Soybeans, Poverty and Inequality in the Brazilian Amazon - with Diana Weinhold & Eustáquio J. Reis

World Development Vol. 52, pp. 132–143, 2013� 2013 Elsevier Ltd. All rights reserved.

0305-750X/$ - see front matter

www.elsevier.com/locate/worlddevhttp://dx.doi.org/10.1016/j.worlddev.2012.11.016

Soybeans, Poverty and Inequality in the Brazilian Amazon

DIANA WEINHOLDLondon School of Economics, UK

EVAN KILLICKUniversity of Sussex, UK

and

EUSTAQUIO J. REIS *

IPEA, Rio de Janeiro, Brazil

Summary. — Using a mixed-method approach of econometric and ethnographic field research, this article examines the social and eco-nomic costs and benefits of increases in soybean production in the Brazilian Amazon. Statistical analysis suggests that increased soyproduction both reduces poverty indicators and raises median rural incomes, but is also associated with increased measures of inequality.Qualitative fieldwork sheds light on the wider political and social causes and consequences of this increasing inequality, pointing to eth-nic differences between large soy farmers and local populations that raise long standing prejudices and tensions.� 2013 Elsevier Ltd. All rights reserved.

Key words — agriculture, poverty, soybeans, Brazil, Amazonia

Final revision accepted: November 27, 2012.

1. INTRODUCTION

Since the 1960s, Brazil’s national and regional governmentshave invested large sums in the development of soybean pro-duction both directly, through research and agriculturalgrants, and indirectly, through the large-scale infrastructuralprojects needed to make the enterprise profitable. Soybeanproducts are currently one of Brazil’s most valuable agricul-tural exports, and a key component of the nation’s continuedeconomic development. With continuing investment in majorinfrastructure projects, particularly in the Amazon region,their importance shows little sign of diminishing.

Observers, however, remain divided about the social andenvironmental effects of this development. On one side manyresearchers fear that the large scale of this plantation crop willboth increase deforestation and displace small farmers,increasing inequality and poverty (Carvalho, 1999; Fearnside,2001). Other researchers, however, argue that most new soy-bean production takes place on land converted from pasture,rather than from forest, and that the growth of supportingenterprises that accompany large soybean production createsjobs and reduces poverty (Brandao, de Rezende, & da CostaMarques, 2005). While all interlocutors are aware that anydecisions concerning Brazil’s economic and developmental fu-ture and its land use involve trade-offs between economic, so-cial, political, and environmental outcomes, the debate at alllevels of society can be extremely impassioned. These argu-ments are often played out in public forums both in the areasmost directly affected by soybean expansion as well as at thegovernmental level and in the national and international med-ia, and this domestic political debate has had real conse-quences in the pattern of Brazilian agricultural expansion.Thus an understanding of the economic consequences of par-ticular agricultural land use patterns is incomplete without acomplementary understanding of how that land use patternimpacts the political discourse.

132

This article contributes to this debate by focussing on the so-cio-economic impacts of soy cultivation in the Amazon. Whilework has been done on the rural poverty effects of the moregeneral increase in Brazil’s agricultural production (e.g., daSilva, Sergio Gomez, & Rodrigo Castaneda, 2009 1) neitherside of this debate has been able to offer much quantitativeempirical evidence specifically related to soy expansion. Somelocalized studies have been done on the environmental impactsof this industry (see Fearnside, 2001; Kaimowitz & Smith,2001), but few recent studies have presented detailed evalua-tions of the claims of increased local inequality and povertyrates. As Fearnside has noted, in order for countries like Brazilto take informed decisions regarding soybean expansion,‘‘what is needed is an honest weighing of costs and benefitsof expanding soybean cultivation, including all social andenvironmental costs” (2001, p. 35).

This article thus presents new empirical evidence on the rela-tionship between soy expansion and poverty. However, weseek not only to measure the impact of the introduction ofsoy production on socio-economic outcomes, but also to bet-ter understand the political impact of this new developmentand the sources of some of the bitter local debates that sur-round the issue. In the process we attempt to reconcile the re-sults of the statistical work with the political debate on theground; to the extent that we find robust evidence that soyproduction does not increase poverty (and strongly suggestiveevidence that it reduces it), how can we explain the continuingstrong local political resistance to soy cultivation on thegrounds of its socio-economic impact? Thus we combine aneconometric analysis of the data with qualitative fieldworkto identify how the economic impacts of soy link to the contin-ued political debates over its impact.

http://dx.doi.org/10.1016/j.worlddev.2012.11.016

SOYBEANS, POVERTY AND INEQUALITY IN THE BRAZILIAN AMAZON 133

2. BACKGROUND

Soy was first grown in significant quantities in Brazil in the1940s and 50s slowly becoming commercially important in thesouthern states of Rio Grande do Sul, Santa Catarina andParana (Brown, Koeppe, Coles, & Price, 2005, p. 462). Grad-ually the Brazilian government, through The Brazilian Agri-cultural Research Corporation (EMBRAPA), wasinstrumental in developing strains of soy that could be grownin other, more northerly, regions of Brazil: first in the cerradoand then in forested areas 2 (Andersen, Granger, Reis, Wein-hold, & Wunder, 2002, p. 78; Brown, 2004, p. 159; Mueller,2003, p. 14). Production then spread steadily north into theLegal Amazon region, particularly in the northern areas ofMato Grosso (Fearnside, 2001; Mueller, 2003) where rela-tively low land prices, soil quality, mechanization-friendlytopography, and gradually improving (though often still veryprecarious) transportation infrastructure made productioncompetitive in international markets (Vera-Diaz, del, Kauf-mann, Nepstad, & Schlesinger, 2008). As this process acceler-ated over the past 10 years, the rapid growth was furtherspurred by the increasing price of soy in both local and inter-national markets (de Souza, De Oliveira, & De Barros Pinto,2010). Figure 1 illustrates the growth of soy in Brazil and inthe Legal Amazon from 1973 to 2010.

As Warnken argues, however, the Brazilian soybean indus-try has been pulled not only by the rapid expansion in worlddemand for soybean products, but also pushed by public pol-icy (Warnken, 1999, p. 3). In the past the crop played animportant role on the national stage in helping Brazil to in-crease export earnings for development purposes. AlthoughBrazil has now evolved economically beyond the days of wor-rying much about the IMF and international debt payments,some commentators still see the crop in these terms. For exam-ple, Steward (2007) notes that the Brazilian government still‘‘supports soy production to generate revenue for payingdown its debt and efficiently utilizing its land area” and that‘‘soybean farmers believe they are national heroes” (Steward,2007, p. 111).

Soybean production, like agriculture generally, requires ac-cess to credit, so it is not surprising that a number of nationaland international banks have become involved. 3 For exampleBNDES (The Brazilian Development Bank) started a programin 2000 that significantly increased credit for agriculturalmachinery. While recent work has emphasized that soy canbe produced profitably on smaller sized ‘‘family farms”, themajority of soy is grown on larger farms (Bickel & Dros,

Figure 1. Soybean a

2003); Zanon, Saes, Corrar, & Macedo, 2010, p. 10 have notedhow large agricultural enterprises such as the Maggi group inMato Grosso have been financed by international banks(2003, p. 19). Multinationals, and particularly Cargill andBunge in Brazil, also provide soy growers with credit, usuallyin the form of seed, fertilizer, and chemicals in return for thesoy harvested (Greenpeace, 2006, p. 2).

Most of this political and economic support for the soybeanindustry has been based on its relatively clear profitability andgrowth potential. Meanwhile, the impact of soy production onthe structure of local economies has received less attention (cf.Warnken, 1999, p. 6). The need for such work is particularlyimportant given the numerous criticisms of the soybean indus-try made by many researchers, activists, and local populations.The main emphasis here has been on the argument that thelarge scale of this plantation displaces small farmers, increas-ing inequality and poverty and, in turn, pushes such individu-als to claim new land, often through deforestation (Carvalho,1999; Fearnside, 2001, p. 27).

