Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2016 Sep 1.
Published in final edited form as: Ann Assoc Am Geogr. 2015 Aug 18;105(5):1026–1040. doi: 10.1080/00045608.2015.1060924

What Drives Indirect Land Use Change? How Brazil's Agriculture Sector Influences Frontier Deforestation

Peter Richards 1
PMCID: PMC4789281  NIHMSID: NIHMS731638  PMID: 26985080

Abstract

From 2000-2005 high returns to soybeans set off an unprecedented expansion of agricultural production across Brazil. The expansion occurred concurrently to a sharp rise in deforestation, leading academics and policy makers to question the extent and means by which the growing agricultural sector was driving regional forest loss. In this article we consider and question the underlying drivers of indirect land use change, namely the potential impact of soybean expansion on beef prices and of land use displacement, via migration. We then present field level results documenting the displacement process in northern Mato Grosso and western Pará States of the Amazon. Our results question the extent to which tropical Amazon deforestation is attributable to land use displacement; however, we argue that the agricultural sector may drive deforestation through other channels, namely through regional land markets.

Keywords: Indirect Land Use Change, Land Use Displacement, Amazon, Agriculture

(1) Introduction

Cultivating grain and oilseeds in the cerrado and Amazon regions1 of Brazil has variously raised hope, concern, and ire from academics, environmental activists, policymakers, and the Brazilian government. The soybean sector has brought economic prosperity to agricultural regions in Brazil, and has been connected to local increases in educational attainment, poverty reductions, and income growth (VanWey et al. 2013, Weinhold, Killick and Reis 2013). Soybean exports have also contributed to Brazil's past decade of economic growth, during which fifteen percent of the country's population, or about thirty million individuals, lifted themselves out of poverty and into the ranks of the global middle class. This economic growth, however, has not come without cost, and agricultural expansion in Brazil has led to questions over the extent to which agricultural led land use change leads to increased carbon emissions and losses in biodiversity. In this article we focus on the impact of Brazil's agricultural sector on tropical deforestation in the Amazon.

Agriculture drives land use change both directly, through the direct conversion of forest or natural land covers to cropland (Rudel et al. 2009, Morton et al. 2006, Lambin and Meyfroidt 2011), and indirectly, through its ability to influence land use change elsewhere (Barona et al. 2010, Richards, Walker and Arima 2014, Arima et al. 2011, Lapola et al. 2010). In this article we focus on the indirect impacts of soybean production in the Brazilian Amazon, a topic that has evolved into a point of both interest and contention in discussions of agricultural impact.

We begin with a review of the literature on soybean expansion in Brazil, and of the work to date on indirect land use change. We pay particular attention to the behavioral and land use frameworks that provide the conceptual basis for indirect land use change, namely the neoclassical migration theories of Sjaastad (1962) and Todaro (1980) and the rent based spatial framework of von Thünen (1966). We then focus on land use displacement, which has been widely cited as a mechanism underlying indirect land use change, through an examination of occupations and migration paths of land-selling and land-purchasing farmers and ranchers in areas surrounding a strategic Amazon highway. We find that the displacement of smallholder farmers or ranchers by soybean farmers may have had a limited role in driving Amazon deforestation in our study region, or across the basin at large. However, we suggest that the agriculture sector may nonetheless affect deforestation decisions more broadly across the Amazon basin, via its ability to influence regional land markets.

(2) The Soybean Boom

Brazil's soybean sector traces back to the 1970s, when Asian and European importers sought to develop new source regions for high protein animal feeds. Initially, Brazilian soybean production was limited to the nation's southern states: Paraná, Santa Catarina, and Rio Grande do Sul. However, in the late 1980s and 1990s, soybean agriculture expanded northward into the cerrado and later the transitional forests of the Amazon region with the colonization of Brazil's interior, and with the adaptation of new seed varieties to the highly weathered soils and low latitudes of the tropics (Warnken 2002, Jepson 2006b, Jepson 2006a, Brown, Jepson and Price 2004). In the following decades, soybean production would also benefit from regional improvements in infrastructure, and by macroeconomic and policy shifts. Economic integration in Mercosul, the reduction of export taxes on primary goods, and the removal of many price supports for domestically consumed crops, for example, opened new opportunities for Brazilian farmers (Helfand and Rezende 2004, Goldsmith and Hirsch 2006). When Brazil floated its real in 1999 it then also sent a shock across its economy that set Brazil's soybean boom into motion (Richards et al. 2012).2

From 2000 to 2005, total soybean production in Brazil increased from 32 to 51 million tons. In the State of Mato Grosso alone the area of soybean production more than doubled. The agricultural boom brought new investment capital to the region and reshaped its economy (VanWey et al. 2013, Weinhold et al. 2013); however, it also reshaped its landscapes. When Amazon deforestation reached a peak at more than 27,000km2 in 2004, academics, environmental interests, policy makers, and even agricultural companies began to fret over the linkages between deforestation and agricultural expansion (Hecht 2005, Morton et al. 2006, Walker et al. 2009b). Yet although the concurrent expansion of agriculture and deforestation suggested a relationship between these two land use changes, linking the two processes was not necessarily straightforward.

During the soybean boom years of 2001-2004 direct conversions of forest to cropland (in Mato Grosso) only amounted to approximately 5,400km2, or about six percent of the 95,000km2 of Amazon forest lost during these years (Morton et al. 2006). During that period, in the cerrado and Amazon regions of Brazil, agriculture was also largely limited to only a handful of regions; in contrast, deforestation was widespread (e.g., see Figure 1). If the soybean boom was driving the enormous spike in tropical deforestation during the early 2000s, it wasn't doing so by clearing new lands for farmland. Several questions thus needed to be answered: if agriculture was driving the concurrent increases in deforestation, then through what means was it doing so, and to what extent? Work in land change science suggested that soybean production might be driving deforestation in Brazil's Amazon not only through direct clearings, but indirectly, through another set of indirect channels, and thus be responsible, indirectly, for a greater percentage of the region's forest loss.

Figure 1.

Figure 1

Open in a new tab

Deforestation and agricultural expansion, 2001-2011. Agriculture is concentrated at the bottom of the image; deforestation is relatively widespread. Deforestation data based on INPE's PRODES dataset (2012); Agriculture data based on NDVI classifications by Spera, et al, (2014).

