Cedula da Terra: Impact Evaluation (2000-2006)

1
Cedula da Terra Impact Evaluation (2000-2006)

• Coordination of study
• José Maria Silveira, University of Campinas
• Marlon Barbosa, SRA/MDA
• Claudia Romano, World Bank
• Henrique Neder Econometric analysis

2
The Cédula da Terra Program (PCT)

• PCT was the predecessor of the Credito Fundiario
  basic features are similar
• PCT was implemented between 1997 and 2002,
  benefiting 16,000 families in 5 states of
  Northeast Brazil.

3
Objectives of study

• Impact of PCT on physical capital, income,
  health, food security, and social capital (not
  presented)
• Compare evolution of PCT to small farmers (not
  presented here)

4
Basics features of evaluation

• Used baseline available from a profile survey
  carried-out in 2000 and a return survey in 2006
  including a new pipeline control group selected
  in 2006
• Data for treatment and pipeline collected with
  same instruments specifically designed for study
• Sample 320 Beneficiaries in 104 settlements
• and 410 control pipeline
• Distributed in five states (BA, CE, MA, PE, MG)

5
Dealing with Selection Bias

• Pipeline comparators partially address problem
  of latent heterogeneity
• Propensity score matching (PSM) using all
  relevant information collected in surveys,
  reduces selection bias due to observed
  characteristics
• Double-difference analysis reduces selection bias
  due to unobserved characteristics with effects
  fixed in time (only impact on income uses
  single-difference due to lack of baseline data on
  income for pipeline)
• The three methods together should significantly
  reduce problems with selection bias

6
Issues to consider about selection bias

• Given that control pipeline was selected from
  groups applying to Program 6 years after
  beneficiaries, there is concern that latent
  heterogeneity is not dealt with satisfactorily by
  use of pipeline.
• Thus, PSM and double-difference are important
  tools for adressing heterogeneity due to
  obervables and to unobserved characteristcs with
  effects fixed in time, respectively.
• Only impact on income uses single-difference as
  recall on pipeline is not reliable

7
PSM

• PSM uses 608 observations after discarding those
  with missing variables
• After common support reduction and matching, the
  paired sample for nearest neighbor is 258
  treatment and 116 pipeline
• After matching all variables are statistically
  similar between groups except 1 at 10 and one at
  5, out of 29 variables
• balancing property test is accepted at 1

8
PSM summary results
9
Comparison of characteristics before and after
matching
10
Graphs with Pscore distribution before matching
11
Graphs with Pscore distribution after matching
N(1)
12
Impact on physical capital double-difference
and matched sample
13
Impact on health and food security

• All methods show no significant impact on
  occurrence of illnesses and on proportion of
  households indicating a situation of security in
  obtaining enough food for family consumption.

14
Impact on income single difference and matched
sample
15
Summary of main results

• Impact on physical capital is highly significant
  and indicates a PCT impact on total assets
  equivalent to 151 of initial capital effect is
  high on productive capital with increase of 154
  over initial value
• Impact on income from agricultural production is
  highly significant and indicates an impact
  equivalent to 315 of initial ag income but
  significance of impact on total income varies
  from null to positive depending on method used.