Global Trends in Conflict-Related Sexual Violence: A Response to Green, Cohen, and Wood

By Andrew Mack

We thank Amelia Hoover Green, Dara Kay Cohen and Elizabeth Jean Wood (henceforth GCW) for their thoughtful critique of the Human Security Report’s analysis of the impact of war on the incidence of sexual violence.

As we made clear in the Report, we believe that Dara Cohen’s dataset, which relies on State Department reports on human rights violations, makes an important contribution to our understanding of wartime sexual violence. Indeed we agree with GCW more than we disagree. But there are some real differences.

GCW say that what they call the “global decline claim” — i.e. that conflict-related sexual violence is likely declining worldwide — is not supported by data.

In fact, as the graph below makes clear, the Cohen dataset itself clearly show that the percentage of conflicts that experienced the two highest categories of rape decreased by about one-third during the 2000s.

This is significant for two reasons.

First, the new millennium is the decade in which we believe that the State Department data on reported conflict-related sexual violence are most likely to accurately represent the trends in actual levels of conflict-related sexual violence. This is because there has been a dramatic increase in reporting on such violence at the UN, other international agencies, the donor community, NGOs and the media during the 2000s.

Second, the 2000s is also the decade in which senior officials in the UN and other international agencies were claiming, without any supporting data, that wartime sexual violence in warfare was increasing.

In the Report we argue that the State Department data from the 1980s almost certainly suffer from chronic under-reporting and thus grossly underestimate the actual level of sexual violence. It defies credibility that the overwhelming majority of conflicts should experience no sexual violence at all in this violent decade.

Under-reporting is a problem that also appears to affect the 1990s. There is a considerably greater percentage of countries with no reported sexual violence at all than in the 2000s. However, we believe that the increase from the mid-1990s up to the year 2001 is influenced by a substantial rise in awareness of the issue of wartime sexual violence. Following the mass rapes in Bosnia and in the genocide in Rwanda, wartime sexual violence received growing attention in the international community and the level of reporting increased.

The trends in reported sexual violence — first up and then down — are very clear in the graph in GCW’s post, reproduced below.

Will the substantial decline in reported conflict-related sexual violence in the 2000s become a continuing trend? No one knows. But it is certainly a major shift from the three-fold increase in reporting that took place in the previous two decades.

Since 1980: More Sexual Violence or More Reporting of Sexual Violence?

As noted by GCW — and as we discuss in the Report — the Cohen dataset shows a very large increase in the incidence of reported conflict-related sexual violence from 1980 to 2000. So it may legitimately be asked why we chose to focus on the 2000-2009 period in which we noted that only 9 percent of the years of active conflict were characterized by the highest level of sexual violence.¹

The reason is simple.

As explained above, the extraordinarily low levels of reported sexual violence in the 1980s were almost certainly a function of a lack of reporting, and not any absence of conflict-related sexual violence.

This was still part of an era in which, as countless feminist and other scholars have pointed out, conflict-related sexual violence was generally ignored in the international community, often being accepted as a “normal” consequence of war — one about which little could be done.

It is true, as GCW point out, that there are some references in State Department reports to wartime rape in this period. But with no reported sexual violence at all in 85 percent of the conflict years in this decade it is clear that these reports were very few and far between.

Indeed just how little reported conflict-related sexual violence there is in the State Department data for the 1980s is apparent in the fact that no countries in the 1980s experience Category 3 conflict-related sexual violence — the highest level — despite the fact that conflicts in this decade were more than twice as deadly on average as those of the 1990s, and more than three times as deadly as those of the new millennium.² Just 5 percent of conflict years are reported as experiencing the serious sexual violence of Category 2, and 11 percent experience Category 1 — low levels of reported sexual violence.

Are the State Department data for the 1980s really credible when they indicate that only 15 percent of conflicts experienced any sexual violence in that period, when in the far less violent 2000s it finds that some 87 percent of conflicts experienced conflict-related sexual violence?

Was there really the huge increase in sexual violence in conflict-related sexual violence in the 1980s and 1990s that the State Department reports suggest, or is what we are seeing simply a huge increase in the reporting of sexual violence?

Since there are no independent sources of cross-national data other than the State Department reports, it is impossible to answer this question definitively. GCW do not argue for or against the case that there has been a real increase in wartime sexual violence.

We argued in the Report that there are good reasons for believing that the State Department’s accounts on sexual violence for the 1980s suffered from severe under-reporting.