(3) Indirect Land Use Change

In the late 2000s researchers began to suggest that soybean expansion in Brazil might not only drive forest loss directly, through the direct clearing of forest for agriculture, but via what has become known as indirect land use change. Indirect land use change occurs when a land use change in one location, say in an expanding agricultural district, leads to a land use change in another location. The channels through which this might occur, however, have not always been clear. To date, research has suggested two principal channels: (1) the occupation of enough pasture to significantly reduce local beef supplies and raise beef profits, which we refer to as a price effect; and (2) the displacement of ranchers and small farmers to frontier regions. We summarize and expand on both of these channels in this section.

The Price Effect

In 2008, articles by Searchinger. et al. (2008) and Fargione, et al. (2008) argued that US and European biofuel policies were not only encouraging ethanol production in the northern hemisphere, but leading, indirectly, to the displacement of soybean production and rangelands to other continents. Their arguments contributed to concern over the impact of biofuel policies on global forest cover, and the possibility that policies designed to reduce C02 emissions were, in fact, leading to broad scale environmental losses (Keeney and Hertel 2009, Hertel, Tyner and Birur 2010, Nassar et al. 2011). Conceptually, their arguments rested on partial or general equilibrium models, where increased corn production (to meet ethanol quotas, for example) was resulting in decreased soybean output and, in turn, raising global prices for food crops such as soybeans. Rising soybean prices were then making soybean production in new agricultural frontiers in Brazil, or elsewhere in South America, viable. This perspective on indirect land use change, or what we refer to here to as a price effect, continues to form the basis for much of the international scale work on indirect land use change (Hertel et al. 2010).

Land use change scientists adopted a similar partial equilibrium approach to explain the linkages between agriculture and deforestation in the Amazon. This work suggested that a rise in soybean prices could result in an initial loss in pastures that, in turn, could lead to a subsequent shock in beef prices and a later expansion of cattle production in frontier areas (Walker 2014). If expanding croplands were indeed leading to an increase in beef prices and expanding cattle herds, then the soybean sector could be, indirectly, responsible for increases in deforestation (Walker et al. 2009b). To date, much of the theoretical interest in modeling the indirect effects of soybean production on land use change dynamics in Brazil has situated this partial equilibrium approach of the international scale work on indirect land use change within a spatially explicit framework. This work has largely been based on a reconfiguration of the rent-based model associated with von Thünen.

In the Thunian rent model production choice is a function of minimizing opportunity costs of production, with land being allocated to whichever land use provides the highest returns (von Thünen 1966, Walker 2001). The Thunian model has long been used as theoretical fabric from which to understand regional scale political-economic and historical change (Cronon 1991, Peet 1969, Walker et al. 2009a, Richards 2011). In research in the Amazon it has been regularly employed to explain the growth of the Amazon's agricultural and pastoral sectors as a function of declining transportation costs or changing market conditions (Walker et al. 2009a, Mann et al. 2010, Bowman et al. 2012, Aldrich et al. 2006).

In a Thunian organization of land uses, a rise in soybean prices would trigger an expansion of soybean production, typically at the expense of a less intensive land use such as pasture or forest cover. This has been well documented in the Amazon. In Mato Grosso State, when soybean prices reached record levels, approximately 5,400km2 of forest and 6,000km2 of pastures were converted to new cropland (Morton et al. 2006). If the magnitude of this pasture conversion had been significant enough to influence regional prices for beef prices and lead to a subsequent expansion of frontier ranching, then the initial agricultural expansion might have led, indirectly, to land use change more broadly across the Amazon.

Unfortunately, while this trajectory is theoretically intuitive, when the magnitude of pasture losses are considered against the magnitude of the Brazilian herd, the occupation of pastures likely had no more than a minimal, if any, impact on regional beef prices. For although, in Mato Grosso, 6,000km2 of pasture were converted to agriculture during the soybean boom years, this only amounted to an area that probably supported no more than 600,000 animals. While this is by no means a small number relative to Brazil's total herd, the scale of this loss was minimal. In total, 600,000 animals amounts to approximately 0.29 percent of Brazil's 200 million-plus animal herd, or 2.3 percent of the Mato Grosso's 26 million head of cattle. Even if local beef prices were highly elastic to supply, any effect would still be clouded by the many other factors that broadly affect returns to beef production and local supplies, from internal trends toward intensification and confinement (e.g. Cohn et al. 2014), to macroeconomic shifts in the exchange rate, changes in global beef demand or supplies from India, Australia, or the United States, or the price of the soybeans and grains used for cattle fattening. We thus turn to the other theoretical approach underpinning recent work on indirect land use change, namely land use displacement.

The Displacement Effect

Land change research has suggested that agricultural change in Brazil might be indirectly driving regional deforestation through so-called land use displacement (Arima et al. 2011, Lapola et al. 2010, Barona et al. 2010, Andrade de Sá, Palmer and di Falco 2013, Gollnow and Lakes 2014). Land use displacement refers to a process that begins with the pushing or displacement of small farmers or ranchers off areas that are converted to large scale agricultural systems (Meyfroidt, Rudel and Lambin 2010). If these displaced ranchers or small farmers proceed to invest their capital and skills in opening new (and perhaps even larger) small farms or ranchlands at the frontier, then Brazil's soybean sector might be leading, indirectly, to deforestation. And if ranchers are not only being displaced, but are arriving to the frontier with more capital, then their ability to clear new land and increase their herd size could be larger than previously.

The displacement process follows from neoclassical migration theory, where migration is a function of the spatial distribution of returns to labor across a set of present and potential work locations, and where individuals maximize their returns to labor by moving to locations where the demand for their skills and capital is highest (Taylor et al. 1996, Massey et al. 1993, Todaro 1980, Sjaastad 1962).3 For displaced small farmers or ranchers selling land in established or consolidating agricultural regions, where land prices are generally higher, purchasing new properties in a frontier area enables rural landowners to increase their access to land.

In theory, land use displacement serves as a mechanism for bringing human and financial capital to the frontier, and thus enables land use change. The migration of displaced small farmers and ranchers thus solves two of the traditional obstacles to agricultural expansion in Brazil, namely the scarcity of capital and labor relative to the region's immense supplies of potentially productive agricultural or pasture land (Ozorio de Almeida and Campari 1995, Sewastynowicz 1986, Bates and Rudel 2004). In this sense, soybean production in agricultural regions reshapes the production decisions in frontier areas by pushing source of human and financial capital to the frontier.