The basic premise of this argument is that large-scale mech-anized agriculture uses less labor per hectare than small-scalefarming techniques. 4 Diegues (1992) describes the effect of themove from coffee farming to soybean cultivation in the south-ern Brazilian states in the 1970s that was precipitated by anumber of problems including ‘‘soil depletion, plant diseases,over-production and lower producer prices” as well as a sys-tematic government-sponsored coffee eradication program:

Between 1970 and 1980 in Parana, the number of farms smaller than 50hectares fell by 109,000 units with a combined loss of 890,000 hectaresin this category. In contrast, farms larger than 1,000 hectares increasedby 450 units with a gain of more than 1,000,000 hectares. . . Many agri-cultural laborers, especially sharecroppers and other tenants, lost theironly source of income. As a result, net migration from the rural areasof Parana reached 2.5 million during the 1970s compared with a netgain of 170,000 in the previous decade (Diegues, 1992, p. 12).

Kaimowitz and Smith note that during the same period RioGrande do Sul lost some 300,000 farms (2001, p. 202). Theyfurther argue that while ‘‘the majority of migrants moved tourban areas. . . a significant number went to the Amazonand cleared forest to grow crops” (2001, p. 202). 5

However others have argued that this kind of reasoningmisses the dynamic argument that, in the long run, large scalemechanized agriculture increases overall productivity andavoids vicious cycles of rural poverty and population growthassociated with low productivity smallholdings and deforesta-tion. In addition, wealth generated by large-scale soybean

rea, 1973–2010.

134 WORLD DEVELOPMENT

farms adds not only to both regional economies and govern-ment revenues but also to the growth of supporting enterprisesthat, in turn, creates jobs and reduces poverty all over the na-tion (Brandao et al., 2005; Mueller, 2003; Mueller & Busta-mante, 2002). It can also be argued that the associatedincreased investment in infrastructure also contributes to gen-eral, local welfare. It has also been suggested that the expan-sion of agriculture, the use of modern technology and theresulting productivity gains, result in the diversification of lo-cal economies (Mueller, 2003, p. 7).

The important point for the current article is that althoughthere is a widespread belief among the public and NGOs aboutthe negative effects of soy, empirical analysis is lacking on bothsides of the debate. Here we make use of a large panel data seton agricultural and economic variables to address the questionof the socio-economic impact of soy. We further complementand augment our econometric analysis with ethnographicfieldwork in a region which has recently undergone a large in-crease in soy production, interviewing local residents, policymakers, and other relevant actors about their views on the im-pact of soy cultivation. These qualitative findings shed light onwhy and how changing local land use patterns shape percep-tions and drive significant political debates in ways that mayseem at odds with the statistical results.

We believe that this dual quantitative–qualitative approachhas many advantages over one-sided methodological alterna-tives for addressing such a politically charged social issue. Inthe first instance, the econometric approach taken here is thatof looking for a general, reduced form relationship betweensoy expansion and changes in inequality and poverty; theunderlying theoretical mechanism is left, for the most part,as a ‘‘black box.” The qualitative fieldwork, on the other hand,produces a more multifaceted and nuanced investigation with-in a single region, but qualitative narratives of residents maybe distorted for various reasons and the external validity ofthe conclusions from a single site may be limited. Brought to-gether, the two approaches can partially redress some of theothers’ weaknesses.

Furthermore, in order to have the maximum policy rele-vance, ideally we would like to help shed light on the wholephenomenon of soy production and economic outcomes,which includes explaining the polarization evident in the de-bate itself. In other words, we want to find an encompassingexplanation; one that explains not only how the relationshipin question operates, but also why local and internationalobservers come to the conclusions that soy is either beneficialor harmful. Statistical analysis alone is unlikely to be able toaddress this latter question, yet for policy makers this maybe crucial.

This article thus proceeds as follows: in Section 2 we de-scribe the data set and outline the quantitative estimationstrategy. Section 3 discusses the empirical results and considersa number of possible limitations to the analysis. Section 4introduces the qualitative case study and summarizes the find-ings, and Section 5 discusses the interactive conclusions fromthe complementary approaches and concludes with sugges-tions for future research.

3. EMPIRICAL STRATEGY

(a) Data

The data available for the current study were extracted froma large database on economic, demographic, ecological, andagricultural variables maintained by IPEA (IPEADATA).

The primary source is the Brazilian Institute of Geographyand Statistics (Instituto Brasileiro de Geografia e Estatıstica,IBGE) Agricultural Census that includes data on agriculturalland use, output, and other economic indicators collected in1970, 1975, 1980, 1985, 1995, and 2005/6. These data are sup-plemented with the IBGE Demographic Census that includesdata on socioeconomic variables such as income, poverty,and inequality for 1980, 1991, and 2000. Poverty is measuredas the proportion of the population below the poverty line, in-comes are measured both as GDP per capita and as medianhousehold income, and inequality is measured using the Theilindex, a weighted average of inequality within subgroups, plusinequality among those subgroups. Finally we make use of theIBGE Municipal Agricultural Survey (Pesquisa Agrıcola Mu-nicipal, PAM) data on crop production and acreage.

IBGE data are published at the municipality (municipio)level, but the number of total municipios, and their bound-aries, changes periodically. Thus in order to be able to com-pare the data over time, municipios are consolidated intoMinimum Comparable Areas (MCAs) to make them consis-tently comparable among Censuses. In the rest of the paperthe term municipality refers to these MCAs, which are our spa-tial unit of analysis. In many cases an MCA is equivalent to acurrent municipio, but in other cases there is no direct map-ping to the current political boundaries. 6 For our period ofanalysis the sample size is 253 MCAs in the Legal Amazon.

4. ESTIMATING STRATEGY

As discussed above, we seek to analyze whether there is sup-port in the data for a systematic relationship between soy pro-duction and the local socio-economic outcomes of interest,income, poverty, and inequality in the Brazilian Amazon.While a number of hypotheses about the underlying mecha-nism driving this relationship have been discussed above, mostof these are rather non-specific with respect to the economics ifnot the politics (i.e., ‘‘large scale production drives out smalllandholders”). Here we remain agnostic as to the underlyingmechanism and treat the question in a reduced form fashion;is there a correlation and if so, what is its sign?

We remain quite cautious about using the term causality; thespread of soy production across Brazil is endogenous, followingan ecological and economic pattern that may be correlated withsocio-economic outcomes of interest for reasons other than anydirect causation from soy to incomes. In the absence of a man-made or natural experiment that introduces some truly exoge-nous variation in soy cultivation, we tackle this question theold fashioned way, with a careful choice of functional formand control variables. Nevertheless we argue that there is goodreason to be cautiously optimistic; inasmuch as soy cultivationis a large-scale mechanized industry that does not depend onlow labor costs, it is unlikely that the choice to plant soy de-pends much on the level of wages itself. In addition, as discussedbelow, we attempt to control for possible unobservable omittedvariables (such as strength of property rights or climate) thatmight both attract soy and be related to poverty.

Our dependent variables, GDP, poverty rates, and Theilindices of inequality, both urban and rural, are measured onlyin 1980, 1991, and 2000. However the 10 year gap betweenmeasurements is not as unfortunate as it may at first seem;poverty and inequality tend to be slow moving variables whoseyear-on-year changes likely contain considerable noise (if theywere measured). Furthermore, the spread of soy into the Ama-zon is a relatively recent (post 1991) phenomenon. The advan-tage to us of using the 2006 Agricultural Census data is that,


prior to their release, agricultural census data were availableonly until 1995. The addition of the new data ensures that,assuming constant annual growth rates, we can interpolatevalues for 2000 from the 1995 and 2006 data. In turn, we inter-polate 1991 agricultural variables from the 1995 and 1985 cen-sus. Note that this necessary interpolation could potentiallygenerate econometric problems if the 1995 data are incorrectlymeasured in a way that is correlated with underlying variablesof interest. Due to changes in the time of year the data werecollected it is highly likely that the 1985, 1995, and 2006 datasuffer a number of differences (see Helfand & Brunstein, 2001for an excellent discussion of this), but it remains unknownwhether these are correlated in a way that would do more thanintroduce additional noise into the analysis.

