Mapping the long-term trajectories of political violence in Africa

This paper introduces a spatially explicit longitudinal sequence analysis of African political violence from 1997 to 2024, identifying six distinct conflict trajectories and their spatial interdependencies to advance conflict research beyond discrete transitions toward a holistic understanding of violence life cycles.

The Gist

Imagine you are a doctor trying to understand why some patients get sick, recover quickly, and others develop chronic, long-term illnesses.

Most previous studies on political violence in Africa were like doctors who only looked at fever spikes. They would say, "Oh, there was a battle here in 2010, and another one in 2015." They focused on the individual moments when violence started or stopped. But they missed the bigger picture: the entire story of the patient's health over time.

This paper, written by a team of geographers, decides to stop looking at just the fever spikes. Instead, they want to map out the entire life story of violence in different places across Africa. They call this "mapping trajectories."

Here is the breakdown of their study using simple analogies:

1. The New Tool: "Conflict Sequences"

Instead of asking, "Is there violence right now?" the authors ask, "What has happened here over the last 27 years?"

They used a massive database of violent events (like battles and explosions) from 1997 to 2024. They broke Africa into a grid of tiny squares (like a chessboard). For every square, they wrote down a "story" of what happened year by year.

• The States: They categorized each year in a square into one of five "moods":
  • Peaceful (NC): Nothing happened.
  • Low & Localized (CL): A few fights, but they were close together (like a neighborhood squabble).
  • High & Localized (CH): Intense fighting, but stuck in one spot (like a siege).
  • Low & Scattered (DL): A few fights, but spread out over a wide area.
  • High & Scattered (DH): Intense fighting, but spread out everywhere (like a wildfire).

2. The Method: "Matching the Stories"

Once they had these stories (sequences) for thousands of squares, they needed to find patterns. They used a computer technique called Optimal Matching.

Think of this like matching socks.

• If you have a sock that goes "Peace -> Fight -> Peace," and another that goes "Peace -> Fight -> Peace," they are a perfect match.
• If one goes "Peace -> Fight -> Peace" and another goes "Peace -> Fight -> Fight -> Peace," they are still similar, just a little longer.
• The computer calculated how similar every single square's story was to every other square's story, then grouped them into six distinct "Conflict Families."

3. The Six "Conflict Families" (The Results)

The authors found that violence in Africa doesn't just happen randomly; it follows six predictable "life cycles."

The "Short-Term" Families (The Colds and Coughs)

These make up about 80% of the violence. They are annoying but usually go away on their own.

• Type 1: The "Flash in the Pan" (Ephemeral):
  • Analogy: A sudden thunderstorm. It rains hard for an hour, then the sun comes out.
  • What happens: Violence starts and stops very quickly (usually within a year). It's often just a local dispute that doesn't turn into a war.
• Type 2: The "Recurring Cold" (Clustered):
  • Analogy: A cold that keeps coming back every winter.
  • What happens: Violence happens in specific spots (like a town or a road) and flares up repeatedly, but it eventually dies down. It's persistent but not permanent.
• Type 3: The "Old Scar" (Past):
  • Analogy: A place that had a bad infection years ago but is now mostly healed, with only an occasional itch.
  • What happens: These places had big wars in the past (like Liberia or Sierra Leone) but are mostly peaceful now, with only rare, small outbreaks.
• Type 4: The "Unstable Stomach Ache" (Unstable):
  • Analogy: A stomach that keeps churning, sometimes getting better, sometimes worse, but never settling.
  • What happens: Violence jumps around between different types of intensity. It's chaotic and hard to predict, but it doesn't usually last forever.

The "Long-Term" Families (The Chronic Diseases)

These are the dangerous ones. Once a place falls into these categories, it is very hard to get out.

• Type 5: The "Chronic Illness" (Enduring):
  • Analogy: A disease that has become part of the patient's daily life.
  • What happens: Violence is intense and clustered in one spot, and it stays there for 5 to 6 years or more. It's like a siege that never ends. This happens in places like Mali, Burkina Faso, and Somalia.
• Type 6: The "Terminal Condition" (Balanced):
  • Analogy: A stalemate where two armies are so evenly matched that neither can win, so they just keep fighting forever.
  • What happens: This is the most persistent type. Violence is intense, clustered, and lasts 8+ years. It happens where two sides are deadlocked, like in parts of the Democratic Republic of Congo.

