Russia’s full-scale war with Ukraine has been going on for exactly three years. Russian society has paid a huge price for the adventure initiated by Vladimir Putin. Approximately 160,000 to 165,000 soldiers, officers, contract soldiers, mobilized soldiers, and other Russians have already died in the war. This number is a statistical estimate of the excess male mortality during the war, obtained by the Meduza and Mediazona teams. It is the result of comparing and analyzing specific documents: the lists of deceased individuals compiled by volunteers since the beginning of the invasion, and the Register of Inheritance Cases, which publishes information about all Russian heirs. The assessment does not include citizens of other countries who fought on the Russian side, including residents of the Luhansk, Donetsk, and other regions of Ukraine. The number is also limited to fatalities and does not include wounded: even severe injuries that do not result in death are not reflected in inheritance statistics or memorial lists. Despite these limitations, this estimate is the most reliable information available about Russia’s total casualties as of today. We will explain how this estimate was obtained, why it is reliable, and what our method reveals about the dynamics of the war.
You can read this report in English here.
What have three years of invasion in Ukraine led to? The main findings of the study
Over the three years of the war, its nature has changed many times, affecting both the dynamics of the sides’ advancements and the losses that accompanied them.
The most general conclusion that can be drawn from the results of the three-year confrontation is that the Russian side’s losses have increased from year to year. During the first three months of the war, the Russian Armed Forces (RF) lost approximately 40-60 soldiers per day, while during the same period last year, they lost 250 soldiers per day. However, this loss dynamics was not reflected on the battlefield. In total, the country lost almost twice as many people killed in each subsequent year as in the previous one.
- In 2022, about 20,000 Russians were killed in the war.
- In 2023, this number increased to approximately 50,000 (47,000 to 53,000) deaths.
- In 2024, the number of casualties approached 100,000.
The last estimate, the least accurate of all, may still be refined in the future due to the specifics of the method, which we will discuss in more detail below. In short, the more time that passes since the start of the conflict, the more data is accumulated, and the more accurately the excess male mortality can be estimated. The least accurate estimates are for the first six months after the analysis date, when the names of the deceased in the Register of Succession are still incomplete. However, even in this case, we can relatively accurately assess the change in the dynamics of loss accumulation.
The graph above shows several waves of loss growth, followed by small declines, but at a higher level. The first wave was associated with the beginning of the invasion, when Russian troops, who were not expecting strong resistance from the Ukrainian Armed Forces (AFU) and were experiencing serious problems with supply and management, suffered significant losses. These losses were particularly high among regular military personnel and young contract soldiers. During this period, the Russian Armed Forces lost approximately 500 soldiers per week, with up to a quarter of them being officers. However, this percentage has significantly decreased in recent years, and it is now less than 3-5%.
After the retreat from Kyiv in April 2022, the rate of casualties in the Russian army decreased by approximately half and only began to increase in the autumn. At that moment, the Ukrainian Armed Forces’ maneuvering actions first take the weak units of the Russian Armed Forces by surprise near Izyum and Balakleya, forcing the enemy to withdraw from the right-bank part of the Kherson region, and then a large-scale Russian offensive begins near Bakhmut. The Battle of Bakhmut reaches its peak in January and February 2023. During this period, the Russian Armed Forces lose approximately 1,500 soldiers per week, or an average of 150 soldiers per day. A significant number of the dead are prisoners, who were previously absent from the front lines.
After the capture of Bakhmut, the rate of casualties slightly decreased to around 100-125 per day during the summer of 2023, when the Ukrainian Armed Forces attempted to launch a counteroffensive in the Zaporizhzhia region. However, this decrease was overshadowed by the significant losses during the previous Bakhmut period. In fact, the losses are at the same level as at the shock moment of the outbreak of war for the Russian Armed Forces.
Finally, in early October 2023, a new Russian offensive begins. And with it, a new period of losses, incomparable in scale to anything that has been observed so far. The most intense weeks of the battle of Bakhmut are inferior in number of deaths to the weeks of lull in the last period. The daily number of casualties in the Russian offensive has almost never been below 200, and sometimes exceeds this average by one and a half to two times. Of the period with the most reliable estimation accuracy (that is, excluding the autumn—winter of 2024), May 2024 turns out to be the record for losses — at that moment, the Russian Armed Forces were advancing in the Kharkiv region, as well as northwest from Avdiivka towards Ocheretino and further to Pokrovsk.
The same high rate of losses remains in the fall of 2024. However, in the last few months of the year, there has been a sharp decrease. This is almost certainly not related to the actual situation on the front lines and is due to the incomplete nature of the data. The decrease in losses in the near future is inherent in the estimation method itself. We will be able to provide more accurate information about this period in about six months.