In general we model the change (or growth) in incomes, pov-erty, and inequality, conditioning on initial (1991) level of thedependent variable and the growth and initial levels of bothsoy and other crops, and other control variables. Our basicestimating equation thus takes the form:

gY it ¼ aþ bY it�1 þ d1Sit�1 þ d2DS þ d3gSit þX

k

kkX kit�1 þ eit

where gY is the growth of the outcome variable (poverty,inequality, GDP, either rural or urban), X is a vector of k con-trol variables, S is our measure of soy acreage (in logs), DS isthe dummy variable for soy in 1991, and gS is the growth ofour soy variable. 7

The initial level of the dependent variable, say poverty, cap-tures all variables (even unobservables) that mattered for deter-mining poverty in 1991. Thus if that region just has morepoverty for some unobservable reason, as long as it caused high-er poverty in 1991 as well, this variable should control for it.

The initial level of soy captures the level effect of soy on thesubsequent change in poverty; in other words, do MCAs thathave more soy initially tend to experience increases or de-creases in poverty over the next 9 years? If so, the observedrelationship could be either due to a time series (causal) rela-tionship between soy and poverty (i.e., as those areas withmore soy, the soy activity actively changed poverty levels overthe subsequent period), or a (spurious) cross sectional rela-tionship (i.e., soy initially established itself in areas with great-er poverty, which subsequently had higher reductions inpoverty for other reasons).

In order to partially address the problem of potential spuri-ous cross sectional relationships we introduce a second soy-re-lated control variable, a dummy for whether the MCA hadany soy in 1991. If there are some unobservable variables thatare correlated with both the evolution of poverty and the like-lihood of having soy acreage in 1991 (for example, proximityto certain infrastructure, or particularly well-functioning prop-erty rights), that would otherwise have created a spurious cor-relation between the level of soy variable and our dependentvariable, these will be controlled for by the soy dummy. Theinclusion of the soy Dummy as a control alters the interpreta-tion of the coefficient on the level soy variable; now it is thatamong the MCAs that already had soy in 1991, those areas thathad greater initial proportion of soy experience greater (or les-ser) change in poverty over the subsequent period.

The third soy variable we include is the growth of soy area.The correlation between this third growth variable and thegrowth of the dependent variable should better capture atime-series effect (although this is a cross section regression);specifically the measured correlation captures whether, ceterisparibus, when soy acreage increases faster in a municipality,does poverty also change faster in that municipality? It is pos-sible that this could be driven by a spurious cross sectional

correlation, but since the levels of the variables are alreadycontrolled for, it is a much harder case to make. Thus we havemore confidence that correlation between two growth rates ismore likely to be representative of a within-municipality(MCA) time series relationship between soy acreage and pov-erty.

The set of control variables (in log-levels and or changes orgrowth rates where noted) include total area of the MCA, lat-itude (a measure of climate), total area in agricultural estab-lishments, total area in pasture, total area in other annualcrops (besides soy), total area in permanent crops, urbanand rural population, and state dummy variables.

Finally, we take into account that agricultural and othereconomic activity is not independent across space; municipal-ities near to each other may share a common climate and geog-raphy and markets are linked. As we are modeling growth andchanges, the potential spatial correlation is less than in levels,but we cannot ignore the possibility. Thus where feasible wereport Moran-I statistics for spatial correlation of the regres-sion residuals (using a spatial weighting matrix which weightsthe five closest neighbors according to distance). The Moran-Istatistic ranges from �1 to 1 where 0 is no spatial correlationand 1 is perfect positive spatial correlation. A low p-value (re-ject the null of MI = 0) thus implies the presence of spatialcorrelation.

5. ECONOMETRIC RESULTS

Most, if not all, of the recent concern about Brazilian soyproduction is due to its explosive growth in the Legal Amazonregion, where the potential environmental effects, such as di-rect or indirect deforestation, could be significant. Further-more, there are number of reasons to suspect that the recentsocio-economic impact of soy in the Legal Amazon could bedifferent than in the south of Brazil. In the first case the initialconditions—the climate and soil quality, the protection (orlack thereof) of property rights, the accessibility of newly con-verted and forest land, the degree of existing poverty and ex-tent of small land holders—are all very different. Saturationeffects should be much more limited. Second, the introductionand growth of soy in the Amazon has been very rapid, and theexponential expansion has been fueled by export and traderather than by domestic demand. Indeed, we find that the rela-tionship between soy and socio-economic outcomes is hetero-geneous and that the patterns that prevail across the wholeBrazilian sample (not reported but available upon request)are not replicated in Legal Amazonia. For these reasons andfor the sake of parsimony then, this analysis focuses specifi-cally on the impact of soy in the Amazon region.

Table 1 presents our baseline results for the effect of soy acre-age on the changes in rural and urban poverty from 1991 to 2000for the 253 MCAs (municipalities) in the Legal Amazon. Theprimary variables of interest with their corresponding coeffi-cients and t-statistics are presented in bold face. Interestingly,we observe that our soy dummy is positive and significant forchange in rural poverty. In other words, in the cross section,those municipalities that had at least some soy production in1991 experienced increases in poverty over the subsequent9 years compared to those who had no soy in 1991. We do notknow why this is; but it suggests that either soybean productiondoes in fact increase poverty, or that there may be significantunobservable (omitted) factors that are correlated both withpoverty and with the presence of soy, and thus whose omission(in the absence of the soy dummy) could create an omitted var-iable bias in the correlation between soy and poverty.

Table 1. Soy and poverty, Legal Amazonia

Dep. Var. Change in rural poverty Change in rural poverty Change in urban Poverty1991–2000 1991–2000 1991–2000

Rural poverty rate, 1991 �0.456*** (�4.3) �0.403*** (�3.31)Urban poverty rate, 1991 �0.429*** (�7.38)Log(area) 0.003 (0.508) 0.007 (1.08) �0.008 (�1.44)Latitude 0.006*** (2.01) 0.005 (1.55) 0.004 (1.27)Log(population), rural 1991 �0.001 (�0.0668) 0.003 (0.23) 0.013 (1.44)Log(population), urban 1991 0.001 (0.132) 0.007 (1.03) �0.023*** (�3.81)Dummy for soya in 1991 0.075*** (2.01) 0.077** (2.13) 0.009 (0.248)

Log(area in soya), 1991 �0.018*** (�2.85) �0.017*** (�2.7) �0.003 (�0.476)

Log(area in other temp crops) 1991 �0.005 (�0.87) �0.006 (�1.62)Log(area in perm crops) 1991 �0.009** (�2.06) 0.000 (0.121)Log(area in pasture) 1991 �0.004 (�0.923) �0.006 (�1.22)Log(total establishment area) 1991 0.006 (0.621) 0.020** (2.21)Change in soya area, %, 1991–2000 0.345 (0.753) 0.054 (0.117) �0.366 (�0.679)

Change in other temp area, %, 1991–2000 �0.104 (�1.24) �0.027 (�0.313)Change in perm area, %, 1991–2000 �0.042 (�0.0475) 0.053 (0.0611)Change in pasture area, %, 1991–2000 �0.106 (�1.37) 0.016 (0.265)Change in establishment area, %, 1991–2000 0.027 (0.92) 0.009 (0.273)Nobs 254 253 253R-Squared 0.5120 0.5448 0.4421Moran-I of residuals �0.042

p-value = 0.155�0.031

p-value = 0.236

Please note: all regressions include state dummies (not shown).Robust t-statistics in parentheses.* Significant at 10%.** Significant at 5%.*** Significant at 10%.


