Clarifying the mediation dilemma: A response to “Sticks and carrots for peace”

Abstract

This brief essay considers the “mediation dilemma” in the light of new analysis by Constantin Ruhe and Iris Volg. Ruhe and Volg’s analysis adds to our understanding of manipulative mediation in two important ways: (a) it demonstrates how an analysis that uses a lens of survival functions clarifies the policy trade-offs beyond what is possible from an analysis that uses a lens of changing hazard rates; and (b) it demonstrates that lighter (nonmanipulative) forms of mediation have a less positive effect on peace stability than in the original analysis. This response also offers important corrections to the conclusions drawn by Ruhe and Volg: (a) ignoring the lens of changing hazard rates misses key ways of testing for the observable implications that arise from the underlying theoretical arguments; (b) Ruhe and Volg misstate some of the theoretical claims made by Beardsley; and (c) almost all of the original implications explored by Beardsley remain supported.

Introduction

In “Sticks and carrots for peace: The effect of manipulative mediation strategies on post-conflict stability,” Ruhe and Volg (2021) present new analyses on the relationship between manipulative mediation and the relapse of interstate crises. The authors conclude that instead of there being a dilemma with regard to a trade-off between mediation’s short- and long-term effects – as developed in Beardsley (2011) – “Manipulative mediation is associated with substantively greater stability compared to unmediated cases.” ¹

This brief response considers the “mediation dilemma” in the light of the new analysis. Ruhe and Volg’s analysis adds to our understanding of mediation in two important ways: (a) it demonstrates how an analysis that uses a lens of survival functions clarifies the policy trade-offs beyond what is possible from an analysis that uses a lens of changing hazard rates; and (b) it demonstrates that lighter (nonmanipulative) forms of mediation have a less positive effect on peace stability than in the original analysis. As a contribution to a positive accumulation of knowledge, Ruhe and Volg’s analysis provides an important clarification of the relationship between mediation style and post-crisis stability.

This response also offers important corrections to the conclusions drawn by Ruhe and Volg: (a) ignoring the lens of changing hazard rates misses key ways of testing for the observable implications that arise from the underlying theoretical arguments; (b) Ruhe and Volg misstate some of the theoretical claims in Beardsley (2011); and (c) almost all of the original implications explored in Beardsley (2011) remain supported.

Important contributions to our understanding

Ruhe and Volg make at least two important contributions that should be highlighted irrespective of the corrections offered below. That is, readers of Beardsley (2011) should also read Ruhe and Volg to understand the effect of mediation style better for two reasons.

First, Ruhe and Volg use depictions of survival functions to clarify the trade-offs that practitioners face in choosing their style of mediation. By presenting survival functions, Ruhe and Volg focus on the likelihood that crises will not recur at any given point between the end of the crisis and 10 years later, based on whether the crisis experienced manipulative mediation. On the contrary, Beardsley (2011) focused on changing hazard rates: variation in the risks of failure at any given point in that 10-year window, conditional on having lasted to that point. The difference pertains to what the reference point is. In Beardsley (2011), the reference point moves over time, so we can see the increasing relative fragility of peace under the condition of manipulative mediation by considering the instantaneous probability of relapse at, say, eight years after the crisis has ended and conditional on the peace lasting to that point. In Ruhe and Volg’s reliance on the presentation of survival functions, the reference point remains fixed, such that we can see that the likelihood of survival of peace after manipulative mediation converges toward that for other crises, but never fully reaches the lower level of peace survival for nonmanipulative mediation cases.

As far as the new analysis is concerned, we do not see a stark trade-off from the standpoint of the practitioner who might be worried about a mediated settlement becoming unconditionally more fragile in the long run. Ruhe and Volg’s analysis clarifies that the cumulative likelihood of crisis recurrence over the next 10 years is about the same regardless of whether manipulative mediation is applied. This is actually similar to a point made by Beardsley (2008: 737): “the results should not be interpreted as suggesting that mediated crises are unconditionally more likely to recur. Recall that unmediated peace arrangements are much more likely to fail in the first few years after a crisis. The key point is that mediation does very well in sustaining short-term peace at the expense of some potential for extremely durable peace.” ² That is, Beardsley (2008, 2011) does not expect the survival curves to cross, only to converge. However, the nuance of this point, which is that the mediation dilemma does not imply unconditional fragility, may have been lost with the focus on changing hazard rates and instantaneous fragility. Importantly, Ruhe and Volg’s analysis clarifies why in many cases manipulative mediation is still worth attempting.

