Similar to Forest conservation and agricultural intensification outcomes of a REDD+ initiative: A quasi-experimental assessment in the Brazilian Amazon
Similar to Forest conservation and agricultural intensification outcomes of a REDD+ initiative: A quasi-experimental assessment in the Brazilian Amazon (20)
Forest conservation and agricultural intensification outcomes of a REDD+ initiative: A quasi-experimental assessment in the Brazilian Amazon
1. Forest conservation and agricultural intensification
outcomes of a REDD+ initiative:
Cauê Carrilho, Carla Morsello
University of São Paulo, Brazil
cauecarrilho@gmail.com
A quasi-experimental assessment in the Brazilian Amazon
2. Introduction: REDD+ and agricultural intensification
Study site: Brazilian Amazon
Empirical strategy: impact assessment
Results, discussion and conclusions
3. REDD+: Reducing Emissions from Deforestation and Forest
Degradation and conservation, sustainable management
and enhancement of carbon stocks.
Curbing deforestation in developing countries to mitigate
climate change
4. United Nations Framework Convention on Climate
Change
Performance-based transfers regulated
by bi- or multilateral agreements
Different on-the-ground interventions,
such as:
Payments for Environmental Services
(PES), alternative livelihood incentives,
law enforcement and tenure clarification.
REDD+ operation
7. Agricultural intensification as a REDD+ strategy
Land sparing (Borlaug
hypothesis)
By fulfilling a certain demand for agricultural goods using
less cultivated area, agricultural intensification (increase in
land productivity) spares land which could then be used for
forest conservation.
8. Agricultural intensification as a REDD+ strategy
In tropical forested regions, forests are commonly converted to low-efficient production
systems. Given that there is plenty of deforested lands, agricultural intensification
seems to be the first choice to increase agricultural yields while reducing forest clearing.
9. Agricultural intensification as a REDD+ strategy
Rebound effect (Jevons'
paradox)
Agricultural intensification might drive more deforestation:
higher agricultural profitability can economically stimulate
farmers to clear more forests for agricultural expansion.
13. REDD+ initiatives are more often reducing
some deforestation (Simonet et al., 2019)
with mixed effects on economic well-being
indicators though (Duchelle et al., 2018)
REDD+ achievements
14. Which mechanisms explain REDD+ outcomes?
REDD+
outcomes
REDD+
interventions
Casual
mechanisms
Why REDD+ initiatives succeed or fail in promoting forest
conservation outcomes?
15. Which mechanisms explain REDD+ outcomes?
REDD+
outcomes
REDD+
interventions
Casual
mechanisms
Why REDD+ initiatives succeed or fail in promoting forest
conservation outcomes?
Agricultural
intensification?
Since agricultural intensification is often
addressed as a REDD+ strategy, we need to
understand whether raising agricultural
productivity contributes to farmers’ gains
while reducing deforestation. In addition,
whether REDD+ initiatives are avoiding the
rebound effect.
16. Research questions
1 Are REDD+ initiatives succeeding in promoting agricultural
intensification (i.e., increase in land productivity)?
2
Is agricultural intensification followed by win-win outcomes, in terms of
reducing deforestation and increasing farmers' yields?
17. Research questions
1 Are REDD+ initiatives succeeding in promoting agricultural
intensification (i.e., increase in land productivity)?
2
Is agricultural intensification followed by win-win outcomes, in terms of
reducing deforestation and increasing farmers' yields?
We estimated short (2 years) and long-term (7 years) effects of a REDD+
initiative on agricultural productivity, farm income and forest cover.
18. The REDD+ initiative
Project Sustainable Settlements in the Amazon
350 smallholders from the Transamazon highway region (Pará, Brazil)
Their main economic activities were cattle ranching and swidden agriculture
Goal: reduce deforestation rates and increase profitability in pasture
and agricultural plots.
Mix of interventions between 2012-2017
Forest protection
• PES – Payments for Environmental Services
• CAR – Cadastro Ambiental Rural
Agricultural production
• Technical Assistance
• Free agricultural inputs
Pictures from:
https://assentamentosustentavel.org.br
19. Data
2010 2014 2019
Before After
Panel data were
collected through
interview survey
along three years
in four treatment
and four control
communities.
20. Treatment and control communities
Treatment communities were
randomly selected among the
communities in which the
NGO intended to implement
the project.
Control communities were
selected based on a pre-
matching procedure to
identify communities with
similar characteristics likely to
influence both initiative
placement and land use and
income outcomes (e.g., forest
cover, distance to the main
road).
1
2
24. Empirical strategy
2010 2014 2019
Before After
Initial effects
Long-term effects
Difference-in-Difference
2012: REDD+ begins 2017: REDD+ ends
ATT = E (y1 – y0|D = 1)
y1: result variable under the treatment
y0: result variable in absence of treatment
D: 1 = household was treated; 0 = household was not treated
The intervention’s impact (i.e.,
participation in the REDD+ initiative)
was estimated by comparing the
changes in outcomes over time
between a treated and a control
group.
25. Empirical strategy
Result variables
Forest cover: forest cover (% of primary and secondary forest in the
household property).
Total farm income: the sum of the household agricultural yields, from crop
and livestock production (both own consumption and trade), obtained in
the twelve months prior to the interview survey.
