Discover more from Pekingnology
A computed estimate of Russian and Ukrainian military casulties
A Simulated Experiment: Whose numbers are likely more reliable?
Some of you might have read How heavy are Russian casualties in Ukraine? on The Economist a few days ago. Now read the following article, originally published in Chinese on the WeChat blog of 复旦大学复杂决策分析中心 Center for Complex Decision Analysis (CCDA), Fudan University on July 25.
Frankly, I don’t totally get the article (not long!), but since it is the work of Prof. 唐世平 Shiping Tang, founder and leader of CCDA, I’m just gonna run it. (Plus, Prof. Tang’s Ukraine as a Solution, written in 2009 but published semi-exclusively in 2022 by Pekingnology, is my most popular post so far :)
The casualty data of troops on both sides published by the U.S. side (mainly by the U.S. Institute for the Study of War) is unreliable.
Russian casualties are likely 22,230-30,995 by 2022-06-24. The Ukraine-published figures on Russian casualties are an overestimate of 13.4%-22%.
The casualties of Ukrainian troops are likely 11,616-12,936 between 2022-02-24 and 2022-04-16. The Russian-published data on Ukrainian casualties in this period is overestimated by 80%-100%.
The final outcome of the war is difficult to predict.
Tang is Cheung Kong Distinguished Professor, Ministry of Education, China; Fudan Distinguished Professor & Dr. Seaker Chan Chair Professor, School of International Relations and Public Affairs (SIRPA), Fudan University.
With five single-authored volumes in English, as the 1st Asian and Chinese scholar to win a major book award in international relations, and as the 1st Chinese scholar to join the editorial board of leading journals in international relations, Prof. Tang is one of the most influential Chinese social scientists internationally.
Princeton University Press will publish his most recent book The Institutional Foundation of Economic Development in September.
(Combat) Casualty Numbers in Russo-Ukrainian Conflict: A Simulated Experiment
The Russian-Ukrainian conflict broke out in the early hours of February 24, 2022, and has been going on for five months now. The dynamics of the battlefield have also changed considerably. The war is now basically a war of attrition, which probably would not end soon.
If we want to arrive at an approximate prognosis of the war, the casualties of both sides are certainly key. We know that both sides in the war typically tend to exaggerate the casualties of their enemies and stay vague about their own casualties - or even not announce them at all.
So can we somehow explore whether the casualty data of both sides are reliable, or at least which side's data is a little more reliable?
We try to answer the question through data modeling, coupled with computer simulations.
From the current stage, the Ukrainian side is still receiving arms assistance from the West. And the Russian side still has a certain advantage in terms of heavy weapons. The future depends largely on whether Western assistance can sustain Ukraine's basic combat power; Ukrainian casualties; Russian weapons depletion; and casualties.
Again, it is particularly important to emphasize that our simulations are designed to test which side's published casualty numbers are more reliable, or almost all of them are unreliable. This is not a prediction of a winner or when the war ends.
As our previous research (in Chinese) emphasized, predicting the winners and losers of wars is difficult (and requires more real-world data), and predicting the end of wars (including when and how they will end) is even more difficult because there is simply no good way.
Based on our previous research on wars, we simulate the daily casualty figures of Russia and Ukraine from February 25, 2022, using the casualty ratio of Russia and Ukraine on the 180th day after the war’s outbreak. We use publicly-available data as the baseline for the simulation, i.e., the casualties of Russia and Ukraine in the 3-4 months before June 24, 2022.
All war-related parameters (e.g., casualty ratios, combat power gains/declines, etc.) for this experiment are based solely on established war research and all the major events since the outbreak, such as ceasefire negotiations, heavy weapons assistance, etc. For the impact of the major events’ impact on the combat power of either side, we multiply the combat power of a side by an upward factor after a particular stage and allow for decay (due to depletion) based on specific data announced in the news.
