Indirect land use change from ethanol production: the case of sugarcane expansion at the farm level on the Brazilian Cerrado

ABSTRACT

The expansion of sugarcane production in the Brazilian Cerrado has resulted in indirect land use change (ILUC), occurring when displaced land uses in one location are reallocated to another. Studies, however, usually identify ILUC at the regional or national level far from the original area of a displaced land use. This study examines the occurrence of ILUC due to sugarcane expansion for ethanol production at the farm level in the Brazilian Cerrado. It fills a gap in the literature by examining socioeconomic, policy and farm-level factors that influence ILUC at the farm level in the Brazilian Cerrado using face-to-face enumerated surveys. Results indicate that ILUC did occur at the farm (producer) scale and farmers who undertook ILUC intensified agricultural production on their farms. Results inform policymakers on how the intensification of agricultural practices may make it potentially difficult to keep protected lands out of production, reducing the environmental benefits from sugarcane-based biofuel production.

KEYWORDS:

• Biofuels
• Cerrado
• farmer
• indirect land use change
• intensification
• local scale

1. Introduction

Increased global demand for biofuel production from agricultural crops has significantly increased and is expected to continue to grow into the future. Brazil is one of the largest suppliers of sugarcane ethanol in the world and is responsible for over half of sugarcane production worldwide (Brazil, 2013). Energy production from sugarcane and its derivatives is the second largest supply of energy in Brazil (Andrade de Sa, Palmer & Engel, 2013). Furthermore, energy from sugarcane is a cornerstone for the Brazilian National Policy on Climate Change, which set a voluntary reduction on greenhouse gasses emissions of 36.1–38.9% of the business as usual by 2020 using 2005 as the base-year (Brazil, Law 12187, 2009). The policy establishes a goal of obtaining 80% of energy power from renewable sources by 2030, and it emphasizes the role of biofuels, especially sugarcane ethanol. This policy aims at reducing greenhouse gas emissions from deforestation, as well as agricultural and livestock production (Brazil, Law 12187, 2009).

In this context, between 2005 and 2013 sugarcane expanded due to a push for greater ethanol production into the Center-West of Brazil, particularly in the states of Goiás (GO) and Mato Grosso do Sul (MS). The Center-West, comprises a significant portion of the Cerrado (tropical savanna), the country’s second largest biome (Shikida, 2013). From 2005 to 2013, 40 new ethanol producing mills were constructed in these two states (Granco, Caldas, Bergtold, & Sant’Anna, 2015). The two states combined (GO and MS) planted over 1.5 million hectares of sugarcane in 2014 and contributed to 15% of total sugarcane produced in Brazil (Institiuto Brasileiro de Geografia e Estatística [IBGE], 2014; Sant’Anna et al., 2016). Further expansion was promoted with the mapping of suitable areas for sugarcane production (12.6 million hectares in GO and 10.8 million hectares in MS) by the Sugarcane Agroecological Zoning, launched in 2010 (Manzatto, Assad, Bacca, Zaroni, & Pereira, 2009).¹ The increased land use toward sugarcane production has directly displaced more traditional crop and livestock production in the region (Nassar et al., 2008).

While direct changes in land use are likely to occur with increased sugarcane production for ethanol in the Cerrado, indirect land use change (also known as a ‘leakage effect’ in the literature) may result, as well. Indirect land use change (ILUC) is defined as taking place when displaced agricultural activities or land use in one location is reconstituted in another (Arima, Richards, Walker, & Caldas, 2011; Barretto, 2013; Lambin & Meyfroidt, 2011; Lapola et al., 2010; Searchinger et al., 2008). For example, increased sugarcane production in the Center-west of Brazil has increased by being produced on land used previously for pasture or soybean production. This may have resulted in pasture and soybean production being undertaken on new lands that had not been under that land usage before, including the potential conversion of native vegetation or forest. A potential driver of ILUC in the Cerrado during this time period was the expansion of sugarcane production for ethanol due to favorable market and policy conditions, such as more affordable agricultural land, good growing conditions for sugarcane, fiscal incentives, credit and investments in infrastructure (de Souza Ferreira Filho & Horridge, 2014; Granco et al., 2015). The displacement of land by these and other market forces moved some displaced agricultural land use activities onto land that was less productive and provided lower management costs (Andrade de Sá et al., 2013; Arima et al., 2011). A potential consequence of the increase in demand for and promotion of the production of sugarcane ethanol in Brazil is the intensification and expansion of production of food and feed crops on existing pasture and crop land, as well as conversion of native forest and vegetation (Barretto, Berndes, Sparovek, & Wirsenius, 2013). These changes may have potentially adverse environmental and biodiversity consequences (Carvalho, De Marco Junior, & Ferreira, 2009; Sawyer, 2008).