Second, by correctly adjusting the data structure, following Jin and Boehmke (2017), to account for time-varying covariates, Ruhe and Volg find that nonmanipulative mediation does not have the same potential for making enduring contributions to peace as found in Beardsley (2011). That is, Ruhe and Volg’s new analyses find similar coefficients to what is found in Beardsley (2011) for the manipulation variable and its interaction with event time, but they do not find that the relationship with respect to nonmanipulative mediation is the same as in Beardsley (2011). This point is understated in their discussion, but should not be lost as an important update to the original study.

The importance of changing hazard rates

Although Ruhe and Volg do well to present survival functions to show convergence in the unconditional likelihood of peace failure, ultimately, their results support the counterfactual claims in Beardsley (2011). The observable implications from Beardsley’s arguments focus on changing hazard rates, and not on the cumulative probability of failure from the time of agreement until some time in the future. The expectation from Beardsley is that agreements formed without manipulative mediation that are able to survive in the short term are the ones that are self-enforcing and, thus, will be more stable in 5 to 10 years time than those that had to rely on manipulation to reach a settlement. If we can divide agreements into those that are self-enforcing and those that are fragile, the likelihood a given peace that has lasted over five years is a self-enforcing type is going to be higher for the cases that did not experience manipulative mediation when compared with those that did. This hypothesis can best be examined by seeing how the hazard rates change over time, such that we can see if the cases with manipulative mediation become increasingly likely to fail the further we get from involvement by the mediator.

Through this lens, Ruhe and Volg’s finding that the survival curves converge in the long run is precisely on par with the expectations of Beardsley (2011). The agreements that were reached via manipulation eventually fall at a faster rate than the agreements that were reached otherwise. This is exactly what Beardsley (2011) expects, based on the work’s theoretical understandings of self-enforcing peace and time inconsistency problems. Beardsley’s (2011: 115) hypothesis – the observable implications of what manipulative mediation is expected to be doing to the underlying bargaining dynamics – remains supported by the Ruhe and Volg analysis.

Working through Figure 1 in Ruhe and Volg illustrates the importance of considering a shifting reference point. With a shifting reference point, we can ask whether peace is more likely to endure from point 0 to point 1 with or without manipulation, and it is clear that the answer is with manipulation – this is the short-term benefit of manipulation. We can then ask whether peace is more likely to endure from point 2 to point 3 with or without manipulation, and it is clear that the answer is without manipulation – this is the increasing relative long-term fragility of manipulation. An analysis that considers how the hazard rate changes over time would correctly consider the cases that failed between time 2 and time 3 (Cases A–D) to be rather fragile, not “stable,” at that time. Moreover, Ruhe and Volg’s example does not consider the potential of manipulative mediation for actually shortening the possible duration of peace in Case E; Beardsley (2011) raises the potential for Case E to be so stable beyond the period of mediator influence precisely because it survived to that point and had an opportunity to become self-enforcing in the absence of manipulative mediation.

Figure 1.

Ruhe and Volg’s (2021) stylized example of fictitious conflict cases with and without manipulative mediation. (a) Fictitious conflict cases and counterfactual failure times with and without manipulation; (b) Probability of new conflict among surviving cases, given the example data in Panel A; (c) Survival of peace with and without manipulation, given the example data in Panel A.

Clarifying the claims made by Beardsley (2011)

Ruhe and Volg misstate Beardsley’s (2011) counterfactual claims in two ways. First, they claim that Beardsley’s argument assumes that “capabilities and characteristics of conflict parties remain the same after conflicts end,” but no such assumption is made. Indeed, Beardsley (2011: 28–30, 41–42, 101–102, 105–108, 114) discusses at length the importance of time inconsistency problems generated by shifting capabilities and resolve, and how manipulative mediation is at greater risk of forcing agreements that do not take into account those potential shifts or develop the means for the crisis actors to renegotiate the bargains in the light of any potential shifts peacefully.

