The presentation analyses the impact of Bolsa Família programme expansion on strengthening the political positions of incumbent politicians at the municipal level.
Local electoral effects of conditional cash transfers in brazil ross van horn
1. LOCAL ELECTORAL EFFECTS OF
CONDITIONAL CASH TRANSFERS IN
BRAZIL
Ross Van Horn
INSTITUTE OF LATIN AMERICAN STUDIES
THE LYNDON B. JOHNSON SCHOOL OF PUBLIC AFFAIRS
UNIVERISTY OF TEXAS, AUSTIN
2. THEORY & BACKGROUND
Bolsa Família Brazil
Impact Evaluations for Bolsa Familia
Inequality (responsible for 25% drop in Gini)*
Poverty ( reduces poverty gap and extreme poverty)*
Schooling
Work
Fertility
Citizenship/social isolation
Electoral Effect
4. PRIOR RESEARCH ON BOLSA FAMÍLIA
ELECTORAL EFFECTS
Relationship between Bolsa recipients and Voting
behavior (survey data, spatial statistical analysis, OLS
etc)
(+) Hunter and Power (2007) “BF single most plausible
explanation” for Lula’s reelection/largely concerned
with shift in traditional voter base for the workers Party
(+) Zucco (2011&2012): Expansion of program
beneficiaries increased incumbent vote share
(+)Others: Marques et al. (2009); Nicolau and Peixoto
(2007); Soares and Terron (2008); Bohn (2011)
(-) “It’s the economy companheiro!” Shikida et al.
National Focus: Lula, Lula e Lula
6. Municipal Mayor Policy Variables
GDP
Log
transformation
B>0
Age Age in Years
B>0
Program Expansion Program Expansion 2004-2008
(Measured as percent change
of beneficiary target set for
each Municipality)
B>0
GDP per capita Log
transformation
B>0
Political
Party
Political Party of Candidate
B≠0
Size of Program Beneficiary Families as
percent of All Families in
Municipality
B>0
Federal Transfers Log
transformation
B>0
Vote
Share
Share of Valid Votes in
2004
B>0
Program Expansion
X
Size of Program
Interaction of two Policy
Variables
B>0
Regions
South
Southeast
Northeast
Dummy
South B≠0
Southeast B≠0
Northeast B>0
Education Dummy
3 Levels : Primary,
Secondary and College
Area ( sq. km) Area of
Municipality
B>0
Density Population /
Area sq Km
B<0
Human
Development
Index
B>0
IDF (vulnerability
index)
B<0
Variable Definitions and Measurement
7. LOGISTIC FUNCTION; ELECTION OUTCOMES
AS DEPENDENT VARIABLE
P(Re-election|X)=
(Municipal Characteristics)
b1 + b2 GDP (log) + b3 GDP per capita (log) + b4 Federal
Transfers (log) + b5 Southeast (dummy) + b6 Northeast
(dummy) + b7 Area (sq. Km) + b8 Density + b9 HDI + b10
Population + b11 IDF
+(Mayor)
b12 i.Political Party (dummy) + b13 Edu medium (dummy) +
b14 Edu High (dummy) + b15 Age + b16 Vote Share (2004)
+(Policy variables)
b17 Program Expansion + b18 Size of Program + b19 Program
Expansion*Size of Program + u
8. LIMITATIONS
Aggregated indicators (ecological fallacy, at best)
Electoral Effect: Political Perception (politician) vs.
Voting (citizens).
Inertia: retrospective voting and the problem of
time
Economic Performance: Municipal GDP, a second
best proxy
Omitted Variables: Characteristics of the
Electorate, campaign expenses, performance in
other social programs etc.
10. PROGRAM EXPANSION & PROGRAM SIZE (CONTINUOUS BY
CONTINUOUS INTERACTION
10 .0227073 .0112094 2.03 0.043 .0007373 .0446773
9 .0193125 .0090796 2.13 0.033 .0015169 .0371082
8 .0154817 .0071701 2.16 0.031 .0014286 .0295348
7 .0113654 .0053619 2.12 0.034 .0008563 .0218744
6 .0071328 .0036332 1.96 0.050 .0000117 .0142538
5 .0029571 .0021595 1.37 0.171 -.0012754 .0071896
4 -.0010005 .0016654 -0.60 0.548 -.0042646 .0022637
3 -.0045941 .0025044 -1.83 0.067 -.0095026 .0003144
2 -.0077118 .0036706 -2.10 0.036 -.0149061 -.0005175
1 -.0102893 .004804 -2.14 0.032 -.0197049 -.0008737
_at
abs0408
dy/dx Std. Err. z P>|z| [95% Conf. Interval]
Delta-method
-.02
0
.02.04
EffectsonPr(Elect)
.1 .2 .3 .4 .5 .6 .7 .8 .9 1
bffamtotalfam
Average Marginal Effects of abs0408 with 95% CIs
-*
n/a
+*
Program Size= 10-30 percent negative
statistical influence
Program Size= 40-60 percent no
significance
Program Size=70-100 percent Strong and
positive influence At 80 Percent coverage, a one
percent increase in coverage
increases probability by almost 2
Percent
11. MODEL COMPARISON: ROC CURVES
0.000.250.500.751.00
0.00 0.25 0.50 0.75 1.00
False Positive Rate
Municipal Model area: 0.6092 Bolsa Familia Model area: 0.6276
Reference
.
chi2(1) = 5.53 Prob>chi2 = 0.0187
Ho: area(xb30) = area(xb40)
xb40 1468 0.6276 0.0160 0.59627 0.65895
xb30 1468 0.6092 0.0160 0.57785 0.64059
Obs Area Std. Err. [95% Conf. Interval]
ROC Asymptotic Normal
. roccomp elect xb30 xb40, graph summary
12. CONCLUSIONS
The electoral effect associated with program expansion in 2006 for
national elections is mirrored at the local level.
Ecological Fallacy: Mayors expand the program had a higher probability
of being re-elected.
CANNOT link to individual voters or beneficiaries of Bolsa Familia.
Municipal-level Correlation
Mayors were rewarded for good governance. The electoral effect can
come from beneficiaries or from non-beneficiaries who see expansion as
beneficial in their municipality.
Zucco (2011): “Solidarity effect”
Non beneficiaries who knew a program participant 18 percent more likely
to vote for Lula
Higer program coverage at municipal level increases non-beneficiary
vote for Lula by 30 percent