Farm productivity: total farm income/cultivated area.
/ha
26. Empirical strategy
2010 2014 2019
Before After
Placebo test
Difference-in-Difference
2012: REDD+ begins 2017: REDD+ ends
ATT = E (y1 – y0|D = 1)
y1: result variable under the treatment
y0: result variable in absence of treatment
D: 1 = household was treated; 0 = household was not treated
DID parallel trend assumption was
confirmed using a placebo test over a
pre-treatment period (2008-2010) in
which no effects were detected.
Forest cover was estimated for 2008
through a retrospective question in the
2010 survey.
2008
27. Empirical strategy
NNM(2X) Nearest-Neighbor Matching estimator, matching each treated household to two of the most similar control households.
NNM(4X) Nearest-Neighbor Matching estimator, matching each treated household to four of the most similar control households.
PSM(kernel) Kernel-based Propensity score Matching, by which we compared households with the closest probability of being treated.
Matching
Matching variables
Normalized
differences
Raw Matched
Forest cover in 2008 (%
of land area)
0.55 0.09
Forest cover in 2010 (%
of land area)
0.52 0.04
Total land area in 2010
(ha)
-0.26 -0.01
Total income in 2010
(BRL)
-0.29 0.06
Household head age in
2010 (years)
0.47 0.12
Household members in
2010 (number)
0.08 0.07
Matching variables
Normalized
differences
Raw Matched
Total farm income in
2010 (BRL)
-0.48 0.04
Forest cover in 2010 (%
of land area)
0.52 0.11
Total land area in 2010
(ha)
-0.27 0.07
Total income in 2010
(BRL)
-0.30 0.19
Household head age in
2010 (years)
-0.48 -0.11
Household members in
2010 (number)
0.08 0.12
Matching variables
Normalized
differences
Raw Matched
Farm productivity in
2010 (total farm
income/farm area)
(BRL/ha)
-0.07 0.05
Forest cover in 2010 (%
of land area)
0.49 0.11
Total land area in 2010
(ha)
-0.25 -0.03
Total income in 2010
(BRL)
-0.28 0.07
Household head age in
2010 (years)
-0.46 -0.11
Household members in
2010 (number)
0.09 0.10
28. Empirical strategy
Result variables
Forest cover: forest cover (% of primary and secondary forest in the
household property).
A potential caveat in our data was the extent to which participants might have over-declared
their forest cover.
We cross-checked household self-reported forest data with remotely sensed data from the
Brazilian Annual Land Use and Land Cover Mapping Project (MapBiomas).
The NGO shared property boundaries of 43 from the 52 treated households in our sample.
Paired t-test and f-test of annual differences revealed that they are not statistically significantly
different in the means, and in standard deviations.
29. Results
2010 2014 2019
Before After
Initial effects
7.80% more of forest cover (6.2 ha)
Non-significant impacts on farm income and farm productivity
DID-matching
estimator
2010-2014
Forest cover (%) Farm income (BRL)
Farm productivity
(BRL/ha)
NNM(2X) 7.80* (4.36) -1,695.39 (6095.59) 1,661.36 (1537.27)
NNM(4X) 8.08* (4.57) 1,145.82 (6143.37) 1,490.37 (1582.70)
PSM(kernel) 10.32** (4.00) 537.27 (7557.73) 1,729.94 (1689.87)
30. Results
2010 2014 2019
Before After
Long-term effects
/ha
28.900 BRL more in annual farm income
3.173 BRL more per cultivated hectare
DID-matching
estimator
2010-2014
Forest cover (%) Farm income (BRL)
Farm productivity
(BRL/ha)
NNM(2X) 6.05 (4.36) 28,900.90** (12550.26) 3,173.98** (1569.17)
NNM(4X) 6.35 (4.29) 32,789.90** (12779.92) 3,158.33** (1557.45)
PSM(kernel) 6.67 (5.31) 37,016.37** (17246.40) 3,186.02* (1677.46)
Non-significant impacts on forest cover
33. Discussion
2010 2014 2019
Before After
Initial effects
/ha
Long-term effects
Agriculture intensification was unlikely a
driving factor for reducing deforestation
34. Discussion
PES was probably responsible for
deforestation reduction.
1) PES contracts were signed in the beginning
of 2013: participants could have reduced
deforestation in 2013 to received payments
in 2014.
2) Most control units were under CAR.
35. Discussion
PES was suspended in 2017 which could explain why deforestation
probably resumed in the long term.
The opportunity cost for conservation was not being compensated anymore: at least part
of the former beneficiaries would return to “business-as-usual” deforestation practices.
36. Discussion
PES was suspended in 2017 which could explain why deforestation
probably resumed in the long term.
The opportunity cost for conservation was not being compensated anymore: at least part
of the former beneficiaries would return to “business-as-usual” deforestation practices.
/ha
Agriculture intensification may have promoted a rebound effect on
deforestation?
If we had found deforestation resuming but no increase in agricultural profitability, we
would have concluded no rebound effect.
OR
37. Conclusions
Agriculture intensification contributed to poverty
alleviation: higher agricultural productivity was
accompanied by more farm income.
PES was probably responsible for
deforestation reduction
If PES had been maintained, long-standing forest conservation and
farm income increases might have followed