A recent paper, “Traumatic injury mortality in the Gaza Strip” in the Lancet, a top medical journal, estimates that 64,260 Palestinians were killed by traumatic injury from October 7, 2023 to June 30, 2024. That’s considerably more than the 37,877 figure cited by the Hamas-controlled Ministry of Health for the same period.
I believe the authors’ claim could be true. With the infrastructure breakdown, mass displacement and confusion in Gaza, counting fatalities and assigning causes of deaths with any precision is impossible. Moreover, I believe the figure matters. The war seems likely to end in a negotiated settlement, and Israel has to weigh the costs of continuing the war—with civilian deaths a major one—against the goals of freeing hostages and degrading the capability for Hamas to rebuild and reattack. If the war is killing 70% more people than Hamas claims, that is a reason to accept somewhat less favorable settlement terms than otherwise. The higher figure also would suggest the ratio of civilian to combatant fatalities is higher than the 1:1 figure implied by Israeli military information, although it would still be far less than any historical urban conflict.
Nevertheless, the authors’ statistical reasoning is deeply flawed. They make no attempt to hide their bias. They refer to Palestinians killed as “martyrs” and have no interest in Israeli or outsider deaths. They call for an immediate and lasting truce, leaving Hamas with fighters and infrastructure in place, and release of all Palestinians (it’s not clear whether they mean all non-citizens of Israel or all non-Jews) in Israeli prisons for any reason. They do also call for Hamas to release Israeli hostages. They make no distinction between Hamas fighters and civilians, nor civilians killed by Israeli versus Hamas actions.
The authors admit about their core data, “Critics have raised concerns about exaggerated numbers of deaths, citing issues such as potential data manipulation, missing identifications, and inconsistencies in demographic classifications,” but in “see no evil” passive voice follow that up with “no literature search was done.” Even the briefest of searches would have turned up thoroughly documented accounts of Hamas massive data manipulation, such as this study from the Henry Jackson Society.
My point here is the statistical method the authors use to inflate the Hamas figures: capture-recapture analysis. The methodology was first developed to estimate fish populations, which explains the name. If you want to know how many trout are in a lake, you might capture, tag and release a sample, say of size 80. You then do a second sample, say of size 100, and see how many of the recaptured fish were tagged in the first sample. If 5 of the recaptured fish have tags, you estimate the total trout population as 1,600. 80 / 1,600 tells you that you tagged 5% of the trout which matches the 5% of your recaptured sample was tagged.
There are important assumptions to justify this calculation. You must know that the trout in the lake remain the same—no births or deaths or trout entering or leaving the lake. The counts and tagging must be accurate. Most importantly, the captures must represent independent random samples of the trout population. If some trout are harder to catch than others, your estimate will be off.
Applying this method to counting deaths from traumatic injury in Gaza is problematic. The Gaza population does not stay the same. An estimated 6% of the population left since October 11, and outsiders have entered, plus there have been births and deaths not due to traumatic injury. The Hamas data is deeply flawed. Matching individuals is not straightforward like looking for a tag—researchers must guess based on name, identification numbers, age and sex, information that is often unavailable or inconsistent or manipulated. Deciding whether a death is due to traumatic injury should be straightforward, but there’s little reason to think Hamas does it honestly. And none of the data sources the authors use are anything like random population samples.
For example, a Hamas survey of the population turned up 7,581 individuals friends or relatives listed as having been killed by attacks, while the authors found 3,190 obituaries or notices posted on-line. 548 individuals were on both lists. From this the authors estimate there were 44,130 traumatic injury deaths using capture-recapture logic.
Leaving aside the incentives for people to report friends and relatives who fled Gaza or disappeared or died of other causes as martyrs, and for Hamas to exaggerate totals, the biggest problem with the methodology is failure to match individuals. The authors tried to exclude duplicate entries but different versions of a name (especially since some records were in English and some in Arabic) or mistakes about age or ID number (if any), and with Hamas known to alter sex and age to disguise fighters as women, children or aged, the totals on both lists will be too high. Moreover, the count of overlap will be too low. If only 7.5% of the names on the lists are duplicates, the capture-recapture method would agree almost exactly with the Hamas figure of 37,877.
How do the authors get from 44,130 to their estimate of 64,260? They do two more capture-recapture analyses using Hamas Ministry of Health counts to get estimates of 52,034 and 79,686. They then do some complex statistical analysis to combine these estimates. But the variations in totals using different combinations of data sources, by the same authors using the same methodology, shows the unreliability of the analysis for this application. In fact, the authors’ claimed 95% confidence interval for the number of traumatic injury deaths—55,298 to 78,525—rules out all three individual analysis results, meaning they’re 95% sure that the estimates used as inputs are wrong. Garbage in, garbage out. The number of traumatic injury deaths might be in the authors’ range, but I would be shocked if they were willing to bet $19 against $1 that it is.
Taking a step back, the official Hamas casualty figure of 37,877 seems plausible for violent deaths of non-Israelis in Gaza from October 7, 2023 to June 30, 2024. A number of sources have come to similar estimates. It could be much higher, and is unlikely to be much lower. It does not include many indirect deaths from lack of medical care and material deprivation. There are many missing people, and many or most of them are likely dead. And there are many wounded.
The main disputes are over what fraction of the casualties are Hamas fighters, and how many of the deaths result from Israeli rather than Hamas actions. To most partisans these make little difference, as they believe either Hamas has all the moral responsibility for initiating the war, using human shields, murdering Palestinians suspected of cooperating with Israel and refusing to release hostages and surrender; or that Israel bears all responsibility for being a settler-colonial state.
But moral responsibility does not drive settlement negotiations. Numbers of civilian deaths should matter and high-quality, unbiased, honest research into the question is valuable. Unfortunately, this paper is the opposite.
