causal forces of deforestation in the brazilian amazon: does size matter? diana weinhold london...

Post on 17-Dec-2015

218 Views

Category:

Documents

1 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Causal Forces of Deforestation in the Brazilian Amazon:Does Size Matter?

  Diana Weinhold

London School of Economics 

Eustaquio ReisInstitute for Applied Economic Research (Ipea)

 Danilo Igliori

University of Cambridge 

III LBA Scientific ConferenceBrasilia, July 2004

Policy Problem

• Maintenance of biologically diverse ecosystems

• Land requirements and the opportunity costs of non-conversion

critical trade-off for developing countries

Motivation

• The economic literature on land use in developing countries has traditionally focused on efficiency and equity issues.

• Farm size is at central stage to discuss variation in the economic performance, and political and social issues in rural areas.

• Causes of deforestation have growing in interest recently in the land use literature.

• With few exceptions (Fearnside 1993, Walker et al 2000) the relationship between farm size and deforestation has not been properly addressed.

Objective

• In this paper we attempt to empirically investigate the existence of the relationship between farm size and deforestation.

• We use some traditional and some more recent econometric results for model evaluation in panel data and a comprehensive data set on land use in the Brazilian Amazon.

• We study the question of whether the size composition of agricultural establishments in the Amazon region plays a role in determining the rate of deforestation.

Data• The data used is part of a database (Desmat) managed by

IPEA/DIMAC (The Directorate of Macroeconomic Studies of the Institute of Applied Economic Research, Brazil).

• This is a data panel for all the municipalities of Brazilian Legal Amazon (AML) including thousands of variables on major economic, demographic and geo-ecological aspects.

• Desmat includes spatially detailed geo-ecological information available in GIS and socio-economic sources, in particular Demographic and Economic Census data observed in 5-year periods from 1970 to 2000.

Explaining Levels

• Is the size composition of establishments correlated with the level of land clearing?

• Dependent Variable: Log of Cleared Land, 1995

• Size composition: shares of private land within different size classes (7 classes, from tiny - < 10 ha, to Giant - >100,000 ha)

Explaining Levels

Control variables• Property rights: shares of private land within

different ownership status (owners, sharecroppers, renters, squatters)

• Area: municipality and establishments• Other: state dummies, distance to state and federal

capitals,percentage of good soil, km of navigable river, and natural vegetation variables.

• Omitted categories:  Large establishment share (>1000 and <5000), and Owner-titled land share

Variables Model 1 Model 2 Model 3 Constant -0.8364

(-0.70) 0.1896 (0.14)

-0.8231 (-0.69)

Log (Municipality area) -0.1218 (-2.70)

-0.0264 (-0.69)

-0.1434 (-3.16)

Log (total establishment area, 1995) 1.0883 (22.57)

0.9053 (27.07)

1.1073 (22.95)

Tiny = share of private area in estabs < 10 ha

1.9427 (4.20)

2.1041 (4.68)

Small = share of private area in estabs >10 and <100 ha

0.9223 (3.27)

0.8571 (3.13)

Medium = share of private area in estabs >100 and <1000 ha

0.2492 (0.67)

0.3998 (1.06)

Exlarge = share of private area in estabs >5000 and <10000 ha

-0.3167 (-0.57)

-0.1880 (-0.35)

Super = share of private area In estabs >10000 and <100000 ha

-0.4189 (-1.03)

-0.3370 (-0.82)

Giant = share of private area In estabs >100000 ha

-0.6751 (-1.87)

-0.5911 (-1.61)

Renter share of private area in estabs rented

-1.2682 (-0.45)

-0.4782 (-0.22)

Sharecr share of private area in estabs w/sharecropping

-1.7473 (-0.99)

-2.7940 (-1.63)

Squatter share of private area in estabs with squatters

0.7702 (2.45)

0.4291 (1.25)

No obs. 257 257 257 R-squared 0.9432 0.9354 0.9448

Dependent Variable: Log of Cleared Land, 1995

Explaining Levels

Results

• In levels it is the small and tiny establishments and squatters that are clearing the most as a proportion of their land.

Explaining Changes

• Is the change of composition of establishments size correlated with changes in the extent of land clearing?

• We then model the change in cleared land from 1985 to 1995 and condition on levels in 1985.

• Dependent Variable = Growth of cleared land, 1985-1995

• Control variables for changes: total private land, shares of estab sizes and ownership.

