LAND INEQUALITY AND CONFLICT IN LATIN AMERICA IN THE TWENTIETH CENTURY

Abstract

In this paper, we study the association between land inequality and civil conflict using a panel of 18 Latin American countries spanning the twentieth century. We find a statistically significant relationship between land inequality and civil conflict onset. This finding is consistent with the ‘land maldistribution hypothesis’ and the view that inequality is important for understanding conflict in Latin America. It is also consistent with recent theories of conflict and the model of revolution applied by Acemoglu and Robinson in their work on political transitions.

Keywords:

• Inequality
• Civil conflict
• Democracy
• Latin America
• land maldistribution hypothesis
• revolution

JEL Codes:

• E0
• O1
• O54

ACKNOWLEDGEMENTS

We would like to thank Martin Paldam, Lant Pritchett and Christian Bjørnskov for comments on an earlier version of this paper. Comments from seminar participants at the World Meeting of the Public Choice Society in Amsterdam in 2007 are appreciated. We thank Toke Aidt and Dalibor Eterovic for sharing their data. We also thank the editor of the journal and two anonymous reviewers for helpful comments and useful suggestions. The usual disclamer applies.

Notes

¹ Some of the data were kindly provided by Toke S. Aidt and Dalibor S. Eterovic.

² The countries are: Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, Uruguay, Venezuela, Costa Rica, Dominican Republic, El Salvador, Guatamala, Honduras, Mexico, Nicaragua and Panama.

³ For example, the data for Angola are available from 1988 only. Angola had a civil war from 1981 to 1999 according to the PRIO definition, meaning that its onset cannot be included. Gabon could be included, but it had no conflicts in the period. Djibouti had one conflict year in 1991. These examples illustrate the more general problem that many conflict onsets would not be included if Sub‐Saharan Africa was added to the data‐set.

⁴ Ideally, unsuccessful coups should also be included, but in practice, such coup attempts are difficult to ascertain.

⁵ Family farms are defined as farms which provide employment for less than five people (including family members), are cultivated by the holder family itself and are owned by the cultivator family or held in owner‐like possession. As such, the measure is intended to distinguish ‘family farms’ from large farms which are cultivated mainly by hired workers (Vanhanen, 2003).

⁶ It should be noted that the COW data‐set is not entirely reliable, see Restrepo, Spagat and Vargas (2006) on Columbia.

⁷ We thus focus on spells of peace, with 1 indicating a failure. Some authors prefer to set ongoing wars to 0 rather than setting them as missing values. This change makes little difference to the results.

⁸ We have used three knots to form the cubic splines. We do not report the peace duration variables, but note that they have high overall significance. The procedure is implemented by the STATA program btscs available at http://www.prio.no/CSCW/Datasets/Stata-Tools/

⁹ Easterly (2007) suggests the wheat–sugar ratio, but this instrument is not time‐varying.

¹⁰ See the Appendix for definitions of control variables.

¹¹ Economic crises are also likely to play a role through their effect on the opportunity costs of conflict, see, e.g., Collier and Hoeffler (2004); Jensen and Gleditsch (2009). Coups have also been argued to be more likely in situations of crises, and less likely when economic conditions are strong (Londregan and Poole, 1990).

¹² Wooldridge (2002, p. 459) notes that interaction can be included in logit models in a straightforward manner when the goal is to assess the sign of the partial effect. In our case, the sign of the partial effect of Polity/democracy is determined solely by the estimated coefficients to Polity and the interaction with Family farms.

¹³ See also Aidt and Jensen (2010) for a test of Acemoglu and Robinson’s ‘threat of revolution theory’. Whereas Przeworski (2009) covers the twentieth century, they cover the nineteenth century, when the voting franchise was actually extended to Western Europe.