Does unemployment drive rebellion?

Maybe not.  There is a new piece out by Eli Berman, Michael Callen, Joseph H. Felter and Jacob N. Shapiro, in Journal of Conflict Resolution, August 2011.  Here is an ungated version, excerpt:

Most aid spending by governments seeking to rebuild social and political order is based on an opportunity-cost theory of distracting potential recruits. The logic is that gainfully employed young men are less likely to participate in political violence, implying a positive correlation between unemployment and violence in places with active insurgencies. We test that prediction on insurgencies in Iraq and the Philippines, using survey data on unemployment and two newly- available measures of insurgency: (1) attacks against government and allied forces; and (2) violence that kills civilians. Contrary to the opportunity-cost theory, we find a robust negative correlation between unemployment and attacks against government and allied forces and no significant relationship between unemployment and the rate of insurgent attacks that kill civilians.

Here is a WQ summary of some additional findings from the paper.


(1) business cycle effects ignored by assumption, and the data points are separated by geography. How does that make sense? Insurgents do move to attack high-value targets, particularly if they perceive (as is likely) that those high-value targets are the occupied ones.

(2) their own speculative mechanisms have a degree of endogenous behavior by the state itself (in mobilizing resources in response to insurgency), so their uncontrolled conclusion's relevance to the merits of those responses are dubious. "A problem Y existed, so the state did X, and the problem Y went away. The problem Y went away, so the state didn't need to do X!".

I can see right from the face of it this is a really bad study. Basically two data points- Iraq and the Philippines. How do they track unemployment in Iraq? I doubt the government keeps accurate statistics - did they do surveys? So you think when the Iraqi answers he's employed by the Mahdi Army they count that as being Employed?

Wonder how it'd look if they added Greece to their data-set.

I'm not sure these are the best subjects, since the Philippines has a southern Muslim insurgency in a predominantly Christian chain of islands, while Iraq has a Sunni/Shia conflict. Neither of these really has any roots in economic factors, and both countries suffer primarily from a lack of social capital (i.e. widespread corruption and rentseeking).

Could use more data points, but I suspect the study is probably broadly correct anyway. The Danegeld tends to encourage rather than discourage the Danes, and (with the obvious exception of OWS) few modern violent politicial movements are actually about unemployment we shouldn't really expect makework gov't employment to "solve" the problem (as oppposed to, in many cases, providing a source of funding for more violence).

They are looking at practically defunct war zones. The correlation between unrest and unemployment should be stronger in more or less functional nations with an unemployment spike.

Unemployment spikes deliberately ignored; they discard nationwide effects.

"Neither of these really has any roots in economic factors"


The Iraqi conflict certainly has economic factors, specifically access to oil revenues and the associated rent-seeking.

Yeah but the related insurgency had much more to do with religious/political issues then people being out of work. And like I said - does being "employed" by one of the sectarian militias count as employment? What is their metric here?

I wonder if unemployment in the US correlates to regime changes.

Two observation test.

reverse causality?

If more attacks take place, government could hire more people to guard against.

Interesting. Rebellion as stimulus.

Plenty of broken windows...

......broken femur fallacy?

not necessarily. just something one might expect. although one may question plausibility.

it might also be that groups choose to attack places of prominence within a country, which would have lower unemployment. the fixed effects may take care of the problem, but a cross-section is being analyzed so it cannot.

Regardless of the levels of violence experienced, all civil disorders are not rebellions.

Rebellion as the expression of forces(s) against force(s) is something quite specific.

People just want to be free.

People Got to Be Free--Lyrics 1968 The Rascals

Another clear problem is that this is probably data with significant spatial autocorrelation that is just unaccounted for. I wonder if they considered a GIS approach.

Comments for this post are closed