4. The Big Discovery: "Contagion"

The most important finding is about neighbors.

Imagine a neighborhood where one house catches fire.

• In the "Short-Term" families, the fire might burn a little and go out, leaving the neighbors fine.
• In the "Long-Term" families, the fire spreads to the next house, then the next, creating a regional inferno.

The study found that violence doesn't just happen in isolation. If a border region has a "Chronic Illness" (Type 5), the violence often spreads across the border to the neighboring country, creating a regional complex of violence. The "life cycle" of violence in one town is deeply connected to the life cycle of the town next door.

5. Why This Matters (The "So What?")

This changes how we should try to fix these problems.

• Old Way: "There is violence here! Send troops to stop it!" (Treating every fire the same).
• New Way: "Let's look at the story."
  • If it's a Type 1 (Flash), maybe we don't need a huge army; just a little local mediation will do.
  • If it's a Type 5 (Chronic), sending troops for a week won't work. We need a 10-year plan involving regional cooperation, because the problem is too big for one country to solve alone.

Summary

This paper argues that to understand war in Africa, we can't just look at the calendar and count the battles. We have to look at the movie, not just the photos.

By watching the whole movie, they found that most violence is short-lived and local. But the violence that really matters—the stuff that lasts for decades—follows a specific, stubborn pattern that spreads across borders. To stop it, we need to understand the pattern, not just the moment.

Technical Summary

Here is a detailed technical summary of the paper "Mapping the long-term trajectories of political violence in Africa" by Radil et al.

1. Problem Statement

Existing research on political violence in Africa predominantly focuses on discrete transitions (e.g., the onset, escalation, or cessation of conflict) using regression-based models or event-history analysis. While these approaches identify correlations between violence and exogenous covariates (e.g., terrain, resources), they often fail to capture the long-term temporal and spatial trajectories of violence.

• The Gap: Current models struggle to distinguish between short-lived, irregular outbreaks and deep, structural transformations in conflict life cycles. They treat violent events as isolated transitions rather than ordered sequences of evolving states.
• The Objective: The authors aim to move beyond isolated transitions to understand the "life cycles" of violence by analyzing how sequences of violent and non-violent episodes accumulate over time and space, and how these trajectories interact across neighboring regions.

2. Methodology

The study employs a spatially explicit longitudinal sequence analysis, adapting a method originally developed in bioinformatics for social science. The analytical framework consists of four main steps:

A. Data Source

• Dataset: Armed Conflict Location and Event Dataset (ACLED), covering Africa from 1997 to 2024.
• Selection: The authors filtered for specific event types: battles, explosions/remote violence, and violence against civilians, excluding riots, protests, and strategic developments.

B. State Space Definition (The SCDi Indicator)

Using the Spatial Conflict Dynamics indicator (SCDi), the authors defined a five-class state space for each grid cell based on the intensity (high/low) and spatial concentration (clustered/dispersed) of events:

1. CH: Clustered/High-intensity
2. DH: Dispersed/High-intensity
3. CL: Clustered/Low-intensity
4. DL: Dispersed/Low-intensity
5. NC: No Conflict

C. Sequence Construction and Similarity Measurement

• Sequences: Each grid cell's history is converted into a temporal sequence of states (e.g., CH-CL-NC).
• Optimal Matching (OM): The authors used the OM algorithm to calculate the "distance" (dissimilarity) between sequences. This involves counting the minimum number of edits (insertions, deletions, substitutions) required to transform one sequence into another.
• Cost Matrix: Unlike standard OM, the authors used an empirical cost matrix derived from observed transition frequencies in the data. Substitutions between states that frequently transition into one another were assigned lower costs.
• Alignment Strategy: Insertion/Deletion costs were set lower than substitution costs. This prioritizes the similarity of motifs (sub-sequence patterns) regardless of their exact timing, allowing for the comparison of places with similar conflict rhythms even if they occurred in different years.