However the coefficient on the log-level soy variable is sig-nificant and negative; although, as discussed above, on averagethose municipalities with soy in 1991 had increases in povertycompared to those that did not (the soy dummy), among thosewith soy, those that had larger proportions of crop areaplanted in soy saw poverty fall further. This is not consistentwith a story in which soy itself is causing the increase in pov-erty captured by the soy dummy, reinforcing the interpretationof the soy dummy coefficient as capturing omitted cross sec-tional unobservables. Indeed, doing the sums for each munic-ipality individually, we find that in all but three Amazonianmunicipalities the net total effect of soy on change in povertyis negative. Furthermore, in regression (3) we look at the im-pact of soy on urban poverty rates and find no statistically sig-nificant effect. This gives us additional confidence that inregressions (1) and (2) we are indeed capturing a rural effectof soy, not just a spurious, general trend in regional poverty.

Although the growth of soy variable is not statistically sig-nificant, the interpretation of this result is not entirely straight-forward. The combination of a positive (in magnitude)coefficient and high standard error suggests that there maybe some areas in which growth of soy and increases in povertyare occurring simultaneously, whereas in others the oppositerelationship prevails. Any observer of either type of locationcould come away with a strong impression of a soy–povertylink, even if that relationship does not hold generally.

Regressions (4)–(6) in Table 2 look at soy and inequality inthe Amazon. Consistent with the literature, we find that boththe log-level and growth of soy are correlated with increasedrural inequality, although the former is not statistically signif-icant and the latter only marginally so (at the 10% level). Also,growth of soy is not significantly correlated with urbaninequality.

The results presented in Tables 1 and 2 only consider the so-cio-economic effects of soy cultivation within a particular

municipality. However, Fearnside (2001, p. 27) and othershave claimed that one of the mechanisms through which theintroduction of soy increases poverty is through the displace-ment of small landowners. If true, these (now poorer) internalmigrants might move to a different municipality and not reg-ister as an increase in their home-municipality’s poverty rate.As we do not have data on individual landowners and theirmovements we cannot check this directly. Nevertheless wecan check to see whether increases in soy acreage in neighbor-ing areas have an impact on municipalities that themselveshave no soy. We run this exercise for the whole of Brazil,not just Legal Amazonia, as there are too few observationsfor meaningful analysis in the latter. Specifically, we calculatea weighted average of soy acreage in a municipality’s five clos-est neighbors. We then examine the impact that neighborhoodsoy acreage has on socio-economic outcomes in our sample ofmunicipalities that themselves have no soy. The results of thisexercise are presented in Table 3. In regressions (7), (8), and (9)we see that those soy-less municipalities whose neighbors havegreater acreage of soy have lower rural poverty rates, higherrural inequality, and higher rural median household income;in other words, the same direction of association that with-in-municipality soy had with these indicators. 8 In sum, whilewe cannot rule out longer-range out-migration of the dis-placed, we find no evidence that increased soy in neighboringmunicipalities increases poverty or incomes in the home region(although there is some impact on inequality, as at home).

There are several alternative interpretations of the spatialresults. If we take them at face value, these results suggestthat our coefficients of the effects of soy on local socio-eco-nomic outcomes may be under-estimated; the ‘‘control” setwith no soy benefits from spillover effects and thus providesa poor counterfactual. On the other hand, there could also besome related omitted variables associated with the soy thathave strong and far-ranging spatial socio-economic effects

Table 2. Soy and inequality, Legal Amazonia

Dep. Var. Change in Theil indexof inequality, rural

Change in Theil index ofinequality, rural

Change in Theil index ofinequality, urban

1991–2000 1991–2000 1991–2000

Theil index, rural 1991 �0.872*** (�10.6) �0.906*** (�11)Theil index, urban 1991 �0.87*** (�15.3)Log(area) 0.026** (2.34) 0.0501*** (2.7) 0.04*** (2.95)Latitude 0.003 (0.642) 0.001 (0.113) �0.007 (�1.38)Log(population), rural 1991 �0.042 (�1.26) �0.038 (�1.14) �0.032** (�2.04)Log(population), urban 1991 0.020 (0.785) 0.021 (0.801) 0.034*** (3.84)Dummy for soya in 1991 �0.093 (�0.918) �0.091 (�0.892) �0.066 (�1.01)

Log(area in soya), 1991 0.015 (1.1) 0.016 (1.2) 0.019* (1.83)

Log(area in other temp crops) 1991 �0.005 (�0.332) �0.006 (�0.802)Log(area in perm crops) 1991 0.003 (0.257) 0.009 (1.58)Log(area in pasture) 1991 0.0296** (2.34) 0.002 (0.185)Log(total establishment area) 1991 �0.060*** (�2.76) �0.026 (�1.36)Change in soya area, %, 1991–2000 4.93* (1.83) 5.13* (1.85) �0.863 (�0.9)

Change in other temp area, %, 1991–2000 �0.183 (�0.976) 0.29 (1.24)Change in perm area, %, 1991–2000 �0.597 (�0.26) �3.6 (�1.27)Change in pasture area, %, 1991–2000 0.0719 (0.414) 0.204* (1.8)Change in establishment area, % total, 1991–2000 �0.0321 (�0.513) �0.117* (�1.69)Nobs 254 253 253R-Squared 0.3070 0.3202 0.6013Moran-I of residuals �0.061

p-value = 0.0420.007

p-value = 0.385


Table 3. Spatial spillovers of soy, all Brazil

Dep. Var. Change in rural poverty rate, Change in Theil indexof inequality, rural

Growth of medianhh income, rural

1991–2000 1991–2000 1991–2000

Rural poverty rate 1991 �0.409*** (�14.2)Theil index, rural 1991 �0.882*** (�35.4)Log median hh income, 1991 �0.53*** (�21.7)Extent of Spatial (Neighborhood) Soy, 1991 �0.144** (�2.05) 0.164** (2.53) 0.537*** (3.1)

Log(area) 0.010** (2.1) 0.0171** (2.02) �0.0402** (�2.55)Latitude 0.015*** (10.1) �0.00399* (�1.66) �0.0493*** (�10.3)Log(population), rural 1991 0.012** (2.08) 0.005 (0.718) �0.0383*** (2.91)Log(population), urban 1991 0.000 (0.0245) �0.000803 (0.21) 0.028*** (4.33)Log(area in other temp crops) 1991 �0.007*** (�3.2) �0.000306 (�0.09) 0.0054 (0.873)Log(area in perm crops) 1991 �0.008*** (�7.48) �0.00122 (�0.713) 0.0213*** (6.88)Log(area in pasture) 1991 �0.023*** (�5.15) 0.00646 (1.11) 0.0374*** (3.37)Log(total establishment area) 1991 0.025*** (3.21) �0.0101 (0.852) �0.0262 (�1.18)Change in other temp. crop area, %, 1991–2000 �0.047** (�2.33) �0.00038 (0.012) 0.13** (2.09)Change in perm. crop area, %, 1991–2000 �0.059 (�1.27) 0.0406 (0.561) 0.282** (2.07)Change in pasture area, %, 1991–2000 �0.007 (�0.424) �0.0348 (1.22) �0.102* (�1.92)*

Change in establishment area, %, 1991–2000 0.001 (0.113) 0.00495 (0.362) 0.0327 (1.19)N. Obs 2612 2612 2612R-Squared 0.480 0.517 0.431



themselves, such as enhanced infrastructure, that could bedriving the outcomes. In other words, perhaps it is not soyitself per se, but rather a whole package of infrastructural,institutional, and economic changes that are more causallyrelated to our outcomes of interest. Indeed, we should keep

in mind that this is in fact what many of soy’s critics areclaiming; not that there is something necessarily bad aboutthe crop itself, but rather with the way it is cultivated andall the associated changes that accompany the arrival of‘‘big soy” to an area.

