Denying Gaza’s death toll doesn’t stand up to scrutiny

Mourners react at the funeral of Palestinians killed in Israeli strikes, Nasser hospital in Khan Younis, southern Gaza, 23 May 2025

Michael Spagat writes in Haaretz on 14 July 2025:

Sergio DellaPergola’s critique of Nir Hasson’s article and the Gaza Mortality Survey (GMS) is error-laden and internally contradictory. He dismisses the GMS as a preprint, ignoring that preprints are standard practice in science and facilitate open peer review. At the same time, he contradictorily deprecates the value of peer review: “Today, the market for scientific publication is broad, generous and diverse, and almost any author can find some platform for what was rejected elsewhere.”

DellaPergola next asserts that the Gaza Ministry of Health’s (GMoH) data “cannot be considered credible because it is itself a political subject.” Dismissing their data out of hand is a common move by those who seek to diminish the scale of civilian suffering in Gaza. However, rather than dogmatically dismissing the value of GMoH data, why not simply investigate them?

The lists of people the GMoH claims have been violently killed in the war are highly transparent and can be scrutinized closely. Israel has the population registry for the Gaza Strip and is, therefore, in an excellent position to investigate the deaths, virtually all of whom are listed by name, sex, age and national ID in publicly available lists. Independent analysts such as myself in a series of essays and Gabriel Epstein, notably here and here, have raked over the data and agree that, following a quality dip between late 2023 and much of 2024, the most recent GMoH lists of violent deaths are credible but incomplete.

DellaPergola has not only failed to seriously investigate the GMoH data; but also failed to understand their methodology. Based on unnamed “Gazan sources,” DellaPergola avers that, “astoundingly,” the GMoH has not even bothered to count bodies arriving in hospital morgues. To the contrary: the primary channel through which the GMoH builds its detailed public records of violent deaths is, precisely, by recording information on (not just counting) bodies in hospital morgues.

The other is from reports made by family members. DellaPergola baselessly declares that the GMoH rubber stamps these reports onto their list “without sufficient investigation of individual cases.” Yet the GMoH has applied such strict criteria for counting fatalities that it even removed hundreds of child fatalities for which strong social media-based evidence of violent death has been found.

DellaPergola’s next point appears to be aimed at the GMS finding that the true number of violent deaths substantially exceeds those recorded by the GMoH. Here he cites Israeli army claims that “residents were warned to evacuate their homes before heavy bombs were launched” so “it’s inconceivable that there are still tens of thousands of buried bodies 21 months after the fighting began.”

But even if his confidence in IDF warnings is justified, the GMoH will still fail to include them on their list if these deaths occurred outside of hospitals and are not reported by families, or if they are reported but then not confirmed by the GMoH. In fact, social media evidence of hundreds of violent deaths that have not made it onto the GMoH lists can be found here and here.

DellaPergola then approvingly cites an article by Professor Michel Guillot and his colleagues which “identified 34,344 people who were killed … between October 7, 2023 and August 31, 2024, based on detailed checks by the Health Ministry.” But the Health Ministry that performed these detailed checks is none other than the GMoH and the data in question is none other than the GMoH’s detailed list of people killed violently in the conflict. One of the main points of Guillot’s article is that the GMoH data is, in fact, credible.

When DellaPergola finally turns to the GMS, he focuses on household composition without first ascertaining how households are defined. For the GMS, a respondent’s household is the people they were living with as of October 6, 2023. DellaPergola considers a hypothetical case of a household of 12 that suffers 1 death and divides during the war into two separate households of 5 and 6 and worries that “the relative completing the survey could report the deceased adult because he or she will think of the original family composition.”

But this is, precisely, what respondents are supposed to do as DellaPergola would have known if he had looked at the questionnaire which is in the public domain. It is true that if both pieces of this pre-October-7 household were selected into the survey then this death would be reported twice. But these will be just 2 among roughly 400,000 households in the Gaza Strip; the chance that they both make it into the survey is minuscule and there are similar (tiny) chances that that both pieces of a split household with 0 deaths could be selected into the sample. In short, this is a trivial problem for the survey.

Next DellaPergola falsely claims that “the Gaza district with the highest ratio of reported victims proportionate to the population (55 per 1,000) is Deir al-Balah, which was not evacuated at the time of the survey.” But a quick analysis of the tables in our publicly available paper reveals that, in fact, households originating in Deir al-Balah experienced among the lowest violent death rate and rate of all causes of the five governorates.

DellaPergola’s next criticism is that “belying expectations, the correlation of cases of violent and non-violent death by district is neither high nor clear, even though it might have been expected that greater destructions would directly or indirectly result in a larger number of evacuees and deaths.” However, comparisons of death rates across governorates are irrelevant because the population moved around during the war and we have information only on the origin of each household, not where they suffered deaths. Therefore, there is no point in searching for such correlations.

DellaPergola then proceeds to an inaccurate critique of our sampling procedures. He quotes us saying that “if the survey does not keep the principle of randomness and mainly reflects willingness to participate, the results are invalid.” Yet our sample is about as random as possible and willingness to participate was barely a factor. We randomly numbered tents from 1 to 100 and a random number between 1 and 10 was selected. If this number was 4 then interviewers approached households 4, 14, 24, etc., through 94. If all agreed to interviews then that sampling point is finished. If there were refusals, which were rare, then the interview team restarted from a different random starting point.

DellaPergola further complains that the participating respondent in each household was not randomly selected as is best practice in opinion polls, wherein each eligible household member will have their own opinion. But this was not an opinion poll, and all eligible respondents would have had the same information about their household.

DellaPergola also claims, falsely, that we did not classify nonviolent deaths by sex and age as even a brief perusal of the paper refutes.

DellaPergola’s next assertion is that “the same statistical rules that apply to the survey’s main subject – deaths – also apply to the secondary subjects, such as prisoners and the missing.” This is not true. A different set of potential biases apply to each variable: Families may believe or hope that loved ones are alive in prison when, in fact, they are dead; some respondents may have misinterpreted our question as to whether a household member was at some point imprisoned.

Moreover, because many people were detained amid displacement, our samples of governorates we could not visit might disproportionately reflect imprisonment. These are all factors that do not apply to death estimations.

DellaPergola concludes by scolding Haaretz for not checking its facts before publishing Nir Hasson’s article. This is ironic to put it mildly, given how riddled with errors his own critique is.

Michael Spagat is Professor of Economics at Royal Holloway University of London. He specializes in the quantitative analysis of armed conflict, including the measurement of war-related death tolls and long-run trends in war fatalities.

This article is reproduced in its entirety