In the 1990s, the shocking atrocities in Bosnia and Rwanda surely meant more reporting of sexual violence than the 1980s. But other wars got much less attention. In the DRC, the “rape capital of the world,” for example, the first major report on wartime sexual violence did not appear until 2002.

We focused on the data for the 2000s because by that time reporting on wartime sexual violence had grown substantially relative to the previous two decades as the issue became more politically visible in the international community.

State Department reports for the 2000s are not only more likely to reflect the reality of conflict-related sexual violence than the data from the 1980s, and likely the 1990s, but they should also be sufficiently comprehensive to reveal core facts, about the variance of levels of sexual violence between countries and between armed groups. Both we and GCW agree that these and other findings are critically important.

The Association Between Battle Deaths and Conflict-Related Sexual Violence

Our Report argued that there was likely an association between the deadliness of armed conflicts and the level of conflict-related sexual violence and that since the end of the Cold War the latter had likely declined along with the former. We never claimed the relationship was linear of course — we could hardly have done so, since we accept the findings of Dara Cohen, Ragnild Nordås and Elisabeth Wood on the large degree of cross-national and intranational variation in the propensity of combatants to perpetrate sexual violence.

GCW point to cases where there is clearly no association between deaths from organized violence and levels of conflict-related sexual violence; it is also possible to point to other cases — like Rwanda — where there clearly is a strong association. But small numbers of individual case studies can’t reveal average effects. Regression analysis of cross-national data can, however, and since the Report was published, Dara Cohen has passed on the results of the regressions she ran on her data for the 2000s. These indicate that there is a positive, though weak association between the incidence of sexual violence and the number of battle deaths. The relationship is not statistically significant however.

Since the regressions were for the period in which we believe the State Department reports are most likely to be accurate we certainly accept this finding. The regression results don’t include sexual violence data for post-conflict years, however, so they do not challenge the core of our argument, namely that conflict-related sexual violence is likely to be reduced substantially after wars end.

Is Ending Wars the Key?

The Report claims that, “since the end of the Cold War, the number and deadliness of armed conflicts has decreased substantially, and with it — we assume — the incidence of conflict-related sexual violence.”

But how do we reconcile this claim with the Cohen data for the 1990s that clearly show an increase in the combined percentage of countries affected by the two highest categories of conflict-related sexual violence during periods of warfare?

Under-reporting may be part of the answer, but the more important reason is that the Cohen dataset, as is, cannot test the proposition that ending wars reduces the incidence of conflict-related sexual violence since it does not collect data on sexual violence in the post-conflict period.

We argued in the Report that it was logical to assume that when conflicts end, conflict-related sexual violence would also end—or at least decline substantially.

When wars end, combatants are rarely demobilized immediately and may continue to perpetrate sexual violence for some time after the fighting has stopped.

With rebel groups, militias and (some) government forces demobilized, large percentages of individuals reintegrated into the their home communities, and with rape camps closed, it would be surprising if, on average, the incidence of conflict-related rape did not decline substantially within a few years.

Indeed it is difficult to think of any arguments to the contrary — i.e., to make the case that three years after a conflict had ended there would have been, on average, an increase in conflict-related sexual violence.

If it is correct that levels of conflict-related sexual violence tend decline after wars end this would explain why any such a decline could not be revealed by the Cohen dataset.

GCW don’t believe that the declinist thesis is the most important one for researchers and policy makers to focus on. We don’t disagree, and our discussion of the possibility that conflict-related sexual violence will decline when conflicts stop, takes up only a very small part of the Report.

We also agree that global trends are of no interest to those directly affected by rising sexual violence in a particular country. But this doesn’t mean that global trends have no policy relevance. If conflict-related sexual violence is increasing or decreasing worldwide, it is important to understand why.

And if it is indeed the case that stopping conflicts has the indirect consequence of stopping or reducing conflict-related sexual violence, then what the UN calls peacemaking (stopping ongoing wars) and post-conflict peacebuilding (preventing stopped wars from restarting) may be important strategies for preventing conflict-related sexual violence.

There is considerable evidence that peacemaking and peacebuilding, while inefficient, are also effective. There are no cross-national data to the best of our knowledge that indicate that any of the current international strategies for reducing war-related sexual violence are effective.

In the 2000s, the data from the Cohen dataset show that the share of conflicts with the two highest categories of conflict-related sexual violence decreased appreciably during periods of warfare. This indicates that something must be reducing sexual violence. Finding out what and why, would seem to be important knowledge for policymakers to have.

¹ We note that—contrary to what GCW suggest—we also report the shares of country years for all other intensity levels (see pp. 25-26 of the Report).

² We thank Dara Cohen for providing these data.