Econometric modeling of the impact of forest conservation policies on the provision of ecosystem services Juan Robalino, Universidad de Costa Rica and CATIE Liberia, 2016
Joint work with a lot of people Alexander Pfaff, Duke University Arturo Sanchez, University of Alberta Catalina Sandoval, UCR Laura Villalobos, Gothenburg University Diego Herrera, University of Vermont Paul Ferraro, Georgia State University Francisco Alpizar, CATIE and EfD a nd many others…
Research question background How is forest related to the provision of ecosystem So, what can we do to protect forest? services? 1) Create protected 1) Reduces the amounts of CO2 in the atmosphere areas 2) Might reduce vulnerability to changes in climate and to 2) Pay landowners to extreme weather events protect their forest And other services like Water and air purification, and scenic beauty…
Why is evaluation important? When evidence is missing… decisions are not based on what works.. despite good intentions, decisions are not optimal Advantages of evaluating Cost effective measures can be identified Generates credibility and increases support and willingness to contribute
Expected impacts of conservation policies on deforestation Protected areas forbid deforestation Payments are incentives to conserve forest
Change in the expected impact The impact of the policy could be reduced due to: Illegal behavior: Illegal deforestation Break the contract Missing the target Parks and payments might be located in areas where no deforestation is going to take place (illustration) Leakage effect People might increase deforestation else where The impact of the policy could also be increased: Propagation and contagion of conservation due to interactions
Simple graphical representation Land Rents f’’ L f Market Park Park
Treatment Effect Factual Contrafactual (Treated) (Untreated) Policy No policy in in parcel X parcel X Treatment effect in Parcel X = The Factual Deforestation Rate - The Counterfactual Deforestation Rate
Estimating counterfactuals Two very common ways of estimating counterfactual deforestation rate: Use areas where no conservations policies have been implemented Use the same area before the policies was implemented
Differences in means Wittemeyer et al. 2008
Before and after comparisons Bruner et al. Science 2001
How do we identify the impact? Ideally, like in the natural science, experiment with random assignment Then, other deforestation drivers are canceled out in expectation Controls for observable as well as unobservable factors However, policies are rarely applied randomly Controlling for observable factors: Regression analysis Matching Strategies
Matching Strategies Treated observations: Untreated Observations: Plots inside National Parks Plots away from National Parks
Advantages and Disadvantages Advantages Reduce the bias due to the lack of random assignment Less dependent on the functional form assumed Disadvantages Unobservables might bias the estimation of the effect Loss of observations (degrees of freedom) Standard Errors
Impacts after bias correction 10 9 8 7 6 5 4 3 2 1 0 Parks in CR (86-97) Brazil (00-04) Acre (00-04) Diff. in means Matching
Impacts after bias correction 10 Parks reduce deforestation but not as much as originally estimated! 9 8 7 6 5 4 3 2 1 0 Parks in CR (86-97) Brazil (00-04) Acre (00-04) Diff. in means Matching
Difference in the impacts We estimated average treatment effects of parks However, treatment effects might vary between parks and within parks We will test if different land characteristics and governance have different effects
Impact according to land characteristics in Costa Rica 16 Avoided deforestation during 86-97 (%) Gentle 14 12 10 8 6 Close 4 Close High 2 Steep Low Far Far 0 Elevation Proximity to San Jose Proximity to National Slope Roads Characteristics of Protected Areas
Impacts by land characteristics in the Brazilian Amazon Avoided deforestation during 2000-2004 (%) 7 Close 6 5 Close 4 High 3 High Low Far Far 2 Low 1 0 Precipitation Fertility Distance to Roads Distance to cities Características
Impacts by land characteristics in the low lands in Bolivia Avoided deforestation 2001-2005 (%) 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Distance to Roads Distance to cities Close Far Characteristics
Minas Gerais 2.5 Low *** Avoided deforestation 1996-09 (%) 2 Low *** 1.5 High *** High 1 *** 0.5 0 Slope Distance to roads Características
So, what did we learn? Protect high deforestation threat areas Forest in plains Forest close to roads Forest close to cities Forest in soils with high fertility But what about levels of restrictions of resource use inside protected areas?
