Back in 1995 I turned down an invitation to get involved in a project funded by Vice President Al Gore’s office to evaluate why states failed. The project proved quite successful (which surprised me), and just last week the Fund for Peace released the eighth edition of their annually published Failed States Index (FSI). [NB: Jay Ulfelder helpfully points out in a Comment below that I have conflated the State Failure project and FFP’s FSI; the latter is apparently only influenced by the former; they are distinct projects.]
Joe Young, writing at Political Violence @ a Glance, argues that FSI is conceptually dead: a binary conceptualization fails to be useful because it misses the variation across space (and over time) in government’s ability to enforce its sovereignty claim. Young proposes the term “ungoverned space” as an alternative, One useful way for an American to wrap her head around this idea is: will the police respond to a call with only one officer, with only two; only when at least two cars are available; or not at all? Interestingly, Young’s proposal reflects an issue that interested Ted Gurr, one of the founders of the FSI project. In his Polity datasets Gurr produced a variable that he called Scope of Government, which was a measure of the extent to which the government was able to regulate economic, political, and social activity throughout the country (that variable was dropped for Polity III and Polity IV). While not exactly the same as a state’s ability to enforce its sovereignty claim, the two concepts would be highly correlated. Of course Gurr’s Scope of Government did not vary sub-nationally, and today scholars are thankfully much more attentive to the value of considering sub-national variation.
The Environmental Change and Security Program at Princeton’s Wilson School also wrote about the FSI today, posting an analysis of the contribution of the components to the index. That exercise unfortunately induces a bit of pain as the authors apparently have little familiarity with measurement theory. This does not render their assessment entirely worthless, but it is chock full of misleading and/or inaccurate claims. I am not feeling the energy to walk through them, so my apologies for the blanket assertion. But the post made me think about the possibility of using a Bayesian latent measurement model constructed from the components of the index (whether the resulting measure of risk to failure would then be used as a dependent or an independent variable). Of course, the FSI does not vary sub-nationally, and is thus evidence of the distance we have to travel on that front.