The Political Economy of Land Conflict in the Eastern Brazilian Amazon

Abstract

An important goal of regional development in the Brazilian Amazon was to enhance social welfare and alleviate dire poverty in other parts of the country by providing land to the poor. Nevertheless, both poverty and landlessness have persisted despite development policies that distributed billions of dollars on highway construction, loans, and outright subsidies. Inequitable land distribution has been held as a prime factor in land conflict across the country. Although episodes of conflict over land are common in Brazilian history, this paper focuses on agrarian issues that arose with the opening of the Amazon frontier in the 1970s. The paper presents a political economy approach that considers the role of hierarchical forces interacting across spatial scales, in creating conditions ripe for land conflict at the local level. The premise is that the Brazilian government, intending to bring about economic and social development, promoted contradictory strategies creating land scarcity. These strategies led to expansion of large ranching operations, creation of conservation units, and demarcation of indigenous reserves, which constrained the pool of land available for small farmer settlement. Empirical analysis employing regression and spatial statistics is used to test the proposed model, advancing previous efforts by applying spatial regression, incorporating improved indicators of conflict and explanatory variables generated by a Geographic Information System (GIS). The findings provide support for some elements of the argument, demonstrating statistically significant relationships between land conflict and land concentration, cattle ranching, and road construction. Finally, a case study analysis of a county in the heart of the land conflict zone is provided, illustrating the interaction of scalar forces, and the articulation of land conflict at the local level.

Key Words:

• Amazonian development
• agrarian reform
• social and environmental justice
• conservation policy

Acknowledgments

I am especially indebted to Robert Walker for his help with the elaboration of the main concepts of this paper. I would also like to thank Stephen Perz for comments on an early version of the manuscript, and Patrick O'Sullivan and Barney Warf for their support of the initial phase of research. I am especially grateful to my Brazilian colleagues Pedro Mourão de Oliveira, Marcos Pedlowski, Eraldo Matricardi, Marcellus Caldas, and Eugenio Arima for their valuable insight on the issues addressed. Finally, Several anonymous reviewers contributed significantly to the paper, although I remain responsible for remaining errors.

Notes

Source: Hoffman (1998).

Note: The data for 1967 came from the Brazilian Institute for Agrarian Reform (IBRA), whose responsibility for land reform later fell within the National Institute of Colonization and Agrarian Reform (INCRA) charge. Data for the remaining years came from INCRA.

Source: CPT, 1996.

Note: The murder count for Pará includes the 19 peasants murdered in Eldorado do Carajás on April 19,1996; excluding this incident, Pará would still have the highest degree of land-conflict-related murders for the nation.

Source: Personal interview with MST representative.

Source: Almeida, 1994.

1. For accounts of the War of Canudos see Os Sertões by Euclides da Cunha (1985); for the Ronco de Abelha and the Ouebra-Quilos rebellions see Crise Agrária e Luta de Classes by Monteiro (1980); and for the Contestado Rebellion see Millenarian Vision, Capitalist Reality: Brazil's Contestado Rebellion, 1912–1916, by Diacon (1991).

2. Many small farmers view indigenous reserves as government land and therefore available for occupation as allowed by the federal constitution. This is especially true in light of the 1997 passage of decree 1775 that opened reserve demarcation for contestation by interested parties.

3. The county, as a geographical unit, is similar to a county in the United States, although Brazilian counties can be substantially larger in terms of land area.

4. This includes data for the county of Eldorado do Carajás, which was created in 1991 from land within the county of Curionópolis, which in turn was created in 1988 from the county of Marabá.

5. Data transformation involved adding a 1 to each value and then performing a Box-Cox power transformation.

6. A tobit model is a regression model for left censored data. Since a significant number of observations had a zero value for their dependent variable, such an approach was appropriate.

7. The construction of the road variable involved the visual inspection of the map Sistema Rodo-Aeroportuário, prepared by the Departamento de Estradas de Rodagem do Pará, 1987.

8. The coefficients were calculated by combining the 15 standard land size classifications in the census to 5 categories: (1) 0 to 49 hectares; (2) 50 to 99 hectares; (3) 100 to 999 hectares; (4) 1000 to 9,999 hectares; and (5) greater than 10,000 hectares.

9. Several counties on Marajó Island are completely under some form of conservation unit.

10. Spatial autocorrelation occurs when variables are correlated with themselves in space (Odland 1988). In the regression context, problems arise when the dependent variables or the error term are spatially autocorrelated. The first case is “substantive” and the second is “nuisance” spatial autocorrelation. When either case occurs, coefficient estimates from OLS are either biased or inefficient (Anselin and Rey 1991).

11. Note in Table 6 (a–c), Table 7and Table 8, once the power transformations were performed, the significance of the Lagrange Multiplier Error for all the models diminished.

12. The war of Araguaia, as it has been termed, was the military siege on the leftist guerrillas from São Paulo who used land reform as the basis for peasant mobilization. This group was viewed as a threat to national security, and consequently all 70 activists, and an undocumented number of peasants, were killed (IstoÉ 2000).

13. The violence intensity (A [Almeida 1994]) is 66; violence intensity (B [Barata 1995]) is 33; and the conflict intensity is 61. The averages for these same variables are 17, 15, and 25, respectively.

14. The county boundaries are for 1997, but the data is aggregated to 1985. Consequently, the blank (white) regions are those counties created since that time.