This work provided the inspiration for UNSFIR, which developed two national datasets for Indonesian conflict; the second dataset (UNSFIR-2), which improved on the first by using provincial as well as national news sources, estimated 10,700 deaths in the 1990-2003 period (Varshney, Tadjoeddin, and Panggabean 2008).
There are also indications that the costs of conflict were higher than the UNSFIR data suggested.
Third, the field-based studies suggested that the full spectrum of significant violent conflict forms were not included in the UNSFIR dataset, in large part because of its exclusive concentation on incidents of collective violence, excluding violence between individuals.
This means that we cannot answer some of the questions the UNSFIR datasets address related to temporal variation.
The Scope and Scale of Violence: Comparing the UNSFIR and KDP/CCN Datasets
In order to compare our data with those of UNSFIR it is necessary to ensure that the definition of what is included and what is not is consistent across datasets.
Indeed, further comparative analysis of the data reveals the extent to which the number of deaths reported in UNSFIR is significantly lower than the number our method revealed.
The UNSFIR dataset severely underestimates the impacts associated with this type of violence.
Yet our data show significant impacts from collective violence in districts that UNSFIR did not find to be particularly conflict-prone, suggesting that violence is distributed more widely across Indonesia.
UNSFIR focuses primarily on those disputes in the bottom right-hand box--communal clashes--and to a lesser extent on those in the bottom left-hand box.
This case is presumably included in the UNSFIR dataset, because the line of causality is fairly clear.
If we include such cases, we find over six times as many conflict-related deaths as does UNSFIR (Table 2).