Economic Dynamics and Forest Clearing A Spatial Econometric Analysis for Indonesia

1. Economic Dynamics and Forest Clearing A Spatial Econometric Analysis for Indonesia David Wheeler Dan Hammer Robin Kraft Susmita Dasgupta Brian Blankespoor Development Research Group World Bank 2012

2. Presentation Outline • Motivation • FORMA: A New Approach to Monitoring Tropical Forest Clearing • Trends in Indonesian Forest Clearing • Model Specification • Data • Econometric Results • Conclusions

3. Motivation • Forest clearing accounts for 15% of annual GHG emissions (WRI 2010). • Most forest clearing occurs in developing countries. • Forest conservation will be difficult as long as forested land has a higher market value in other uses. • Actual success of compensation schemes will depend on program designs tailored to the economic dynamics of forest clearing. • Economic returns to forest clearing vary widely over space and time (RFF 2011).

4. The Case of Indonesia • Forest Clearing in Indonesia is heavily driven by palm-oil and wood-processing exports to fast-changing Asian Markets. • Availability of Monthly database for forest clearing at 1 km resolution since 2005 from FORMA (Forest Monitoring for Action).

5. FORMA • Constructs deforestation indicators from MODIS- derived data on the incidence of fires and changes in vegetation color as identified by the Normalized Difference Vegetation Index. • Calibrates to local deforestation by fitting statistical model to the best available information on actual deforestation in the area. • Incorporates biological, social and economic diversity by monitored territory into blocks and fitting the model to data for the parcels in each block. • FORMA is calibrated using the map of forest cover loss hotspots (FCLH) for 2000- 2005 published by Hansen et al. 2008. • Applies the fitted model to monthly MODIS indicator data for the period after 2005. • Output • A predicted forest-clearing probability for each 1 sq km parcel outside of previously-deforested area as identified in the FCLH map- for each month. • Selection of parcels with probabilities exceeding 50%. • An index of forest clearing from the above.

6. Large-Scale Forest Clearing in IndonesiaDecember 2005-December 2010 Indonesia’s natural forest area in 2000 was 951,160 sq. km. (WRI, 2010)

7. Annualized Forest Clearing in Indonesia Top 5 Provinces in January 2007

8. FORMA/Indonesia Sumatra

9. FORMA/Indonesia Sumatra

10. FORMA/Indonesia Riau Province

11. Forested in 2000 270 km 167 mi Forest in 2000

12. Cleared by 2005 (Hansen) 270 km 167 mi Forest in 2000 Cleared 2000 - 2005

13. Probability 50 – 60% 60 – 70% 70 – 80% 80 – 90% > 90% Forest in 2000 Cleared 2000 - 2005 FORMA 12/2005

14. 1/2006

15. 3/2006

16. 6/2006

17. 9/2006

18. 12/2006

19. 3/2007

20. 6/2007

21. 9/2007

22. 12/2007

23. 3/2008

24. 6/2008

25. 9/2008

26. 12/2008

27. 3/2009

28. 6/2009

29. 9/2009

30. Probability 50 – 60% 60 – 70% 70 – 80% 80 – 90% > 90% Forest in 2000 Cleared 2000 - 2005 Cleared by 10/2009 270 km 167 mi

31. Changes in Forest Clearing, 2006-2011* By Kabupaten Increase No Change Decrease * January - August

32. Model: Building Blocks • Proprietor/ occupant of a forested area considers the relative profitability of maintaining/ clearing the area. • In each period, the agent compares the present-value profitability of sustainably harvested forest products with the clear-cut value of forest products and the cleared land’s present value profitability in its best use (e.g., plantation, pasture, smallholder agriculture, settlement). • Forest clearing dynamics are different in cases where commercial exploitation rights are well- or poorly-defined. • Determinants of forest clearing highlighted in prior research: Population Scale & Density, Distance from Markets, Quality of Transport Infrastructure, Agricultural Input Price, Topography, Precipitation, Soil Quality, Zoning of Land.

33. Model Specification π = Expected relative profitability of forest clearing pe = Vector of expected prices for relevant products (palm oil, sawlog) qe = Vector of expected demands for relevant products (palm oil, sawlog) n = Rupiah-denominated input cost per unit of output t = Transport cost per unit of output (mean travel time to nearest city of 50,000+) c = Communications cost per unit of output (coverage by mobile phone networks) ie = Expected interest rate xe = Expected exchange rate (rupiah/dollar) g = Quality of governance from investors’ perspective (KPPOD index) r = Regulatory quality (KPPOD index) u = Officially-designated use (protected forest, palm oil plantations, timber plantations, logging concessions) h = Population density y = Unskilled wage rate w = Precipitation (forest-burning is more difficult when rainfall is heavier) s = Slope of the terrain (mean slope, std. deviation) Expectations: π’(pe)>0, π’(qe)>0, π’(n)<0, π’(t)<0, π’(c)<0, π’(ie)<0, π’(xe)>0, π’(g)>0, π’(r)<0, π’(h)>0, π’(y)>0, π’(w)<0, π’(s)<0

34. Data

37. Findings • Significant roles for lagged (+) changes in product prices, demands, (+) exchange rate and (-) interest rate. • Highly variable lags: • < 1 year for product prices. • around 1 year for product demands and exchange rate. • close to 2 years for real exchange rate. • Significant roles for (+) communication infrastructure, (+) zoning for palm oil plantations, and physical factors: (+) uncleared forest in 2000, (-) terrain slope and (-) rainfall. • Insignificant roles for local governance quality, access time, population density, poverty rate, protected area status and zoning for timber plantations.

38. Conclusions • Forest clearing is an investment highly sensitive to • Expectations about future forest product prices & demand; • Changes in the cost of capital; • Relative cost of local inputs; • Cost of land clearing. • Opportunity cost of forested land fluctuates widely with changes in international markets, local weather conditions and decisions by financial authorities about exchange and interest rates. • Forest conversation programs are unlikely to succeed if they ignore the economic dynamics of forest clearing.