Food, fuel, and the hidden margins of capital

Abstract

Changing demands for food and biofuel have reshaped global agricultural outlooks and pushed production in areas long considered too marginal for the widespread production of agricultural commodities. Perhaps no region has drawn more attention to the environmental impacts of expanding agricultural production than the Amazon. Considerations of indirect land use change have brought a new dimension to discussions related to the impacts of agriculture on the region's forest cover. In this article, I provide a location-based conceptualization of indirect land use change that brings to light the intra-regional movement of capital and skills between the cattle and agriculture sectors. The article suggests that amid rapid increases in rents for soy production and land prices, ranchers face strong incentives to relocate their operations to forest regions.

Keywords:

• Amazon
• soybeans
• ranching
• land use
• land cover change
• Brazil

1. Introduction

Look abroad through natures range,

Nature's mighty law is change;

Robert Burns (From ‘Duncan Gray’)

Changing diets and demands for food will dramatically reshape agricultural markets in the coming decades (Tilman et al. 2001; Naylor et al. 2005, 2007; FAO 2010). A larger, wealthier, and increasingly urban population that consumes more processed food, more meat, and more fuel will heighten demand for agricultural lands. Some of the largest such growth is foreseen for the production of meat and cereal crops (FAO 2010). Given that the United States has limited reserves of agricultural land on which to expand production, much of the new agricultural output is expected to come through new and more efficient production abroad. As one of the world's largest reserves of relatively unexploited land, the scrub forests of the cerrado and the tropical forests of the Amazon are likely to play an important role in provisioning the world's food and energy in the coming decades (Borlaug 2008; Pingali, Raney, and Wiebe 2008).

Energy production will also yield a new dimension to the growth of global agricultural land. Biofuel consumption has already tripled since 2000, and estimates suggest that current levels will double by 2020 (FAO 2009). A principal argument for increased biofuel production rests on its potential to reduce greenhouse gas emissions by between 10% and 90% (FAO 2009). However, the emissions savings associated with biofuels have been the subject of some debate. For although it has been suggested that ethanol and biodiesel production represent a move toward carbon neutrality, it has also been argued that any net carbon savings may be compromised through the production process and through other unseen effects on carbon reservoirs, such as indirect land use change ¹ (ILUC; Fargione, Hill, Tilman, Polasky, and Hawthorne 2008; Searchinger et al. 2008). For example, if new incentives for ethanol crops in the American Midwest lead to deforestation in the Amazon or Indonesia, carbon savings associated with American ethanol or biodiesel production may be compromised.

Understanding the nature and impact of effects derived from ILUC is complicated by difficulties in measuring and conceptualizing the impact broadcasted from changing land use in one location on another, potentially distant location. In this article, I take up this issue; I focus on the conceptual linkages between the expansion of agriculture and the advancement of the cattle-led frontier in the Amazon. I draw attention to the role of place- and use-specific resources as drivers of ILUC in a region of critical economic and ecological importance, namely the Brazilian Amazon. To this end, I advance from recent work considering ILUC through general equilibrium models (Keeny and Hertel 2009; Kretschmer and Peterson 2010) and consider the ILUC process as driven partly through the relocation or spatial relegation of use-fixed capital. This framework positions pasture-led forest loss and an advancing agricultural frontier within an adaptation of a rent-based model (Walker et al. 2008, 2009; Lapola et al. 2010). Integral to this analysis is the ‘fixing’ of land managers to their respective production strategies and considerations of migration incentives. I also consider the relative impact of capital redistributed to a capital-scarce frontier, where the marginal environmental impact of its relocation is larger. In this sense, I interpret ILUC as a response not only to price or supply changes, but also to a more diverse set of regional or even local-scale production interests, including use-fixed investments and inter-sector capital transfers.

The presentation begins with a review of recent attempts to conceptualize and estimate ILUC. Much of the existing work on ILUC has so far focused on aggregate land use changes and land use trajectories within the United States, or the application of multi-region, multi-sector, general equilibrium models such as the Global Trade Analysis Project. In this literature, with the notable exceptions being models developed by Lapola et al. (2010) and Barona, Ramankutty, Hyman, and Coomes (2010), Brazilian land use change is considered only anecdotally, and then only largely as a theoretical point of impact (where the ILUC effect occurs) associated with changes in land use in the American Midwest. In Section 3, I focus on the evidence suggestive of an indirect impact of soy production, which occurs vis-à-vis a land use interaction with existing pasturelands. This provides the genesis for Sections 4–6, where I outline a conceptual model of regional ILUC. I consider two sub-processes of interest, namely the migration and spatial redistribution of human capital tied to the agricultural and cattle industries and an inter-sector financial transfer associated with property sales and land use change. I suggest that rising commodity prices result in windfall profits, higher land rents, ² and increased property values. On the ground in the Amazon this entails the ‘pushing’ or dispelling of ranching operations to the lower-valued regions at the frontier. The process is facilitated by the development of use-fixed capital in both the ranching and agriculture sectors and a financial windfall obtained by ranchers as they liquidate appreciated pasturelands and reinvest the proceeds to open new land.

The final sections of this article focus on ILUC and land use dynamics within a strategic transport corridor of the Amazon, in northern Mato Grosso, Brazil. Section 7 draws on agricultural production data and land prices to consider relocation incentives for displaced ranchers. The following section provides insights from a field survey and key informant interviews, conducted in 2008 and 2009, to consider the relocation of production strategies at the farm or agent level. The article closes with a discussion of ILUC as a component of carbon accounting and offers a critical evaluation of current agro-environmental policy in the Amazon, including the so-called soy moratorium, initiated in 2006.

