kwaw s. andam international food policy research institute
DESCRIPTION
What are the Effects of Land Use Restrictions on Local Communities? Evidence from an Impact Evaluation of Costa Rica’s Protected Areas. Kwaw S. Andam International Food Policy Research Institute. Perspectives on Impact Evaluation: Approaches to Assessing Development Effectiveness - PowerPoint PPT PresentationTRANSCRIPT
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
What are the Effects of Land Use Restrictions on Local Communities?
Evidence from an Impact Evaluation of Costa Rica’s Protected Areas
Kwaw S. AndamInternational Food Policy Research Institute
Perspectives on Impact Evaluation: Approaches to Assessing Development Effectiveness
31 March - 2 April 2009, Cairo, Egypt
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Acknowledgements
Co-authors: Paul Ferraro (Georgia State Univ.) and Margaret Holland (Univ. of Wisconsin-Madison)
Evaluation Office, Global Environment Facility (GEF)
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Research Question
How different would socioeconomic outcomes have been in the absence of protected areas?
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Policy Context
Protected Areas:
• Most widely used conservation tool
• Role in climate change policy?
• More planned for future
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Motivation
Strong debate: how do protected
areas affect local people?
Most studies focus on
environmental impacts only
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
The Evaluation Challenge
Protected Areas placed selectively
Selection biasSelection biasLack of common supportLack of common support
Millennium Ecosystem Assessment (2005):
“Many protected areas were specifically chosen
because they were not suitable for human use.”
Empirical evidence: e.g. Costa Rica (Powell 2000; Sanchez-Azofeifa);
Nepal (Hunter & Yonzon 1993) ; Australia (Pressey 1995) ; United States
(Scott et al. 2001) ;
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Therefore, an evaluation must…
1) Objectively measure indicators of human welfare
2) Measure indicators before and after establishment of protected area
3) Measure indicators in both “treated” and “control” areas
4) Measure baseline characteristics that affect both location of protected areas and how indicators change over time
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Study site – Costa Rica
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Costa Rica’s Protected Areas
Protected before 1980
Protected after 1980
Forest cover 1960
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Approach
Estimate impact of protected areas established before 1980 on changes in census tract-level socioeconomic indicators between 1973 and 2000
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Key question
How different would socioeconomic
outcomes have been in the
absence of protected areas?
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Data
Quantitative indicators of change in socioeconomic outcomes (infrastructure, assets, poverty indices)
Control variables including land use productivity, forest cover (1960), accessibility (distance to markets, ‘road-less volume’), baseline (1973) indicator
Measures near time of and after establishment of protected areas
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Matching Methods
Select control communities similar to communities near protected areas (treated) in terms of pre-protection characteristics
Key assumption: without protection (and conditional on control variables), control and treated communities would, on average, have similar socioeconomic outcomes in 2000
Treatment: at least 20% protected before 1980
• >>> any remaining differences in outcomes are due to protection
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Variable Sample Mean ValueProtected Segments
Mean Value Control
Segments*
Diff Mean Value
Avg. Raw eQQ‡
Forest Area in 1960 (km2) UnmatchedMatched
15.81615.816
0.8338.618
14.9837.198
14.8976.543
High Productivity Land ◘ (km2)
UnmatchedMatched
0.8850.885
0.5401.012
0.345-0.127
0.2990.191
Medium Productivity Land (km2)
UnmatchedMatched
1.9221.922
0.4021.671
1.5200.251
1.4670.263
Medium-Low Productivity Land (km2)
UnmatchedMatched
4.3754.375
0.6503.584
3.7250.791
3.5940.719
Roadless Volume (km3) UnmatchedMatched
319.040319.040
21.322172.510
297.718146.530
291.960133.220
Distance to City (km) UnmatchedMatched
53.83653.836
35.34553.219
18.4910.617
18.4354.476
Poverty Index in 1973 UnmatchedMatched
1152411524
8574.20012016
2949.800-492.000
2950.100964.230
Percent of houses in bad condition in 1973
UnmatchedMatched
18.34318.343
15.01719.573
3.326-1.230
3.5041.313
Percent of houses in slums in 1973
UnmatchedMatched
1.4541.454
1.9421.486
-0.488-0.032
0.6030.150
Percent of houses without telephones in 1973
UnmatchedMatched
99.30499.304
93.60799.468
5.697-0.164
5.8100.144
Percent of houses without electricity in 1973
UnmatchedMatched
4.3654.365
3.0153.832
1.3500.533
1.2560.544
Percent of houses without access to water supply in 1973
UnmatchedMatched
39.46239.462
16.32140.458
23.141-0.996
23.0641.248
◘ Low productivity land is the omitted category.* Weighted means for matched controls.‡ Mean difference in the empirical quantile-quantile plot of treatment and control groups on the scale in which the covariate is measured.