In Brazil, the nationwide expansion of sugarcane production of both ethanol and sugar came primarily from converting cropland (~32%) and pasture (~66%), with the remaining 2% coming from natural vegetation (World Bank, 2011). With increasing land prices in the areas in which sugarcane expanded, there was market pressure to move displaced crop and pasture production to other regions, despite increases in agricultural productivity (World Bank, 2011). Simulating the expansion of sugarcane from 2005 to 2025, De Souza Ferreira Filho and Horridge (2014) found that every hectare of sugarcane expansion in the South Central part of Brazil results in a 0.14 ha decline in forest or natural vegetation and 0.47 ha decline in pasture along the agricultural frontier in the Amazon region of Brazil. Displacement of agricultural activities and resulting ILUC can lead to increased deforestation, loss of natural wildlife habitats, impacts on biodiversity, lower carbon sequestration from the land and increased greenhouse gas emissions that may be as great as direct land use change (Allan et al., 2015; Andrade de Sá et al., 2013; Arima et al., 2011; Fargione, Hill, Tilman, Polasky, & Hawthorne, 2008; Searchinger et al., 2008). For example, Carvalho et al. (2009) find that crop-dominated landscapes are more fragmented in the Cerrado and provide smaller numbers of fragments that can support threatened mammal species. They indicate that expansion of sugarcane production on prior pasture land in the Cerrado could further fragment those lands, leading to potential habitat and biodiversity losses. The environmental and biodiversity implications of ILUC are likely a result of both sugarcane expansion and agricultural intensification in the region. Barretto et al. (2013) indicate that agricultural intensification coincided with agricultural expansion in agricultural frontiers, such as the Cerrado, during the 1990s and 2000s.

Analysis of indirect land use change has mostly taken place at the global scale with some studies conducted on a more regional level (Andrade de Sá et al., 2013; Arima et al., 2011). Many studies at the global level can identify if ILUC is occurring in Brazil, but cannot identify the place of occurrence with any specificity beyond the regional or national level (Arima et al., 2011). Sub-national studies focus on a detailed spatial scale, providing evidence of the displacement of agricultural production due to sugarcane expansion, leading to deforestation and conversion of natural vegetation to pasture (e.g. Andrade de Sá et al., 2013; de Souza Ferreira Filho & Horridge, 2014). These studies show that the displacement of agricultural production by sugarcane expansion can result in a cascading effect, where ILUC spreads across a landscape due to migration of agricultural activities farther away from the original displacement of the agricultural activities in question. Thus, ILUC can occur hundreds of kilometers from the area of displacement over time (e.g. Andrade de Sá et al., 2013). This cascading effect begins in the local area where the agricultural production was originally displaced with local producers. At the local level, ILUC can occur if farmers begin to convert other agricultural land or natural vegetation to displaced agricultural activities. Barretto et al. (2013) find that areas of cropland expansion in Brazil usually occur within short distances of each other.

Palmer and Owens (2015) argue that to fully be able to deal with ILUC as a result of biofuels expansion, researchers and policymakers need to understand the place-specific context of factors that lead to ILUC. Analysis at the local or farm level may have implications for understanding socioeconomic, policy and farm level factors that may impact ILUC, such as the Sugarcane Agroecological Zoning policy, access to rural credit and risk-averse behavior (Andrade de Sá et al., 2013; Granco et al., 2015). For example, understanding factors that impact ILUC at the farm level may help with designing sustainable biofuel and agricultural policies and initiatives that further limit deforestation, protect natural habitats and ensure the integrity of Brazil’s protected areas (e.g. Rausch & Gibbs, 2016).

The purpose of this paper is to examine socioeconomic, policy and farm-level factors influencing indirect land use change at the farm level as a consequence of the expansion of sugarcane production for ethanol in the Brazilian Cerrado. Using a set of intensive face-to-face enumerated surveys, the paper uses qualitative assessments and regression techniques to examine ILUC at the farm scale. The paper contributes to the literature on bioenergy and ILUC in several novel ways: (i) by examining ILUC at the farm level in Brazil; (ii) examining demographic and behavioral characteristics of farmers that may influence ILUC at the farm level due to the expansion of agricultural production for biofuels; and (iii) providing farm and local level recommendations for policymakers for helping to reduce deforestation, degradation of natural habitats and threats to protected areas in Brazil to provide more sustainable biofuel feedstock production.

2. Local indirect land use change

Gnansounou, Panichelli, Dauriat, and Villegas (2008) identify four different types of indirect land use change (ILUC): (i) spatial ILUC; (ii) temporal ILUC; (iii) use ILUC; and (iv) displaced activity/use ILUC. The two primary types of ILUC occurring at the local level in this study are (i) spatial ILUC and (ii) temporal ILUC. Spatial ILUC occurs when the production of sugarcane as a new land use pushes the previous land use to another location (Gnansounou et al., 2008). At the local or farm level, this may occur when sugarcane production displaces a prior land use (e.g. soybean production) that is still profitable or viable to another location locally or on the farm (e.g. lands with native vegetation or pastureland) to maintain the current level of production of that prior land use. This type of ILUC is not a result of crop rotations. Temporal ILUC occurs when land was cleared for a prior land use that is later used for sugarcane production (Gnansounou et al., 2008). At the local or farm level, this could occur on land that was originally cleared for another use (e.g. pasture) that may have become degraded over time and is then put into sugarcane production. ILUC may lead to more intensive agricultural production on the land (e.g. displaced soybean production being produced on degraded pastures). By intensification, we mean land use activities that require relatively greater amounts of inputs (e.g. tillage, pesticides, fertilizers, fuel); are more capital-intensive; and increase yield per unit of land than the prior land use. Turner, Plevin, O’Hare, and Farrell (2007) warn that environmental consequences such as loss of ecological functions, threats to biodiversity, increases in greenhouse gas emissions and nonpoint source pollution may result from agricultural intensification. Barretto et al. (2013) found that agricultural intensification in the agricultural frontier of the Cerrado has led to agricultural area expansion, with more land (both native vegetation and other land uses) being placed into agricultural production. The primary focus here is on spatial ILUC.

