Discussion Paper
No. 2017-83 | October 09, 2017
Alexander Pfaff, Juan Robalino, Eustaquio J. Reis, Robert Walker, Stephen Perz, William Laurance, Claudio Bohrer, Steven Aldrich, Eugenio Arima, Marcellus Caldas and Kathryn R. Kirby
Roads & SDGs, tradeoffs and synergies: learning from Brazil’s Amazon in distinguishing frontiers
(Published in The Sustainable Development Goals—Assessing interlinkages, trade-offs and synergies for policy design)


To inform the search for SDG synergies in infrastructure provision, and to reduce SDG tradeoffs, the authors show that road impacts on Brazilian Amazon forests have varied significantly across settings. Forest loss varied predictably with prior development – both prior roads and prior deforestation – and in a spatial pattern suggesting a synergy between forests and urban growth in such frontiers. Examining multiple roads investments, the authors estimate impact for settings of high, medium and low prior roads and deforestation. Census-tract observations are numerous for each setting and reveal a pattern, not consistent with endogeneity, that confirms our predictions for this kind of frontier. Impacts are: low after relatively high prior development; larger for medium prior development, at the forest margin; then low again for low prior development. For the latter setting, the authors note that in such isolated areas, interactions with conservation policies influence forest impacts over time. These Amazonian results suggest ‘SDG strategic’ locations of infrastructure, an idea they suggest for other frontiers while highlighting differences in those frontiers and their SDG opportunities.

JEL Classification:

O12, O13, H23, H41, Q23, Q24, Q56


Cite As

[Please cite the corresponding journal article] Alexander Pfaff, Juan Robalino, Eustaquio J. Reis, Robert Walker, Stephen Perz, William Laurance, Claudio Bohrer, Steven Aldrich, Eugenio Arima, Marcellus Caldas, and Kathryn R. Kirby (2017). Roads & SDGs, tradeoffs and synergies: learning from Brazil’s Amazon in distinguishing frontiers. Economics Discussion Papers, No 2017-83, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2017-83

Comments and Questions

Anonymous - Referee Report 1
October 24, 2017 - 08:18
This paper addresses a well-known tradeoff between development and environment related SDG dimensions, namely the impact of road infrastructure expansion on land cover change at tropical forest margins in the context of the Brazilian Amazon. The paper’s conceptual model departs from von Thünen’s theory of decreasing agricultural rents as distance from market centers increases. The authors expand the classic model by considering varying levels of prior development at forest frontiers along the distance gradient. This leads to the prediction (and empirical confirmation) that deforestation outcomes of road construction are highest at intermediate levels of development with implications for SDG sensitive infrastructure planning; for example, siting new roads in areas with high prior development levels can sometimes minimize the development-environment tradeoff. I have enjoyed reading the paper and recommend it for publication. Below I provide a few comments that the authors may want to consider in revising the manuscript. • Can we think of the distance-development gradient as a cross-sectional representation of the forest transition curve? If so, I am wondering what mechanisms are at play when the hypothesized inverted U-shaped relationship with deforestation is strong (e.g. national context characterized by low development, commodity export orientation, weak environmental governance) versus weak (e.g. high development, domestic market dependent agriculture, strong environmental governance). I suggest expanding the discussion section a bit to characterize settings in which the proposed conceptual model is most useful to inform transport infrastructure planning. My gut feeling is that a country’s general development level (i.e. reliance on agriculture) and quality (including ambition) of environmental governance are the key mediators of the tradeoff relationship between road building and forest cover. • The Brazilian model of Amazon colonization differs quite a bit from that of other tropical forest margins, i.e. important primary roads where build to connect urban centers, including for geopolitical reasons, with suitability for agriculture often not being among the top priorities. What would be the implications of higher levels of endogeneity of roads’ locations for the model and the spatial patterns depicted in Figure 2 – perhaps also something to expand on in the discussion. • The regression equation could be justified with a bit more detail: Why is “prior deforestation” not entering the model in % terms like the dependent variable or vice versa? Why does “prior roads” enter in t-1 and “prior deforestation” in t? Given that the dependent variable contains deforestation in t, wouldn’t a t-1 form be preferable? Does the subscript, “t, t-1” represent the variable in first difference? • What is the benefit of running a WLS regression as opposed to including the size of census tracts in an OLS approach? Doesn’t this result in parameter estimates being biased towards relationships in large tracts? Also, using a dependent variable in % form may have unfavorable distributional implications? Why are you not using the first difference of absolute changes in deforested areas? • Please number equations • Regression tables: Please include standard errors in parenthesis (instead of p-values) and indicate significance levels using stars. Where robust or clustered standard errors calculated? • Some of the references are incomplete, e.g. Foster and Rosenzweig (2003)

Anonymous - Referee report 2
October 30, 2017 - 10:30
The authors model the impacts of road construction on deforestation depending on prior development. This is a very interesting topic and from my point of view is worth to be published. However I do have a few points that the authors should consider to improve the paper: - The authors justify their study as providing the means for a more sustainable road planning. However, I think that in areas where we have low prior development and according to their model rather low impacts on deforestation, deforestation might be more critical to the natural functioning of the ecosystem. In these areas, which are the very valuable in terms of ecosystem functioning, road effects (apart from deforestation) would be much more devastating than in the intermediate developed regions. This should be discussed at the end of the paper.- The paper would certainly benefit from some language editing and proof reading as there are some grammatical mistakes and incorrect or unclear sentences- In the introduction the authors often mention tradeoffs but without explicitly naming the tradeoffs- “Rural roads may generate less of the kinds of benefits that tend to be counted officially, yet more of the benefits that tend not to be” Can you give an example?- Please use equation numbers and refer to them- In the deforestation formula some words are crossed out- Try to avoid starting sentences like “Table 4´s second column…” Make a statement what the results show, refer to the actual content of the columns and write (Table 4) behind the statement.- Overall the captions of tables and figures are not self-explaining. Please provide more information here- In the tables you write yes in many cells but it is unclear what that means? Is it telling us that it is significant? Please provide the info in the table header or/and more detailed values in the table cells- Figure 3: What does high or low variation mean? Variation of what?

Alexander Pfaff et al. - Reply to referee reports
December 06, 2017 - 09:31
see attached file