2. Indirect land use change

Conceptually, ILUC begins with the reality that ethanol and biodiesel crops (e.g., soy, corn) require large tracts of farmland. Increasing production to meet both new biofuel and existing food demand will require cropland expansion, typically at the expense of less-intensive land production strategies or other non-agricultural land covers (Turner, Lambin, and Reenberg 2007a; Lubowski, Plantinga, and Stavins 2008). Greater demand for cropland, and presumably higher crop prices, will result in a rise in bid-rents and subsequently, the spatial realignment of land uses as the production of less-intensive goods (such as beef) is relegated to more marginal lands. Searchinger et al. (2008), for example, estimate that increasing US ethanol production to 56 billion liters will result in the conversion of 1.8 million hectares of land to cropland. Assuming that some of the land to be brought into production currently holds stores of carbon, these conversions could effectively offset any carbon savings attained through a switch from fossil fuel to biofuel consumption.

The debate regarding ILUC has become particularly controversial with the adoption of the Renewable Fuel Standard by the Environmental Protection Agency and the Low Carbon Fuel Standard by the California Air and Resource Board, both of which consider greenhouse gases produced at various stages of the fuel production practice. The accounting of ILUC as a component in the carbon footprint for biofuels has been particularly contentious (Simmons 2008, 2009). The Low Carbon Fuel Standard, for example, stands to diminish or negate the potential carbon savings associated with biofuel production in the American Midwest, a fact which has created a controversy around how to measure ILUC, given the potential importance of CO₂ accounting and the difficulties associated with measuring spatially diffusive causality (Gnansounou, Panichelli, Dauriat, and Villegas 2008; Babcock 2009a; Hertel et al. 2010; Tyner, Taheripour, Zhuang, Birur, and Baldos 2010).

The problem has been partly addressed through the application of multi-sector, multi-region general equilibrium models such as the Global Trade Analysis Project (Kretschmer and Peterson 2010; Tyner et al. 2010). This approach, however, is fundamentally aspatial in its conception and estimation. Although perhaps adequate over the long term, from a shorter time horizon the place- and use-specific allocation of resources will distort land use change at the regional and local levels. This effect, which includes linkages between land use systems and farm-level incentives specific to location, remains less than fully understood.

The magnitude and mechanisms by which the loss of forest here is linked indirectly to the expansion of crop production is of critical importance to current efforts toward carbon accounting. This article takes a spatial approach to understanding ILUC, and accounts for use-fixed capital investments and the spatial mobility of human capital in the agriculture and ranching industries. ³ Specifically, I focus on two principal questions. First, how does the expansion of agriculture on old pastures, via a cascading spatial land use effect, result in the conversion of forest to pasture? Second, do ranchers or other land managers adapt their production capacities to economic opportunities for a place at a given time, or do they relocate spatially to maximize their production?

These questions are critical to understanding ILUC and form central pillars to the conceptual model developed in this article. Following Walker et al. (2009), I advance a framework whereby farmers and agribusinesses migrate to regions where and when economic conditions for soy or other plantation crops are more favorable than those for beef. As these farmers and support businesses relocate, they bring with them skills, knowledge, and investment capital that will facilitate production in these locations. A consequence of this migration is the spatial redistribution of use-fixed human and financial capital. I suggest that the spatial and sectoral redistribution of these investments not only allows for the redistribution of production, but also amplifies land use change by displacing lesser intensive production strategies to critical environmental regions. In this sense, the true impact of the expansion of crop production in the Amazon is not only through its effect on prices for beef or soy, but rather through the displacement of use-fixed human capital and the liquidation and transfer of financial capital previously tied to old ranchlands.

3. ILUC and Brazil

During the past two decades, no nation increased its areas of agricultural production and pasture more than Brazil. Today more than 70 million cattle (Figure 1), one-third of Brazil's cattle herd, graze in the Amazon; the Amazonian state of Mato Grosso alone now produces 30% of Brazil's annual soy crop, or 8% of global output (Margulis 2004; Walker et al. 2008; IBGE 2010b).

Figure 1. Growth of the soy and cattle industries in the Brazilian Amazon, 1990–2008.

The low cost of implementation of beef production and transporting the cattle industry's inputs and outputs into and out of distant points of production has allowed the formation of new pastures in even many of the Amazon's most distant frontiers. Not surprisingly, linkages between cattle ranching and deforestation have been widely documented (Hecht and Cockburn 1989; Rudel and Roper 1996; Faminow, Dahl, Vosti, Witcover, and Oliveira 1999; Margulis 2004; Walker et al. 2008) and the vast majority of Amazonian forest loss, approximately 80%, is attributed to clearings to make way for new pastures (Margulis 2004; Brandão, Rezende, and Marques 2006). Although in certain regions of the Amazon, smallholder production continues to account for a significant portion of deforestation (Walker, Moran, and Anselin 2000; Palm, Vosti, Sanchez, and Ericksen 2005; Labarta, White, and Swinton 2008), the focus of this article is on areas where large ranching and agriculture operations predominate.

Although evidence does suggest that commodity-driven agriculture has and perhaps will expand directly into forest (Steininger et al. 2001; Morton et al. 2006; Galford et al. 2008), much of the soy expansion has come through encroachment into established pasturelands. The cost of converting pasture to agriculture is far less than that of converting moist tropical forest, a process that usually takes years before soy can be grown (USDA 2007; Jepson, Brannstrom, and Filippi 2010). Consequently, deforested land is typically first cleared for pasture, used until the soils are degraded, and, if the opportunity exists for crop production, converted to farmland. The pattern of land use transitions, first from forest to pasture, then from pasture to agriculture, has been further cemented with the imposition of a ‘moratorium’ on clearings for new soy production, an industry-led policy initiative established in 2006 and recently extended to 2014. Implicit in this article is the argument that the moratorium has cloaked the true impact of soy by eliminating the crop's most visible linkage to environmental degradation, namely the direct conversion of forest to crop fields.