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Results
1 2 3 4Outcome % Houses in
bad conditionPoverty Index
% Houses in slums
% Houses without telephones
Matching Estimators (Effect of protection on change in outcome 1973-2000)Covariate matching – Mahalanobis
-9.3***(2.5)
-987.6***(248.9)
-3.4***(0.8)
1.2(1.5)
Covariate matching – Mahalanobis with calipers
-3.0**(1.2)
-500.2***(134.6)
-1.3***(0.5)
1.3(1.3)
Replicating Conventional Methods (Effect of protection on outcome in 2000)
Difference in meansN treated(N available controls)
399(16002)
399(16002)
399(16002)
399(16002)
*** significant at .01; ** p<0.05; * p<0.10
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Additional Results
1 2 3
Outcome Unsatisfied basic needs (3 or 4)
% Houses without electricity
% Houses without water supply
Matching Estimators (Effect of protection on outcome in 2000)
Covariate matching – Mahalanobis
-14.1***(3.0)
-20.7***(4.6)
-10.7**(4.4)
Covariate matching – Mahalanobis with calipers
-8.5***(1.3)
-9.4***(1.5)
1.7(2.0)
Replicating Conventional Methods (Effect of protection on outcome in 2000)
Difference in means 8.7*** 16.8*** 31.1***
N treated(N available controls)
399(16002)
399(16002)
399(16002)
*** significant at .01; ** p<0.05; * p<0.10
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Caveats
Average effects only (adverse effects on subgroups may still be present)
Costa Rica may be very different from other nations
Indicators may not capture other aspects of poverty
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Ongoing work (1)
Census segments change over time from 1973 to 2000
Need to account for changes in unit of analysis over time
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
1973: 4,694 segments
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
2000: 17,264 segments
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Ongoing work (1)
Use simple areal weighting to reconcile 1973 segments with disaggregated segments in 2000, and analyze at level of 1973 segments
Findings confirm analysis with disaggregated segments
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Ongoing work (2)
Spillover effects: do effects of protection spillover onto nearby ‘unprotected’ communities?
Findings: None to small positive spillover effects
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Ongoing work (3)
Test sensitivity of results to hidden bias (unobservable characteristics that affect likelihood of protection and socioeconomic outcomes)
Test using Rosenbaum bounds tests So far, mixed results
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Summary
No evidence that protected areas had negative effects
Protected areas may have had positive effects
But conventional methods (erroneously) imply the opposite
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Discussion
Why differences in estimates?• Simply comparing post-protection outcomes
does not account for pre-existing differences correlated with poverty indicators
How does protection have positive effects?• Ecotourism• Protected area infrastructure• Initiatives to ease deforestation pressures• Ecosystem services?
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Thank you!
Kwaw S. AndamInternational Food Policy Research Institute
Addis Ababa, EthiopiaEmail: [email protected]: ++251.617.2507
www.ifpri.org
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Appendix
More comparisons with conventional estimates
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Replicating conventional methods1 2 3 4
Outcome % Houses in bad condition
Poverty Index % Houses in slums
% Houses without telephones
Matching Estimators (Effect of protection on change in outcome 1973-2000)
Covariate matching – Mahalanobis
-9.3***(2.5)
-987.6***(248.9)
-3.4***(0.8)
1.2(1.5)
Covariate matching – Mahalanobis with calipers
-3.0**(1.2)
-500.2***(134.6)
-1.3***(0.5)
1.3(1.3)
Replicating Conventional Methods (Effect of protection on change in outcome 1973-2000)
Difference in means 2.8*** 129.1*** 1.2** 23.4***
OLS2.4***(0.8)
709.6***(114.8)
0.6*(0.3)
11.3***(1.5)
N treated(N available controls)
399(16002)
399(16002)
399(16002)
399(16002)
*** significant at .01; ** p<0.05; * p<0.10