Table 4. Soy and incomes, Legal Amazonia

Dep. Var. Growth of medianhh Inc, rural

Growth of medianhh Inc, urban

Growth of realGDP, rural

Growth of realGDP, urban

1991–2000 1991–2000 1991–2000 1991–2000

Log(median hh income), rural 1991 �0.383*** (�3.81)Log(median hh income),urban 1991 �0.541*** (�8.41)Log(GDP), rural 1991 �1.02*** (�15)Log(GDP), urban 1991 �0.918*** (�7.03)Log(area) �0.007 (�0.216) 0.054** (2.41) �0.072 (�1.27) 0.011 (0.248)Latitude �0.003 �0.015 �0.010 �0.009

(�0.252) (�1.54) (�0.419) (�0.444)Log(population), rural 1991 �0.046 (�0.824) �0.038 (�1.24) 0.496*** (5.66) 0.227*** (3.06)Log(population), urban 1991 0.020 (0.587) 0.093*** (5.23) 0.117* (1.93) 0.812*** (6.04)Dummy for soya in 1991 �0.239 (�1.58) �0.036 (�0.319) �0.544** (�2.31) �0.574*** (�2.68)

Log(area in soya), 1991 0.043* (1.82) 0.024 (1.29) 0.12*** (2.93) 0.111*** (2.99)

Log(area in other temp crops) 1991 �0.006 (�0.264) �0.003 (�0.178) 0.002 (0.0702) �0.089*** (�3.08)Log(area in perm crops) 1991 0.032* (1.69) 0.018 (1.6) 0.042* (1.72) 0.041** (1.98)Log(area in pasture) 1991 0.035* (1.67) 0.015 (0.822) 0.128***2.14(3.2) �0.039 (�0.819)Log(total establishment area) 1991 �0.035 (�0.799) �0.077** (�2.14) 0.3*** (3.58) 0.0291 (0.352)Change in soya area, % total 1991–2000 6.11** (2.01) 1.24 (0.885) 8.18** (2.38) 3.99** (1.99)

Change in other temp area, % total 1991–2000 0.26 (0.646) 0.48 (1.23) 0.844 (1.17) �0.523 (�1.01)Change in perm area, % total, 1991–2000 2.98 (0.826) �2.68 (�0.745) �1.67 (�0.261) 18*** (2.91)Change in pasture area, % total, 1991–2000 0.205 (0.693) 0.236 (1.32) 0.223 (0.402) �0.249 (�0.896)Change in establishment area, % total, 1991–2000 �0.049 (�0.384) �0.138 (�1.22) 0.072 (0.284) 0.12 (0.743)Growth of urban population, 1991–2000 0.09 (0.618) 0.842*** (5.99)Growth of rural population, 1991–2000 0.191 (1.53) 0.132* (1.65)Nobs 253 253 253 253R-Squared 0.4598 0.4853 0.8822 0.8479Moran-I of residuals �0.040

p-value = 0.167�0.058

p-value = 0.0760.161

p-value = 0.000.038

p-value = 0.124



Finally, in Table 4 we look at the association between soyproduction and incomes in Legal Amazonia. In regressions(10) and (11) we find that both levels and changes of soy arepositively and significantly correlated with increases in medianrural household income, but not with median urban householdincome. Both soy variables are also strongly positive and sta-tistically significant for both rural and urban total GDP (percapita, as we are controlling for rural and urban population)in regressions (12) and (13).

As mentioned above, for all regressions except those inTable 3 we additionally check for the presence of spatial

Table 5. Soy and incomes, Lega

Dep. Var. Change in rural poverty

1991–2000

Spatial lag of dep. variable �0.272** (�2.34)Dummy for soya in 1991 0.070 (2.07**)Log(area in soya), 1991 �0.017 (�2.88***)Change in soya area, % total 1991–2000 0.312 (0.71)Nobs 253R-Squared 0.5566Moran-I of residuals 0.014

p-value = 0.310

Please note: all regressions include state dummies + other regressors from TaRobust t-statistics in parentheses.* Significant at 10%.** Significant at 5%.*** Significant at 10%.

correlation in the residuals. We find no evidence of spatialcorrelation in the models of rural and urban poverty change(regressions 2–3), urban inequality change (regression 6), ormedian rural income growth (regression 10), but weak evi-dence for rural inequality change (regression 5), median ur-ban income growth (regression 11), and urban GDPgrowth (regression 13). Finally, we find strong evidence ofspatial correlation in the model of rural GDP growth (regres-sion 12). In Tables 5 and 6 we repeat the main regressions,including a spatially lagged dependent variable. The resultsfor the spatial lag and key soy-related explanatory variables

l Amazonia, with spatial lag

Change in urbanpoverty

Change in Theil indexof inequality, rural

Change inTheil index of

1991–2000 1991–2000 1991–2000

�0.351*** (�3.1) �0.295* (�1.82) .0996 (1.18)0.0167 (0.491) �0.0931 (�0.969) �0.0666 (�1.02)�0.0032 (�0.559) 0.0184 (1.45) 0.0187* (1.82)�0.464 (�0.941) 5.11** (2.04) �0.785 (�0.805)

2553 253 2530.4621 0.3346 0.60350.045

p-value = 0.093�0.002

p-value = 0.474�0.023

p-value = 0.301

bles 4 and 5 (not shown).

Table 7. Soy and incomes in the Amazon, summary of main quantitative results

Regressor: Dep. Var.: Soy dummy Level area of soy Growth of soy area Spatial correlation?

Change rural poverty Positive Negative No effect NoChange urban poverty No effect Negative No effect NoChange rural inequality No effect No effect Positive MaybeChange urban inequality No effect Positive No effect NoGrowth rural median hh Inc. No effect Positive Positive NoGrowth urban median hh Inc. No effect No effect No effect MaybeGrowth rural GDP Negative Positive Positive YesGrowth urban GDP Negative Positive Positive Maybe

Table 6. Soy and Incomes, Legal Amazonia, with spatial lag (cont.)

Dep. Var. Growth of medianhh Inc, rural

Growth of medianhh Inc, urban

Growth of realGDP, rural

Growth ofreal GDP, urban

1991–2000 1991–2000 1991–2000 1991–2000

Spatial lag of dep. variable �0.307* (�1.82) �0.151 (�1.47) 0.124 (1.41) �0.0736 (�0.779)Dummy for soya in 1991 �0.214 (�1.56) �0.0288 (�0.253) �0.548** (�2.32) �0.622** (�2.43)Log(area in soya), 1991 0.0444** (2.15) 0.0231 (1.23) 0.12*** (2.94) 0.118*** (2.7)Change in soya area, % total 1991–2000 5.29* (1.92) 1.25 (0.892) 7.97** (2.31) 4.04* (1.95)Nobs 253 253 253 253R-Squared 0.4717 0.4897 0.4786 0.4178Moran-I of residuals 0.016

p-value = 0.293�0.017

p-value = 0.3610.135

p-value = 0.0000.046

p-value = 0.087

Please note: all regressions include state dummies plus other regressors from Table 6 (not shown).Robust t-statistics in parentheses.* Significant at 10%.** Significant at 5%.*** Significant at 10%.


are found in regressions (14)–(21); we find that the basic re-sults on the soy variables are not changed. Looking at theMoran-I statistics for the residuals, we find that includingthe spatial lag eliminates the spatial correlation (of the 5-closest, which is what we tested for) for the change in ruralinequality (regression 16). However the other weak effects(regressions 19 and 21) remain, and there is still strong evi-dence of spatial correlation in the model of growth of ruralGDP (regression 20). In this latter model, the evidence sug-gests that spatial correlation is true residual correlation; thespatial lag of the dependent variable carries almost noexplanatory power or statistical significance. With spatialcorrelation in the residuals, the standard errors in model 12may be biased (and the coefficients could be biased as wellin model 20 as we have included a lagged spatial dependentvariable in the presence of spatial correlation of the residu-als), so some caution exercised in interpreting these results.Nevertheless, especially for the other models, we find it reas-suring that the main results on the soy variables are robust tothe inclusion of the spatial terms.

