Published on: 9 April 2024
On 24 March 2021, Islamist militia affiliated with Islamic State launched a coordinated attack on the city of Palma in Mozambique. Battles between the Islamist militants and security forces lasted for about two weeks before state forces regained control of the area.
In the aftermath of the attack, anecdotal reports of killings appeared, such as those at the Amarula Hotel,1https://www.bbc.com/news/world-africa-56557623 but a comprehensive fatality estimate did not materialize.2https://www.zitamar.com/over-1-190-killed-in-2021-palma-massacre-new-survey-finds/ In an effort to paint a more accurate picture of the scale of the assault, investigative journalist Alex Perry3Perry is an award-winning writer with credentials in investigative journalism. His previous work on Boko Haram’s use of beheadings was used by the International Criminal Court. For more on Perry’s work, see https://www.alex-perry.com/biography/ conducted a survey of more than 13,000 households to gather information from locals who lived through the attack.4https://www.alex-perry.com/palma-massacre/ Recognizing the value of Perry’s data on the Palma attack, ACLED has incorporated this new data into its dataset to provide a more accurate fatality estimate – bearing in mind all the general considerations on fatality estimates.
Before incorporating Perry’s survey data into the ACLED dataset, ACLED evaluated the raw data and consulted Perry on the methodology employed for his research. ACLED also engaged other experts on the context, including ACLED’s partner Zitamar News.5https://acleddata.com/about-acled/our-partners/local-data-collection-partners/#1604955233388-dfb6b5b5-736d The sections below lay out the steps and considerations ACLED undertook to incorporate Perry’s data.
Key Changes
Correction Period and Geographical Scope: The data corrections span from 24 March to 5 April 2021, the period over which the attack took place.6https://www.zitamar.com/over-1-190-killed-in-2021-palma-massacre-new-survey-finds/ Corrections were limited to events in Palma district, with the exception of one event in Macomia.
Publication Date: The corrected data were published on 8 April 2024.
Summary of changes to ACLED data after incorporating Perry’s survey data:
- Fatalities: Based on Perry’s survey, 664 additional civilian fatalities were added to ACLED’s original 137 fatalities, bringing the total fatalities of the assault to 801 fatalities (including both civilians and combatants).
- Events: A total of 16 events were updated to incorporate the new fatality information. Also, one new ‘Abduction/forced disappearance’ event was coded.
- The notes of the updated events have a standard phrase as follows (bolded):
- On 29 March 2021, members of the Mozambican Defense and Security Forces supported by Dyck Advisory Group (DAG) helicopters clashed with members of an Islamist militia in Palma (Palma, Cabo Delgado), in an attempt to retake the area. At least 42 civilian fatalities added based on the Perry survey.
- All events updated with Perry’s survey data have “Alex Perry” listed in the Source column.
- The notes of the updated events have a standard phrase as follows (bolded):
How ACLED incorporated Perry’s survey data
Perry’s survey data identified a total of 1,402 victims, including 1,193 who were killed or are missing (presumed dead), and a further 209 who were kidnapped.
- Of these, ACLED only included fatalities reported as shot or beheaded, as these were backed by either witness testimony or the identification of bodies. Additionally, these fatalities had enough information to be attributed to specific events already coded by ACLED.
- The majority of fatalities excluded from Perry’s data consists of fatalities from drowning and missing people whom Perry presumed dead. ACLED considers accidental drownings as indirect fatalities and therefore does not include these in its data. The accounts of the missing people could not be verified as fatalities or be attributed to existing ACLED events or coded as new events and were thus not included.
- Some additional fatalities were excluded due to potential double counting. ACLED identified potential duplicates by comparing information about the victims (e.g. names, age, gender, location, cause of death).
In order to attribute Perry’s fatality data to specific ACLED events, ACLED compared Perry’s data against the events already coded to specific locations within the date range of the attack. The fatalities from Perry’s data were then integrated into the nearest corresponding ACLED event. Since Perry’s data does not include information about whether or not the victims were killed as part of an Armed clash or an Attack, ACLED is not able to code new events based solely on Perry’s fatality data. Accordingly, ACLED excluded some fatalities from Perry’s survey that were reported in Pemba which ACLED was unable to attribute to the attack on Palma.
