Do conservation incentives increase the effectiveness of protected areas?

Size: px

Start display at page:

Download "Do conservation incentives increase the effectiveness of protected areas?"

Rebecca McCormick
6 years ago
Views:

1 Introduction Do conservation incentives increase the effectiveness of protected areas? Elı as Cisneros, Jan Bo rner, Stefano Pagiola and Sven Wunder ZEF Bonn, ZEF Bonn, CIFOR Rio de Janeiro, World Bank Washington Intern workshop on evaluating forest conservation Initiatives, December 10, 2013 Elı as Cisneros, Jan Bo rner, Stefano Pagiola and Sven Wunder

2 Context Protected areas reduce deforestation worldwide Nelson and Chomitz (2011) Multiple-use reserves work better than strictly protected reserves Nelson and Chomitz (2011) Protected areas with PES schemes reduce deforestation versus non-protected areas Honey-Rosés et al (2011)

3 Research question Do conservation incentives increase protected area effectiveness?

4 Case study Bolsa Floresta Program Sustainable Amazon Fund (FAS) Worlds largest conservation incentive program (10 M hectares) In 15 sustainable use reserves With one reserve being the first certified Brazilian REDD project Inhabitants are highly reliant on forest and fish resources See also: Börner et al (2013)

5 Bolsa Floresta Program components Bolsa Floresta Familia (family component) Conditional cash transfer Bolsa Floresta Social (social component) Improve public services Bolsa Floresta Associação (association component) Communities allocate funds freely Bolsa Floresta Renda (income component) Production line investments

6 Bolsa Floresta s potential impact mechanisms on welfare and conservation Improved living conditions Cash transfers and development programs Reduced internal pressures Opportunity costs of rule-compliance Monitoring and enforcement Reduced external pressures Building local conservation alliances See: Börner et al 2013

7 Study design Spatial information on sustainable use reserves Sustainable use reserves with Bolsa Floresta Yearly data on deforestation Yearly data on set of controls

Introduction Sustainable use reserves and deforestation Unit of analysis: 20 x 20 km grid cells (2007-2011) Source:

8 Introduction Sustainable use reserves and deforestation Unit of analysis: 20 x 20 km grid cells ( ) Source: Calculations from J Schielein (ZEF) based on data from PRODES project (INPE) Elı as Cisneros, Jan Bo rner, Stefano Pagiola and Sven Wunder

9 Introduction Bolsa Floresta Program Unit of analysis: 20 x 20 km grid cells ( ) Source: Calculations from J Schielein (ZEF) based on data from IBAMA and FAS Elı as Cisneros, Jan Bo rner, Stefano Pagiola and Sven Wunder

10 Introduction Base sample I 100% protected areas in 2007 (5% tolerance) 658 controls I 100% BFP treated areas in 2011 (5% tolerance) 125 treated I Buffer around treated excluded I Brazilian border cells excluded Elı as Cisneros, Jan Bo rner, Stefano Pagiola and Sven Wunder

11 Empirical Strategy Average treatment effect on the treated (ATT): E[Def 1i Def 0i BFP i = 1] Conditional independence assumption: E[Def 0i X i, BFP i = 1] = E[Def 0i X i, BFP i = 0] Impact estimation: Regression, Matching mean comparison, Post-matching treatment regression

12 Control variables (G) Grid cell characteristics Initial forest (2007), past deforestation ( ), market distance, remoteness indices, land use classes (2008) (G + N) Neighboring cells characteristics Neighboring initial forest, neighboring past deforestation (G + N + B) Border characteristics Neighboring sustainable use reserve status Sources: PRODES project & TerraClass - INPE, IBAMA, SIPAM, IBGE

13 Selection on observables Matching estimation Post-Matching estimation Treatment prediction on observables Log-likelihood estimation (logit); Dep variable: Dummy on BFP treatment status (1) (2) (3) Init forest 3520*** (0585) 0330 (0907) 0330*** (0907) Past deforestation -0270*** (0047) -0138*** (0042) (0042) Market distance -0042*** (0005) -0062*** (0006) -0063*** (0006) Distance to rivers -0012*** (0003) -0017*** (0003) -0017*** (0003) Distance to roads 0027*** (0002) 0028*** (0002) 0028*** (0002) Agricultural area -0005** (0002) -0004** (0002) -0004** (0002) Pasture area -0005*** (0001) -0005*** (0001) -0005*** (0001) Secd vegetation 0002*** (0000) 0002*** (0000) 0002*** (0000) Neigh init forest 0063*** (0000) 0063*** (0012) Neigh past deforest -0003*** (0077) -0003*** (0000) Neigh US reserve (0311) Controls group G G + N G + N + B