We expect that the out-migration of small farmers and ranchers from soybean producing regions in Mato Grosso would be highest when returns to soybean production are highest. During periods of high returns farmers will have a greater incentive to expand, given the increase rent potential of agricultural land, and given that farmers will have more capital on hand for investing. Consequently, land should appreciate more rapidly during periods when agriculture is expanding. As land appreciates in value in areas around agricultural districts, including areas presently occupied by small farmers or ranchers, landowners in these regions may seek to capitalize the value of their land assets by selling. A subset of these owners may then seek to invest this capital, and their already acquired skills and network connections, in re-establishing their production methods in new locations.

In theory, the 6,000km2 of ranchlands displaced in Mato Grosso by soybean production during the soybean boom might have led to the widespread displacement of small farmers and ranchers to regions across the Amazon Basin. If a significant subset of these small farmers and ranchers relocated to forest areas and re-established their production at the expense of the forest, then a significant and substantial fraction of the 160,000km2 of Amazon forest lost during the last decade might be attributable, indirectly, to the growth in regional soybean production. Unfortunately, and despite the widespread citation of land use displacement as a mechanism underlying regional scale indirect land use change, there is little field level evidence to support assertions that this process is indeed occurring. In this research we sought to provide this evidence. Assuming that displacement would be most pronounced where agriculture and cattle expansion were occurring within relative proximity to forest loss, we thus turned our attention to Brazil's BR-163 region, the location of both one of the fastest growing agricultural districts, and some of the highest rates of forest loss in the Amazon.

(4) Field interviews in the BR-163 Region

Our analysis concentrated on a 250 × 800km corridor of Brazil's federal highway BR-163, a road that stretches across the States of Pará and Mato Grosso. This specific study area stretches from the scrub forests of the cerrado, near Lucas do Rio Verde, through the transitional Amazon-cerrado forests of Sorriso and Sinop, and to the high forests of the Amazon of western Pará State (see Figure 2a).

Figures 2a-3c.

Figures 2a-3c

Open in a new tab

Area of interest, showing major cities and rivers (left, 2a); 2b (center). Deforestation and agricultural expansion; 2c. Approximate locations of property purchases identified in the interviews, categorized by purchaser occupation.

During the past decade soybean production in this region doubled, nearly all of it at in north-central Mato Grosso (see Figures 2b and 3a, Spera et al. 2014). The cattle herd in the Mato Grosso portion of the BR-163 region also doubled; farther north, in Novo Progresso, in the State of Pará, the herd quadrupled, to 636,000 animals (IBGE 2013). Deforestation, as we show in Figures 2b and 3b, was also exceptionally high here. Approximately 20,000km2 of humid tropical forest were cleared in this area since 2000, or slightly more than ten percent of forest loss in the entire Amazon. There was nowhere else in the Amazon where both soybean expansion and deforestation occurred at such magnitude, and in such close proximity, as in the BR-163 corridor. If land use displacement was occurring, there was nowhere where it should have been more observable than here. We thus went to this region in 2011 and 2012 prepared to document land use displacement. We specifically focused on finding those who we predicted would be most likely to be affected by the process, namely land purchasers.

Figures 3a-3b.

Figures 3a-3b

Open in a new tab

Cross sections of agricultural area and remaining forest cover in the BR-163 corridor, for 2000-2010. Y-axis is distance from the southern extent of the study area in Figures 3a-3c. Area refers to total area used for soybean production or in forest cover. Agriculture is concentrated in the southern extent of the study area. The highest rates of forest loss in this region, from 2000-2010, were concentrated in Mato Grosso.

To document land use displacement we targeted recent land purchasers. To first identify land purchasers, and to contextualize our structured interviews with land purchasers we conducted unstructured interviews with key officials in the public and private sectors (e.g., at the mayor's office, farmers’ and ranchers’ unions, and agricultural supply stores). In addition, we solicited documents from recent land sales from cartorios, or local title offices, which included the names and addresses of both land purchasers and land sellers. Where possible, we sought out the buyers indicated in the cartorio documents, or from the discussions with the key public and private officials, for interview. While it was often difficult to arrange interviews due to travel costs, schedule difficulties and remoteness, we succeeded in identifying and interviewing 54 recent land purchasers.

We specifically targeted owners of properties larger than 250 hectares, and where purchases were made since the year 2000, or since the commencement of the soybean boom.4 We focused on purchased properties under the premise that any displacement effect requires a transfer of control over a parcel, and that a sale and purchase of a property is the clearest indicator of a permanent transfer of control.5 Our sampling strategy follows that of theory-based sampling, or specifically, purposive sampling (Curtis et al. 2000), given that our sampling strategy follows our hypothesis that land use displacement, if occurring, is most likely to be linked to in and out-migrations. In total we interviewed 21 ranchers, mostly in western Pará or in the very north of Mato Grosso, and 33 farmers, all of whom planted soybeans (see Figure 2c).

The mean size of the purchased properties associated with the interviewed owners was 3,249 hectares. The largest purchase was a massive latifundia ranch and forest property of 39,000 hectares. The smallest purchase was a 121 hectare parcel. For properties consisting of cropland, the mean planted area was 1,238 hectares. For cattle ranches, the mean pasture size was 1,127 hectares. Nearly every property also included significant areas of forest reserve. The mean forest area across the sample was 2,148 hectares.

Of the 54 land purchasers we interviewed, 31 had changed land use after purchasing their properties. Fifteen of the land changes consisted of forest clearings for pasture. Eleven purchasers had cleared forest for agriculture; and eight had converted pasture areas to cropland.6 We found no instances of lands purchased in cropland or pasture that were then subsequently changed to a less intensive land use, i.e., cropland left to pasture or pasture left fallow or to revert to secondary forest.

In 23 of the interviews the purchaser had moved after the purchase. Thirty-one of the purchases were made by landowners already in the region. Many of the purchasers who had purchased their properties in the early 2000s had migrated to the region (see Figure 4). Few of the interviewees had purchased their properties during the middle of the decade, when returns to agriculture dipped with the revaluation of the real. We interviewed a number of farmers and ranchers who had purchased properties since 2008; however, nearly all of these more recent land purchasers already owned properties in the region.