For example, it has been reported that "as of April 11, the United States and its allies had provided Ukraine with about 25,000 air defense weapons systems and 60,000 anti-tank weapons systems." This suggests that the Ukrainian side has received U.S. supplies of armaments to help it offset the damage caused by Russian airstrikes.
(Figure 1, from Wikipedia, numbers by 2022-06-24)
First, based on Figure 1, it is easy to see that the casualty figures for both sides vary significantly from different data sources.
In this regard, let's assume five scenarios:
the U.S. release of casualty figures on Russian forces is relatively accurate;
the Ukrainian official release of casualty figures on Russian forces is relatively accurate;
the U.S. release of casualty figures on Ukrainian forces is relatively accurate;
the Russian official release of casualty figures on Ukrainian forces is relatively accurate;
the Ukrainian official release of casualty figures on Ukrainian forces is relatively accurate.
Combining the baseline casualty data obtained from all relevant parties' data and adding the adjustment to the combat power of both sides as a result of the major events (armament deliveries, tactical changes), we conduct a simulation of the number of casualties on both sides for six months (2022-02-24 to 2022-08-24). Each simulation round was calculated 1000 times.
Here are the results (in a table)
Based on the simulation results, it can be found that if the data provided by the Ukrainian government or the U.S. side is used as a benchmark, the Russian side has an extremely low chance (no more than 30%) of winning the war 180 days after the outbreak.
[Pekingnology: The 1st, 3rd, and 5th rows.]
This conclusion still holds after several rounds of different parameter adjustments. Therefore, the casualty data on both sides published by the U.S. side (mainly by the U.S. Institute for the Study of War) is unreliable.
[Pekingnology: This study means that the chances of a Russian Victory on the 180th day, based on the facts on the ground, obviously couldn’t be that low, so the study reasons that the scenarios in the 1st, 3rd, and 5th rows should be unreliable. The 1st and 3rd scenarios refer to US data on Russian and Ukrainian casualties are accurate, respectively.]
If, on the other hand, the casualty figures published by the Russian side on the Ukrainian side are used as a benchmark, the probability of a Russian victory is substantially high (36.8% + 46.6% = 83.4%).
[Pekingnology: The scenario in the 4th row, i.e. Russian data on Ukrainian casualties are accurate.]
However, judging from the speed of the Russian advance in the first and second phases of the war, the Russian side has not been progressing very well. Therefore, the Russian data may have exaggerated the casualties on the Ukrainian side.
In this regard, we conducted a reverse extrapolation based on the existing simulation framework, i.e., we assumed that the final victory rate of both sides is known, and starting from that assumption we extrapolated the recent casualties in reverse. It was found that if we assume a victory rate between 40% and 60% for either side, we can instead obtain a relatively stable range of casualty estimates for them.
The Ukrainian side announced a total of 34,530 Russian casualties on 2022-06-24. According to our simulations, the range is likely to be 22,230-30,995, which means that the figures announced by the Ukrainian side overestimate Russian casualties by 13.4%-22%.
On 2022-04-16, the Russian government announced a total of 23,367 Ukrainian soldiers had been killed in action. According to our simulations, the number of Ukrainian casualties from 2022-02-24 to 2022-04-16 should be more reasonable at 11,616-12,936. Therefore, for this period, the Russian data overestimates the number of Ukrainian casualties by 80%-100%.
By "relatively stable range of casualty estimates", we mean that the casualty figures of both sides multiplied by the relevant war coefficients for different stages and units of the conflict are exactly offset, i.e., both sides reach a relatively reasonable and balanced state of troop deprivation. And after recalibrating the casualty data according to the newly obtained casualty estimation interval for both sides, the simulation is then conducted.
We find that the chances of victory for either side are between 40% and 60%, supported by roughly the same parameters. All in all, the final outcome of the war is difficult to predict.
The most popular Pekingnology newsletter, authored by Prof. Tang in January 2009 but semi-exclusively published here in February 2022: