Bad Neighbors: Failed States and Their Consequences

Abstract

State failure reflects the collapse of a sovereign state, and has been hypothesized to destabilize an entire region. We assess the negative effects of state collapse, focusing particularly on the spatial diffusion of these consequences. We argue that the instability, unrest, and civil war that increase the risk for state collapse are not limited to the failed/collapsed state; states neighboring—or located within close distance of—a failed state are also likely to experience subsequently higher levels of political instability, unrest, civil war, and interstate conflict. We also evaluate the likelihood of state failure itself diffusing to other states. Specifically, we test the proposition that state failure causes political turmoil in nearby states to a greater extent than in distant countries. We do so by including a distance-weighted measure of state failure and by evaluating the effect of collapse in contiguous states. We conclude that state failure/collapse itself is not contagious, but some of its most negative consequences do indeed diffuse to other states.

Keywords:

• fragile states
• political violence
• spatial diffusion
• state collapse
• state failure

A previous version of this paper was presented at the 40th North American Meeting of the Peace Science Society (International), November 10–12, 2006, Columbus, Ohio, and we wish to thank a number of conference participants for their comments.

Notes

For example, see Saleyhan and Gleditsch (2006) for a discussion of the regional clustering of civil war and the effects of refugee flows as one important factor in the diffusion of civil conflict. Note also the debate in the analysis of civil war, over the effects of ethnicity; see Fearon and Laitin (2003) or Elbadawi and Sambanis (2000) for results that minimize the effects of ethnicity on civil war.

The Civil War variable contains all conflicts designated “Type 3” in the PRIO Dataset on Armed Conflict. These are described as internal armed conflicts, and range in intensity from minor (with 25 battle deaths in a conflict) to intense (1000 battle deaths in a given country-year).

All Type 2 conflicts (interstate armed conflict) in the PRIO Dataset on Armed Conflict.

Collier and Hoeffler (2004) have shown that low national income is a major factor in the occurrence of civil war. Thus, while a variety of possible economic measures could be used, we have opted for more parsimony, and included only national income and economic openness.

We chose to lag independent variables one year as this provides us with the most rigorous test of our hypotheses regarding diffusion and contagion of the consequences of state failure; the one-year lag reflects the “minimum incubation period” for the spread of the effects of state collapse. We expect longer lags to strengthen our findings.

We estimated the models using both random effects and fixed effects approaches, and the two approaches yielded substantively similar results. Under these circumstances, random effects models have the virtue of being more efficient, as well as preserving information from observations that fixed effects models must necessarily exclude (Hsiao, 2002). Alternative approaches to analyzing diffusion processes may include spatial lag and spatial error models. However, models for spatially referenced panel data (e.g., Franzese and Hays, 2007) are in their infancy, and we leave the use of those models for future work on the effects of state collapse.