Volatility & Risk Predictability Index

Overview

The Volatility & Risk Predictability Index supports early warning and risk management by providing practical information for monitoring conflict environments. The index tracks positive deviations (increases) from baseline violence levels to assess the frequency and intensity of conflict surges. Rather than predict conflict rates far into the future with limited applicability, the index evaluates the stability and frequency of the patterns of high and low violence rates in specific areas. The watchlist notes which first-order administrative divisions (e.g. states, provinces, governorates, etc.) are at ‘extreme risk’ of violence. These regions are of greatest concern. Violence rates are calculated using the number of battle events, explosion/remote violence events, violence against civilians events, and riot events, as well as the excessive force against protesters sub-event type of the protest event type, excluding peaceful protest events and protest with intervention events.

Using the Volatility & Risk Predictability Index

This index quantifies the predictability and stability of the overall conflict rate for each first-order administrative division in the ACLED dataset. Conflict patterns may be relatively consistent in areas with high levels of violence, and programming and planning can adapt to address stable conflict trends. Other areas may have comparably high violence levels yet unstable patterns, with intermittent conflict spikes in already violent spaces. Still other areas may register low baseline violence rates yet experience frequent intense bouts of conflict. In these cases, programming must account for the instability of conflict patterns, and the likelihood for the political environment to quickly shift. 

The interactive dashboard tracks the volatility of conflict by first-order administrative division, relative to a baseline level of conflict, in order to assess a basic risk level, outlined in the table below. These levels indicate the likelihood of a future spike in violence in the given first-order administrative division.

Volatility/Baseline	Low baseline (<1.5 events/wk)	High baseline (>1.5 events/wk)
Low volatility (fewer than 6 weeks out of the past year see violence spiking 2 or more standard deviations above the ADM1 baseline)	Low risk	Consistent risk
High volatility (6 or more weeks out of the past year, violence spiked 2 or more standard deviations above the ADM1 baseline)	Growing risk	Extreme risk

A baseline level of violence is calculated from the average number of violent events per week, based on the last three years. A baseline level of violence of 1.5 events per week on average or higher is a ‘high baseline’ first-order administrative division, in which violence is common (right-hand cells in table above).

Volatility is calculated based on the number of times the number of violent events in a given week surpasses two standard deviations over the baseline level of violence (how often spikes in violence are reported). Violence within two standard deviations over the baseline level of violence in a given week is within expected bounds; violence over two standard deviations over the baseline level of violence in a given week, however, represents an anomaly (a spike in violence). When six or more weeks in a given year (or over 11.5% of weeks in a year) register such volatility (violence more than two standard deviations over the baseline), that first-order administrative division is classified as one with ‘high volatility.’

The graph in the dashboard depicts the number of events each week over the course of the prior year for a particular first-order administrative division (navy line on graph); the left-hand Y-axis represents the number of events each week, while the right-hand Y-axis denotes how many standard deviations from the mean that event count represents. The orange line on the graph represents the baseline level of violence in that first-order administrative division.

Places that register a high baseline of violence, where violence is common, and are highly volatile as violent spikes occur more frequently, are areas of ‘extreme risk’. Those that register a low baseline of violence, yet have highly volatile and common spikes, are classified as areas of ‘growing risk.’ These two contexts are of greatest concern for early warning, especially the former (which are the first-order administrative divisions that are flagged on the watchlist each week).

Areas that register a high baseline of violence, but spikes are not common and so conflict is not volatile, are classified as ‘consistent risk’. Those that register a low baseline of violence and where conflict is not volatile are classified as areas of ‘low risk.’  How can you assess whether a volatile subnational region is located in a country where overall violence levels are rising? Use the Conflict Change Map to better understand in what countries violence has recently spiked.