Despite the transfer of pastures to crop production, total pasture area has increased, not decreased. In fact, statistical evidence suggests a correlation between the expansion of soy and the expansion of the extensive Amazonian cattle industry (Almeida de Menezes and Piketty 2007; Barona et al. 2010). Thus, the argument has emerged that changes in the extent of soybean production at the periphery of the Amazon are directly related to pasture development in the region's interior. Such considerations were strongly supported by Barona et al. (2010) who noted the parallel movement of cattle and soy industry ‘centroids’ from 2000 to 2006 and put forth notions of a so-called compensatory deforestation effect associated with the soy boom. I argue that this parallel expansion is due not only to similar responses to favorable economic conditions for both goods, but also to a displacement effect whereby ranches simultaneously reap a financial windfall and seek to reinvest their human capital upon spatial-economic relocation, a process that involves their relocation to more marginal lands and the loss of forest cover. In this sense, I focus on ILUC as a by-product of the soy expansion in the Amazon and develop a conceptual outline by which pasture to soy conversions act as drivers of pasture-led forest loss.

4. Land rents and the ripple effect

In this section, I consider the role of land rents in providing land managers incentives to relocate and change their land uses, and recognize the broad spatial effects that emerge (or ripple) at regional scale as these decentralized, independent land managers make their land use decisions. By adopting a framework based on regional land rents and their ties to global commodity markets (Walker et al. 2008), I filter individual, agent-based decisions (e.g., Giddens 1984; Chowdhury and Turner 2006) through a spatial structure established by distance and freight costs. This model is referred to both implicitly and explicitly in research addressing the importance of transportation systems as drivers and determinants of land cover and land use change (Smith 1984; Cronon 1991; Chomitz and Gray 1996; Angelsen 2007; Kaimowitz and Angelesen 2008). It aligns not only with recent work on ILUC (Turner, Plevin, O'Hare, and Farrell 2007b; Lubowski et al. 2008) but also with a growing interest in the flows or ‘frontiers’ of capital and capital-intensive production systems in Latin America and the Amazon basin and the subsequent effect of such flows on regional land cover (Hecht 2005; Jepson 2006; Bernardes 2007; Walker et al. 2009).

The rent-based model dictates that many of the diverse factors that condition land use are linked to the respective capacity of a parcel to generate rents or otherwise sustain a livelihood. The present work thus incorporates rent-seeking behavior, or the search for profitable production, as its principal mechanism, rather than any static land use model. Thus, although the notorious concentric circles of the primeval version of spatial rent theory may be unacceptable as tools for explaining land use organization and distribution (Barnes 2001), the underlying principle continues to serve as a viable conceptual basis by which to understand and theorize the role of location and transportation costs as a component of profit potential and production choices (Peet 1969; Kaimowitz and Angelesen 2008). Given that the Amazonian agricultural frontier, perhaps more than any other region, is limited by transport costs (as opposed to production capacity), the rent-based framework is particularly applicable as a means by which to conceptualize the cascading spatial-economic effects that drive ILUC.

For the present article, an economic frontier is considered to occur where incentives for two or more land uses are at equilibrium. This frontier moves when dynamic economic conditions, for example, a price increase, a development in transportation infrastructure, or a favorable variation in the exchange rate, change the rents for a production strategy on a given parcel of land. These new opportunities for profit ‘push’ the frontier outward or extend the zone of possible production to new spaces (Walker et al. 2008). For simplicity, consider a two-product system with an intensive and extensive crop within an abstract landscape. Here, two land use frontiers, an intensive and extensive frontier, form at various distances from a market center, providing that two conditions are met. First, the products must vary in their rent potential prior to factoring in transportation costs. Second, the production strategy capable of generating higher rents must have higher per product unit transportation cost (Dunn 1967; Walker 2001; Walker and Solecki 2004; Angelsen 2007). The point of equilibrium between intensive and extensive production that thus emerges is the intensive frontier. As described in Walker (2001), I consider as the extensive frontier the fringe of the uncultivated wilderness, whether forest or subsistence agriculture or another ‘non-productive’ space, or where rents for all production possibilities equal zero.

In accordance with the concept of ILUC, expansion of crop production at the intensive frontier is ultimately felt at the extensive frontier, where displaced ranching operations reemerge at the cost of forest cover. This suggests that as rents for crop production in the Amazon change, two land cover change processes occur, and that they are linked. The first is intensification, where rents for agriculture exceed those of ranching, and lesser intensive grazing lands are converted into more intensive production practices (such as agriculture). The second process, and the one of great interest to current considerations of ILUC and carbon accounting, is deforestation. The mechanisms by which the expansion of a more intensive land use results in the expansion of the less-intensive one, however, remain unclear. The following section develops these conceptual underpinnings, namely through considerations of the displacement and mobility of use-fixed capital and an inter-sector value transfer.

5. Use-fixed capital and agent mobility

In this section, I address producer displacement and link spaces of production to the human agents responsible for their economic land use. Land use change (or transitions) can occur only when the knowledge and skills for a new production strategy must be invested at the location of production. Consequently, I consider the migration or movement of a land manager or agent with whom these skills are invested as contingent for land use change. Thus, production may be profitable but only when the necessary inputs, including human capital, are relocated to or developed in the new location. Naturally, the production agent or the land manager is one such critical input.

For skilled workers, whether in agriculture, ranching, or another profession, human capital is embodied within the individual agent. As land may contain use-value in the form of structural improvements or other investments, so too does the manager of that land, who possesses a set of fixed skills and experiences that increase the marginal value of their possession of land, so long as it is dedicated to a certain production strategy. In this sense, skills are investments in the land manager that increase the value of their work. These skills, however, are specific to a means of production or, in the present context, a particular land use. Just as ranchers are invested with the skills, know-how, and contacts that facilitate ranching, successful farmers have developed the skill assets critical to agriculture. Given their investments, farmers and ranchers thus have fixed incentives to pursue their production strategy of habit. Although this use-fixed human capital and other forms of fixed capital (equipment, fencing) are not the only determinants of land use, these investments do incentivize farmers or ranchers, whether producing sugar beets in Michigan or soy in Mato Grosso, to produce their crop of interest in the face of potentially uncertain economic or market conditions.