Where did these data come from? And why can we trust them?
The assessment method, based on the analysis of inheritance cases and lists of deceased individuals, was developed by Meduza and Mediazona in a series of studies published since mid-2023 (1, 2, 3). It is most fully described in the first of these texts (please refer to the methodological appendix at the bottom), which also includes the corresponding code. However, since then, the counting method and its specific implementation have been repeatedly updated, so we will try to provide a very brief description of the analysis logic below.
Since the beginning of the full-scale invasion, a joint team of volunteers and journalists from Meduza and the BBC Russian Service has been compiling lists of Russian military personnel who have died. Until now, only the list creators had access to these lists, but on the third anniversary of the war, the lists were fully opened, and all the data can now be accessed through a special project by Meduza. The lists are compiled based on official reports from local authorities, schools, obituaries on social media, and photographs from cemeteries. As of today, there are 95,322 names on the list. The deceased are usually identified within a few weeks of their deaths. However, there are instances where the death of a military personnel remains unknown for a year or two. The lists do not include citizens of other countries, so the study does not cover volunteers from the self-proclaimed Donetsk and Luhansk People’s Republics, military personnel from North Korea, or foreign mercenaries. The lists are almost exclusively composed of men, with only 25 women listed, representing less than 0.03% of all casualties. Based on indirect evidence, the proportion of women in the Ukrainian Armed Forces’ casualties is significantly higher.
- Register of Inheritance Cases (RND) — this is the public registry of the Notary Chamber, which contains data on open inheritance cases. With his help, the testator’s relatives can get information about the notary who opened the case. Opening a notary’s office is not the only way to inherit, and not all deceased persons have property that requires such registration. On average in Russia is opening cases for about half of the deceased. This proportion strongly depends on age: the older a person is, the more likely they are to have their own apartment or other property that requires inheritance, and the more often they end up in the RND.
- Under Russian law, heirs have 180 days to enter into an inheritance in a simplified manner. After this period, the inheritance is recognized only by the court. Although there are such cases, historical analysis of the registry shows that more than 90% of cases are opened within a 180-day period (most often in the first week after death or the last week before the deadline). This means that if a person has been dead for more than six months, the RND contains relatively complete data.
- With a representative sample of RND (and taking into account that there are almost no women among the losses of the Russian Armed Forces), we can calculate the number of excess inheritance cases that appeared in the registry after the start of the war. This number is the difference between the actual number of cases and the value that would have been expected if there had been no war. This approach is similar to the way demographers calculate excess mortality during epidemics or heatwaves. Two important differences are that, first, only the number of excess cases can be predicted based on the RND data, not the number of deaths themselves. To convert one to the other, additional information is needed (more on this later). Second, since there are almost no women among the Russian Armed Forces’ casualties, the number of excess cases among women can be used as an additional reference point for predicting the number of excess cases among men. In this way, it is possible to calculate the excess male mortality, which is not affected by factors that affect both women and men equally.
- The inheritance file indicates not only the date of death, but also the date of its registration in the registry office information system. There is usually not much difference between them: in the vast majority of cases, death occurs in Russia is registered within one to three days. Rare exceptions are associated with deaths on holidays and weekends, violent deaths and deaths in remote areas. However, for the military personnel who are listed in the RND, this rule does not apply: many of the deaths in the sample have a significant delay of several weeks or even months. On average, as we know from the поименные списки, half of the combat deaths are registered two weeks after the death. We refer to these cases as “late deaths.” An analysis of trends in the opening of cases among them allows us to further reduce the excess male mortality in the oldest age groups (40+), where the share of military deaths is minimal and the background expected non-war-related mortality is maximal.
- In the illustration below, you can see a visualization of the complex intersection of these death groups: the total number of Russians who died during the war (only men are shown for simplicity); those with open inheritance cases; those whose deaths were related to combat operations; those who were identified and listed by volunteers; and finally, those whose death registration date differed significantly from the actual date of death. Separately, there are citizens of Ukraine and other countries who are fighting on the Russian side. They are not included in the final assessment, nor are they listed by name.
- As we mentioned above, the excessive number of inheritance cases for young men makes it possible to identify an analysis of long—term trends – this approach is understandable and well described in the scientific literature. However, it allows you to get the number of excess deaths only among those who have already been included in the RND, that is, among the owners of some kind of inheritance. However, not all dead soldiers, as we know from the roll-call lists, are included in the register at all. Not immediately after death, not even years later. Such people, as can be assumed, did not have either relatives or, more likely, real estate that would require registration, despite their existence in Russia’s “military mortgage” program.
