Coding of fatalities in Gaza since 7 October 2023

Challenges in coding fatalities in Gaza since 7 October 2023

In ACLED’s disaggregated event-based dataset, fatalities are recorded when they can be attributed to specific events, based on the information reported (for more, see this ACLED Knowledge Base Fatalities Q&A article). Journalists and aid organizations have had great difficulty reporting from Gaza since 7 October 2023.¹ This, coupled with the intensity of the conflict, means that information on fatalities attributed to specific incidents is limited and incomplete. The Israeli Defense Forces (IDF) report aggregate fatality estimates for certain operations. In addition, fatalities are reported by the Palestinian Ministry of Health (MOH) as an aggregate number for the entirety of Gaza for a period spanning a day to several days. Regional experts, international human rights organizations,² and journalists³ have regarded the MOH as a reliable source on civilian fatality counts in Gaza. 

In order to adapt to this information environment and avoid significant undercounting of fatalities in Gaza, ACLED decided to incorporate aggregate fatality counts reported by IDF and MOH into the dataset.

ACLED does not typically use fatality numbers based on aggregated statistics, such as fatality records from morgues or hospitals, as these records include “indirect deaths” — which ACLED does not code — that happen as an indirect effect or consequence of the conflict (e.g., famine, disease, or structural lack of access to health care), but not directly due to armed violence.⁴ However, the MOH directly attributes fatalities reported for Gaza to “Israeli aggression.”⁵ Another issue with aggregate fatality counts is the deconfliction with fatalities reported to specific events that are also collected by ACLED. 

To prevent double-counting of fatalities, ACLED needs to be able to deduce which fatalities it already counted and to which events these likely belong before it adds (the remainder of) the aggregated fatality number to its dataset. For IDF-reported fatalities this means that ACLED codes claims that are specific enough to do so, while ignoring broad statements such as “1,000 Hamas operatives have been killed since 7 October.” 

For example, the report On 18 Dec, IDF claimed to have killed more than 600 combatants in operations in Shujaiyya over the past two weeks is processed as follows: The total of fatalities for this time period and location based on individual (non-aggregated) reports is established and subtracted from the IDF claim of 600. The resulting number of fatalities are distributed equally across the 16 “Armed Clash” and “Explosions/Remote violence” events between IDF and Hamas militants coded with “Casualties unknown” for that location and time period. The distribution of such fatalities is mentioned in the Notes for transparency and identification.⁶ 

The MOH does not distinguish between civilian and combatant fatalities, and as such there is uncertainty over the fatality number breakdown between civilians and Palestinian armed group combatants. Additionally, the MOH list has included some fatalities that may be attributable to Palestinian armed groups or are undetermined.⁷ Conversely, the IDF tends to report fatalities as Hamas operatives, but it is unclear to what extent these figures take into account the distinction between nearby civilians, people working with Hamas, and Hamas combatants.⁸ ACLED is not able to verify such claims on the ground and thus relies on the information as reported by these sources. Because of ACLED’s approach to attribution, the number of Hamas combatants killed will be closer to IDF claims than MOH claims.

Fatalities attributed to specific events account for the vast majority of fatalities ACLED records for armed clashes and a portion of fatalities recorded for remote violence. Based on the information ACLED has collected on these fatalities, we deduce that the remainder of unattributable IDF- and MOH-reported fatalities are related to events for which an unspecified, but likely high, number of fatalities was reported. The overwhelming majority of these events are “Air/drone strike” and “Shelling/artillery/missile attack” events in urban areas of Gaza. This includes any errant missiles fired by Palestinian groups from inside Gaza. 

Because the IDF reports are more specific than those of the MOH, allowing for better deconfliction with already recorded events, ACLED processes IDF-claimed fatalities first. MOH-claimed fatalities are included as far as unattributed fatalities remain. This approach makes it relatively more likely that IDF claims of Hamas combatant fatalities are recorded compared to treating all unattributed fatalities as civilians (as would be the case if they were sourced from MOH).⁹ This approach also makes it impossible for ACLED-recorded fatalities to drop below the MOH estimate. However, because reporting from non-MOH sources gradually became more comprehensive over 2024, ACLED fatality figures pre-MOH inclusion have been close to, or exceeding, MOH estimates since May 2024, making the contribution of the MOH less significant over time. While ACLED’s approach to fatality estimates has the shortcomings and biases listed above, we believe it to be the best approach given the limited information available (for more, see the ACLED Knowledge Base - Fatalities Q&A article and the question What are the known biases of fatality numbers?).

How are MOH fatality counts for Gaza incorporated in ACLED data?

This section explains how MOH fatality counts for Gaza were incorporated into ACLED data for the period of 7 October to 22 November 2023. Fatality counts from more recent dates are incorporated into the data following the same methodology.

1. ACLED compiled MOH press releases from 7 October to 22 November that included information on fatality numbers. The MOH published these press releases at irregular intervals; i.e., the first press release was published on 7 October, and subsequent reports were released on 11 October, 12 October, 15 October, and so on.
   • The total number of fatalities reported by the MOH from 7 October to 22 November is 14,854, which is higher than the total number of fatalities ACLED recorded for specific events for the same period based on other sources.
2. To incorporate the fatalities reported by the MOH into ACLED data, the difference in fatalities attributed to specific events that ACLED recorded between the dates of subsequent MOH press releases was subtracted from the difference in fatalities reported by the MOH over the same period.¹⁰ The remaining fatalities were then evenly redistributed across events recorded between the day of the last report and the day before the most recent report where an unspecified but likely high (i.e., over 10) number of fatalities was reported. This is expressed by the following formula:

P = date of preceding MOH report
Q = date of next available MOH report
R = date of next available MOH report − 1 day
N = number of events recorded between P and R with unspecified, but likely high (i.e. over 10), death tolls

• For example, as of the MOH’s 3 November 2023 press release, the death toll in Gaza since 7 October was 9,257, an increase of 732 fatalities from the previous press release on 31 October. Over the same period (i.e., from 31 October to 2 November, both days inclusive), ACLED recorded a total of 694 fatalities attributed to specific events. The difference between the increase in fatalities reported by the MOH and the increase in the reported fatalities ACLED recorded over the same period, 38, was then evenly distributed across all events from the same period where an unspecified but likely high number of fatalities had been reported.

3. Events in which these fatalities were redistributed have a standard explanation in the notes in square brackets, as follows: 
   On 1 November 2023, Israeli military airplanes struck targets in Gaza City (Gaza City, Gaza Strip). Several persons were killed and injured. [Palestinian Ministry of Health reported 9257 fatalities up to 2 November 2023. 38 fatalities not attributed to specific events split across 8 events between 30 October - 2 November. 5 fatalities coded to this event.]

These changes were first applied to the data on 4 December 2023, resulting in an addition of approximately 8,000 fatalities for the period of 7 October to 22 November 2023. ACLED continues to incorporate fatality reports from the MOH in the same manner for more recent dates, and will also adjust fatality counts for specific events should more reliable information emerge.