The table below includes countries with the highest predicted risk of conflict, i.e. predicted probability of ‘either’ conflict over 0.300 (or where conflict occurs in over 30 percent of the simulations), and/or observed conflict in 2012. These predictions are from the article Predicting Armed Conflict, 2010-2050, and data on observed conflict are from the UCDP/PRIO Armed Conflict Dataset (version 4, 2013).
Highest predicted risk of conflict and/or observed conflict in 2012
|Country||Predicted conflict||Observed conflict||Classification|
|Congo, Dem. Rep.||0.350||0.251||0.600||Minor||TP|
|Central African Republic||0.273||0.167||0.439||Minor||TP|
Conflict is defined in accordance with the UCDP/PRIO dataset as contested incompatibility between a government and an organized opposition group causing at least 25 battle-related deaths during a calendar year (Themner and Wallensteen 2011). The three conflict levels are ‘no conflict’ or less than 25 battle-related deaths reported in a year; ‘minor conflict’ or between 25 and 999 battle-related deaths per year; and ‘major conflict’ with more than 1,000 battle-related deaths per year.
The classification column at the right of the table shows the relation between predicted and observed conflict in 2012. True positive (TP) implies that the country had observed conflict in 2012 and was also expected to have conflict according to our estimates (with predicted probability of ‘either’ >= 0.300). False positive (FP) means that the country did not have conflict in 2012, although it was predicted to (‘either’ >= 0.300). True negative (TN) implies that the country was neither expected to have conflict or had conflict (‘either’ < 0.300). False negative (FN) means that the country had observed conflict in 2012, but was not predicted to (‘either’ < 0.300).
The classification table below summarizes our predictions for 2012 compared to the observed conflict situation. It includes the countries in the table above as well as the countries classified as ‘true negatives’ (TN). Conflicts are considered to be predicted if conflict occurs in more than 30 percent of the simulations, i.e. where predicted probability is over 0.300.
Classification of predicted and observed conflicts in 2012
|Observed conflict||Observed non-conflict|