Definitions

Outcomes, Levels of analysis, and the forecasting periodization used in ViEWS

Outcomes (dependent variables)
ViEWS applies a “divide and conquer’ strategy to the forecasting problem. Using the calendar month as the temporal unit, the system analyses separately three different outcomes of fatal political violence, as defined [1] and recorded by the Uppsala Conflict Data Program (UCDP):
1

State-based violence (sb)

Armed conflict between two or more actors – of which at least one is the government of a  state [2] – over a contested incompatibility [3] that concerns government and/or territory.
2

Non-state violence (ns)

The use of  armed force between two or more organised armed groups, neither of which is the  government of a  state
3

One-sided violence (os)

The deliberate use of  armed force by the  government of a  state, or by a formally organised group [4] against civilians. 
The three outcomes are assessed using two different fatality thresholds:
  • At the country level, the system generates predictions of the risk that 25 or more people will be killed per outcome in a given country and month.

  • At the sub-national level , the system forecasts the risk of at least one (1) person being killed per outcome, month, and 0.5×0.5 decimal degree location (measuring approximately 55x55km at the equator, see Levels of analysis  below). 
Notes: [1] Please note that in order for a given conflict event to be included in the final and annual UCDP-GED dataset that the ViEWS forecasts are evaluated against, the conflict dyad at hand must have resulted in at least 25 battle-related deaths over the course of the concerned calendar year. This criterion is not applied in the UCDP-Candidate data that informs the ViEWS forecasts on a monthly basis in addition to the GED data, and has therefore been excluded from the outcome definitions above. When evaluating the forecasts against GED data, it is nevertheless implicitly applied. More about the UCDP-Candidate dataset, and its differences from the GED data, can be found in the 2021 presentation article of the UCDP-Candidate dataset.  

[2] In line with UCDP coding procedures, the government of a state is defined as the party controlling the capital of the state, whether or not the party is the de jure holder of power.

[3] An incompatibility is by the UCDP defined as a stated challenge over the governmental power or over a specified territory.

[4] Armed groups are here defined as any non-governmental group of people that have announced a name for their group and that uses. 
Levels of analysis
The forecasts for political violence are presented at two levels of analysis: the country-month level and sub-national PRIO-GRID-month level. 
Temporal unit: Calendar months 
Geographic scope: Africa

The country-month level (25+ BRDs)

The country-month level uses the calendar month as the temporal unit of analysis and countries as the spatial unit. Within the geographic scope of Africa, the set of countries used is defined by the Gleditsch-Ward country code, while the geographic extent of the countries is determined by the latest version of CShapes. Please note that the choice of country set and delimitations thereof was made on methodological grounds and does not reflect the views or opinions of the ViEWS team. 

The PRIO-GRID-month level (1+ BRD)

At the PRIO-GRID-month (pgm) level, the spatial units are derived from the PRIO-GRID, a standardized spatial grid structure consisting of quadratic grid cells that jointly cover all areas of the world at a resolution of 0.5 x 0.5 decimal degrees, approximately 55×55 km around the equator. Calendar months are used as the temporal unit.

The actor level (forthcoming)

An actor level is currently under development.
Forecasting periodization
In order to train and calibrate the forecasting models used in ViEWS, all available data are split into two sets of data partitions. One is used for testing the ViEWS models, and the other for true forecasting.
This periodization, as well as the evaluation and true forecasting procedures, are described in detail in the  Data setup and notation online appendix to  ViEWS’ 2021 Special Data Feature in Journal of Peace Research. 
Data partitioning in ViEWS. Source: Hegre et al. ViEWS2020: Revising and evaluating the ViEWS political Violence Early-Warning System. Journal of Peace Research. 2021; 58(3):599-611.
Source: Hegre et al. ViEWS2020: Revising and evaluating the ViEWS political Violence Early-Warning System. Journal of Peace Research. 2021; 58(3):599-611.
The evaluation periodization is used to test the models and ensembles. In this periodization, the available data are split into three partitions: a training, calibration, and testing period. The periodization as a whole runs from the first month of available UCDP-GED data on the outcomes of interest (January 1990) up until and including the last month of data in the most recent UCDP-GED release. If the last release covers the year of 2018, the evaluation periodization ends in December 2018. If the last release covers 2019, it runs up until and including December 2019.  In the forecasting periodization, used for the true/actual forecasting, the data are split into four partitions: a training, calibration, ‘predictor updating’, and forecasting period. The 36-month calibration period uses the same cut-off date as the end of the evaluation periodization, i.e., the last month of available UCDP-GED data. The calibration period is here followed by a ‘predictor updating period’. In between releases of the annual GED-data, this period runs from the last month of available UCDP-GED data up until and including the month of the last UCDP-Candidate data release. During this time, the ViEWS system is informed by the latter as a monthly substitute for the annual UCDP-GED data. One month is consequently added to the ‘predictor updating period’ each month until a new release of UCDP-GED data is made available, upon which both the evaluation and forecasting periodizations are updated accordingly. The ‘predictor updating period’ is followed by the rolling 36-month period for which the ViEWS system generates its monthly forecasts. 
The table above shows an example of the two periodizations as of 1 January 2020. At this point in time, the last UCDP-Candidate data release was that of December 2019 (covering the month of November 2019), whereas the last UCDP-GED release covered the year of 2018. The table is taken from the ViEWS’ 2021 Special Data Feature in Journal of Peace Research, in which both of these periodizations, and the forecasting procedures based on them, are described in detail. See the article, and the  Data setup and notation online appendix (A) for further details. 

Please note that as each annual UCDP-GED data is released, the periodizations above shift by one year. 
Notations:
  • Calendar time:  τ 

  • month_id: Internally and in any documentation of the ViEWS system, we refer to individual calendar month as a numeric identifier determined by a counter that started on 1 in January 1980. December 2019 is consequently identified as month_id 480. 

  • Subscripts:  The subscripts used in both the evaluation and forecast columns describe the concerned partition (the training, calibration, and the forecasting procedures based on them, are described in detail. See the article, and the Data setup and notation online appendix (A) for further details.