HR: 11:20h
AN: B42A-05 [Abstracts]
TI: Effects of Roads on Deforestation in the Brazilian Amazon: new evidence using census-tract data and
addressing endogeneity
AU: * Alexander, P
EM: ap196@columbia.edu
AF: Columbia University, 2910 Broadway
Hogan Hall 104, New York, NY 10025
United States
AU: Reis, E
EM: ejreis@ipea.gov.br
AF: IPEA/Rio, Avenida Presidente Wilson, Rio de Janeiro, RJ 20020
Brazil
AU: Robalino, J A
EM: jar101@columbia.edu
AF: Columbia University, Department of Economics
420 West 118th Street, New York, NY 10027
United States
AU: Bohrer, C
EM: bohrer@vm.uff.br
AF: University Federal Fluminense, Departamento de Geografia, Rio de Janeiro, RJ 20020
Brazil
AB:
Regional landscape projections constitute a valuable tool for identifying policies that can address the development needs of
Amazonian populations while minimizing loss of ecosystem integrity. Changes in transport costs in particular are expected to
affect both livelihoods and deforestation, and there are significant planned policies which would involve investments to
lower such costs.
Recent work has advanced the literature in terms of both its questions and its empirical approach. Specifically, more has
been learned about the development gains attendant to loss in forest area. Also more "dynamic" and more appropriate
approaches have been taken to infer factors' effects. One particular result of interest from this work has been that roads'
effect depend on the setting. Most controversially, it suggests that more roads could lower deforestation rate in cleared
areas.
In reconsidering that idea, this paper addresses the greatest limitation upon the previous research. Specifically, the use of
census-tract-level data over time provides over 20 times the observations. This permits much greater control for
unobservables that could easily drive any previous results. Further, it permits separate regressions for different settings
to directly address effects of roads.
The results suggest that setting does indeed matter and that roads' effects may fall with clearing. However, it does not
appear that lowering the cost of transport ever lowers rates of deforestation. While to this point we have results only for
our first time period, this is a strong important result.
The second part of the paper considers the possibility that the effects of roads are overestimated. The reason for this is
endogeneity of roads, i.e. they are not randomly but purposefully located. Should their location be driven by factors that
would themselves lead to greater forest clearing, empirical results appearing to indicate effects of roads could be showing
effects of those factors.
Project team members have been studying what lead to the creation of parts of the road network, and we examine empirically
what land-use settings are the most likely to give rise to new roads. We then use this result to set up cleaner comparisons
of locations with and without new roads, controlling for other factors by comparing locations that are similar in terms of
those factors. Early results find that this lowers estimated effects of roads but they remain quite significant.
DE: 9360 South America
DE: 6329 Project evaluation
DE: 6334 Regional planning
DE: 0400 Biogeosciences
SC: Biogeosciences [B]
MN: 2004 AGU Fall Meeting