D. Clustering and Spatial Analysis

• Clustering: Ward's hierarchical clustering was applied to the OM distance matrix to group grid cells with similar sequences. A 6-cluster solution was selected (plus a "No Conflict" category) based on dendrogram inspection.
• Spatial Validation: The authors used multitype join count statistics (Cliff & Ord, 1981) to test for spatial autocorrelation, verifying whether specific trajectory types cluster geographically more than expected by chance.
• Metrics: The study calculated the Mean Violence Stopping Time (MVST) for each cluster to estimate the typical duration of violence before returning to a "No Conflict" state.

3. Key Results

The analysis identified six distinct conflict trajectories, categorized into two main branches: Shorter-Term (Types 1–4) and Longer-Term (Types 5–6).

A. Shorter-Term Trajectories (Types 1–4)

These account for roughly 80% of conflict-affected cells and are characterized by rapid returns to peace.

• Type 1 (Ephemeral): The most common (n=2,105). Violence emerges and dissipates quickly (MVST: 1.23 years). Often transitions from Clustered/Low (CL) to No Conflict (NC). Found on the peripheries of conflict zones.
• Type 2 (Clustered): High recurrence of intense violence (CH→CH self-transition: 60%). MVST: 3.67 years. Concentrated in active theaters like the Sahel and eastern DRC.
• Type 3 (Past): Locations with historical conflict but little recent recurrence. Dominated by brief flare-ups returning to peace.
• Type 4 (Unstable): Complex oscillations between dispersed and clustered states before resolution.

B. Longer-Term Trajectories (Types 5–6)

These represent entrenched, high-intensity conflicts that are highly resistant to resolution.

• Type 5 (Enduring): Characterized by persistent CH states (73% self-transition). MVST: ~5.9 years. Found in Mali, Burkina Faso, Lake Chad Basin, and Somalia.
• Type 6 (Balanced): The rarest but most persistent (n=166). Defined by abrupt escalation to CH and extremely low probability of de-escalation (only 8% transition to NC). MVST: 8.41 years. Common in the eastern Great Lakes region and parts of Nigeria.

C. Spatial Patterns

• Spatial Clustering: Trajectories are not randomly distributed. Shorter-term types are scattered, while Longer-Term types (5 and 6) form contiguous belts across borders.
• Borderlands as Nodes: Enduring conflicts (Types 5 and 6) heavily concentrate in transborder hotspots (Sahel, Lake Chad, Great Lakes), confirming that border regions act as long-term conduits for conflict diffusion and interdependence.

4. Key Contributions

• Theoretical: Reframes conflict not as a series of discrete events but as evolving life cycles. It demonstrates that persistence is path-dependent: once a locality enters a high-intensity clustered state (CH), it is statistically unlikely to de-escalate quickly.
• Methodological: Successfully adapts sequence analysis and optimal matching to political violence. This allows researchers to distinguish between "noise" (short irregularities) and "systemic change" (sustained cycles) and to compare local trajectories across space.
• Empirical: Provides a granular typology of African conflict, revealing that while most violence is short-lived, a specific subset of "enduring" and "balanced" conflicts drives the continent's long-term instability.

5. Significance and Policy Implications

The study argues that "one-size-fits-all" interventions are ineffective because different trajectories require different logics:

• Ephemeral Conflicts (Type 1): Heavy-handed intervention may be counterproductive; focus should be on preventing escalation.
• Unstable/Clustered Conflicts (Types 2 & 4): Require repeated stabilization efforts targeting local political economies and key infrastructure.
• Enduring/Balanced Conflicts (Types 5 & 6): These are self-reinforcing and cross-border. They require regional cooperation, integrated peacebuilding, and long-term governance reforms rather than unilateral military solutions.

Conclusion: By shifting the analytical lens from transitions to trajectories, the paper offers a framework for better early warning systems (e.g., integrating trajectory types into ViEWS) and more targeted peacebuilding strategies that account for the temporal rhythms and spatial interdependencies of violence.