Table 7 provides a concise summary of the primary quanti-tative results. Controlling for spurious cross sectional correla-tion between soy location and incomes, we find no robustevidence that increased soy cultivation increases poverty with-in an MCA; to the contrary our evidence suggests that if any-thing poverty declines. We do find increased soy cultivation isassociated with increased rural income inequality, but the rela-tionship is only marginally statistically significant. Finally,even controlling for total agricultural area and crop composi-tion, increases in soy cultivation are associated with strong in-creases in rural median household income and rural GDP percapita, indicating improvements in both the welfare and pro-ductivity of rural populations. In fact we find both level andgrowth effects for rural median household income and GDP,and for urban GDP as well.

6. QUALITATIVE FIELDWORK

Given the relatively robust econometric finding of signifi-cant economic growth effects of soy, the existence of relativelystrong local political resistance (on economic grounds, as op-posed to environmental grounds which we do not address inthis article) in some areas raises some interesting questions.The statistical work identifies overall average relationships;in the case of poverty perhaps the effects are heterogeneous.Alternatively there could be unobservable omitted variablebias or measurement error. Or it could be that the econometricresults are correct but the political debate is fueled by otherconsiderations. Either way, from both an econometric as wellas a policy perspective, it would be useful to better understandthe source of the beliefs and political debates surrounding thesoy boom in the Amazon. To that end, in parallel with ourquantitative research, we conducted qualitative fieldwork inareas of soy production in the Brazilian Amazon.

The qualitative part of this research began with visits tovarious parts of the states of Mato Grosso and Para to ob-serve and discuss soy production and the issues that sur-round it. This initial survey emphasized the differences thatexist across the region in terms of the history and currentreality of soy production, from the transition areas whereforest is being transformed into agricultural land, throughvarious forms of small-scale farming to vast, industriallyrun land holdings. The area that usually receives the mostattention, particularly in the national and international med-ia, is the point at which the transition is being made to soyagriculture, either from standing forest (rarely) or, more usu-ally, from other forms of agriculture, particularly pasture-land. In such areas there tends to be a political and socialpolarization between different groups of actors and tensionsoften flare up into violence (Nepstad, Stickler, & Almeida,2006, p. 5; Simmons, 2004).


In contrast to those areas undergoing this early stage of ra-pid transformation stand those areas that have longer estab-lished soy farms, particularly in the central areas of MatoGrosso. Here the farms are measured in thousands of hectaresand focus on industrial and technological efficiency, with landthat was once cerrado (savannah) or rainforest levelled intoendless fields of monoculture. While each farm will have per-manent residents, in general the local population is agglomer-ated into more urbanized spaces. These settled areas, whilereceiving less media attention, have been the subject of severalpolicy and quantitative studies (e.g., Bickel & Dros, 2003;Mueller, 2003; Zanon et al., 2010) and some qualitative work(e.g., Fisher, 2007).

Between these two examples lie those areas where soy hasbeen introduced and the initial conflicts have died down butlarge-scale farming has not come to dominate. In some casesthis may represent a transition stage before larger farms buyup less profitable small farms (Zanon et al., 2010). In othersit may be a more stable equilibrium in which various geolog-ical, climate, social, or political factors limit the extension offarm sizes or the production of soy itself. Given our researchobjective of investigating the qualitative experience of agricul-tural (specifically soy) transformation, although there has been

Figure 2. Field

Figure 3. Area of Cultivated C

relatively less attention given to such areas, this region seemedto provide a unique opportunity to glean insights into theassociated social transformations and interactions that occurafter initial conflicts have died down, but while actors are stillable to discuss the nature of the conflicts themselves. Thusalthough we cannot claim that the qualitative results are nec-essarily comprehensive or representative of other regions inthe Amazon, we hope that research in this region provides in-sight into some of the sources of conflict that have arisen inconjunction with the rise of soy production. To this end, afterthe initial survey, it was decided to focus our research on theSantarem region of Para state (see Figure 2).

As discussed above, the region was chosen for closer analysisbecause it offers something of a microcosm of some of the cen-tral political and economic issues surrounding increased soy-bean cultivation. Specifically a port for the internationalexport of soy was opened by Cargill in 2003. While large-scaleagriculture has been used in the area since the influx of migrantslinked to the construction of the Trans-Amazonian highway inthe 1970s, the period since the port’s opening has seen a large in-crease in the production of both soy and rice (see Figure 3).While the overall production of grains in the area is still dwarfedin comparison to the industrial agricultural output of parts of

work site.

rops, Santarem 1989–2007.


Mato Grosso, nevertheless the increase has been associated withvarious social and political changes in the area.

Having chosen to focus on the Santarem region 8 months offieldwork was conducted over an 18-month period in 2009 and2010 mainly in and around the rural town of Belterra, about50 km outside the city of Santarem. Formal and semi-formalinterviews with individuals related to the soy industry werecomplemented with longer-term participant observation inthe local community. The population under study included re-cent newcomers to the region who were mainly focused onlarge-scale agriculture, and older populations that had a vari-ety of livelihoods, including both small and large-scale agricul-ture and private and state employment, as well as theunemployed and retired. An effort was made to seek out differ-ent members of society according to age, ancestry, and socio-economic status, with interviews focused explicitly on attitudesto the environment, local development, agriculture, and gov-ernment policies. Qualitative information from the interviewsabout the local land issues and development was supple-mented through the use of GPS data on landholdings and,where possible, the discussion of such data with interviewees.

7. ETHNOGRAPHIC FINDINGS

The primary aim of Cargill’s port was to provide an alterna-tive export option for large-scale soy farmers in the state ofMato Grosso. The port’s development was linked to a govern-ment-backed project to pave the BR-163 that runs from Cuiabato Santarem. As this road building project has been stalled forvarious economic and political reasons Cargill has been keento promote the production of soy and rice in the immediate San-tarem region in order to make use of the new facilities.

The incentives offered by Cargill, as well as loans from the re-gional development bank, Banco da Amazonia, local and na-tional political encouragement (Almeida & Santos, 2010) andthe availability of relatively cheap land encouraged farmersfrom southern Brazilian states to move to the area. Fisher(2007) notes that the federal government deliberately targetedimmigrants from southern Brazil in the belief that ‘‘their previ-ous technical and administrative experience would allow themto transform financial resources into ‘physical capital’ of thekind desired by planners” (Fisher, 2007, p. 352). Indeed, soy isa relatively expensive crop to grow, needing a high level ofinvestment in fertilizers, pesticides and machinery—it is gener-ally held that it is not economically viable to grow soybeanson plots of less than 500 hectares (but cf. Zanon et al., 2010).This means that while many local existing residents and land-owners have been interested in, or attempted to, grow soy theyhave seldom found it to be profitable. On the other hand, newlyarrived migrants from the South, with significantly greaterfinancial resources (and/or benefitting from better weather con-ditions), have been more successful at profitably farming soy.

Belterra itself sits on a flat plateau with a climate and soilswell suited to cultivation, 9 a general lack of surface water,however, has meant that certain forms of agriculture, and par-ticularly cattle ranching, are difficult to sustain. With sufficientmonetary investment in deep wells, however, soy and rice canbe grown efficiently. This has meant that since Cargill’s initialinterest in the area there has been a steady transfer of landfrom cattle ranching and small-scale family plots to larger-scale soy. The sellers of land then move either to the larger cit-ies of the region including Santarem but also including Belemand Manaus or to the small local, rural towns of Belterra,Tabocal, or Mojuı dos Campos. While such moves have beenunderpinned by financial reasons, interviews with people who

had sold land, or were still keen to do so, often focused on thequalitative differences between rural and urban living. Specif-ically they emphasized the lack of state resources in ruralareas, particularly in terms of education and healthcare provi-sioning, issues of particular importance for families with youn-ger children. While incoming farmers may recognize theseissues they often have means of circumventing them includingby leaving school age children with relatives in southern Bra-zil, boarding them in local cities or through ownership of reli-able forms of private transport.

