Datasets

Reliable data and sources on security incident and casualty recording in Somalia was limited and difficult to obtain.

Among other sources, data on security incidents were drawn substantially from publicly available datasets from the organisation Armed Conflict Location and Event Data Project (ACLED).11 ACLED collects data on violent incidents in Somalia, among other countries, coding each incident with the time and place, type of violent incident, the parties involved and an estimate on the number of fatalities. It also provides relevant breakdowns per region, actor, type of incident, and period. The information is collected in a database that is accessible via registration, searchable and kept continuously up to date.

ACLED codes security incidents as follows: battles, explosions/remote violence, violence against civilians, riots, protests, and strategic developments.12 For the purpose of this report the following type of events were included in the quantitative analysis and graphs of the incidents: battles, explosions/remote violence, and violence against civilians. Additionally, ACLED codes actors involved in security incidents as follows: Actor1 is the ‘[o]ne of two main actors involved in the event (does not necessarily indicate the aggressor)’ and Actor2 is the ‘[o]ne of two main actors involved in the event (does not necessarily indicate the target or victim)’,13 while ‘[i]n most cases, an event requires two actors, noted in columns ‘Actor 1’ and ‘Actor 2’. However, the ‘Explosions/Remote violence’, ‘Riots’, ‘Protests’, and ‘Strategic developments’ event types can include one-sided events’.14

Characteristics and potential limits for COI use of ACLED data are:

1. data primarily come from secondary sources such as media reports. Lack of or under-reporting might critically affect the depiction and the assessment of the situation on the ground. In the context of Somalia, for instance, the fact that large portions of the country are under the control of Al-Shabaab severely hinders media reporting activities. Adding to this, the fact that mobile phones are not allowed in these same areas, because prohibited by the group, further signals the militant’s group complete monopoly on information;

2. ACLED’s database only provides figures of direct deaths or fatalities per event,15 not of injured persons and casualties; indirect deaths resulting from security events, which in the relevant literature range in average between 3 and 9 more deaths per fatality, are also not taken into consideration;16

3. ACLED’s database does not distinguish between civilian and non-civilian fatalities;17

4. all ACLED fatality figures provided in this EUAA COI report are estimates based on ACLED’s methodology. In ACLED’s codebook their methodology is explained;18

5. the coding of events as ‘Battles’, ‘Explosions’, or ‘Violence against civilians’ is independent from their actual impact on civilians. While acts of ‘violence against civilians’ are always co-coded as ‘civilian targeting’,19 ‘explosions’ may be or may not, depending on whether civilians were the main or only target;20 ‘battles’ instead, never are,21 despite the fact they are often about civilian targets in Somalia.22

When focusing on the involvement of specific actors within certain regions, the drafters based their analysis on all those incidents, where ACLED coded the relevant actor either as ‘Actor1’ or as ‘Actor2’. This approach aims to illustrate the general level of involvement of the respective actors in the conflict without distinguishing between Actor1 and Actor2, as these categories, according to ACLED's methodology, do not indicate any differentiation in terms of content/semantics.

In various instances, particularly at general country level, whenever other sources on security incidents were available over the reference period, ACLED’s data have been corroborated/contrasted with other data.

In terms of UNHCR PRMN displacement figures, the drafters worked with the publicly available data at UNHCR’s Operational Data Portal.23

In terms of IDP population distribution across regions, the drafters worked with the publicly available harmonised IDP figures from UNOCHA, as of October 2025.24

In terms of data on evictions, the drafters relied on data available via NRC’s Eviction Information Portal.25 NRC defines ‘forced evictions’ as the ‘removal against their will of people from their homes and/or land which they occupy, without any form of legal or other protection’.26