was introduced in different
countries at vastly different dates
•
The check on the effect identification in DiD model setup for de jure and de facto institutional
development used to examine whether the different types of electoral law enforcement are driven by
country-related time trends that could undermine the causal effects
•
Strategy similar to Besley and Burgess (2004) paper on the effects of regulation on firm performance
across Indian states
•
Reconstruct and re-estimated the first stages for de jure and de facto political institutions as follows:
Z
Z
De Facto
i ,t
2012
j
 Electoral 
'
  i  Di    t  Dt  i  t  ˆ1  
  Xi ,t   wi ,t
Provision
i  Argentina
t 1800

i ,t
Venezuela
•
j
 Electoral 
'
  i  Di    t  Dt  i  t  ˆ1  
  Xi ,t   vi ,t
i  Argentina
t 1800
 Provision i ,t
Venezuela
De Jure
i ,t
2012
The effects remain stable and do not appear to be driven by country-specific time trends
6 april 2017
31
Counterfactual Scenario
•
What happened to Latin America‘s long-run development if it had U.S-style de
jure and de facto political institutions in place since independence?
•
A counterfactual scenario constructed by taking the baseline response
coefficients on de jure and de facto political institutions, developing an
institutional time-series under alternative institutional design, and
decomposing the scenario into observed and unobserved component
•
Several different types of institutional design taken into account apart from the
U.S such as Western Offshoots (Australia, Canada) and Western Europe (UK,
France, Spain, Portugal)
• Two additional scenarios considered:
1. Adoption of US-style de jure/de facto political institutions upon the abolition of
slavery
2. Adoption of US-style de jure/de facto political institutions upon initial suffrage
extension
6 april 2017
32
Counterfactual Scenario
•
Having US-style de jure and de facto political institutions since independence would
yield large-scale economic gains with per capita income gap behind the U.S shrinking
by a fifth
•
US-style de jure and de facto institutional design since independence would have
made Latin America initially much richer than the U.S
•
Starting in 1850s and continuing after 1900s, every Latin American country would
fall behind the U.S per capita income both in absolute and relative terms
•
Under U.S-style de jure and de facto political institutions, Argentina in 2012 would
be as rich as Germany, Chile and Uruguay as rich as Spain, and Mexico and
Venezuela as rich as Czech Republic
•
De jure and de facto political institutions would obviously not prevent Latin America
from its ultimate decline behind the U.S but would mitigate the magnitude of decline
considerably
•
Having Latin-style institutions in place since independence would make Latin America
poorer or no better off in the long-run
6 april 2017
33
Counterfactual Scenario
Table: Summary of Counterfactual Scenario
Country
Real GDP Per Capita in
2012
Implicit
Gain
Observed
Counterfactual
Country
Equivalent
Argentina
10,875
20,561
Germany
Bolivia
3,280
9,482
Brazil
Chile
Colombia
7,015
15,204
7,625
Mexico
Ratio of GDP Per Capita
to the U.S GDP Per Capita
Implicit
Gain
Observed
Counterfactual
89%
34%
65%
31 p.p.
Thailand
189%
10%
30%
20 p.p.
10,787
16,451
9,520
Poland
Spain
Hungary
53%
8%
24%
22%
48%
24%
34%
52%
30%
12 p.p
4 p.p
6 p.p.
8,142
14,871
Czech
Republic
82%
26%
47%
21 p.p.
Peru
Uruguay
6,354
12,738
11,172
16,534
Armenia
Spain
75%
29%
20%
40%
35%
52%
15 p.p
12 p.p
Venezuela
9,644
13,352
Czech
Republic
38%
30%
42%
22 p.p.
6 april 2017
34
Counterfactual Scenario
Figure 6: Counterfactual Colombia with US-style
De Jure/De Facto Political Institutions Since the
Abolition of Slavery
Figure 5: Counterfactual Argentina under U.Sstyle De Jure and De Facto Political Institutions
Colombia
1800
1850
1900
1950
2000
Real GDP Per Capita (1990 -K)
0
0
5000
10000
Argentina
1800
1850
1900
1950
2000
Actual GDP Per Capita
U.S. GDP Per Capita
GDP Per Capita
Counterfactual GDP Per Capita with U.S-style institutional design
Counterfactual under US-style De Jure/De Facto Institutions upon the Abolition of Slavery
6 april 2017
35
Counterfactual Scenario
Figure 8: Counterfactual Chile with U.S-style De
Jure and De Facto Political Institutions Since
1800
Figure 7: Counterfactual Mexico with US-style
De Jure and De Facto Political Institutions
15000
0
0
5000
10000
Real GDP Per Capita (1990 -K)
Real GDP Per Capita (1990 -K)
15000
10000
5000
Chile
Mexico
1800
1850
1900
1950
2000
Actual GDP Per Capita
GDP Per Capita - US-style De Jure/De Facto Political Institutions upon Suffrage Extension (1917)
1800
1850
1900
1950
2000
GDP Per Capita
GDP Per Capita with U.S-style De Jure/De Facto Institutions Since 1800
6 april 2017
36
Counterfactual Scenario
Brazil
0
Real GDP Per Capita (1990 -K)
5000
10000
15000
Figure 8: Counterfactual Brazil
1800
1850
1900
1950
2000
Real GDP Per Capita
U.S Counterfactual since Abolition of Slavery (1888)
U.S Counterfactual
UK Counterfactual
6 april 2017
37
Counterfactual Scenario
Argentina‘s Counterfactual
Rosario, Argentina
Lüneburg, Niedersachsen, Germany
6 april 2017
38
Counterfactual Scenario
Persistent Effects of Good vs. Bad Institutional Choices, Part I.
San Jose, Costa Rica
Guatemala City, Guatemala
6 april 2017
39
Counterfactual Scenario
Long-Run Effects of Bad Institutional Choices – Rochina Favela in Rio de Janeiro
6 april 2017
40
Conclusion
• Institutions matter for long-run growth and development but it less clear how and which types of
institutions matter for long-run growth
• The timing and introduction of franchise extension, abolition of wealth and literacy-based voting
qualifications and abolition of slavery for a sample of 9 core Latin American countries is significantly and
robustly related to the dynamics of de jure and de facto institutional development
• More pluralist and participatory de jure and de facto political institutions help explain Latin America’s longrun economic development
• Extractive institutional development in post-independence Latin America has its roots in the colonial
political power of major rent-seeking coalitions and major landowners deriving political rents from denying
access to the political and economic opportunities to the non-elites
• Such pattern of institutional development critically hindered the path of long-run economic growth and
development, and prevented Latin America from embarking on the path of economic performance
comparable to the United States and Northwestern Europe
• Hence, de jure and de facto political institutions both independently and endogenously influence the path
of economic growth and development over time with sizeable effects on long-run growth
• The effect of de jure and de facto political institutions appear to be causal
6 april 2017
41