ILUC is place specific, in that the causes of ILUC will be unique to and differ by location, making any general causal linkages at a macro-level potentially flawed without taking into consideration analyses at the local or regional level (Palmer & Owens, 2015). Both market (e.g. greater foreign demand for ethanol) and policy factors (e.g. Sugarcane Agroecological Zoning) at the local (e.g. fiscal incentives) and macro-level have driven the expansion of sugarcane production in the Cerrado (Granco et al., 2015; Naasar et al., 2008). This has resulted in spatial ILUC within the states of GO and MS. Market and policy factors have resulted in changes in price signals to farmers that induce land use changes at the farm level, including spatial ILUC. Looking at the history of agricultural production, agricultural expansion over time and the discussion here, spatial ILUC is not a new phenomenon globally and has been occurring for a long period of time due to different market and policy forces. The contribution here is from the assessment of spatial ILUC as a result of the expansion of sugarcane production at the farm level in GO and MS.

3. Materials and methods

3.1. Area of study

The geographic area of study consists of two states, Goiás (GO) and Mato Grosso do Sul (MS), in Brazil, which are located in the Brazilian Cerrado, which covers 98% of the area (34 million ha) of GO and 60% of the area (21 million ha) of MS (Ministerio do Meio Ambiente, 2014). Figure 1 provides an overview of the study region and location of sugarcane production and sugarcane mills within the study region. The study region (and Cerrado) exhibits a wide variety of vegetation patterns, including open grassland, closed woodland and intensive cropping systems (Klink & Machado, 2005). Advances in agriculture in the region have allowed farmers to overcome soil deficiencies that inhibited crop expansion, transforming the Cerrado into a breadbasket (Jespon, 2006). The primary agricultural enterprises in this region include cattle, soybean, corn and sugarcane production. Pastureland accounted for 26 million ha of land use, while soybean and sugarcane production accounted for 5 million and 1.4 million ha in 2013, respectively (Brazil, 2015). The states of GO and MS are at the forefront of the recent sugarcane expansion in the country, with over 60 sugarcane mills in 54 counties across the two states (Brazil, 2015). Figure 1 illustrates the significant expansion of sugarcane production between 2005 and 2013 in the study region, both in land use and construction of new sugarcane mills.

Figure 1. Study Region, including counties where field work was conducted, as well as sugarcane production and mills in Goiás and Mato Grosso do Sul in 2005 and 2013 (Granco, 2017).

3.2. Data

Data were collected using face-to-face enumerated surveys with landowners and farmers in 22 Municipios (e.g. counties) in the Brazilian states of Goiás (GO) and Mato Grosso do Sul (MS). Survey design was based on studies conducted in Quirinopolis, GO (Picanço Filho & Marin, 2012, 2012a; Picanço Filho, 2010). The survey was tested with experts and farmers within the study region prior to its application in the field. The counties surveyed in each state were chosen based on: (i) geographic location of sugarcane production in 2012 using the National Institute for Space Research (INPE) Canasat Project (Rudorff et al., 2010); and (ii) sugarcane production growth obtained from the Brazilian survey of county-level agricultural production – PAM (Instituto Brasileiro de Geografia e Estatı´stica, 2014a, 2014b).

We contacted landowners and farmers from sugarcane growers associations, rural syndicates and the Goiás and the Mato Grosso do Sul Federation of Agriculture and Livestock (FAEG and FAMASUL) to participate in the survey. The survey provides information on participants’ demographics, farm characteristics, landownership, sugarcane production and contracts, perceptions of mills’ interaction with the local community, and land use. For the purposes of this study, we asked a detailed set of questions that tracked land used for sugarcane production and other uses over time at the farm level to assess how the expansion in sugarcane production changed land usage at the farm level. We specifically collect data on sugarcane acreage and distribution; expansion in area planted to sugarcane since 2010; and impact on displaced agricultural activities due to expanded sugarcane area. Surveys were conducted in 2014 from June to July. A total of 148 landowners and farmers were interviewed over a 60 day period. Farms’ remoteness, difficulty in locating farms and farms’ large sizes made time a serious constraint on data acquisition, limiting data collection efforts.²