Figure 4.

Figure 4

Open in a new tab

Number of interviewed land purchases, classified by year of purchase and as local purchases or in-migrations.

Every one of the in-migrating land purchasers (e.g., those who moved after their land purchase) succeeded in increasing their land area after selling their previous holding(s) (Table 1). Those who purchased land in Pará generally had less land in their previous locations than their counterparts who traveled to Mato Grosso. However, in-migrants to Pará achieved a greater increase in access to land after purchase. This should be expected, given the relatively low cost of land in western Pará. Many of the purchases were followed by investments in land use change. Our principal interest, however, lay in whether these purchases were also leading to displacement.

Table 1.

Examples of pre and post-migration property sizes for in-migrants.

County Year Purchased Previous New (Difference) %Change
Sinop 2000 350 500 150 43
Claudia 2001 0 280 280 n/a
União do Sul 2001 420 6000 5580 1329
Castelo do Sonhos 2001 73 3235 3162 4331
Santa Carmen 2002 720 2400 1680 233
Castelo do Sonhos 2002 36.3 1500 1464 4032
Guarantã do Norte 2002 314 624 310 99
Novo Mundo 2002 0 900 900 n/a
Sinop 2002 200 900 700 35
Santa Carmen 2003 1000 1900 900 90
Santa Carmen 2003 180 900 720 400
Novo Progresso 2003 157 1635 1478 941
Novo Progresso 2003 73 550 477 653
Novo Progresso 2003 580 1936 1356 234
Novo Progresso 2003 26 193 167 642
Novo Progresso 2003 400 5000 4600 1150
Novo Progresso 2004 0 779 779 n/a
Castelo do Sonhos 2004 0 850 850 n/a
Novo Progresso 2006 100 400 300 300
Claudia 2007 85 360 275 324
Novo Progresso 2008 73 330 257 383
Claudia 2010 38 121 83 218
Novo Progresso 2011 0 726 726 n/a
Averages
Purchases in the State of Pará are listed in bold. Overall 209 1392 1183 857
Mato 301 1353 1052 308
Grosso
Pará 126 1428 1301 1407
Open in a new tab

If displacement was occurring, and was connected to the purchase and occupation of existing pasture areas by soybean farmers, then we would expect that farmers purchasing new properties would indicate that they had purchased their lands from ranchers, and that the previous owners had relocated to a frontier region. Alternatively, those ranchers that were migrating to and purchasing properties in western Pará might indicate that they had recently sold properties at the periphery of an agricultural area, or that their old ranch was now being farmed for soybeans. In each interview we thus questioned the land buyer on their previous locations, and on the present day location of the land seller.

(5) Does displacement drive indirect land use change?

If displacement of ranchers or small farmers was occurring in the study region, we expected to find that ranchers buying new ranches in Pará State were being pushed or displaced from newly created cropland areas, or that farmers were displacing former ranch owners. Either finding would have supported the hypothesis that ranchers are being displaced by the expanding agricultural sector. What we found instead was that land purchasers were either already residing near to where they were purchasing these new properties, or they were migrating to the region from elsewhere across Brazil (Figure 5). Many of the farmers we interviewed thus originated from either the agricultural districts in southern Brazil, or the cropping districts around Lucas do Rio Verde or Sorriso. They were arriving from areas where high cropland values or small properties made acquiring large agricultural properties difficult or even impossible. To expand production, they had relocated their operations to Mato Grosso. We observed a similar pattern for ranchers. Ranchers purchasing properties in the region, mostly in the northern end of BR-163, were also either already located locally, or were arriving from consolidated ranching regions elsewhere in Brazil (or in a few cases, from Paraguay). While these findings do broadly support suggestions that farmers and ranchers will migrate to increase their access to land, they do not suggest either that ranchers are arriving from new soybean regions. In fact, none of the ranchers we interviewed in western Pará had arrived from the agricultural regions in Mato Grosso.

Figure 5.

Figure 5

Open in a new tab

Origin location of interviewed ranchers and farmers.

In each interview we also questioned each land purchaser on the present-day location and occupation of the former property owner.7 If those interview subjects that were aware of the status of the previous owner were to suggest that the previous owner had sold their land and relocated to more marginal frontier regions, whether within or beyond our study region, then we could confirm the so-called displacement effect. However, again, despite identifying and interviewing subjects that we hypothesized would be most likely to be tied to land use displacement, we could not clearly document this effect. Instead, we found several different processes at hand. In cases where farmers purchased land from other farmers, many of the sellers had continued on in their occupations by relocating to the periphery of the agricultural frontiers. Some moved to the agricultural districts expanding outward around Sinop and Sorriso; others moved into the cerrado regions of western or central Mato Grosso, or to the Matopiba region, in the northern cerrado area that lies to the east of the Amazon biome. Others moved to Santarém, Parà, western Mato Grosso, and in one case, the south (for family reasons). However, while these displacement patterns follow the general thesis of our expected migration channels (to increase access to land), they did not suggest that these land sales were tied to displacement or deforestation.

For those who purchased land from non-farmers, the pathways of the prior landowners was less clear. In some instances the purchaser had purchased the property from a timber company. In these cases it was difficult to pinpoint how resources tied to timber extraction were being redistributed, spatially; however, interviews suggested that timber companies had moved on to less exploited regions in northwestern Mato Grosso, or to southeastern Amazonas State. In other cases land purchasers had bought land from local ranchers. Given that we expected that ranchers were being displaced by expanding farmland, and that these ranchers might e responsible for regional deforestation, we expected to find that land selling ranchers had moved north to re-establish their ranches. What we found instead were stories of ranches being sold, but that few, if any of these ranchers were moving northward. In some cases ranchers sold due to family issues, for example, a divorce had mandated that a couple's assets be sold and divided. In others an owner living in São Paulo or Cuiabá had passed and his children did not share the same interest in maintaining an Amazon ranch. We also found stories of previous owners simply shifting their assets out of ranching. In one case the previous owner had sold their rural property to purchase an auto dealership. In another, an ex-rancher had become involved in politics and local government. In these cases the former land owners had effectively abandoned ranching as an occupation. While the circumstances and particular stories of each land seller varied, two common threads united the interviews. First, if the former owners did not sell under financial distress, or due to family reasons, then they had sold to capitalize on the increased value of their investments. Second, in none of these cases did the land purchase lead directly to a land use displacement.