As with other forms of capital, human capital is subject to market demand and is liable to appreciation or depreciation (Sjaastad 1962). When and where the demand for a specific set of skills and experiences is altered by changing social or economic conditions the value of those skills will rise or fall. If economic conditions favor crop production, then the demand for ranching skills will fall, at least in this specific location. Skill assets, however, in contrast to other assets, cannot be readily converted to fungible capital through divestiture. Yet they can be relocated. Where economic conditions lead to the local depreciation of a given skill set in one region, the individual may find that the only way to capitalize on their invested skills is through migration. Depending on the potential and demand for ranching in other locations, ranchers may thus find a strong incentive to relocate. In the case of the Amazon frontier, land managers with skills in extensive, frontier ranching migrate from a point of displacement (from the intensive frontier) toward the extensive frontier. In this sense, these economic ‘opportunists’ will move to take advantage of profitable opportunities as they emerge (or dissolve) across space. In any case, if the agent skilled in a lesser intensive production strategy (e.g., frontier ranching) is no longer competitive in his location they must either accept a lower (or zero) wage, retrain, or relocate (Figure 2 and Table 1). Both of the displacement trajectories incur a cost but bestow a benefit. Each is considered in turn.

Table 1. Principal options faced by ranchers after displacement

Option	Retraining	Relocation
	Convert ranching operation into farm or otherwise intensify	Ranchers sell off all land and relocate, presumably to an area of lower marginal productivity
ILUC	Insignificant	Significant

Figure 2. Potential land use trajectories pursued by displaced rancher.

5.1. Retraining

Retraining involves the acquisition of the skills and experiences necessary to adopt a new production strategy. In the Amazon's many agricultural regions, commodity crop production often offers the potential for profits beyond those accessible to frontier ranching. This is particularly so where the development of the commodity network reduces transaction costs for the acquisition of inputs and the transport of harvested commodity crops. The cost of retraining or of acquiring the skills necessary to shift from beef to soy production, a potentially risky endeavor, however, may be substantial. Any conversion from pasture to soy is also likely to require a significant investment in the land, an investment which must be made without the financial windfall (a concept that will be developed in greater detail in the following section) that ranchers might receive through the sale or rental of their land. Cultural ties to professional practices, in addition to the costs involved in the retraining process and difficulty in securing the requisite investment capital, may also contribute to the notion that ‘uma vez gaúcho, sempre gaúcho’ (‘once a gaucho, always a gaucho’). Nevertheless, for younger producers with an extended employment horizon, retraining may present a particularly attractive opportunity.

5.2. Relocation

Relocating incurs the cost of reestablishment and the potential loss of valuable professional and social networks. Two displacement trajectories of particular interest to ILUC concern relocation. In the first, ranchers reap a moderate financial windfall that allows for the improvement or opening of land within existing properties. In this scenario, a portion of the former ranchland is converted to agriculture, whereas the remaining land continues to be used for ranching. The process may include the opening of new forest areas to accommodate a relocated herd. Implicit here are the intensification of production on existing ranchlands and the conversion of nearby forest areas for production.

Per a second relocation scenario, ranchers reap the full financial windfall from the sale or rental of their land and relocate (and potentially expand) their production strategies at the frontier. In this case, the redistribution of human and financial capital directly results in the extension of the frontier. This displacement–relocation scenario is by no means specific to extensive frontier ranching, and displaced smallholder colonists very well may likewise liquidate their cleared and appreciated properties and relocate to new colonization initiatives.

Supposing that the producer will choose the displacement trajectory that offers the most prodigious economic return, some displaced producers will seek to reap an advantage from their accrued skills and relocate when or where relocation offers the greatest opportunity for profit. More critically, assuming that some portion of displaced ranchers (or small farmers) at the intensive margin follow the relocation scenario, then the displacement of ranchers and ranches by agriculture can be interpreted as an indirect driver of pasture-led deforestation.

What separates the Amazon condition from much of the United States (and much of the rest of Brazil) is the relative mobility of its farmers, a factor likely tied to the still recent development of the Amazon as an expanding agricultural frontier. Given that much of the Amazon was colonized only as recently as the latter half of the past century, and that soy has only come to prominence in the past two decades, few Amazonian farmers and ranchers were born in the region. Consequently, more producers have fewer ties to the land or may be more open to repeated settlement opportunities as the frontier (and economic opportunity) progresses. In this sense, the open nature of the frontier here means that the sale of land may not mean the sacrifice of a way of life. Indeed, by moving again, frontier ranching can likely be duplicated, whether through spontaneous occupation of land at the fringes of the frontier or via the purchase or acquisition through other means of low-cost land in more distant regions.

In the Amazon, land has traditionally been relatively accessible financially, if not physically. The supply of skills and resources to develop the land's agricultural potential, however, is comparatively limited. Consequently, the marginal utility of human and financial capital outshines that for land. As frontier skills are displaced through the encroachment of more intensive agricultural land uses, these skills will be made available and spatially redistributed to the frontier, where both use-fixed human and financial capital are relatively scarce.

Land appreciation, paired with producer migration, is of particular importance to understanding ILUC, given that relocation not only mobilizes human capital, but also converts fixed capital sunk in land (e.g., the land value) assets to fungible capital for spatial redistribution. Appreciation thus enables relocating (or displaced) agents to liquidate sunk capital, which facilitates the establishment of their operations in new locations.

6. Financial windfall and the displacement effect

The liquidation of capital through the selling of old pastures amounts to what is referred to here as a financial windfall for ranchers. It also amounts to a transfer of wealth from the soy industry (or its investors) to the cattle industry, which can subsequently be used to finance the purchase of new properties and the relocation of existing operations. Thus the transfer of old pasture land to incoming farmers allows for the spatial redistribution of capital to the Amazonian frontier, the relative absence of which has long stood as a major impediment to the region's economic development.