For these financial and practical reasons soybean farming inthe Santarem region has come to be dominated by families fromthe southern Brazilian states. Such farming families tend to be ofNorth European heritage and their physical appearance andcultural practices tend to differ from the older local populationswhose ancestry includes elements from indigenous Amazonian,Southern European, and African (via Brazil’s Northeast region)populations. These differences are noted and often commentedon by members of both groups in the region. More than physicaldistinctions, individuals tend to emphasize the cultural differ-ences that exist in the working and social practices of the twogroups. One focus is on the distinctions between homes and liv-ing styles. While local populations tend to have smaller housesand spend more time sitting outside them in their surroundinggrounds, immigrants from the south bring a particular style oflarge house, with surrounding verandas that provide privacyfrom the outside. Members of older local populations portraythemselves as very sociable in comparison to the more recentarrivals from Brazil’s southern states. However people fromthe South tend to portray such sociability as a form of lazinessthat they contrast with their own work ethic. Such tensions obvi-ously link to older histories of colonialism, as well as the relativeseparation of the Amazonian region from the rest of Brazil.They are also linked to the relative wealth of individuals andfamilies coming from the South.

These differences play out in both physical and cultural fea-tures of the two populations, and in the current context the field-work found that these tensions tended to find a particularexpression around the issue of increased soybean cultivation.Even as local residents found employment on these larger farmsor benefited from their custom, the underlying racial and polit-ical tensions continued to feed animosity. These feelings havealso been encouraged by the local media, religious groups,and local and international NGOs. For example, an article ina local paper in Santarem describes confrontations betweenincoming soy farmers and ‘‘local” Santarem populations duringa demonstration organized by Greenpeace against soybeanfarming. It reports one of the farmers as saying ‘‘we who arecoming here are bringing development for you, you are Indians,stupid and lazy” 10 (Gazeta de Santarem—May 24, 2006).

Here the particular context of the Cargill port played animportant role, in that the port itself became a relatively sim-ple, public, and photogenic symbol of the soy industry as awhole. In the first instance there were particular issues sur-rounding the building of the port, specifically the fact that itwas alleged to have been built on a site of archeological impor-tance 11 and without proper consideration of its widerimpacts. 12 These issues were then linked by its opponents tomore general concerns about the expansion of soy farmingin the Amazon. Beyond these issues, however, the port, as aspecific and identifiable target lent itself to the publicity gener-ating tactics of confrontation used by the international envi-ronmental NGO Greenpeace. The region thus became amicrocosm of the debates surrounding soybeans with the mix-ing up of various economic, social, and environmental issues.While the specific nature of the well-publicized events involv-


ing various local, national, and international actors during2003–07 was particular to Santarem, comparison with otherareas suggests that the social and ideological divisions visiblein Santarem parallel those found across the region. By the timeof our fieldwork in 2009 and 2010, the earlier dramatic publicdivisions and manifestations had receded, leaving behind a lessobvious but arguably more fundamental division between dif-ferent groups that we believe is representative of the region asa whole (cf. Fisher, 2007 for Mato Grosso).

Here we return to the fact that even as our quantitative anal-ysis shows the rise of median incomes and local GDP associ-ated with soybean production and, most importantly, alowering of poverty levels, it also supports the view that in-creased soy production increases inequality. This result is con-sistent with our qualitative observations that soybeanproduction in Legal Amazonia is mostly controlled by wealthylandowners on large farms. Thus, even as all levels of localpopulations benefit economically from the growth of soybeanproduction, large landowners accrue, or are perceived to ac-crue, relatively more gains.

From our field observations and deeper discussions with indi-viduals from different groups in the area it appears that it is localperceptions of this growing inequality that fuel much of the longterm opposition to the increase in large-scale soy farming in theAmazon region. We believe that the emphasis on inequality islinked to other social and political factors, and in particularthe fact that most large-scale farmers are immigrants to the re-gion from the south of the nation. In the context of our quanti-tative findings, these tensions help to explain some of thecontinued animosity felt toward soybean cultivation even asits appears to have the effect of lowering poverty levels.

Interestingly, the actions of local, national, and interna-tional groups against the Cargill port, in conjunction with re-search and action in other regions, as well wider argumentsagainst the environmental and perceived economic issues sur-rounding soy agriculture, have led to a reduction of govern-mental and economic support for farmers. In the firstinstance a moratorium on the purchase of soy grown on newlydeforested areas was ratified on July 24, 2006 by various na-tional and international companies and environmental NGOsas well as local representatives. 13 More generally there alsoappears to have been a reduction in governmental supportfor the industry and a concomitant reduction in availablefinancing for farmers (Stickler & Almeida, 2009). During ourfieldwork the effects of this latter change were already beingfelt with many of the smaller local soy producers complainingthat they were no longer able to find loans to support theinvestment in seeds, fertilizers, pesticides, and equipment that

accompanied the start of the planting season. As such in 2010several farmers that we had met the previous year had stoppedlarge-scale agricultural production, deciding instead to lookfor non-agricultural jobs in the public and private sector. Be-yond the economic and personal impacts that these newrestrictions are having on individuals, the positive economicoutcomes our quantitative work has associated with increasedsoybean cultivation suggest there may be even wider negativerepercussions from this new institutional position.

8. DISCUSSION

In this article we have provided both quantitative and qual-itative evidence rejecting the hypothesis that the introductionof soy into the Amazon has increased poverty. On the con-trary, our evidence is suggestive of poverty reducing effects,and strongly indicative of positive income gains. We controlfor possible spurious cross sectional correlation between soyand incomes, and find that increases in soy cultivation areassociated with strong increases in rural median household in-come and rural and urban GDP per capita.

Both the quantitative evidence as well as ethnographic field-work also point to an association between increased soy produc-tion and inequality. While the quantitative work indicates onlyincome inequality, the qualitative analysis provided a muchmore nuanced and informative characterization of the natureof the inequality, pointing to ethnic differences between largesoy farmers and local populations that raise long standing prej-udices and tensions. Eight months of ethnographic fieldwork inone of the earlier ‘‘hotbed” areas of political tensions over soyproduction led us to the qualitative conclusion that a significantportion of the negative publicity and local beliefs about soy heldby a subset of the population can be traced to this perceived (andactual) unequal gains of soy expansion between different ethnicgroups with historical rivalries.

Although this study specifically focussed on the socio-eco-nomic consequences of soy production, the findings could alsobe of importance in light of current discussions about the rolethat various deforestation reduction schemes, including car-bon trading schemes and international payments for environ-mental services, might play in the future protection of theforests. In calculating future payments for forest protection,not only must the services of the standing forest be consideredbut also the lost opportunity costs of using its land for otherpurposes. In such situations clear understandings of the eco-nomic and social benefits and costs of land use in the Amazonwill be of vital importance.

NOTES

1. This FAO study shows that for all of Brazil the boom in agriculturalproduction between 1995 and 2006 can be linked to a 16% reduction inrural poverty (Silva et al., 2009, p. 103).

2. Geographic limits remain on the use of these new strains of soy basedon their continued need for a relatively sustained dry season (Fearnside,2001, p. 34).

3. A recent study by Profeta, Gomes, Lırio, and Costa (2010) shows a clearpositive link between levels of rural credit and Brazilian soybean exports.

4. Fearnside adds that on large-scale farms in Legal Amazonia employ-ment tends to be given to workers from outside the region (usually fromthe same southern states as the farm owners) rather than to localpopulations (Fearnside, 2001, pp. 24–25).

5. Lopez Villar et al. (2007, p. 24) have similar findings in Argentinawhere they argue that the introduction of soy has contributed to theacceleration of land consolidation. ‘‘During the 1990s, the number offarms in the Pampas area decreased from 170,000 to 116,000, while theaverage size of farms doubled.”

6. As the municipio, and hence MCA, are politically defined spatialentities. In regions where population is dense there tend to many MCAsand they tend to be smaller. Where population is sparse, however, such asin some parts of the Amazon, the MCAs are very large.

7. Note that the regression maps out a nonlinear relationship betweensoy and the outcome variable.


8. The spatial impact on urban poverty, median household income andurban inequality was not statistically significant. The effect on rural GDPwas not statistically significant but the effect of greater spatial soy onurban GDP was positive and significant, results available upon request.