14 Selection on observables Matching estimation Post-Matching estimation Mean comparison of deforestation Mean Mean Difference log Diff Control Treated T - C T - C Deforestation (ha) *** -0812*** (404) (035) (928) 0078 Observations Groups

15 Selection on observables Matching estimation Post-Matching estimation The effects of Bolsa Floresta on deforestation Mean comparison after matching: ln Yearly Deforestation Controls group G Without calipers With calipers (075 SD) Estimate Pairs/ Estimate Pairs Treated P-score matching -0147** 4, ** 4,750 (0058) 625 (0058) 625 Mahalanobis matching -0197*** 3, *** 2,875 (0065) 625 (0064) 575 Controls group G + N P-score matching , ,775 (0049) 625 (0047) 615 Mahalanobis matching -0125* 3, *** 2,825 (0068) 625 (0068) 565 Controls group G + N + B P-score matching , ,775 (0048) 625 (0047) 615 Mahalanobis matching , ,600 (0068) 625 (0059) 535

16 Selection on observables Matching estimation Post-Matching estimation Matching covariate balance Before and after matching on covariates G + N, without caliper Neigh past def Neigh ini forest Secd vegetation Pasture Agriculture Dist road Dist river Dist market ast deforestation Ini forest (2007) Unmatched Matched (p score) Matched (mahalanobis) Standardized difference in means Source: Calculations from J Schielein (ZEF) based on data from IBAMA and FAS

17 Selection on observables Matching estimation Post-Matching estimation Post-Estimation strategy Average treatment effect on the treated (ATT) with panel data: E[Def 1i,t Def 0i,t X i,t, BFP i,t = 1]

18 Selection on observables Matching estimation Post-Matching estimation The effects of Bolsa Floresta on deforestation Weighted cluster-robust OLS estimates; Dep variable: ln Yearly Deforestation Before After Matching Matching (1) (2) (3) (4) Bolsa Floresta treatment -0112** ** -0155** (0056) (0069) (0067) 0079 Controls G Yes Yes Yes Yes Controls N Yes Yes Yes Yes Controls B Yes Year effects - - Yes Yes Clouds - - Yes Yes R-sq N Groups

19 Main findings : Deforestation decreased where the Bolsa Floresta Program is implemented 12-20% yearly avoided deforestation corresponds to ha between 2007 and 2011 ICDP/PES schemes can additionally conserve forests within protected areas Evidence of detrimental effects at the borders of reserves

20 Main findings : Deforestation decreased where the Bolsa Floresta Program is implemented 12-20% yearly avoided deforestation corresponds to ha between 2007 and 2011 ICDP/PES schemes can additionally conserve forests within protected areas Evidence of detrimental effects at the borders of reserves Further research: 20x20 km grid cells Analysis on buffer areas

21 Main findings : Deforestation decreased where the Bolsa Floresta Program is implemented 12-20% yearly avoided deforestation corresponds to ha between 2007 and 2011 ICDP/PES schemes can additionally conserve forests within protected areas Evidence of detrimental effects at the borders of reserves Further research: 20x20 km grid cells Analysis on buffer areas Caveats: Selection on unobservables (ex RDS, APA, AM)

22 Do conservation incentives increase the effectiveness of protected areas? ZEF Bonn, ZEF Bonn, CIFOR Rio de Janeiro, World Bank Washington Intern workshop on evaluating forest conservation Initiatives, December 10, 2013 defpicpng

Sven Wunder CIFOR. Payments for environmental services (PES) conditions for success

Sven Wunder CIFOR. Payments for environmental services (PES) conditions for success Sven Wunder CIFOR Payments for environmental services (PES) conditions for success Structure I. PES: definition and concept II. III. IV. Cases of implementation Challenges of design Conclusions and perspectives