It is hardly controversial to suggest that the farther one progresses toward the frontier, the less expensive land becomes (larger supply and lower demand) and the scarcer the stocks of human or financial capital (see Figure 3). Consequently, the environmental impacts of capital invested at the frontier are amplified in comparison to its impact on already more productive spaces. Even if the rate of appreciation for pasture areas at the intensive margin is comparable to the rate of appreciation of land at the frontier, the owner of an equal-sized parcel at the intensive margin will see a greater sum increase in land values. This additional value, once liquefied, can be employed to obtain and bring into production a larger parcel at the frontier. In this sense, as property values rise in areas of crop expansion, the opportunity cost to ranchers remaining in these areas will grow. ⁴ A financial windfall obtained by liquefying (or renting) appreciating land assets facilitates the relocation and potential expansion of both the agent and the industry by breaking down a principal economic constraint, namely access to capital in relation to land at the frontier. I note here that even where credit markets are functioning properly (which may be unlikely at the frontier), for land to be brought into production the skills and knowledge of how to apply financial capital must also be present. In effect, the pairing of the financial windfall with use-fixed human capital results in the simultaneous spatial redistribution of these resources to the frontier.

Figure 3. The proportion of capital inputs to land in relation to distance from a theoretical market or point of consumption.

The windfall capital effect can be illustrated within the adaptation of the rent model elaborated earlier. Again, assume a two-product system, where maximum rents and average freight costs are differentiated and where an abstract landscape is partitioned into an intensive and extensive zone, and an area of uncultivated wilderness. Assume also that within this landscape the value of land is proportional to its potential for generating rents. Ceteris paribus, the closer the location of a parcel to a point of export, the lower the transportation cost. This surplus, derived from the location premium, is ultimately obtained by the landowner via higher rental charges (or land values). In this model, a financial windfall can be estimated based on the spatially contingent surplus and area transferred (shaded area on the x-axis, Figure 4) from the encroaching farming sector to displaced ranchers (who retire or relocate). A change in rents for soy (shown as R _soy′ in Figure 4)pushes outward the area of viable agricultural production, allowing for the conversion of pasture to crop production. The value of land transferred from pasture to soy (shaded area, Figure 4) is redistributed to ranchers, potentially to be reinvested at the extensive margin, or frontier.

Figure 4. Capital transfer is shown using a rent-based model.

Because land prices are higher at the point of intensification (where pastures are converted to crop production) than those at the extensive margin (where forests are cleared for cattle), the spatial utility of the displaced capital will be higher at its point of relocation. Practically speaking, the value of land sold at the intensive margin can be used to obtain a much larger parcel of land at the extensive margin. I also note that displaced capital may be redirected to the frontier on the prospect that appreciating land values will make an ‘unproductive’ investment profitable at a later date (Walker 2001).

Per the conceptual elements developed here, ILUC is driven by the realignment of a land use hierarchy founded on the allocation of land according to profit potential and economic marginality. I thus suggest that displacement occurs when new incentives for intensive land uses results in the relegation of less-intensive production strategies to more marginal land. Displacement, however, is not benign. Given that opportunities for ranching may exist elsewhere, where land can be acquired cheaply, the dislocation of human capital tied to ranching will increase availability of this valuable capital and create new opportunities in new spaces of production. Additionally, the transfer (and liquidation) of capital from the soy industry to ranching via the sale of appreciated land endows the ranching industry with a valuable base of fungible capital that facilitates relocation. Because marginality on the Amazon frontier is, in many places, a factor of access and transaction costs, the redistribution of capital to migrating frontier ranchers also results in the redistribution of capital to once inaccessible forest regions.

7. ILUC on BR-163, the ‘Soy Highway’

In the remainder of this article, I illustrate the two conceptual components developed in the previous section through an analysis of events and changes occurring in northern Mato Grosso.

I focus heavily on federal highway BR-163, also known as ‘Cuiabá-Santarém,’ or even simply the ‘Soy Highway,’ given the prolific movement of soy along the southern reaches of the road. BR-163 is also considered one of Brazil's most contentious locations, a designation stemming from its strategic positioning in an area of both critical ecological importance and minimal institutional control, and recently, heated discussions regarding whether to pave or not to pave (Fearnside and Lima de Alencastro Graça 2006; Fearnside 2007). Nearly 1800 km in length, BR-163 transects the southern basin, a region that includes extensive tracts of moist tropical forest, fields of soy, corn, and cotton, and cattle pastures. Completed (but never fully paved) in 1976, its construction was part of a larger plan to draw the Amazon region economically and politically under the influence of the federal government.

Today the asphalt ends several kilometers south of the Mato Grosso–Pará state border, north of which the comparatively well-maintained highway of Mato Grosso gives way to nearly 800 km of dusty track and derelict bridges. The region's inaccessibility has weakened institutional control (particularly in northern Mato Grosso and western Pará states) and resulted in a history checkered by conflicts, particularly over land titles and land rights. The dusty track nevertheless remains the only road between Mato Grosso and the Amazon River. However, its state of disrepair ensures that nearly all of the beef and soy produced in northern Mato Grosso must travel more than 2000 km south to reach points of consumption (for beef, in São Paulo state) or export (for soy, at Santos, also in São Paulo state, or in nearby Paranagua) in the south. Given the distinctive distribution of production strategies along the BR-163 corridor (Figure 5), defined largely by the distribution of transport costs, as well as the rapid expansion of both the intensive (soy) and extensive (cattle) frontiers, and the appreciation of croplands, pasture, and forest tracts (Table 1), the corridor provides a useful context within which to examine and assess the functional mechanisms of ILUC.

Figure 5. Cattle production (at left) has expanded at the northern reach of the BR-163 in Mato Grosso; soy production (at night) has expanded in the southern municipios, closer to points of export in São Paulo and Paraná states. Changes shown for the period 1998–2008.