Acre State in the Brazilian Amazon impacts according to level of restriction 4 Avoided deforestation 2000-04 (%) 3.5 3 2.5 2 1.25 1.5 ** 1 0.5 -0.63 0 -0.5 -1 Sustainable use Integral protection
What about leakage effects? Rents f’’ L f Merkets Park
Empirics Factual Counterfactual (Treatment) (Untreated) No National Plot Plot National Park X X Park Treatment effect in plot X = Factual Deforestation Rate - Counterfactual Deforestation Rate
Leakage effects on 86-97 deforestation Far from the park entrance Close to the park entrance 12 12 Close to Roads Close to 10 10 Roads 8 8 6 6 4 4 Far from Far from Close to Roads Roads Roads Far from 2 Far from 2 Roads Close to Roads Roads 0 0 0-5 Km 5-10 Km -2 0-5 Km 5-10 Km -2
Previous evidence from CR shows that parks’ impacts on wages are positive close to the entrances of the parks 20 18.68 14.47 15 9.57 Local and immigrant 10 8.46 workers Local workers 3.94 5 0 Close to Park Close to the Far from the -1.08 Entrance Entrance -5
… and that close to entrances, females benefit the most… 18 16.46 16 14 12 Females 10 8 6.47 6 Males 4 2 0 By Sex
Park and forest effects on vulnerability to climate • Evidence of the effects of forest on floods – Tan-So et al. 2016 in Malaysia (in the wet season) – Pacay et al. 2015 in Honduras (in the dry season) • Effects of protected areas on diseases – PA are negatively correlated with malaria, acute respiratory infections and diarrhea (Bauch et al. 2015) • Effects of protected areas on natural shocks – In Mexico, they reduce exposure but if exposure occurs, they do not reduce the adverse effects (Roman et al. 2016)
Simple Model of PES Payments increase the returns of forests Rents The reduction of forest will be in the segment between f y f’ All landowners between f’ y L will try to enroll land in the program P f’ L f Market
Impacts after bias correction 3 2.5 2 1.5 1 0.5 0 PES in Mexico (03-06) Alix- PES in CR (97-00) PES 97-00 in CR (00-05) PES 00-05 in CR (00-05) Garcia et al. 2012 Diff. in means Matching
Impacts after bias correction 3 2.5 2 1.5 1 0.5 0 PES in Mexico (03-06) Alix- PES in CR (97-00) PES 97-00 in CR (00-05) PES 00-05 in CR (00-05) Garcia et al. 2012 Diff. in means Matching
Payments’ effects by land characteristics Avoided deforestation 2000-05 (%) 4 Gentle 3.5 Low 3 2.5 2 1.5 High 1 Steep 0.5 0 Slope Distance to San José Characteristics
Payment effects by offices 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 Cañas Limón Nicoya Palmar Norte Pococí Sarapiquí San Carlos San José Oficinas de FONAFIFO
Spillover effects Leakage effects Evidence for Mexico (Alix-Garcia et al. 2012) In poor municipalities, deforestation increases next to enrolled parcels In less poor municipalities, deforestation decreases next to enrolled parcels There might be behavioral reasons too What if payments are only given to landowners that are going to deforest? Experiment in Costa Rica where people are excluded from payments
Behavioral spillover effects (Alpizar et al. 2015) Experiment: one hour survey to landowners After 30 minutes, we gave them 10 dollars and ask for a donation for an environmental NGO At the end, we gave them 10 dollars more and ask for a donation again, but we provide incentives or exclude from those incentives We test three rules Exclude those that gave a lot and include those that gave little Exclude those that gave little and include those that gave a lot Randomly choose who gets the subsidy
Behavioral spillover effects Effect on those who receive 800 600 Effects on contributions 400 Net effect 200 0 -200 Effects on those excluded -400 -600 Subsidy to those that Subsidy to those that Subsidy to those Additionality rule Reward rule Exogenous rule gave little gave a lot randomly chosen
Poverty impacts of PES What happens when PES coverage increases by 10%? Impact of PES on poverty 0.020 0.02 0.016*** 0.015 0.015 Poverty (%) 0.010 0.01 0.005 0.005 -0.013*** 0.000 0 Total effect -0.005 -0.005 High Low -0.010 -0.01 -0.015 -0.015 Slope *** indicates significance at 1%
Recommend
More recommend