Discussions with local recent land purchasers offered insight into their motivations, and why agricultural expansion may not be displacing ranchers and small farmers. First, soybean production benefits greatly from returns to scale, and production on small properties may not be financially viable. Compiling a large property from a community of smallholder farmers would involve significant and likely prohibitive, transactions costs. Instead, in-migrating farmers sought to carve new farms from either the degraded forests of exhausted timber concessions, or in some cases, from patrimonial properties owned by absentee landowners, where properties were large and documents were registered. For those seeking to maximize access to land, occupying degraded timber tracts provided an attractive option, as these properties could be acquired relatively cheaply. Second, farmers stressed the need for titles and property documents, the papers that facilitate access to credit, whether for crop inputs or machines.

Certainly, our set of 54 interviews with land purchasers is not exhaustive, and more work must be conducted before we can fully rule out land use displacement as the principal underlying factor in the correlation between agricultural expansion and deforestation in the Amazon. However, our findings and interpretations were broadly confirmed and supported by discussions with key officials in our unstructured interviews. They are also supported by an analysis of migration trends from Brazil's 2010 census, which we discuss in the following section.

(6) Migration

To support our field observations, we compared our interview findings against migration trajectories recorded in Brazil's 2010 micro-scale census data (IBGE 2010), the Brazilian equivalent of the US long form census questionnaire. The micro-scale questionnaire covers more than 20 million respondents, or approximately one-tenth of Brazil's population, and includes information on where each respondent lived five years earlier, in 2005. To support our interview findings, we turned to the census data to provide additional information on the magnitude and direction of out-migration from soybean producing regions, and of in-migrants to western Pará. If out-migrants from soybean regions are relocating to areas of high deforestation, we expect that some of these migrants will be captured in the micro data from Brazil's census. Alternatively, if in-migrants to counties of high deforestation are arriving from expanding soybean areas, this too should be observable in the population subset recorded in the census data.

Of the more than 20 million respondents to the long form questionnaire, more than 1.6 million indicated that they had lived in another municipal district in 2005. Approximately 4,000 of these had left one of the soybean producing counties in our study area.8 In contrast, 6,000 respondents had moved into these counties, or that net migration here was positive. We obtained similar results when we consider only rural residents; net migration was positive, indicating that more rural residents were arriving to the soybean regions than were leaving them. In Figures 6a and 6b we show the direction of in and out-migrants from the soybean districts in Mato Grosso. The majority had moved westward or south, but not into western Pará. Counties in northern Mato Grosso and western Pará, including Novo Progresso, sent more migrants to the soybean districts than they received. While this data by no means capable of confirming that the displacement of smallholder farmers from the soybean regions farther north into the frontier is not occurring, it is broadly supportive of the interview findings, namely that any such effect, if occurring, is probably not occurring on a significant scale.

Figures 6a-6d.

Figures 6a-6d

Open in a new tab

Total and rural net-migration into soybean counties (6a-6b) and Novo Progresso, PA (6c-6d). Each dot represents a single in or out migration. Numbers indicate net-migrations. Positive numbers indicate that more migrants had left the shaded county(ies) for that location. Thus more migrants arrived from than left for Novo Progresso for the soybean counties at the southern extent of the study area.

Where then are migrants to northern Mato Grosso and western Para arriving from? As an example, we map the locations of in and out-migrants from Novo Progresso, Pa, in Figures 6c and 6d. Most of the in-migrants were arriving to Novo Progresso from elsewhere in Pará, or from northwestern Mato Grosso. Few were arriving from the soybean districts in our study area. In fact, net-migration to Novo Progresso from these areas was negative (see table 2). Again, we stress that although this data cannot confirm that small farmers and ranchers are not arriving to Novo Progresso from the new and expanding soybean regions farther south, the data are broadly supportive of our interview findings.

Table 2.

Rural and Total Net-Migration from Novo Progresso, Pará and the soybean districts of the BR-163 region.

Total Migrations Rural Migrations
Soybean Regions Novo Progresso Soybean Regions Novo Progresso
In-Migration 6246 410 1088 85
Out-Migration 3796 500 965 102
Net-Migration 2450 −90 123 −17
Open in a new tab

Author's calculations, based on 2010 long form census questionnaire (IBGE 2010)

Soybean region includes ten municipios: Claudia, Feliz Natal, Ipiranga do Norte, Itauba, Lucas do Rio Verde, Nova Ubirata, Santa Carmem, Sinop, Sorriso, and Vera

Consequently, while we acknowledge the limitations of drawing conclusions from a limited set of interviews and from the limited data of the long form census questionnaires on migration, we hope that these findings will give pause to researchers seeking to explain the linkage between the cattle and soybean frontiers as a function of migration and displacement. However, while we cautiously argue that displacement is not drive regional deforestation, we suggest that soybean production nonetheless occupy an important role in driving regional deforestation in the Amazon. We elaborate on this in the ensuing section.

(7) What Does Drives Indirect Land Use Change?

A key aspect of indirect land use change in Brazil is that the soybean sector must by some means be capable of altering land use decisions at a regional scale. In the general or partial equilibrium approaches, this influence was transmitted through beef prices, as local pastures were converted to soybeans. However, we showed that area of beef production lost to soybean production was minimal when viewed relative to Brazil's total beef supply. In the land displacement approach, soybean expansion influenced regional land use outcomes by sending forth both human and financial capital into frontier areas. In this article, we sought evidence of this process through interviews with in-migrating ranchers and farmers in an area of rapid land use change. Our extensive search in the BR-163 region, however, did not yield evidence of land use displacement. Yet field work, and both the structured and unstructured interviews, did suggest that another factor was linking agricultural change in Mato Grosso to deforestation; namely land appreciation.

For indirect land use change to take place, a change in the agricultural sector must have an impact on land use decisions being made in the frontier. We theorize that this effect occurs through land prices. Economics research has long recognized spatial correlation in land and real estate values (Basu and Thibodeau 1998, Dubin 1992). A shock or sudden change in land values, for example, affects not only value in one location, but also influences values across a neighborhood or landscape (Harding, Rosenblatt and Yao 2009). In Mato Grosso, record returns to agricultural production in the early 2000 not only affected local cropland values, but raised land values more broadly across the region.

