The causes, dynamics, and outcomes of armed conflict are usually studied from an aggregated
country-level perspective. This body of research has been successful in identifying a handful
of factors that in general distinguish between peaceful and unstable societies. For example,
we now know that intrastate conflicts occur most frequently in large, poor, ethnically diverse,
and oil-rich societies (Hegre and Sambanis, 2006). Despite all its merits, however, this
research is not well suited to evaluate micro-level explanations for social unrest – to which
theories of civil war often refer (e.g. Gurr, 1970). Even so, it is not uncommon in these studies
to draw inferences about local determinants and individuals’ motivation. For example, some
argue that the powerful empirical association between low per capita income and outbreak of
civil war is driven by low opportunity costs of joining a rebellion in such countries (Collier
and Hoeffler, 2004). While unemployment and poverty may well be important risk factorsi
,
this deduction nonetheless constitutes an ecological fallacy as it offers a micro-level
explanation for an observed country-level pattern. More generally, an unfortunate but often
unnoticed limitation of the habitual state-centeredness is the implicit assumption that conflicts
have a uniform spatial impact on the conflict-ridden societies.ii A brief look at the
contemporary world, with its many peripheral insurgencies, shows that this is not a very
attractive assumption. In fact, geography exerts a pervasive influence on armed conflict. In
Colombia, vast forests and inaccessible rural hinterlands, coupled with an unusually favorable
resource endowment, have made the current guerrilla war the longest uninterrupted intrastate
conflict in the post-World War II period (Harbom and Wallensteen, 2007). And it is hard to
imagine that the insurgencies in Cabinda, Angola or Assam, India would have lasted for
decades, had they not enjoyed local support and occurred in some of the most outlying areas
of the affected countries.