- So, even among the officers who died back in 2022, only 68% of people are found to have hereditary cases in the RND. This is an indicator for the youngest group, those who were 20-24 years old at the time of death. With age, the proportion of people who are investigated increases, but even in the best-case scenario, not everyone is investigated. For individuals over the age of 35, the rate is 80-85%.
- Consequently, the discovery of n excess hereditary cases during the desired period for men in the desired age cohort indicates a higher number of deaths than n. To obtain this number, we need to multiply the excess cases by the conversion rate from case to death. The analysis of the group of “early” victims, whose relatives had the most time to reach a notary, allows us to calculate it. In the case of victims who died in 2022, it took at least two years to open a case, which is four times longer than the law requires. It is reasonable to assume that a significant number of new cases will not be opened for these individuals, which means that the conversion factors calculated for them can be used for further calculations.
- Having obtained the number of excess inheritance cases and the transfer coefficients for cases to death, we can finally calculate the total number of deaths. The number will be correct only if the composition of the deaths during the war was homogeneous, in the sense that the average death in 2022 and the average death in 2024 are equally likely to be included in the RND. However, we know for sure that the composition of those fighting on the side of the Russian Armed Forces has changed a lot: for the first six months or a year, Russia lost mostly military personnel; then a significant number of mobilized joined them; at the beginning of 2023, thousands of prisoners fought near Bakhmut; and from the middle of 2023, a massive campaign was launched to attract volunteers, the main incentive for participation in the war, for whom money became important.
- The social status of these groups, their financial situation, and, consequently, their likelihood of entering the RND, are very different. If we don’t take into account the fact that prisoners are 2.5 to 3 times less likely to become heirs than regular military personnel, on whom the death rates are based, we will significantly underestimate the final estimate.
- To solve this problem, we will use another well-known research approach that comes not from demography but from medicine and social sciences: the balancing of case-control groups. In our study, we use officers as a control group, as we are confident that they are homogeneous. On average, the probability of an officer being in the RND depends only on their age and rank (possibly their region), but it is unlikely to depend on whether they died in 2022 or 2024. The relative probability of an officer being in the RNDThe RND for all other dead can be calculated by comparing them with officers: for each non-officer, you need to select a representative of the control group so that their dates of death and age cohorts match, as well as, preferably, additional features such as the region of residence. After making a comparison, we get a “table of doubles”, which allows us to estimate the ratio of the average probabilities of hitting the RND for the experimental and control group in a particular week of the war. This ratio is a separate adjustment that takes into account the contribution of prisoners or the influx of volunteers, which has transformed the Russian Armed Forces into a completely different professional army than the one that fought near Kyiv in early 2022.
- So, after calculating the excess mortality among men, we must multiply it by the coefficients of transfers to death and the adjustments for differences between different military groups. This will provide us with the final result.
- For example, we can take a specific (first) week of the war and a specific age group — men between the ages of 20 and 24 (as described in the first text). Among the cases in the RND that were opened for those who died between February 28 and March 6, there are 130 men, 76 of whom have a significant delay between death and its registration. There were only 16 female heirs who died this week, and all of them were “early.” The linear five-year pre-COVID trend of late deaths for men of this age predicts an average of two deaths, which means that 74 of them are excess. Given that only 69% of the deaths in the name lists were “late” this week, we can expect 109 excess deaths. From the same lists, it is clear that only 68% of the deceased military personnel in this group were eventually assigned to inheritance cases, which means that the resulting number should be multiplied by 1.46. This results in 160 excess deaths. In this case, the correction for the difference between officers is equal to one and does not change anything, as almost all of the deaths in the first week were members of the control group. In the first text, we estimated the number of deaths in the same week in the same group to be 155, which is an exact match within the margin of error. It should be noted that the method has been improved many times since then. Initially, for example, we did not take into account the delays in registering deaths. The match in values demonstrates the reliability of the overall approach, despite the changing details of the counting method.
- The last six months, and especially the last three months since the start of data collection in the RND, are incomplete, as many relatives of the deceased have postponed their visit to the notary to the last day. Although this behavior can be partially compensated for (as described in our first text), it introduces a significant error (which can be estimated based on trends among groups that were not directly affected by the war). To estimate mortality during this period, we have to rely on the RND data itself, using the operational estimate described in our last text instead. The operational estimate is based on a simple linear relationship between the number of deaths and the number of names on the list by a specific day of the war. It predicts the total number of deaths (although it does not guarantee the “correct” distribution of deaths by week).
- Experience shows that this approach is quite accurate: by early July 2024, we predicted 120,000 deaths based on the operational model. After analysis based on real RDD data, the estimate was reduced to 115,000, which is significantly less than the range of random error.
Analysis of the prospects for war and Donald Trump’s peacekeeping efforts
