IV.- DISEÑO METODOLOGÍA DE CÁLCULO DEL VALOR DEL AGUA DE RIEGO
5.3. Metodología de aplicación del MVC
5.3.2 Descripción de las actividades principales
The second major observable effect of truces surrounds violence against civilians. In truce, rebel organizations should lose control of their soldiers, particularly the soldiers they recruited during the truce who are attracted by material benefits and safety. This should result in more civilian abuse and extortion. Some of the subtler forms of civilian abuse may be difficult to detect, but killings of civilians are tracked fairly consistently in conflicts around the world. When rebel soldiers disobey orders with civilians, stealing and settling personal scores, they are more prone to kill civilians who get in their way. Therefore, we should see an uptick in incidents in which rebel soldiers kill civilians.
The key outcome, then, is civilian deaths from rebel one-sided violence, which should in-crease after a truce. Here again UCDP events data provides an opportunity to witness these trends, whether in or out of open conflict. I counted up the civilian deaths in any events coded as occurring between rebels and civilians (i.e. not including “collateral” civilians killed in clashes between gov-ernment and rebel forces). While this certainly results in an undercount of rebel violence against civilians — excluding non-fatal events — it is still a strong indicator of civilian abuse over time.
In Column 2 of Table 7.1, I laid out the per-year averages of civilian deaths before and after a truce. Before they were offered a truce, the 19 truce cases had very similar levels of civilian violence as other separatist conflicts, averaging about 30 civilians per year killed by rebels. Sur-prisingly, the first few years of ceasefire show a dramatic drop in civilian deaths, down to just an average of just 6 recorded civilian deaths. Within a few years, though, the trend reverses com-pletely, shooting up to an average of 174 civilians killed per year after the fifth year of truce. After
just a few years of truce, rebels kill five times as many civilians per year as they do before the truce. Put another way, before the truce these rebel groups killed seven government soldiers for every civilian they killed. More than five years after truce, they killed three civilians for every government soldier.
Table 7.3 tests these results more systematically. Rebel OSV numbers are events, so I use a zero-inflated negative binomial model to measure the frequency of violence against civilians.5 As with the rebel fragmentation data, the primary purpose of Models 1 and 2 is to add in conflict-level fixed effects.6 These fixed effects adjust for overall differences in violence between movements, allowing the comparisons to be purely over time, comparing before and after a movement. This helps to account for any selection effects that might result from particularly violent movements being overrepresented in the data either before or after a truce.
The regression results in Table 7.4 are dramatic and confirm exactly the trends visible in the raw numbers. All else equal, rebel abuse of civilians increases 60% after a truce — even as battlefield violence falls dramatically. This happens in spite of the substantial decrease in civilian abuse in the years immediately following a truce. After five years of truce, as rebel groups return to fighting and the organizational effects of truce set in, violence rises much higher. The results from Model 2 show that these long-post-truce years involve more than four times as many civilian killings as prior conflict years.
As with rebel fragmentation, I then run two additional models (Models 3 and 4) which bringing in other separatist conflicts’ over-time trends for comparison. In general, rebel movements in a conflict tend to abuse civilians more often over time. When accounting for this trend, the effect of truces on rebel violence might actually be negative overall because of the dramatic drop in
5A Poisson model is inappropriate because the outcome, government deaths, is overdispersed. This is because government deaths tend to be clumped together in years with more rebel victories.
6In this case, I do not control for battlefield violence because it might unfairly discredit truce environments. If truces reduce violence against civilians in part by reducing battlefield violence, this would be counted as a gain by the international community, and perhaps a sign of rebel strength because rebels are reducing violence against civilians with their battlefield violence. However, battlefield violence may influence whether violence is likely to be recorded at all, so I include battlefield violence as a predictor for the first stage (whether there is any violence against civilians at all).
Table 7.3: Effect of Truce on Rebel Violence Against Civilians
Dependent variable:
Rebel OSV
(1) (2) (3) (4)
After Truce 0.523∗∗∗ −0.235∗∗∗
(0.023) (0.027)
> 5 Years After Truce 1.605∗∗∗ 1.289∗∗∗
(0.022) (0.025)
Year of Conflict 0.064∗∗∗ 0.028∗∗∗
(0.001) (0.001)
Conflict FE Yes Yes Yes Yes
N (Conflict-Years) 227 227 726 726
N (Conflicts) 19 19 63 63
Note: ∗p<0.1;∗∗p<0.05;∗∗∗p<0.01
Zero-inflated negative binomial regression with conflict fixed effects. The unit of analysis is conflict-years in which there was any violent events. Models 1-2 include the 19 separatist con-flicts that experienced both pre- and post-ceasefire periods in the dataset (1989-2016), and Models 3-4 add in all other separatist conflicts.
civilian killings in the first five years (Model 3). Accounting for this trend, however, discards one potential effect of truces on civilians: truces keep rebel group alive when they might otherwise collapse. So if longer-lived rebel movements kill more civilians, truces may result in more civilian deaths simply by keeping conflicts going long enough for rebels to abuse civilians more and more over time. In in the long run, however, rebel groups in truce commit much more violence against civilians than would be expected, even when accounting for these effects (Model 4). After the first few years of truce, when there tends to be little violence, new recruits begin to abuse civilians in large enough numbers to show up in the events data. After five years in truce, rebel groups commit well over than three times as much violence against civilians per year as would be expected, even accounting for these factors.
This is generally consistent with the hypothesis (H5) that rebel organizations tend to lose con-trol of their soldiers, and therefore to abuse civilians, during and after a truce. This is not only a human tragedy but a sign of massive organizational failure. Because separatist rebels tend to operate in fairly homogeneous minority regions, they disproportionately kill civilians in their own ethnic group. Groups that kill potential civilian supporters tend to discredit themselves in the community and weaken their position vis-´a-vis the government.