Consider first the variation and appreciation of land prices for forest, pasture, and arable land in municipios (roughly equivalent to a US county), along BR-163 (Table 2). A review of land prices suggests that land values are a function of land cover (spatially fixed investments) and distance to São Paulo. Predictably, land prices decrease with distance to São Paulo. From 2002 to 2005, a period of high prices for soy and cattle production, land prices appreciated rapidly. It is hardly controversial to suggest that the rise in land prices in the Amazon stems at least partly from favorable conditions for agriculture and ranching. This is in stark contrast from 2008 to 2009, when prices for land stabilized, and when farmers suffered from declining soy prices and a relatively unfavorable exchange rate (Table 3). Per the theoretical models outlined in the earlier sections, the appreciation of land, combined with the spatial gradient in land prices, incentivizes the relocation of low-intensity land uses such as frontier ranching to the extensive margins.

Table 2. Land prices in the BR-163 corridor, 2005 (US$/0.01 km²), percent change from 2002 in parentheses

Municipio	Forest	Pasture	Arable	Distance (km) to São Paulo
Nova Mutum	–	1724 (206)	3696 (200)	1842
Sinop	–	–	3203 (249)	2020
Guarantã do Norte	160 (38)	716 (319)	903.4 (109)	2249
Alta Floresta	343 (82)	685 (64)	–	2353
Note: Prices are converted to US$ using 2010 exchange rate and do not account for fluctuations in rate of exchange: 1 BR Real = US $0.57.
Source: FNP (2005).

Table 3. Land prices in the BR-163 corridor, 2009 (US$/0.01 km²), percent change from 2008 in parentheses

Municipio	Forest	Pasture	Arable	Distance (km) to São Paulo
Nova Mutum	–	1653 (18)	4237 (−4.9)	1842
Sinop	–	–	2926 (5.5)	2020
Guarantã do Norte	399 (23.4)	855 (−7)	1710 (11.1)	2249
Alta Floresta	456 (27.1)	855 (−2)	–	2353
Source: FNP (2010).

As suggested in Section 6, the rapid appreciation of land prices and the spatial differentiation of pasture and forest values provide an incentive for the relocation of ranchers to the northern, lesser-developed areas of the region. To illustrate these incentives, consider, for example, the land values displayed in Table 2 as the basis by which to calculate the financial windfall described earlier, where the windfall is equitable to the price received for pastureland sold at the intensive margin. Assume that land managers choose their production strategy and location to maximize profit. If relocating and reestablishing ranching operations is believed to be more profitable than remaining in the current location and either continuing ranching or retraining for soy production, then the rancher at a point of displacement will relocate.

Consider then land prices from Nova Mutum and per-hectare profits for ranching and soy production in neighboring Sorriso, two municipios located at the intensive margin (Figure 6). Costs for soy production (2006) in Sorriso are estimated at US $527 per hectare (CEPEA 2007), producing 3000 kg per hectare. In 2006, a 60 kg soybean bag in Sorriso was valued at US $14.82, resulting in an average income of US $741 per hectare and sum profit as US $214 per hectare. Obtaining these profits requires converting pastures to arable land, a cost estimated by farmers to be as high as US $570 (1000 Reais) per hectare. However, at about the same time, pastures in neighboring Nova Mutum (land prices were not available for Sorriso) were valued at approximately US $1724 (FNP 2005). Assuming that forest clearing costs could largely be offset through timber sales on land to be cleared, a new ranch could be established for little more than the price of land (Margulis 2004). In Alta Floresta, a municipio 500 km to the north, forest land can be acquired for US $343. There, per hectare profits for ranching were estimated at US $79 per hectare. Although per hectare profits may have been slightly higher for ranching in Nova Mutum, these data were not available; figures for Alta Floresta were substituted.

Figure 6. Location of Alta Floresta, Sorriso, and Nova Mutum in northern Mato Grosso.

Starting with these figures for land prices and profits, the net present value (NPV) of future rents (Equation (1)(1)) is given by

(1)

In Equation (1)(1) R is rents, W the financial windfall acquired through the sale of land, C the cost of newly acquired land (for relocation) or land use conversion costs, r the discount rate (0.05), and t the year. To account for potential life cycle implications, NPV was calculated for three different employment time horizons: t = 5, 15, and 30 (years). Note that this analysis does not account for perceptions or attitudes toward risk, which incentivizes to various degrees remaining and maintaining current land use practices. The results suggest that economic incentives during times of appreciation and spatial differentiation of land prices (particularly when this differentiation is attributed to transportation costs) are highly structured to favor producer relocation.

Amid the incentives outlined in Table 4, several land use trends were apparent in Nova Mutum and Alta Floresta. In the former, despite continued deforestation (and the probable opening of new pastures), from 2003 to 2005 cattle populations declined to 104,897 from 109,244 heads. At the same time, soy production dramatically rose 36%, from 2450 to 3338 km². Farther north, in Alta Floresta, the cattle population expanded by 95,600 heads and nearly 482 km² of forest was cleared (IBGE 2010a, b; INPE 2010).

Table 4. Per hectare NPV for ranchers in Nova Mutum, Mato Grosso, for potential production strategies

				Net present value
Option	Rents	Costs	Windfall	5 years	15 years	30 years
Remaining	79	0	0	342	820	1214
Retraining	214	570	0	356	1651	2719
Relocating	79	343	1725	2067	2545	2939
Note: Values listed in US$ (5% discount rate).

8. Farm-level observations

Indications of land prices and cattle and soy production alone by no means confirm that the increase in production in Alta Floresta has necessarily come through the relocation of displaced producers from Nova Mutum, per se. However, it is suggestive of dynamic structural incentives linking intensification at the intensive margin (e.g., at Nova Mutum) and the movement of the extensive margin (at Alta Floresta). To support these findings, additional observations must be made at the farm level. To this end, a series of key informant interviews and a survey were conducted along much of the Santarém-Cuiabá corridor in 2008 and 2009.

Results suggested that (1) production (and producers) is highly mobile, that (2) expansion is limited through access to capital and exports costs, that (3) large-scale agriculture is a follow-in (or, secondary) land use, as rarely do farmers directly convert forests into arable fields for agriculture, and that (4) few ranchers turn to farming, or that land managers ‘displaced’ by agriculture pursue their former land use strategy after displacement. I consider each in turn.