As seen in Table 1, 91 out of 142 (64%) of the surveyed farmers had either produced sugarcane or rented land for sugarcane production at some time, while 80 (56%) of them still grew sugarcane. The rest of the respondents grew other crops or were cattle producers. The farmer population surveyed does not represent the entire farmer population in these states in Brazil. Rather, respondents represent the group of commercial farmers most likely to be approached by mills to supply sugarcane or to rent out their land for sugarcane production. The sampled group primarily consists of large commercial farmers, which are part of associations, rural syndicates and/or cooperatives involved in sugarcane production. Members of organizations tend to manage commercial farms usually larger in size. Hence, the average farm size in our sample is 1466 ha, while that of the 2006 Agricultural Census³ is 415 ha (IBGE, 2006). This difference is due to the census comprising a much larger number of smaller farms than the survey. Nevertheless, the percentage of male farmers in the census is close to that of the survey. In the census, 92% of farmers are male, while in our survey, 96% of the respondents were male. In terms of education, our survey has a higher percentage of farmers with high school and college degrees than the census. In our survey, 37% of the respondents had completed high school and 28% college. In contrast, the census reported 4% of farmers completed high school and 3% college. The survey also reported a higher average sugarcane production value and yield than the Companhia Nacional de Abastecimento (CONAB, 2013). CONAB (2013) reports an average yield of 70.3 tons/ha for the study region, while survey respondents reported an average yield of 87.7 tons/ha.

Table 1. Descriptive Statistics for Survey Data of Farmer Respondents on Sugarcane production, land use and management.

Category	Mean	Standard Deviation^a	N
Percentage (as a decimal) of farmers who have grown sugarcane.	0.64	–	142
Percentage (as a decimal) of farmers currently growing sugarcane.	0.56	–	142
Mean area in hectares.	240	393
Percentage of farmers producing sugarcane that converted land to sugarcane production by land type.
Crop land	0.58	–	91
Pasture	0.64	–	91
Native land	0	–	91
Percentage (as a decimal) of farmers who indicated that the following reasons were ‘important’ for their decision to grow sugarcane.
Higher profits	0.74	–	91
Contract with the local mill	0.64	–	91
Higher risks associated with other crops	0.35	–	91
Production costs	0.27	–	91
Pest management problems	0.10	–	91
Inside the Sugarcane Agro-Ecological Zoning	0.36	–	91
Farmers that increased sugarcane area since 2010
Percentage (as a decimal) of farmers	0.52	–	91
Mean hectares of land converted to sugarcane production.	167	329	91
Percentage (as a decimal) of sugarcane area that displaced or replaced prior land uses for farmers who expanded production since 2010. The percentage (as a decimal) of farmers who discontinued the displaced land use or moved it to a new area is provided for each land use category. Replacement or displacement of the land use is indicated in parenthesis after each description.
Pasture	0.51	–	47
Production lost (replacement)	0.60	–	42
Bought new land to put into pasture production (displacement)	0.33	–	42
Converted other land on-farm to pasture production (displacement)	0.024	–	42
Land planted to soybeans	0.43	–	47
Production lost (replacement)	0.52	–	29
Bought new land to put into pasture production (displacement)	0.10	–	29
Converted other land on-farm to soybean production (displacement)	0.21	–	29
Land planted to other crops	0.036	–	47
Production lost (replacement)	0.33	–	6
Bought new land to put into other crop production (displacement)	0	–	6
Converted other land use on-farm to other crop production (displacement)	0.17	–	6
Cerrado or natural vegetation	0.043	–	47
New land (rent or buy)	0.021	–	47

^aStandard errors are only provided for continuous statistics.

Summary statistics for sugarcane production and associated land use is provided in Table 1 and discussed in the next section of the paper. Table 2 provides summary statistics for sugarcane producers who increased the area planted to sugarcane on their farm in 2010 for the analysis of farm, socioeconomic, policy and factors impacting indirect land use change on commercial farms in the study region. While Table 2 provides overall summary statistics for the farms used in the ILUC analysis, there were some differences between the farms in MS and GO. Surveyed farmers used in the ILUC analysis in MS compared to farmers in GO were on average larger in size (1858 ha vs. 1394 ha), rented less land (approximately 6% of total farm land vs. 21% of total farm land), were half as likely to purchase crop insurance (25% and 55% of farmers purchased crop insurance in MS and GO, respectively), and were located closer to the mill (17 km vs. 24 km). The other summary statistics for the farms in each state were very close to the summary statistics for the study region.

Table 2. Descriptive statistics for factors impacting indirect land use change decisions by farmers surveyed (N = 47).

Variable	Definition	Mean (Frequency)	Standard Deviation
Dependent Variable
ILUC	Equal to ‘1’ if a farmer made a land use decision resulting in indirect land use change on their farm and ‘0’ otherwise.	0.17	–
Explanatory Variable
Percent Rented	Percentage of farm land rented.	16.1	28.3
Farm Size	Size of the farming operation (ha).	1552	1836
Risk Avoider	Equal to ‘1’ if the farmer identifies themselves as very cautious or tends to avoids risks and ‘0’ otherwise.	0.62	–
Insurance	Equal to ‘1’ if the farmer has purchased crop insurance and ‘0’ otherwise.	0.45	–
Sugarcane Sales	Percentage of gross farm sales from sugarcane production.	52.2	74.6
College	Equal to ‘1’ if the farmer is a college graduate and ‘0’ otherwise.	0.40	–
Zoning	Equal to ‘1’ if the farmer identified the Sugarcane Agro-Ecological Zoning as an important reason for producing sugarcane and ‘0’ otherwise.	0.34	–
Income from Farm	Percentage of household income from the farming operation.	82.2	26.1
Distance to Mill	Distance (km) to the closest sugarcane mill.	22.0	14.9

^aStandard errors are only provided for continuous statistics.