Rising land values influence landowners’ decision processes both locally and regionally. A landowner who acknowledges the appreciation of their land will be more likely to either invest in it to maximize its value or to sell. For example, if a homeowner's home value were to double or triple in the space of two to three years, they would be more likely to invest in maintaining or improving the property, given that the marginal returns to the investment (say a new roof) would rise. For landowners in the Amazon, a similar process occurs. As land values rise in Mato Grosso or Pará, or as the value difference between cleared land and forest land increases, landowners face a heightened incentive to risk the potential legal repercussions of land clearing, or absorb the actual cost of the clearing process. Thus not only does land appreciation incentivize in-migration, which brings in new financial and human capital to the Amazon, but it also encourages landowners to invest in their properties. In the Amazon the first investment in a rural property is often deforestation. As one rancher who had recently deforested a portion of their property and planned to then sell an earlier deforested portion remarked during an interview: “deforestation, if you can get away with it, is the most profitable activity in the Amazon.” Not coincidentally, deforestation in Mato Grosso increased with the rise in land values in the early 2000, but fell as land values declined or stabilized (Figure 7).

Figure 7.

Figure 7

Open in a new tab

Mean nominal cropland values (in 2002$Rs), annual change in soybean cropland (ha), and area deforested (km2) between 2002-2010, for Mato Grosso.

Our field interviews, which focused on land sales as a driver of displacement, did not directly address the impact of land prices on deforestation, and more research will be needed to confirm the role of land appreciation in driving the increase in deforestation. Nevertheless, much of the research in the Amazon has already connected land speculation and appreciation of rural properties to land use change in this region (Fearnside 1983, Chomitz and Thomas 2001, Ozorio de Almeida and Campari 1995, Rudel and Horowitz 2013, Bowman et al. 2012). Other work has broadly recognized that landowners clear land as a means by which to secure ownership (Alston, Libecap and Mueller 1999, Aldrich et al. 2011). In this article we step beyond this past work by arguing that the deforestation increases in the Amazon during the early years of the last decade were not only tied to speculation and appreciation, but specifically to speculation tied to the success of the agricultural sector. Rising returns to soybean production led to higher land prices across the region, including for pasture or forest properties. As land values rose, landowners solidified their claims on land by deforesting, or opened their land to increase their property values.

(8) Conclusion

In this research we sought to find evidence of land use displacement through the in and out migration of land purchasing and land selling agents in a region of strategic interest to the agricultural and environmental sectors. We pursued our analysis by framing rural-rural migration as a function of agent level neoclassical migration decisions made within a Thunian spatial structure. We then interviewed land purchasers on their previous locations and occupations, as well as the present day location and occupation of any prior owners. Findings from field interviews were broadly congruent with our theoretical position, with farmers and ranchers migrating to maximize the returns to their skills and financial capital by seeking larger properties, both for immediate production or as investments. Field interviews, however, were more notable for what they did not find: clear evidence of land use displacement. Analyses of migration patterns using census data also suggest that net-migration from soybean regions into the relatively dense forested areas in western Pará was negative.

If our research on rural migration dynamics did not clearly reveal the direct displacement of ranchers or small farmers from agricultural frontiers to areas of rapid deforestation, the interviews did recognize the importance of spatial differentiation in land values, and of land appreciation in driving in-migration. The land buyers who migrated into the region had largely done so during the early 2000s, when appreciation was fastest, and when land values in the south greatly overshadowed those in Mato Grosso. The process not only brought new investment capital to the region, but it also contributed to the increased activity in the region's land market. Land appreciation during this period also reshaped land use decisions in frontier areas, and increased incentives to landowners to clear new lands, or to potentially capitalize on the fast increasing value of their land assets. Perhaps there is nowhere where this process was clearer than in the BR-163 region, where not only did agriculture expand widely in Mato Grosso, but the expansion was pared with plans to improve the highway conduit between the soybean districts and forested areas in western Pará. Agricultural expansion and the new highway project combined to accelerate land speculation and appreciation in the region, and thus heightened landowners’ incentives to clear land to increase their property values.

In this article we thus argue that the greater impact of an expanding agriculture sector on land use change lies not in the occupation or displacement of cattle pastures, but rather in their impact on land markets in the Amazon, and the subsequent motivation of landowners to invest in their increasingly valuable properties. We hope that more case study research will emerge in the ensuing years to support or refute the theoretical positioning and findings of this article, both in terms of understanding the relationship between land appreciation and deforestation, and more broadly, the relationship between agricultural changes and regional forest loss.

Footnotes

1

In this article we refer variously to the cerrado and Amazon Biomes, and the Legal Amazon. The Amazon Biome encompasses the humid moist forests to the north and south of the river. The cerrado area comprises the scrub upland forests of central Brazil. Much of the agricultural expansion in the Legal Amazon, a political designation that includes areas of both cerrado and humid Amazon forests, has taken place in either cerrado regions, or in the transitional areas of the Amazon Biome.

2

The real was originally introduced in 1994 at a value equal to 1$US. Between 1996 and 1998 the exchange rate was closely controlled by Brazil's central bank.

3

The migration of farmers and ranchers from Brazil's southern states has already been closely tied to the expansion of agriculture in the region. While this research did not aim to map or capture the impact or influence of a migration network, its impact was clearly evident from interviews with recent migrants.

4

We sought the owners of property purchases of this size, however we include results from several interviews of new property purchases that were less than 250ha.

5

We recognize that rental agreements present another form by which land control is transferred. Indeed, across the Amazon, many ranchers enter rental agreements with farmers, partly as means of pasture recuperation and improvement. However, rental agreements do not necessarily lead to the wholesale capital transfer that is implied in a land purchase. For while the skills are displaced (e.g., the rancher is no longer ranching at that location), the rancher does not receive the financial windfall for potential investment in purchasing and opening new lands that would be derived from a land sale. This would limit the magnitude of any displacement effect.

6

We classified a parcel as having gone through a land use change if a land use change of magnitude took place after the purchaser took possession. This includes both land converted from forest to pasture and land converted from one land use to another, typically from pasture to croplands. We included in the process the “cleaning” of juquirão, or the lower level vegetation of logged and degraded forests or secondary growth on unmaintained pastures.