8.1. Mobility

Frontier ranchers and farmers are highly mobile. This not only contributes to the rapid expansion of both the extensive and intensive frontiers, as suggested in the conceptual framework outlined in the previous sections, but is consistent with findings that suggest a tendency for multiple production locations per frontier farmer or rancher (Branford and Glock 1985). Farmers, particularly those at the edge of the soy frontier in northern Mato Grosso and western Pará, were largely recent arrivals (within the past 10 years); migration to the region was driven by a desire to test economic opportunities for soy or other crops in new locations. However, although many suggested that they had only recently arrived in the region, this was not so for their ‘arrival’ to the soy industry. The region's new farmers had worked in soy production for years and had honed their skills elsewhere, whether in the nation's agricultural strongholds in the south, in other regions of Mato Grosso, or even as far away as the Brazilian agricultural colonies in Paraguay.

8.2. Constraints to expansion

Farmers suggested that their expansion has been largely constrained by access to financial capital and transport costs. Given the relative cost of inputs, including the preparation of new fields and the purchase of machinery required to establish a large-scale farming operation, lack of access to affordable credit is a prominent obstacle to expansion. ⁵ For producers located along the unpaved regions of BR-163 in western Pará or in areas of northern Mato Grosso, one-half of the farm-gate price for soy was often destined to pay for its transportation. Given that the 2000 km journey from northern Mato Grosso to the port at Santos must be made by truck and trailer (as opposed to far more efficient rails or waterways), transportation costs will continue to constrain the region's agricultural and pasture lands.

8.3. Pastures in transition, or soy as a follow-in land use

As a third observation, I add that none of the farmers surveyed saw themselves as ‘pioneers’ or that their arrival was not seen as a part of a first wave of colonization. None claimed to have opened large tracts of land, and they viewed themselves as preceded by an earlier cohort of ranchers and small farmers. The farms that they had carved together were created through the improvement of old or ‘dirty’ pastures, or from smallholder farming communities. Rarely were they cleared from standing forests; where forest clearing did occur, it was said to have been done largely in residual form, and aimed only for the expansion of existing fields or to accommodate the machinery employed in cultivation. Such findings are largely in agreement with recent suggestions that soy expands into pasture, and rarely does so through encroachment into forest (Brown et al. 2007; Jepson et al. 2010; Lapola et al. 2010). This is particularly significant, as it supports the notion that soy production is largely a follow-in land use that takes advantage of the investments and infrastructure that emerge after colonization.

8.4. Displacement

A final set of observations derived from this work suggests that, although soy may not be a principal driver of direct forest loss, new production displaces ranchers and, in some cases, smallholder farmers. Farmers noted that the previous owners or managers of their land followed one of the displacement trajectories described in Section 5. Displaced small farmers or ranchers moved into ‘the city,’ where they used their financial windfall to purchase homes and reap access to amenities such as electricity, running water, and medical facilities. Other former owners had also effectively retired, often to the states from which they had previously immigrated.

Some of these displaced land managers appear to be replicating their production strategies in other locations, as indicated by discussions with area researchers. Evidently, poorer residents obtained a financial windfall from land sale and moved to regional assentamentos (state-supported settlement initiatives for small farmers) where they once more took up their previous land use strategies. In other cases, displaced ranchers sold a portion of their lands and relocated to newly cleared portions of the remainder. Still others sold their entire properties and used the windfall to resettle on the agro-pastoral frontier. Thus, both interviews and discussions repeatedly referred to the presence of a displacement effect both within the soy industry itself and through the displacement of ranching (and timber) operations. Thus, this fieldwork suggests a displacement effect associated with the expansion of soy production. Nevertheless, the results agree with the concept that ILUC occurs as producers identify costs associated with maintaining their production location and identify opportunities to maximize rent potential in a new location. The observations also support the framework presented earlier, particularly through evidence that land use change tends to occur with change in ownership, and by extension, the notion that the soy expansion has not largely come through the maturation of human and financial resources once invested in ranching.

9. Conclusion

Agriculture has taken on a prominent role in discussions regarding Amazonian deforestation during the past several decades. However, the focus on proximate drivers of deforestation in the region has minimized our understanding of the indirect linkages between the expansion of crop production and forest loss. Although recent work in the Amazon is rapidly beginning to bring these issues to light, only recently have the indirect effects of the expansion of agriculture begun to be recognized (Barona et al. 2010). Lapola et al. (2010), for example, recently suggested that sugarcane ethanol and soy biodiesel in Brazil could account for one-half projected indirect deforestation (121,970 km²) by 2020. In this article, I have called attention to this process at the regional level and have outlined a framework by which to conceptualize the cascading spatial effects that emerge through land use displacement.

The conceptual mechanisms developed and illustrated here have focused on the spatial redistribution of capital that accompanies the relegation of lesser-intensive land uses. They affirm that although cattle ranching, as a lesser-intensive land use, remains the principal industry responsible for the clearing of forest cover, new demand for soy, whether for processed foods, poultry and swine, or biodiesel production, continues to act as a driver behind its expansion. Given the prevailing incentive structure for agricultural expansion (e.g., via expansion into old pastures), new soy fields are unlikely to directly result in forest loss. Indirectly, however, via the displacement of other land uses and their associated capital, an expansion of agriculture in the Amazon will continue to manifest itself through ILUC in the form of a decline in forest cover.

Any consideration of the carbon footprint associated with the increase in soy production in the Amazon thus depends on understanding the ILUC process. This process rests on the appreciation of land and the opportunity costs of continuing current practices when economic incentives suggest that relocating is a more profitable alternative. Where and when agricultural rents have been high and land has appreciated rapidly, as it has in Mato Grosso for much of the past decade, ranchers will continue to face strong incentives to relocate their production to the frontier. In this sense, the dislocation of valuable capital and skills by soy results in an advancement in the cattle frontier via the ILUC process and, potentially, a deleterious impact on forest cover as the cascading economic effect breaks on the margins of the forest.