3.3. Farm level modeling of ILUC

3.3.1. Conceptual model

Andrade de Sá, Palmer, and Engel (2012) and Andrade de Sá et al. (2013) show that a necessary condition for the displacement of agricultural activities due to sugarcane expansion is that the output for the displaced agricultural commodity faces a relatively inelastic demand. This is likely the case for soybeans or corn, which are food staples, feed crops and primary export commodities in Brazil and worldwide. Thus, there is a market incentive for farmers to continue planting these crops, even as they consider new potential enterprises on-farm. We model farmers’ land use decisions about discontinuing production of a displaced land use or agricultural production activity or moving it to another physical location to maintain that activity, following Andrade de Sá et al. (2013).

We assume that farmer i is considering replacing land use k on their farm with a displaced land use j as a result of increased sugarcane production on-farm. This decision will be based on the expected profitability of land use option j on the given land relative to the profitability of land use k. We assume that the expected relative profitability of replacing land use k with land use j is given by: = . is a vector of farm characteristics that impact production and land allocation decisions for the farm (e.g. land tenure arrangements, available land, purchasing crop insurance, farm sales, household income from the farm and proximity to a sugarcane mill). is a vector of other socioeconomic factors that may influence a farmers’ land allocation decisions (e.g. risk aversion and education). Many of the factors in both and have been shown to impact the decision of a farmer to undertake a particular agricultural activity (Pannell et al., 2006). Wu (1999) shows that the purchase of crop insurance will impact cropping mix on the farm, potentially toward high earning and risker crops. Following the discussion in Granco et al. (2015), the proximity of a sugarcane mill may heavily influence land allocation on-farm by increasing the relative profitability of growing sugarcane as proximity to the mill increases. is a vector of policy variables that may impact land allocation decisions at the farm level. For example, if a farm is in the Sugarcane Agroecological Zone, then a farmer may deem it as an advantage to devote more land toward sugarcane production and less to other agricultural land uses, reducing the likelihood a farm that may continue a displaced agricultural activity.

Farmer i will replace land use k with the displaced land use j if . That is, if the expected profitability of converting to land use j is greater than remaining with land use k.

3.3.2. Empirical model

What is observed (and asked about) in the survey is if spatial ILUC occurred due to an increase in sugarcane production on-farm since 2010. That is, if a prior land use activity (e.g. pasture, soybean production or other crop production) on land currently under sugarcane production was moved to a new area of land on the farm to continue that land use activity. A binary variable (ILUC) was created using the survey data to indicate if a farmer had converted a portion of land not under sugarcane production to a displaced agricultural activity due to an expansion in the area planted with sugarcane since 2010 (Table 2). Observationally, farmer i will convert their land from land use k to the displaced land use j if:

(1)

Now assume that

(2)

where are parameters to be estimated, and is a mean zero IID error term. The explanatory factors are those identified in section 3.3.1, and summary statistics for them are provided in Table 2. Given the potential multicollinearity between gross farm sales and household income from the farm, gross sales are replaced with the percent of gross farm sales from sugarcane production to bring in the influence of the extent of sugarcane production on the farm. We recognize that market factors (e.g. crop prices) will play a significant role in shaping land use decisions, but given the cross-sectional nature of the study and lack of availability of spatially explicit market data, these factors were unable to be modeled. Assuming that is distributed extreme value Type 1, the model given by Equations (1) and (2) can be estimated as a logistic regression model (Train, 2009). Marginal effects for impact of a given explanatory factor on the probability of ILUC occurring on-farm are calculated following Greene (2012) with asymptotic standard errors estimated using the delta method (Greene, 2012).

4. Results and discussion

A qualitative and quantitative assessment of survey findings and ILUC modeling is presented. The qualitative assessment examines the descriptive statistics of the responses provided in the survey, while the quantitative assessment examines factors impacting ILUC on-farm.

4.1. Sugarcane expansion at the farm level and ILUC

The more recent expansion of sugarcane for ethanol production in Brazil has occurred in the Cerrado region, primarily in the states of GO and MS. Of the approximately 1.8 million hectares of land added into crop production in both states between 2005 and 2013, 53% of it was for sugarcane production (Granco et al., 2015; IBGE, 2014). Much of the expansion in sugarcane production in this region came from repurposed pastureland (69.7%) and cropland (25%) (Adami et al., 2012). Granco et al. (2015) suggest that mismanaged and degraded pasture and an outbreak of foot-and-mouth disease in 2005 in MS may have helped to precipitate expansion of sugarcane production into the region. Manzatto et al. (2009) and Granco et al. (2015) indicate that the Sugarcane Agroecological Zoning in 2009 encouraged conversion of degraded pastureland to sugarcane production to avoid competition with cropland for food production. Around the same time, soybean rust was also indicated as a contributing factor to sugarcane expansion (Sant’Anna et al., 2016). In addition, Granco et al. (2015) state that favorable agricultural conditions, agricultural policies, fiscal incentives and investments in infrastructure have helped to sustain the expansion of sugarcane production for ethanol. This was partially evidenced by the construction of over 40 new mills in the region during the time period to meet increasing demands for hydrous ethanol (Granco et al., 2015; Newberry, 2014).