7

In ten of these interviews the land purchaser was unsure of the activities or location of the old owner.

8

The ten municipal districts include Claudia, Feliz Natal, Ipiranga do Norte, Itauba, Lucas do Rio Verde, Nova Ubiratã, Santa Carmem, Sinop, Sorriso, and Vera.

References

  1. Aldrich S, Walker R, Simmons C, Caldas M, Perz S. Contentious Land Change in the Amazon's Arc of Deforestation. Annals of the Association of American Geographers. 2011;102:103–128. [Google Scholar]
  2. Aldrich SP, Walker RT, Arima EY, Caldas MM, Browder JO, Perz S. Land-Cover and Land-Use Change in the Brazilian Amazon: Smallholders, Ranchers, and Frontier Stratification. Economic Geography. 2006;82:265–288. [Google Scholar]
  3. Alston LJ, Libecap GD, Mueller B. Titles, Conflict and Land Use: The Development of Property Rights and Land Reform on the Brazilian Amazon. University of Michigan Press; Ann Arbor: 1999. [Google Scholar]
  4. Andrade de Sá S, Palmer C, di Falco S. Dynamics of indirect land-use change: Empirical evidence from Brazil. Journal of Environmental Economics and Management. 2013;65:377–393. [Google Scholar]
  5. Arima E, Richards PD, Walker R, Caldas M. Statistical Confirmation of Indirect Land Use Change in the Amazon. Environmental Research Letters. 2011;6 [Google Scholar]
  6. Barona E, Ramankutty N, Hyman G, Coomes OT. The role of pasture and soybean in deforestation of the Brazilian Amazon. Environmental Research Letters. 2010;5:024002. [Google Scholar]
  7. Basu S, Thibodeau TG. Analysis of spatial autocorrelation in house prices. The Journal of Real Estate Finance and Economics. 1998;17:61–85. [Google Scholar]
  8. Bates DC, Rudel TK. Climbing the “Agricultural Ladder”: Social Mobility and Motivations for Migration in an Ecuadorian Colonist Community*. Rural Sociology. 2004;69:59–75. [Google Scholar]
  9. Bowman MS, Soares-Filho BS, Merry FD, Nepstad DC, Rodrigues H, Almeida OT. Persistence of cattle ranching in the Brazilian Amazon: A spatial analysis of the rationale for beef production. Land Use Policy. 2012;29:558–568. [Google Scholar]
  10. Brown JC, Jepson W, Price KP. Expansion of Mechanized Agriculture and Land-Cover Change in Southern Rondônia, Brazil. Journal of Latin American Geography. 2004;3:96–102. [Google Scholar]
  11. Chomitz KM, Thomas TS. Geographic patterns of land use and land intensity in the Brazilian Amazon. Citeseer; 2001. [Google Scholar]
  12. Cohn AS, Mosnier A, Havlík P, Valin H, Herrero M, Schmid E, O'Hare M, Obersteiner M. Cattle ranching intensification in Brazil can reduce global greenhouse gas emissions by sparing land from deforestation. Proceedings of the National Academy of Sciences. 2014:201307163. doi: 10.1073/pnas.1307163111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cronon W. Nature's Metropolis. Norton; New York: 1991. [Google Scholar]
  14. Curtis S, Gesler W, Smith G, Washburn S. Approaches to sampling and case selection in qualitative research: examples in the geography of health. Social Science & Medicine. 2000;50:1001–1014. doi: 10.1016/s0277-9536(99)00350-0. [DOI] [PubMed] [Google Scholar]
  15. Dubin RA. Spatial autocorrelation and neighborhood quality. Regional science and urban economics. 1992;22:433–452. [Google Scholar]
  16. Fargione J, Hill J, Tilman D, Polasky S, Hawthorne P. Land Clearing and the Biofuel Carbon Debt. Science. 2008;319:1235–1238. doi: 10.1126/science.1152747. [DOI] [PubMed] [Google Scholar]
  17. Fearnside PM. Land-use trends in the Brazilian Amazon region as factors in accelerating deforestation. Environmental Conservation. 1983;10:141–148. [Google Scholar]
  18. Goldsmith P, Hirsch R. The Brazilian Soybean Complex. Choices. 2006;21:97–103. [Google Scholar]
  19. Gollnow F, Lakes T. Policy change, land use, and agriculture: The case of soy production and cattle ranching in Brazil, 2001–2012. Applied Geography. 2014;55:203–211. [Google Scholar]
  20. Harding JP, Rosenblatt E, Yao VW. The contagion effect of foreclosed properties. Journal of Urban Economics. 2009;66:164–178. [Google Scholar]
  21. Hecht S. Soybeans, Development, and Conservation on the Amazon Frontier. Development and Change. 2005;36 [Google Scholar]
  22. Helfand S, Rezende G. C. d. The impact of Sector-Specific and Economy-Wide Policy Reforms on the Agricultural Sector in Brazil: 1980-1998. Contemporart Economic Policy. 2004;22:194–212. [Google Scholar]
  23. Hertel TW, Tyner WE, Birur DK. The Global Impacts of Biofuel Mandates. The Energy Journal. 2010:75–100. [Google Scholar]
  24. IBGE . 2010 Demographic Census. Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística); 2010. Sample Population (Amostra Pessoas). [Google Scholar]
  25. IBGE . Produção Pecuária Municipal (Municipal Cattle Production)1974-2012. Instituto Brasileiro de Geografia e Estatística [Brazilian Institute of Geography and Statistics]; São Paulo: 2013. [Google Scholar]
  26. Jepson W. Private agricultural colonization on a Brazilian frontier, 1970-1980. Journal of Historical Geography. 2006a;32:839–863. [Google Scholar]
  27. Jepson W. Producing a modern agricultural frontier: Firms and cooperatives in Eastern Mato Grosso, Brazil. Economic Geography. 2006b;82:289–316. [Google Scholar]
  28. Keeney R, Hertel TW. The Indirect Land Use Impacts of United States Biofuel Policies: The Importance of Acreage, Yield, and Bilateral Trade Responses. American Journal of Agricultural Economics. 2009;91:895–909. [Google Scholar]