Another claim is that the authors imply Beardsley’s (2011) viewpoint is consistent with Luttwak’s (1999) claim that it is often better to let the parties fight it out. However, this is not consistent with the conclusions articulated by Beardsley, who writes that “this book is not a call for general pessimism regarding third-party assistance in conflict management. Instead, I contend that, even when the trade-off is unavoidable, intervention can still be worthwhile because the short-term benefits can outweigh the long-term costs” (Beardsley 2011: 11). Further: “That mediation more often leads to a decrease in long-term peace does not necessarily make it a bad bet. Even when outside involvement makes the long-term prospects for peace grim, which is certainly not always the case, mediation can be strictly better – to the disputing parties as well as to the international community – than the alternative of escalating hostilities. My argument and analysis in this book thus urge prudence, not pessimism, when considering mediation as a conflict management vehicle” (Beardsley 2011: 178).

Preserving the implications

In combining the insights from both quantitative and qualitative analyses, Beardsley (2011: 183–190) draws seven implications:

1. First, mediation should be used sparingly if major vulnerabilities are likely to arise if implementation fails.

2. Second, third parties need to be aware of issues related to legitimacy.

3. Third, outside actors should intervene more carefully when the disputants could benefit from using mediation for ends other than peace.

4. Fourth, potential third parties should hesitate to become involved in a peace process if coordination with regard to implementation is likely to prove difficult.

5. Fifth, the use of leverage itself is not actually a source of long-term instability; it is the attenuation over time of that leverage that increases the propensity for renegotiation.

6. Sixth, low-key involvement by a single mediator can work well in the long run.

7. Seventh, mediation is worth the gamble when there are considerable advantages and few disadvantages.

These implications stem from the theoretical arguments that were confirmed by the evidence obtained from the changing hazard rates according to which it was found that mediated peace processes – especially manipulative mediation – tend to produce peace settlements that do well in the short term and then experience higher hazard rates over time. In the light of Ruhe and Volg’s analyses, the first implication from Beardsley (2011) should be broadened to include any peace process that stands to leave the actors extremely vulnerable during the implementation phase, given that the authors find the cumulative rate of failed implementation is similar for cases with and without manipulative mediation. Based on the Rwandan genocide that followed the Arusha Accords, Beardsley’s point, however, is that manipulative mediation has a greater potential for generating false hopes of protection and, thus, existential consequences for relapse when the actors are extremely vulnerable as a result of the false hope. Another exception is the sixth point, because the new findings do not support the evidence that mediation without manipulation has a positive relationship with peace stability.

Conclusion

Ruhe and Volg are correct in pointing out that Beardsley’s (2011) focus on changing hazard rates has the potential to overstate the extent to which there is a true “dilemma” from a policy standpoint. By focusing on the potential for unconditional durable peace from the standpoint of the end of the previous crisis, Ruhe and Volg clarify that peace processes with manipulative mediation are not less likely to produce survivable peace over a 10-year window than other peace processes. Moreover, by correcting the structure of the data, they show that nonmanipulative mediation does not have much efficacy in reducing crisis recurrence. That being said, much of the original hypothesis testing and interpretation that follows from Beardsley’s (2011) analysis remains.

When considering Ruhe and Volg alongside a corrected interpretation of Beardsley (2011), the expectation should be that the low short-term probabilities of failure and the increase in the hazard rates in the long term balance out when considered through a lens of survivable peace over a 10-year window. Practitioners would still do well to consider the underlying possibility that manipulative mediation will reduce the potential for the actors to reach a self-enforcing peace – manifested in hazard rates increasing over time – so that they can both better manage the risk of conflict relapse in the periods after mediation and make prudent choices about whether costly manipulative mediation is worthwhile if, ultimately, the expected survivability of peace converges toward what is expected without manipulation.