Table 1 provides descriptive statistics for farmers surveyed in GO and MS. Data on the questions referenced in Table 1 were missing for six out of the 148 farmers surveyed, narrowing the sample size to 142.⁴ Of the farmers surveyed, 64% of the farmers had grown sugarcane on their farm and 56% of the farmers were growing sugarcane as of 2014. The mean area of sugarcane planted on farms was 240 ha. Farmers grew sugarcane on land that was previously used as cropland (58%) and pastureland (64%) prior to conversion to sugarcane, substantiating the findings by Adami et al. (2012). Farmers primarily chose to grow sugarcane to obtain higher profits, followed by the ability to contract with a local sugarcane mill, having land within the Sugarcane Agroecological Zone, to avoid risks associated with other crops, and lower production costs. The ability to contract with a mill in their local area that promised to maintain land productivity and not degrade the land was also tempting for farmers. Thus, the presence of a sugarcane mill seems to have been a precipitating factor in expanding sugarcane production at the farm level.

Of the farmers who had grown sugarcane (N = 91), 52% of them have expanded their sugarcane production since 2010 (or the past 5 years from the time the survey was administered in 2014).⁵ The mean increase in sugarcane planted area was 167 ha. As seen in Table 1, new sugarcane production replaced or displaced pastureland, cropland planted to soybeans, other cropland and native vegetation (Cerrado) on 51%, 43%, 4% and 4% of the land converted to sugarcane production by farmers, respectively. Only 2% of the farmers expanded sugarcane planting on new land purchased or rented by the farmer. For much of the displaced agricultural activities, production was lost. Farmers did not continue the displaced land use for pasture, soybean production and production of other crops 60%, 43% and 33% of the time, respectively. This is in light of the fact that total planted area devoted to production of other major cash crops (e.g. corn and soybeans) in the region were increasing in both states at the same time (Granco et al., 2015). For pastureland, 33% of farmers who had displaced pasture due to the expansion either purchased or rented land to retain pastureland on the farm.

A farmer who continued to undertake displaced agricultural activities, following the expansion of sugarcane for ethanol production, could replace another land use on their farm or purchase new land and replace the old land use on that new land, resulting in spatial indirect land use change. ILUC captured here did not include land use changes that were due to crop rotations, and displaced agricultural activities that were placed on new lands were only counted as ILUC if the agricultural activity replaced a different agricultural activity on that land. For displaced pastureland, only 2.4% of the farmers converted lands on their farm or new land from a land use to pasture that was not used prior as pasture. For cropland planted to soybeans, 21% of the farmers converted lands on their farm or new land from a prior land use to soybean production. For cropland planted to other crops, 17 % of the farmers made a decision that resulted in ILUC. Most of the ILUC occurring at the farm level was for crop production instead of pasture and resulted in pastureland often being converted to soybean production. About 50% of pastureland in the Cerrado is considered degraded due to poor management and productivity and has been targeted as potential land for crop production under the Sugarcane Agroecological Zoning (Bustamante et al., 2012; Granco et al., 2015). Pastureland that is being converted for crop production may result in more intensified production on the land. While this change in land use provides increased agricultural production to meet food demand (a potentially more efficient use of the land), it may lead to environmental and biodiversity consequences. Barretto et al. (2013) found that agricultural intensification in agricultural frontier areas, such as in parts of the Cerrado, resulted in additional agricultural expansion. Increased use of inputs for crop production may result in additional runoff of nutrients and sediment into water bodies. In addition, more intensified production may negate part of the reduction in greenhouse gas emissions and increases in carbon sequestration from sugarcane and ethanol production (Fargione et al., 2008; Sawyer, 2008). Carvalho et al. (2009) found that landscapes that have a greater area planted to crops reduce habitat for threatened mammal species in the Cerrado. These findings help further substantiate evidence of ILUC as a result of displaced pastureland occurring much farther from the area of sugarcane expansion (Andrade de Sá et al., 2013; Barretto et al., 2013). This may be a consequence of problems in livestock markets and degraded pastureland being better suited for crop production as previously mentioned.

4.2. Farm level modeling and ILUC

Estimation results for the logistic regression model examining factors influencing ILUC at the farm level are provided in Table 3. Overall the fit of the model is good with a McFadden Pseudo R² of 0.54. For the farms that had exhibited ILUC on their land in the study region, being a risk avoider; purchase of crop insurance; having a college education; and a higher percentage of household income that came from the farm significantly impacted the likelihood of having ILUC on the farm due to an expansion of sugarcane production on-farm.

Table 3. Logistic regression results for assessing farm, socio-economic and policy factors impacting indirect land use change farmer decisions.