9. Evidence exists of relatively settled forms of agriculture from beforecontact with Europe and across the plateau there are many instances ofancient, apparently anthropogenic, terra preta, soils, where centuries ofhuman use have helped to enrich the soils (Bechtold, 1982).

10. ‘‘nos que estamos vindo aqui trazer desenvolvimento pra voces, vocessao ındios, burros e preguic�osos.”

11. Research conducted in conjunction with Cargill’s EnvironmentalImpact Study noted that there was no evidence of archaeological remainsin the area of the port (Cargill, 2010, chap. 7, pp. 25–26).

12. Debates over Cargill’s legal obligations to conduct an EnvironmentalImpact Study were the central cause of delay in the construction andopening of the port.

13. http://www.abiove.com.br/english/ss_moratoria_us.html.

REFERENCES

Almeida, R. H. C., & Santos, M. A. S. D. (2010). AmbivalenciasGovernamentais e a Expansao da Fronteira Agrıcola da Soja: O Casode Santarem, Para. Presentation to Sociedade Brasileira De Economia,Administrac�ao E Sociologia Rural, Congresso 48�. Available at: http://www.sober.org.br/palestra/15/1257.pdf.

Andersen, L. E., Granger, C. W. J., Reis, E. J., Weinhold, D., & Wunder,S. (2002). The dynamics of deforestation and economic growth in theBrazilian Amazon. Cambridge, UK: Cambridge University Press.

Bechtold, G. (1982). Terra Preta do Indio: Anorganisch-chemische Kennz-eichnung eines brasilianischen Anthrohumoxes [Thesis]. Germany:Bayreuth University.

Bickel, U., & Dros, J. M. (2003). The impacts of Soybean cultivation onBrazilian ecosystems. WWF Forest Conversion Initiative.

Brandao, A. S. P., de Rezende, G. C., & da Costa Marques, R. W. (2005).Agricultural growth in the period 1999–2004, outburst inSoybeans areaand environmental impacts in Brazil, Rio de Janeiro.

Brown, J. C., Koeppe, M., Coles, B., & Price, K. P. (2005). Soybeanproduction and conversion of tropical forest in the Brazilian Amazon:The case of Vilhena, Rondonia. Ambio, 34(6), 462–469.

Brown, L. R. (2004). Outgrowing the Earth: The food security challenge inan age of falling water tables and rising temperatures. New York, NY:W.W. Norton.

Cargill (2010). Estudo de Impacto Ambiental, Terminal Fluvial de GraneisSolidos da Cargill Agrıcola S.A, Available at: <http://www.sema.pa.gov.br/download/RIMA%20CARGILL_2010.pdf>.

Carvalho, R. (1999). A Amazonia rumo ao ‘‘Ciclo da soja”. AmazoniaPapers No. 2. Amigos da Terra. Sao Paulo.

de Souza, G. R., De Oliveira, S. C., & De Barros Pinto, L. (2010). AInfluencia De Premio, Cambio E Prec�os No Mercado Externo Sobre OPrec�o Da Soja No Brasil. Presentation to Sociedade Brasileira DeEconomia, Administrac�ao E Sociologia Rural, Congresso 48�. Availableat: <http://www.sober.org.br/palestra/15/1084.pdf>.

Diegues, A. C. (1992). The social dynamics of deforestation in the BrazilianAmazon: An overview: Discussion paper No. 36. UNRISD.

Fearnside, P. M. (2001). Soybean cultivation as a threat to the environ-ment in Brazil. Environmental Conservation, 28(1), 23–38.

Fisher, W. H. (2007). Surrogate money, technology, and the expansion ofsavanna soybeans in Brazil. In: A. Hornborg, J. R. McNeill, J.Martinez-Alier (Eds.), Rethinking environmental history: World-systemhistory and global environmental change (pp. 345–360). Lanham, MD:Altamira Press.

Greenpeace (2006). Eating up the Amazon. London: Greenpeace.Helfand, S., & Brunstein, L. (2001). The changing structure of the Brazilian

agricultural sector and the limitations of the 1995/96 agricultural census.Revista de Economia e Sociologia Rural, 39(3), 179–203.

IPEADATA. Economic database of the Instituto de Pesquisa EconomicaAplicada. <http://www.ipeadata.gov.br/>.

Kaimowitz, D., & Smith, J. (2001). Soybean technology and the loss ofnatural vegetation in Brazil and Bolivia. In A. Angelson, & D.Kaimowitz (Eds.), Agricultural technologies and tropical deforestation(pp. 195–211). Wallingford, UK: CAB International.

Lopez Villar, J., Freese, B., Bebb, A., Bassey, N., Amendola, C., &Ferreira, M. (2007). Who benefits from GM crops?: An analysis of theglobal performance of GM crops (1996–2006). Friends of the EarthInternational.

Mueller, C. C., & Bustamante, M. (2002). Analise da Expansao da Soja noBrazil [Analysis of the expansion of soybeans in Brazil]. MIMEO,Department of Ecology, National University of Brasilia (UNB) (inPortuguese).

Mueller, C. C. (2003). Expansion and modernization of agriculture in theCerrado—The case of soybeans in Brazil’s Center-West. Universidadede Brasılia, Department of Economics Working Paper 306.

Nepstad, D. C., Stickler, C. M., & Almeida, O. T. (2006). Globalization ofthe Amazon soy and beef industries: Opportunities for conservation.Conservation Biology, 20(6), 1595–1603.

Profeta, G. A., Gomes, M. F. M., Lırio, V. S., & Costa, L. V. (2010).Credito Rural E Seus Efeitos Nas Exportac�oes Brasileiras De Soja.Presentation to Sociedade Brasileira De Economia, Administrac�ao ESociologia Rural, Congresso 48�. Available at: <http://www.sober.-org.br/palestra/15/374.pdf>.

Simmons, C. S. (2004). The political economy of land conflict in theEastern Brazilian Amazon. Annals of the Association of AmericanGeographers, 94(1), 183–206.

Steward, C. (2007). From colonization to ‘‘environmental soy’’: A casestudy of environmental and socio-economic valuation in the Amazonsoy frontier. Agriculture and Human Values, 24, 107–122.

Silva, J. G., Sergio Gomez, E., & Rodrigo Castaneda, S. (2009). BoomAgrıcola y persistencia de la pobreza rural. FAO. Available at: <http://www.rlc.fao.org/es/prioridades/desarrollo/boom/pdf/boomagri.pdf>.

Stickler, C., & Almeida, O. T. (2009). Finaciamento internacional para osetor agro-industrial no Mato Grosso: uma oportunidade paraconservac�ao?. Novos Cadernos NAEA, 11(1), 51–84.

Vera-Diaz, M., del, C., Kaufmann, R., Nepstad, D., & Schlesinger, P.(2008). An Interdisciplinary model of Soybean yield in the AmazonBasin: Climatic, edaphic, and economic determinants. EcologicalEconomics, 65(2), 420–431.

Warnken, P. F. (1999). The development and growth of the soybean industryin Brazil. Ames, Iowa: Iowa State University Press.

Zanon, R. S., Saes, M. S. M., Corrar, L. J., & Macedo, M. A. (2010).Produc�ao De Soja No Brasil: Principais Determinantes Do TamanhoDas Propriedades. Presentation to Sociedade Brasileira De Economia,Administrac�ao E Sociologia Rural, Congresso 48�. Available at:<http://www.sober.org.br/palestra/15/38.pdf>.

ScienceDirectAvailable online at www.sciencedirect.com

http://www.abiove.com.br/english/ss_moratoria_us.html

Soybeans, Poverty and Inequality in the Brazilian Amazon - with Diana Weinhold & Eustáquio J. Reis

Documents

Transcript of Soybeans, Poverty and Inequality in the Brazilian Amazon - with Diana Weinhold & Eustáquio J. Reis