Response by Curtis Bell and Scott Wolford

Emily Ritter and Adrian Florea have conducted a pair of rigorous, thoughtful replications of our article Oil Discoveries, Shifting Power, and Civil Conflict, focusing respectively on reproducing our theoretical and empirical models. Theyve contributed to what we hope will be a productive modeling dialogue" (Myerson 1992) about the link between oil resourcesboth proved and exploitedand civil conflict. In this response, we focus on summarizing what we can learn from their proposed extensions, though we also clarify a point of confusion in the interpretation of the theory, which we use to discuss where the contributors efforts can take the modeling dialogue in the future.

First, Ritter proposes an extension of the theoretical model. She shows that giving the government the option to repress after discovery but before rebels can attack narrows the conditions under which civil war occurs, requiring a larger prospective shift in power to overcome the rebels present disincentives to fightwhich now include the exogenous costs of war and the endogenous effects of repression. This occurs even as the reasons behind the outbreak of civil war and the models empirical implications remain the same. This is notable in two ways. First, the possibility of preemptive repression points to an additional empirical factor that can attenuate the relationship between oil discovery and civil war: pre-existing coercive capacity at the time of discovery, which allows the state to disrupt preventive civil wars before they start. Second, as she notes, it can help explain otherwise puzzling instances of repressionparticularly those that might appear to occur out of nowhere or outside the Law of Coercive Responsiveness (Davenport 2007)because they are preventive rather than reactive (see, inter alia, Danneman and Ritter 2014, Ritter 2014). Further, an empirical model based on this insight, we suspect, would find a curvilinear relationship between oil discovery and repression; repression will spike in response to moderately sized oil discoveries but decline (to be replaced by the outbreak of civil war) as discoveries grow so large that repression cannot keep the lid on rebellion.

The incentive to repress in this extended model also sheds light on government incentives in the baseline modelto take steps to avoid war in the present so that gains from oil reserves can be realized. In that sense, Floreas remark that anticipated wealthreduces governments motivation to negotiate a peaceful deal is a misreading of what is an admittedly subtle mechanism. Anticipated wealth actually makes governments very willing to negotiate and make concessions today; it renders them patient, because they will be strong enough to renege on those concessions later, once new oil wealth is realized as increased military power. It is not the government but potential rebels that lose the incentive to negotiate in the shadow of anticipated oil wealth. This process perhaps tempts the government to repress, but not to draw itself into an all-or-nothing civil war before it grows strong. The governments willingness to make concessions but inability to commit to them in the long run is precisely what renders the rebels unwilling to accept even the most generous concessions. [1]

Next, Florea conducts a thorough empirical replication, reproducing the core published result on our original sample and one carefully adjusted for some quirky codings of civil conflict in the UCDP data. In each case, newly proved oil reserves are associated with an increased probability of civil conflict in the poorest states, or those for whom the future shift in relative power will be greatest. The bulk of his contribution, though, focuses on a test that uses not civil conflict onset but civil conflict presence as the outcome variable, on the logic that the shift in power takes time to realize and therefore may incentivize conflict as the shift looms on the horizon (Powell 2012). Rather than drop years of ongoing civil war, he keeps them in the data, where a positive outcome indicates that a civil conflict either begins or does not stop. In Model 3, which selects on those states with per capita GDP of less than $12k/year, an increase in proved reserves is associated with the presence of fighting in the following year. This is an intriguing pattern, thoughas Florea notesit remains more suggestive than dispositive.

By our reading, there are two questions underlying his basic insight. First, do oil discoveries during civil wars tend to lengthen them? This is perhaps closest in spirit to Floreas empirical model, though it might call for an interaction with a lagged dependent variable. Another option is to observe wars over time and record the instances of new oil discoveries during the fighting. [2] Second, do civil wars that begin as the result of oil discovery last longer, or perhaps end differently, than other civil wars? This requires a different empirical model that samples on wars themselves, perhaps coding the size of new proved reserves in states on the eve of the civil wars outbreak. This new model might help illuminate when and how the fighting stops: once the oil wealth is realized or the ability to exploit is eliminated, once the rebels are defeated, or once the rebels capture it (or the state apparatus to control it).

We are not yet convinced, however, of the need to sample on government-rebel group dyads, unless our theory can be extended to modeland finddependencies across observations of multiple rebel groups fighting the government at the same time. Our hunch is that a looming adverse shift in power makes any rebel group eager to fight, which suggests that country-year might be the appropriate unit of analysis. If victory will allow the winner of the war, government or rebel group, however, to exploit the vanquished, it might be associated with a proliferation of rebel groups fighting to secure their own survival. Such a theoretical model, as well as empirical models to assess these relationships between war duration and the number of rebel groups, awaits further exploration.

In sum, Ritter and Florea have each made significant contributions to what we hope turns out to be a fruitful research agenda, extending our theoretical and empirical models to explain both state repression and the duration of civil conflict. Replication is valuable as a policing mechanism, to encourage careful work and to allow for the correction of mistakesin both theory and empiricsbut replication can also lead to new, creative, and unanticipated insights, which both contributors have done in this case.

[1] See Powell (2004) for a characterization of this mechanism in several contexts through a general inefficiency condition.

[2] It is also possible that oil discovery is unlikely during civil war, which would push observed effects—even if true—towards zero.

Discuss this Article
There are currently no comments, be the first to post one.
Start the Discussion...
Only registered users may post comments.
ISQ On Twitter