Modeling Political Violence Risk in an Increasingly Unstable World

October 9, 2026 / 4 min read

World events have raised the profile of political violence (PV) in insurance. The land area exposed to conflict globally has roughly doubled between 2021 and 2026, with the conflicts in Ukraine and the Middle East the main areas of concern for the insurance sector. Attacks on commercial shipping in the Strait of Hormuz, the Red Sea, and the Black Sea have reached unprecedented levels. While war has been the primary concern recently, insured losses from strikes, riots, and civil commotion (SRCC) have been among the most significant in the past decade, outstripping terror losses by ten to one.

Insurers can no longer assess the peril of political violence and terror by judgment alone. The critical question is whether it can be modeled rigorously and what that looks like in practice.

outside-protest

Predicting war before it starts

Verisk's Predictive War model, launched this year, is a machine learning model that covers every country. It estimates the probability of a war — defined as 1,000 or more deaths per year, a threshold highly correlated with battle damage and therefore insurable loss — occurring within a country's borders in the following 12 months. The model uses proprietary measures of each country’s characteristics to assess the probability of future conflict.

Alongside it sits an AI-powered Geopolitical Relations index. This tool measures the diplomatic relationship between every pair of countries by the nature of their diplomatic exchanges, their military and territorial disputes, and their capacity to project power. It’s not predictive; rather, it provides a structured, consistent view of current conditions and a foundation to analyze potential scenarios.

Before hostilities escalated in Iran at the beginning of this year, Verisk’s model had assigned Iran a 67% probability of war.

SRCC: From global foresight to a US catastrophe model

While war may be occupying insurers' minds currently, the $10 billion of SRCC losses since 2019 should not be forgotten.

Verisk’s Global SRCC Predictive Scores estimate future SRCC severity across about 50,000 administrative areas worldwide, using a dozen predictors selected from roughly 200 candidates. Scores are validated against Verisk’s PCS^® loss data and historical global SRCC events, and updated quarterly as conditions evolve. For transparency, every score is accompanied by its underlying drivers and frequently a supporting narrative, so users can explain each location’s score. The model is trained to predict the types of civil unrest that cause insurable losses. Only about 1% of civil unrest events are SRCC events.

For the United States, Verisk has developed a full probabilistic SRCC catastrophe model with frequency and severity resolved to the ZIP code level, and focused on commercial and municipal property. By perturbing the inputs of the predictive scores, the model generates 500,000 potential futures containing some 1.9 million SRCC events, a classic catastrophe model event catalog. Combined with damage curves and the financial engine that accounts for exposure value and policy terms, it is a powerful tool for estimating losses from SRCC events in the United States.

From data to decisions: three workflows

Three examples illustrate how clients apply the capabilities underlying Verisk’s scores:

Sharper accumulation management. The conventional ring-accumulation approach orders exposure by exposed limit. Scoring those same exposures against SRCC or war hazard transforms the picture as both the exposure and hazard are being assessed. This approach surfaces locations that genuinely warrant attention and would otherwise be missed.
Faster, more consistent underwriting. Exposure and hazard categories can be combined into structured rules, with automatic approval for low-risk combinations, automatic decline for the highest, and human review for cases in between. This allows underwriters to concentrate their judgment where it’s needed most while achieving consistency across portfolios, geographies, and time.
Insights that exposure data alone can’t produce. Clients can model losses across return periods with full policy terms applied, re-examine historical events, run realistic and extreme disaster scenarios, and conduct counterfactual analysis of live events, as Verisk did for the 2025 Los Angeles protests, assessing potential losses if the observed protest footprint had turned violent.

Complementing the modeling and scoring capabilities, McKenzie Intelligence Services, recently acquired by Verisk, provides high-resolution, real-time event monitoring from damage estimates following the 2024 riots in France’s New Caledonia territory to live tracking of 2026 strikes on shipping and land-based assets in the Middle East. Clients get a real-time view of how unfolding events affect their exposures.

The road ahead

There’s more to come: the predictive war model is being refined to subnational levels, and the SRCC catastrophe model will be added to Verisk Synergy Studio in 2027.

Political violence is volatile, human, and analytically demanding, but it’s quantifiable. With validated, transparent, forward-looking models spanning war, conflict, and civil unrest, insurers can bring the same rigor to PVT that they bring to natural catastrophes: consistent underwriting, hazard-aware accumulation management, and a genuine view of the tail.

Reach out to a Verisk representative to learn more about SRCC modeling that goes beyond the data.

Learn more

Modeling Unrest: Quantifying Political Violence as Instability Grows

Predicting war before it starts

SRCC: From global foresight to a US catastrophe model

From data to decisions: three workflows

The road ahead