  29. Lambin EF, Meyfroidt P. Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences. 2011;108:3465–3472. doi: 10.1073/pnas.1100480108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Lapola DM, Schaldacha R, Alcamoa J, Bondeaud A, Kocha J, Koelkinga C, Priesse JA. Indirect land-use changes can overcome carbon savings from biofuels in Brazil. Proceedings of the National Academy of Sciences of the United States of America. 2010;107:3388–3393. doi: 10.1073/pnas.0907318107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Mann ML, Kaufmann RK, Bauer D, Gopal S, Vera-Diaz MDC, Nepstad D, Merry F, Kallay J, Amacher GS. The economics of cropland conversion in Amazonia: The importance of agricultural rent. Ecological Economics. 2010;69:1503–1509. [Google Scholar]
  32. Massey DS, Arango J, Hugo G, Kouaouci A, Pellegrino A, Taylor JE. Theories of International Migration: A Review and Appraisal. Population and Development Review. 1993;19:431–466. [Google Scholar]
  33. Meyfroidt P, Rudel TK, Lambin EF. Forest transitions, trade, and the global displacement of land use. Proceedings of the National Academy of Sciences. 2010;107:20917–20922. doi: 10.1073/pnas.1014773107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Morton DC, DeFries R, Shimabukuro YE, Anderson LO, Aral E, del Bon Espirito-Santo F, Freitas R, Morisette J. Cropland expansion changes deforestation dynamics in the southern Brazilian Amazon. Proceedings of the National Academy of Sciences of the United States of America. 2006;103:14637–14641. doi: 10.1073/pnas.0606377103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Nassar AM, Harfuch L, Bachion LC, Moreira MR. Biofuels and land-use changes: searching for the top model. Interface Focus. 2011;1:224–232. doi: 10.1098/rsfs.2010.0043. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Ozorio de Almeida AL, Campari JS. Sustainable Settlement in the Brazilian Amazon. Oxford University Press; New York: 1995. [Google Scholar]
  37. Peet JR. The Spatial Expansion of Commercial Agriculture in the 19th Century: A von Thunen Interpretation. Economic Geography. 1969;45 [Google Scholar]
  38. Richards P. Cotton, Soy, and the Final Atlantic Forest Frontier. The Professional Geographer. 2011;63:1–21. [Google Scholar]
  39. Richards PD, Myers RJ, Swinton SM, Walker RT. Exchange rates, soybean supply response, and deforestation in South America. Global environmental change. 2012;22:454–462. [Google Scholar]
  40. Richards PD, Walker RT, Arima EY. Spatially complex land change: The Indirect effect of Brazil's agricultural sector on land use in Amazonia. Global Environmental Change. 2014;29:1–9. doi: 10.1016/j.gloenvcha.2014.06.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Rudel TK, Horowitz B. Tropical Deforestation: Small Farmers and Land Clearing in Ecuadorian Amazon. Columbia University Press; 2013. [Google Scholar]
  42. Rudel TK, Schneider L, Uriartec M, T. BL, II, DeFries R, Lawrence D, Geoghegan J, Hecht S, Ickowitz A, Lambin EF, Birkenholtz T, Baptista S, Grau R. Agricultural intensification and changes in cultivated areas, 1970-2005. Proceedings of the National Academy of Sciences. 2009;106:20675–80. doi: 10.1073/pnas.0812540106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Searchinger T, Heimlich R, Houghten RA, Dong F, Elobeid A, Fabiosa J, Tokgoz S, Hayes D, Tun-Hsiang Y. Use of U.S. Croplands for Biofuels Increases Greenhouse Gases through Emissions from Land-Use Change. Science. 2008;319:1238–1240. doi: 10.1126/science.1151861. [DOI] [PubMed] [Google Scholar]
  44. Sewastynowicz J. “Two-Step” Migration and Upward Mobility on the Frontier: The Safety Valve Effect in Pejibaye, Costa Rica. Economic Development and Cultural Change. 1986;34:731–753. [Google Scholar]
  45. Sjaastad L. The Costs and Returns of Human Migration. The Journal of Political Economy. 1962;70:80–93. [Google Scholar]
  46. Spera SA, Cohn AS, VanWey LK, Mustard JF, Rudorff BF, Risso J, Adami M. Recent cropping frequency, expansion, and abandonment in Mato Grosso, Brazil had selective land characteristics. Environmental Research Letters. 2014;9:064010. [Google Scholar]
  47. Taylor EJ, Arango J, Hugo G, Kouaouci A, Massey DS, Pellegrino A. International Migration and Community Development. Population Index. 1996;62:397–418. [PubMed] [Google Scholar]
  48. Todaro M. Internal Migration in Developing Countries: A Survey. National Bureau of Economic Research, Inc; 1980. pp. 361–402. [Google Scholar]
  49. VanWey LK, Spera S, Sa R. d., Mahr D, Mustard JF. Socioeconomic development and agricultural intensification in Mato Grosso. Philosophical Transactions of the Royal Society of London. Series B. 2013:20120168. doi: 10.1098/rstb.2012.0168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. von Thünen JH. The Isolated State. Pergamon Press; Oxford and New York: 1966. [Google Scholar]
  51. Walker R. Urban sprawl and natural areas encroachment: linking land cover change and economic development in the Florida Everglades. Ecological Economics. 2001;37:357–369. [Google Scholar]
  52. Walker R. Sparing Land for Nature in the Brazilian Amazon: Implications from Location Rent Theory. Geographical Analysis. 2014;46:18–36. [Google Scholar]
  53. Walker R, Browder J, Arima E, Simmons C, Pereira R, Caldas M, Shirota R, de Zen S. Ranching and the new global range: Amazônia in the 21st century. Geoforum. 2009a;40:732–745. [Google Scholar]
  54. Walker R, DeFries RS, Vera-Diaz M. d. C., Shimabukuro Y, Venturieri A. The Expansion of Intensive Agriculture and Ranching in Brazilian Amazonia. In: Keller M, editor. Amazonia and Global Change. American Geophysical Union; Washington D.C.: 2009b. pp. 61–82. [Google Scholar]
  55. Warnken PF. The Development and Growth of the Soybean Industry in Brazil. Iowa State University Press; Ames, Iowa: 2002. [Google Scholar]
  56. Weinhold D, Killick E, Reis EJ. Soybeans, poverty and inequality in the Brazilian Amazon. World Development. 2013;52:132–143. [Google Scholar]

RESOURCES