Kidnap for Ransom (K4R) Observatory

Scope. Built from articles collected on kidnapping, abduction and hijacking keywords. Ransom was not a collection term. Behaviours are derived inductively from what reports describe; ransom appears where the evidence puts it, not as an organising frame.

This monitor reconstructs a decade of kidnapping in Nigeria from the media record — 1,748 incidents drawn from 13,191 newspaper and broadcast reports between 2014 and 2026. It reads the crime three ways: as geography and tempo, as a human toll, and as a repeatable sequence of behaviours. Every figure describes what was reported, not the true incidence, which is higher and unevenly covered.

Key points

1,748 media-reported incidents across all 36 states and the FCT, 2014–2026.
Killing is the most frequently reported behaviour (43.4% of reports), ahead of release (40.5%); an explicit ransom demand appears in 17.5%.
Use the navigation bar to move between geography, tempo, people, anatomy, coverage, behaviour and infrastructure.

The story in five findings — read across, then explore each in depth

The scale

A decade-long rise; killing appears in 43.4% of reports, an explicit ransom demand in 17.5%.

The geography

Incidents cluster on highways and across the northwest; reporting skews south.

The people

Single victims dominate by count, but mass abductions drive the human toll.

The sequence

A repeatable script: seizure into forest, then an economic exit — or a lethal one.

The gap

Schools are 0.5 km away; police 24.5 km. Deaths rise where the state is farthest.

Key indicators

Rates across reported incidents

Kidnapping is not evenly spread. It concentrates along highways and in the northwest, and the map below plots every geocoded incident against the states that carry them. Because the record is media-derived, the pattern also reflects where reporting is densest — a caveat that matters most when comparing North and South.

Key points

Incidents cluster on roads and highways and across the northwest banditry belt.
The South shows more reported incidents, partly a media-density effect; the North carries more mass, community-targeted abductions.
Click any state to filter the map — the filter also reshapes the Anatomy charts.

Where it happens

Geocoded incident points

Filter

Incident map

Each dot is a geocoded incident; rings scale to a state's count. Click a state to filter.

outcomereleaseddeceasedunknown

States by incident count

Top 15. Click a bar to filter.

The crisis has a rhythm. Reported incidents rose sharply across the decade, cluster in particular months, and rise and fall with the intensity of news coverage. This page separates the underlying trend from the reporting signal.

Key points

Reported NKM grew steeply across the decade, with peaks in 2021 and 2024–25.
Incidents show seasonal structure, concentrated in particular months.
Coverage volume tracks — and at times amplifies — the incident count.

The rhythm of the crisis

Volume · region · season

Coverage corrected. An earlier version of this chart showed sharp collapses in 2020 and 2022. Those were gaps in the news scrape, not falls in kidnapping. The series here is rebuilt from the complete raw corpus of 79,084 scraped articles (21,605 kidnap-relevant, linked into 18,550 events), which restores 2020 and 2022 to plausible levels. Volume still reflects media attention, not verified incidence.

Reporting volume by month

Kidnap-relevant articles per month, from the complete raw corpus.

North vs South, by year

Share of coverage. The northward tilt is the clearest directional signal.

Seasonality — month of year

Articles by calendar month, all years combined.

Who carries the story — news sources

Articles by outlet (top 14).

Behind the counts are victims. Most reported incidents involve a single captive, but mass abductions — of schoolchildren, worshippers, travellers — drive the human toll. Gender is inferred from the language of each report rather than from named individuals, so it is read as a group-level pattern.

Key points

Single-victim incidents dominate by count, but mass abductions account for a large share of people taken.
Incidents are predominantly coded male, with a substantial mixed-gender share reflecting community raids.
Gender here is an inferred distribution, not a determination about any individual.

Who, and how many

Gender · victims per incident

Gender profile of incidents

Inferred from the language of reports, not from stated gender. Explicit gender appears in only 152 of 1,748 incidents (8.7%); this inference covers all of them but leans masculine where reports carry no gender marker. Name-based inference on the same incidents gives 71–74% male against 88.7% here, so read this chart as an upper bound on the male share.

Victims per incident

Single victims dominate, but mass abductions drive the human toll.

How does an incident end, and where does it begin? Most kidnappings are economic and resolve through a negotiated release; a minority end in death. The setting is overwhelmingly opportunistic — roads, schools, farms, homes — places easy to reach and hard to defend.

Key points

Of resolved incidents, about 81% end in a release and 19% in a reported death.
The most common settings are schools, towns and roads.
Use the state filter on the Geography page to see these break down regionally.

Anatomy of an incident

Outcome · setting

How incidents end

Release status where reported.

Where victims are taken

Abduction setting nearest the event.