• other control variables and omitted variables as in the analysis for levels

Variables Model 1 Model 2 Model 3 constant 0.8603

(0.10) 2.1286 (0.25)

-1.3654 (-0.17)

Log(cleared land,1985) -4.5913 (-6.38)

-3.7828 (-6.20)

-4.1020 (-5.89)

Log (Municipality area) 0.3755 (0.74)

0.4657 (1.18)

0.1429 (0.33)

Growth of total establishment Area, 1985 – 1995

8.7006 (12.97)

7.0566 (14.79)

9.0469 (12.90)

Log(total establishment area in 1985)

4.1865 (4.03)

3.2624 (4.94)

3.9812 (4.12)

Change in Tiny share, 1985 - 1995

8.2549 (2.27)

10.5373 (2.71)

Change in Small share, 1985 - 1995

2.4569 (0.90)

1.4786 (0.57)

Change in Medium share 1985 - 1995

-2.3446 (-0.67)

-1.0732 (-0.31)

Change in Exlarge share 1985 - 1995

-3.4477 (-0.76)

-2.1218 (-0.49)

Change in Super share 1985 - 1995

-3.7032 (-0.97)

-3.2415 (-0.91)

Change in Giant share 1985 - 1995

-10.1914 (-2.25)

-11.4055 (-2.30)

Tiny share, 1985 3.5362 (0.74)

1.9054 (0.40)

Small share, 1985 5.0644 (2.05)

4.0580 (1.67)

Medium share, 1985 -0.6826 (-0.22)

-0.8366 (-0.26)

Exlarge share, 1985 -3.7399 (-0.67)

-2.4220 (-0.43)

Super share, 1985 -2.0576 (-0.65)

-0.9762 (-0.32)

Giant share, 1985 -4.9973 (-1.18)

-5.0556 (-1.05)

Change in Renter share 1985-1995

3.7721 (0.18)

6.7894 (0.35)

Change in Sharecropper Share, 1985-1995

-16.0937 (-1.36)

-19.6620 (-1.45)

Change in Squatter share, 1985-1995

9.2743 (3.52)

7.4037 (2.81)

Renter share, 1985 10.4524 (0.51)

19.8466 (1.04)

Sharecropper share, 1985 -21.2954 (-1.11)

-23.0401 (-1.13)

Squatter share, 1985 8.0450 (2.50)

7.9108 (2.28)

No obs. 257 257 257 R-squared 0.7522 0.7487 0.7785

Dependent Variable = Growth of cleared land,

1985-1995

Explaining Changes

Results

• This more or less confirms the first result,

• Caution: there is evidence that the reason squatters are associated with more clearing has to do with the fact that they operate on very small scales.

Causality

• The model for changes has the additional benefit that, controlling for the initial levels of the variables, many of the hidden omitted-variable problems so common in levels will be eliminated from the analysis.

• This does not mean that omitted variable problems have been eliminated entirely, however, for again, a correlation in changes could be due to other, unmeasured and/or omitted variables.

Causality 2

• we adopt some recent econometric techniques for panel model evaluation to test for a type of Granger-causality.

• Essentially we shall test whether knowledge about the size composition of establishments will help us predict, out-of-sample, the future rate of deforestation (and vice versa).

• We use panel model evaluation techniques suggested by Granger, C.W. and L. Huang (1997). Evaluation of Panel-data models: Some Suggestions from Time-series, unpublished manuscript, UC San Diego.

Causality 3Applications for the Amazon

• Andersen, L., C. Granger, E. Reis, D. Weinhold, and S. Wunder (2003). The Dynamics of Deforestation and Economic Growth in the Brazilian Amazon, Cambridge University Press: Cambridge.

• Weinhold and Reis (2003) . Land Use and Transportation Costs in the Brazilian Amazon, Working paper, AAE, UW-Madison.

Causality 4

• In our first evaluation attempts we find no evidence of causality.

• This suggests that the correlations found are just that and we should be careful before making policy decisions with the hope of altering the rate of land clearing.

Conclusions

• Our cross-section modelling using regression analysis for levels and changes in cleared area suggests correlation between size and deforestation.

• Tiny establishments apparently clear more land proportionally.

• These results don’t pass a more rigorous causality test.

• We should be cautious and extra attention should be paid to trying to figure out the underlying causal mechanisms between establishment size and deforestation.

• There is large scope for further research on the matter.

top related