This leads to several points of conclusion. First, this article suggests that intensification, often held out as a means by which to combat forest loss and implicit in any discussion of ILUC, serves indirectly, as a driver of deforestation. This is contrary to the long-standing theory, sometimes referred to as the Borlaug hypothesis, which suggests that an increase in yield ultimately results in land sparing and a reduction of the total land needed for agriculture (Borlaug 2008). The findings of this analysis are closely aligned with recent work suggestive of a linkage between intensification and forest loss and, more generally, an increase in land brought into production (Vosti, Witcover, Oliveira, and Faminow 1998; Angelsen, van Soest, Kaimowitz, and Bulte 2001; Keys and McConnell 2005; Kaimowitz and Angelesen 2008; Rudel et al. 2009).

Second, given the low cost of land, particularly at the frontier, it follows that the supply constraint of least resistance in these areas is through the expansion of land, not capital. Per the displacement mechanisms outlined earlier in this article, investment in production at the intensive margin transfers capital to less-extensive land uses, and thus breaks down the capital constraint to their expansion. Given the high ratio of land to capital at the extensive frontier, the marginal impact of capital investments is extremely high, which amplifies the effect on regional forest cover. Consequently, the land use effect attributed to displaced cattle may be significantly higher than the area of displacement.

Finally, this article also offers a direct challenge to the understanding that the Amazon's agricultural boom causes only minimal impact on forest cover. It brings into question the effectiveness of a soy moratorium, a policy trumpeted by a consortium of producers, agribusinesses, and non-government organizations and designed to deter direct encroachment. Here I argue that a major ecological impact of soy lies not with its direct encroachment, but rather with the spatial redistribution of use-fixed human capital and the transfer and liquidation of financial capital to lesser-intensive land uses. Thus, although any moratorium on products derived from newly deforested land may mitigate land clearings during peak boom periods (e.g., when economic conditions become so favorable that soy production could conceivably encroach directly into forest regions), it is unlikely to offer any real release from the pressures derived from ILUC. This being said, discussions with farmers in the region suggest that there is little recognition of their industry's potential ILUC impacts. Indeed, farmers surveyed or interviewed as part of this research were adamant that their work brought not only economic benefits to their municipios in the form of increased per capita income and access to new amenities and resources, but also greater environmental stewardship. To an extent, these perceptions appear both sincere and correct. Farmers invest heavily in their land and many take pains to ensure that the necessary steps are taken to comply with the region's strict environmental laws. Given that the importers of soy are acutely aware of the value of maintaining a minimum level of eco-friendly standards to satisfy increasingly environmentally conscious consumers, there are incentives to operate in a manner deemed environmentally sensitive. Nevertheless, I suggest that the structure of the soy expansion, whether done ‘responsibly’ (without direct forest clearings) or ‘irresponsibly’ (with direct forest clearings) by the individual producers, wields a significant and adverse impact on the basin's overall forest cover via ILUC.

Incentives for soy and beef are unlikely to abate in the coming decades. The same can also be said for efforts to reduce carbon emissions. Given that cutting deforestation rate is a cost-effective means by which to reduce CO₂ emissions (Angelsen et al. 2009; Nepstad et al. 2009), policy-makers will continue to battle to balance the need to maintain minimum levels of forest cover with new food production. Without fail, farmers and ranchers in the Amazon will have an important role to play in provisioning the world with affordable and abundant supplies of food and fuel. This role, however, will continue to test the Amazon as a reservoir of carbon and biodiversity. Steps will need to be taken to ensure that the expansion of intensive production does not result in the displacement of less environmentally responsible production strategies. Most importantly, policy-makers must recognize the linkages between multiple, spatially divergent land uses, or many otherwise worthy efforts to reduce CO₂ emissions may simply be diverting, not abating, significant environmental impacts.

Acknowledgements

This article is based on research conducted partly through work developed with the assistance of the Social Science Research Council's Dissertation Proposal Development Fellowship. The article also acknowledges research supported in part by NSF project (BCS-0620384) ‘Collaborative Research: Globalization, Deforestation, and the Livestock Sector in the Brazilian Amazon.’ The views expressed here are those of the author and do not necessarily reflect those of NSF. Dr. Robert Walker contributed greatly to the formation of this article through his comments on earlier drafts and through germane discussions pertaining to the conceptual developments presented within. I also thank Dr. William McConnell, Dr. Scott Swinton, Dr. Soren Anderson, and several anonymous reviewers; their thoughtful comments greatly improved the text. Any errors remain the sole responsibility of the author.

Notes

1. Direct land use change refers to the land use effect associated with activities being carried out on a parcel of land itself. ILUC refers to the indirect effects of direct land use change. Specifically, it refers to the displacement or leakage of existing production activities and the relocation of these operations, often to more marginal locations. For a more complete understanding of the differences between direct land use change and ILUC, see Gnansounou et al. (2008).

2. Rent refers to the potential profit surplus specific to a given parcel of land. This is generally written in equation form as π _k =  q_kp_k  – v_kk_k  – f_kd, where rents at a given location (π _k ) are defined as a function of produce generated per unit of land (q_k ), output price (p_k ), the cost of capital (v_k ), the cost of inputs (k_k ), and unit shipping costs (f_k ) at distance from point of consumption (d). Increases in either output or output price have a positive relationship with overall rents, whereas an increase in the costs of capital or inputs required results in a negative impact. The transport cost associated with the good in question determines the slope of the ‘rent curve,’ which at some distance from the market or point of consumption diminishes to the point where production is no longer profitable.

3. I define as use-fixed capital, investments where utility is specific to a single production strategy or set of closely related strategies.

4. The actual cost of relocating may be minimal. The cost of opening forests for pastures is often offset by selling the wood located on the property (Margulis 2004).

5. The mean farm size of the subjects interviewed was nearly 2000 hectares, or approximately 20 km².