Not every incident receives equal attention. This page treats the corpus as a record of attention — how many outlets carried an incident (spread), how long coverage persisted (persistence), and how much was written (volume). The events that held national attention stand apart from the everyday churn.

Key points

A small number of incidents achieve wide spread and long persistence — the ones that held national attention.
Most incidents are covered briefly, by few outlets.
The state comparison contrasts scale (how many incidents) against lethality (how often fatal).

The shape of coverage

Volume · spread · persistence

Volume, spread and persistence of reporting

Each bubble is an incident thread. X = persistence (days from first to last article); Y = spread (distinct outlets); size = volume (articles). Threads that persist and spread widely are the events that held national attention.

State comparison

Scale vs lethality

States by scale and outcome

X = incidents recorded; Y = share ending in death; bubble size = median ransom demanded; colour = region. Upper-right states are both frequent and lethal.

Beneath the statistics is a repeatable script. Using Behaviour Sequence Analysis, each report is decomposed into coded behaviours and ordered by a canonical model of how a kidnapping unfolds, so the diagram shows which behaviour reliably follows which. Two resolution pathways dominate: a negotiated economic exchange, and a lethal one.

Key points

The opening is perpetrator-driven: arrival, gunfire, seizure, transport into forest.
Resolution splits into an economic chain (demand → negotiate → pay → release) and a lethal chain (shot → killed → body found → police informed).
The codebook is editable — regroup behaviours and re-render the thematic rings.

Thematic structure

Behaviours by phase of the event

Chronology of behaviours

Inner ring = phase of the incident; outer ring = behaviours coded within it, sized by how often they appear across reports. Hover for detail.

Behaviour codebook — edit & regroup

Rename phases, move a behaviour to a different phase, or drop it. Apply to re-render the rings.

The behavioural sequence

Lag-one state transition diagram

State transition diagram — how a kidnapping unfolds

Chronology runs left to right. Arrows are antecedent → sequitur; thickness follows the standardised-residual key. Grey nodes are perpetrator behaviours; pale nodes are victim or state responses. Derived from — media-reported incidents (canonical event ordering).

SR ≥ 6

Where an incident happens relative to infrastructure shapes its outcome. This page overlays incidents on GRID3 Nigeria's official layers — police, health, schools, water, roads, settlements and LGA socioeconomic vulnerability — to reveal a stark infrastructure gap and its association with lethality.

Key points

Social infrastructure is ubiquitous (schools 0.5 km, health 0.7 km) but policing is sparse (24.5 km, OSM primary).
Lethal incidents sit farther from police and in more socioeconomically vulnerable LGAs.
Kidnapping is roadside: a median of 1 km to the nearest major highway.

Proximity to infrastructure

GRID3 Nigeria layers

Police stations from OpenStreetMap (—, used as the indicative primary because it is denser near incidents); health facilities (—), schools (—), water points (—) and the LGA socioeconomic vulnerability score are from GRID3 Nigeria. The GRID3 official police layer is compared in the addendum below. Incidents sit at geocoded town centroids, so distances are indicative — read the pattern, not the metre.

The infrastructure gap

Median distance from an incident to the nearest facility of each type. Social infrastructure is ubiquitous; security infrastructure is not.

Filter incidents

Incidents against the police network

Crosses are police stations; dots are incidents, shaded by distance to the nearest station.

How far is help?

Distance to police, by outcome

Median km to nearest station.

Distance to police — distribution

Incidents by distance to the nearest police station (km).

States by policing distance

Median distance to nearest station, states with ≥8 incidents.

Addendum · GRID3 official police layer

Incidents by LGA vulnerability band

GRID3 socioeconomic vulnerability score of each incident's LGA (1–5).

Lethality by vulnerability band

Share of resolved incidents ending in a reported death, by LGA vulnerability band.

Roadside exposure

Distance to the nearest trunk or primary road. Kidnapping is a roadside phenomenon.

Settlement density & lethality

Buildings within 5 km of each incident (GRID3 settlement extents), and death rate by built-up band.

Explore the geocoded incident record directly. Sort any column, search and filter, page through the results, and download whatever view you build.

Incident explorer

sortable · searchable

1,140 geocoded incidents shown. The full 1,748-record dataset (44 fields) is on the Data page or via the API.

Every dataset behind this monitor is downloadable. Aggregate tables and the geocoded incident points are generated in your browser as CSV, Excel or JSON; the full standardised incident record is in the open data package, or straight from the API when connected.

Download the data

CSV · XLSX · JSON

Licensed CC-BY-4.0. Media-visible incidents only; not a census. Cite: African Leadership Centre, KCL (2026), NKM-NGA v1.0.0.