Variable	Coefficient Estimate (Standard Error)	Marginal Effect^a (Standard Error)
Intercept	−30.03**(13.75)	–
Percent Rented	0.032(0.024)	0.0022Q(0.0017)
Farm Size	0.00061(0.00045)	0.000042(0.000031)
Risk Avoider	3.68(2.36)	0.18***(0.064)
Insurance	4.19**(2.14)	0.23***(0.064)
Sugarcane Sales	−5.95(4.10)	−0.41(0.29)
College	6.13**(2.89)	0.31***(0.058)
Zoning	2.11(1.76)	0.14(0.096)
Income from Farm	0.22*(0.11)	0.015***(0.0079)
Distance to Mill	−0.079(0.068)	−0.0055(0.0048)
Fit Statistics
Log likelihood		−9.93
McFadden Pseudo R^2		0.54
AIC		39.9
Number of Observations		47

^aMarginal effects are calculated at the means of the explanatory variables. Asymptotic standard errors are estimated using the delta method (Greene, 2012).

*10% level of statistical significance, **5% level of statistical significance, ***1% level of statistical significance

Risk played a role in potentially influencing ILUC on the farms studied. Being a risk avoider, which indicated people who are very cautious or try to avoid most risks, increased the likelihood of ILUC occurring on their farm by 18%. Highly risk averse farmers found new areas for displaced agricultural activities that provided much higher income with certainty than other or replaced agricultural activities. This increased the likelihood that the farmer engaged in ILUC. Most of the farmers that exhibited ILUC replaced pastureland with displaced soybean production, as this cropping enterprise provides higher returns with a more certain market in that region. Furthermore, farmers who had purchased crop insurance were 23% more likely to engage in ILUC. Even if soybean production was moved to less productive land to be able to optimize sugarcane production, crop insurance provides a way to reduce the potential yield and market risk of soybean production on less productive and more variable lands. Thus, at the farm level, ILUC due to sugarcane expansion could potentially result in more intensified cash crop production on some less productive lands, increasing the expansion of mechanized intensive crop production in the region.

If a farmer had a college education, it increased the likelihood of ILUC on the farm by 31%. More educated farmers may have the resources to invest in making their farming operation more productive, improving farm performance. Studies have found linkages between improvements in human capital and improved farm performance (Hansen & Greve, 2015; Kilpatrick, 2000; Xayavong, Kingwell, & Islam, 2016). In this regard, the farmer may recognize that displaced agricultural activities that provide a higher profit margin than other land uses on the farm will maximize profit, use resources more efficiently and provide needed diversification as a risk management strategy. This may result in further agricultural intensification of production on existing or new land (e.g. moving from pasture to soybean production) and potential adverse effects, such as degradation and loss of native vegetation and changes in biodiversity (Allan et al., 2015). Farmers with a higher level of household income derived from the farm were 1.5% more likely to engage in ILUC for each 1% increase in household income coming from the farm. Thus, farmers that rely on the farm as their primary source of income will likely keep enterprises on the farm that provide the highest income generation potential to support their household. This result indicates that ILUC is more likely to occur on farm households that are highly dependent on the farm for their standard of living.

Of interest is that the percent of gross farm sales coming from sugarcane production and the proximity to a sugarcane mill did not significantly impact the likelihood of ILUC on-farm. Granco (2017) finds that area planted to sugarcane increases as distance to the sugarcane mill increases. While the sign of the coefficient on the distance to the mill confirms this result, the marginal effect is not statistically significant. Harvested sugarcane must be processed within 72 h to avoid substantial sugar loss, limiting that the distance sugarcane can be transported (Neves, Waak & Marino, 1998). The mean distance to a mill of surveyed farms was 22 km, which places most of the farms within the needed radius to ensure no or very limited quality loss, potentially resulting in a lack of statistical significance. A potential reason for the percent of gross farm sales from sugarcane production not having a significant impact is that sugarcane production placed on degraded or less productive land may have resulted in that land not being replaced due to its low productivity. While increases in sugarcane production displaced agricultural production, the production of corn and soybeans in the region was increasing, as well (Granco et al., 2015). Thus, dynamics in the soybean market are likely to have played a greater role in influencing ILUC at the farm level, being placed on more fertile land that was purchased or rented.

5. Conclusions

The purpose of this paper was to examine socioeconomic, policy and farm-level factors influencing indirect land use change (ILUC) at the farm level as a consequence of the expansion in sugarcane production for ethanol in the Brazilian Cerrado. The paper provides a descriptive and logistic regression analysis of ILUC due to the expansion of sugarcane production at the farm level in the Brazilian states of Goiás (GO) and Mato Grosso do Sul (MS), using a rich data from a set of face-to-face enumerated surveys with commercial farms in the region. The study conducted provides a novel contribution to the literature by examining ILUC at the farm level due to increased biofuels production and examining socioeconomic, policy and farm factors that may influence ILUC.

Results show that ILUC does occur at the farm level as a result of the expansion of sugarcane production within the study region. Many studies focus on ILUC at a global or regional scale from biofuels production and the resulting ILUC far from the originally displaced agricultural activity (e.g. Andrade de Sá et al., 2013). This study provides evidence that while the cascading effect from the displacement of agricultural activities due to expanded biofuel feedstock production may result in ILUC far from the area of original displacement, ILUC can still occur in the immediate area where agricultural activities are being displaced. This seems to be the case with sugarcane expansion in GO and MS. Barretto et al. (2013) find that expansion of crop production usually occurs within close proximity to existing crop production in the Cerrado, providing support for ILUC occurring at the local level. Much of the observed ILUC took place when soybean production was displaced. Many of the farmers surveyed who had ILUC on their farm converted pastureland to soybean production that was displaced due to the sugarcane expansion. Thus, policymakers who are examining ILUC as a result of increased demand for ethanol and agricultural biofuel feedstocks should not only focus on the effects of ILUC, such as deforestation, far from the originally displaced activities, but should recognize that ILUC may occur in the immediate area from displaced agricultural activities. In addition, this may result in more intensified production in the local area, which is what seemed to have occurred in the states of GO and MS from 2005 to 2013 (Granco et al., 2015). As previously mentioned, this may have adverse environmental impacts and threaten wildlife habitats and biodiversity in the Cerrado biome. Furthermore, it may reduce the intended environmental benefits associated with using biofuels over traditional fossil fuel alternatives (Fargione et al., 2008).

Examining factors that may influence ILUC at the farm level, the analysis of survey data showed that being a risk avoider, use of crop insurance, having a college education and deriving higher levels of household income from the farming operation increased the likelihood of experiencing ILUC on-farm. The farm situation may result in more intensified production on more land to maintain farm livelihood, resulting in ILUC at the farm level. De Souza Ferreira Filho and Horridge (2014) indicate that efforts to promote biofuel expansion in Brazil may hinder enforcement efforts to limit further deforestation in the country. While much of this effort has focused on the Amazon, the Cerrado is Brazil’s second largest biome with a rich biodiversity. Agricultural expansion and intensification in the Cerrado due to increased demands and policy for greater ethanol production have significantly increased in the recent past, putting further environmental pressure on this biome (Granco et al., 2015). Further intensification of agriculture, partially through ILUC, may result in further agricultural expansion (Barretto et al., 2013), potentially impacting ecosystem service provision and putting local biodiversity at further risk (Carvalho et al., 2009). The intensification of agricultural practices may make it potentially difficult to keep lands that need to be protected out of production. In addition, the intensification of agricultural production as a result of ILUC will likely lessen any benefits of reductions in greenhouse gas emissions and carbon sequestration from promoting sugarcane-based biofuel production (Fargione et al., 2008).

The findings from this paper should provide insight into potential local ILUC in other similar situations. In India, 39% of the sugarcane produced is used to produce ethanol. In addition, sugarcane production has been expanding in other countries of the world, including China, Australia and South Africa. Increased demand in these other countries has been primarily driven by increased sugar demand, but similar pressures as described in this paper may still result at a local level with potential adverse consequences (Lisboa, Butterbach-Bahl, Mauder, & Kiese, 2011). Furthermore, increases in the demand for biofuels for other feedstocks, such as corn and cellulosic sources, will have impacts on local and global markets. These will have direct impacts on land use at the local level. For example, rapid increases in corn and soybean prices, partially due to increased demand for biofuel feedstock production, has resulted in significant conversion of grasslands and wetlands in the Western Corn Belt of the United States, threatening local wildlife and increasing potential for further environmental degradation (Wright & Wimberly, 2013). While direct land use change is likely to result, local ILUC is also likely to occur, potentially exacerbating adverse consequences from increased pressure for intensified and expanded agricultural production across the landscape.

While we provide a first attempt at examining ILUC at the local or farm level, future research can take this much farther. A more in-depth analysis of ILUC at the local level can be obtained by revisiting and expanding the number of properties surveyed to provide a validation dataset that can be coupled with satellite data and land use maps to conduct an ILUC on-farm and neighborhood analysis over time. Furthermore, additional socioeconomic, policy, market and geographic factors could be brought in to the analysis of ILUC by adding a temporal dimension onto the cross-sectional analysis conducted in this study.

Acknowledgement

This work was supported by the National Science Foundation [NSF BCS-1227451 and NSF SMA-1359082]. Any opinions, findings and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the National Science Foundation [NSF BCS-1227451 and NSF SMA-1359082].

Notes

1. See Granco et al. (2015) for a detailed analysis of the expansion of sugarcane production in Brazil.

2. The survey sample was obtained from farmer contacts provided by sugarcane growers associations, rural syndicates, the Goiás and the Mato Grosso do Sul Federation of Agriculture and Livestock (FAEG and FAMASUL). In addition, additional farm contacts were made when provided by farmers of these organizations who were survey respondents. Given time constraints, limited manpower, travel distances and geographic considerations, all farmers on a contact list could not be contacted. Thus, a response rate for the survey could not be calculated.

3. The 2006 Agricultural Census is the most current census.

4. There were 142 usable responses for calculating descriptive statistics in Table 1, but only 91 of these produced sugarcane and 47 increased their planted area to sugarcane since 2010. Thus, giving the different sample sizes used to calculate the descriptive statistics in Table 1.

5. The year 2010 was used in the survey to limit biases from memory recall farther back than 5 years.