This article consists of the script for a series of two YouTube videos in which I present my analysis of the of Soviet and German forces on the Eastern Front in World War Two. The first video focuses on comparing the strengths and losses of the major combatants and comparing their relative effectiveness. The second video presents a computer simulation of the Eastern Front which I use to test various counterfactuals aimed to investigate how important different factors were in determining the outcome of the war.
The powerpoint slides for both videos are available here.
An excel spreadsheet of the data can be found here.
A detailed outline of the methods and sources used can be found here.
Part 1: Eastern Front Strengths and Losses
The Eastern Front of the Second World War was a titanic military conflict between the world’s two most powerful totalitarian states – Nazi Germany and the Soviet Union. Over the course of four years, armies of millions of men and thousands of tanks and aircraft fought over a battlefield extending thousands of miles, from the Baltic in the North to the Black Sea in the south, and from the Volga in the east to the Elbe in the West. While the Eastern Front has long been the subject of scholarly and popular attention, major questions remain to be addressed.
In this video series I shall consider two major questions. First, how many casualties were there on the Eastern Front? Recent archival work has provided critical new evidence overturning older figures, and indicating that the Eastern Front was more deadly than previously thought. This has implications for assessments of the relative effectiveness of the major combatants, as well as for appraisals of the critical factors in determining the outcome of the conflict. Second, how close did Germany come to victory in the east? In particular, could Germany have defeated the Soviet Union even after the failure of Operation Barbarossa? Many scholars and commentators have argued that the failure of the German invasion of 1941 to achieve its objectives meant that Germany would be unable to achieve victory against the much greater combined manpower and industrial capacity of the allies. Traditionally, however, these claims have been based on highly subjective intuitions, with no clear methodology for validating such counterfactuals.
In this first video, I will first present a novel dataset containing estimates of German and Soviet losses, personal strength, tank strengths, and manpower reserves for every month of the war in the east. In the second video in this series, I will use my dataset to develop a computer simulation of the Eastern Front, which permits a rigorous quantitative analysis of various counterfactuals to be made. My results indicate that contrary to common belief, the Soviet Union came very close to defeat owing to near depletion of manpower reserves. The three key factors which ensured Soviet victory were the failure of the Germans to properly mobilise and prepare for the initial invasion in 1941, the disastrous German defeat at Stalingrad, and the increasing diversion of German manpower and equipment away to combat the Western Allies from late 1942 onwards.
Before examining the statistical data that I have compiled, we will first review the history of the campaigns on the eastern front. The conflict began on 22nd June 1941 with the launch of Operation Barbarossa, the German invasion of the Soviet Union. Three army groups smashed through the Soviet forces and advanced thousands of miles into Soviet territory in the space of only a few months. Despite tremendous Soviet casualties and rapid German advances, the Germans were unable to achieve their objectives of capturing Leningrad and Moscow. In the subsequent winter counteroffensive, Soviet forces pushed the Germans hundreds of miles back from Moscow, while continuing to build up their force strength to over 5 million. In the spring of 1942, the Germans launched an offensive in the south aimed at capturing the oilfields of the Caucasus. Once again, the Germans made rapid advances, but were unable to reach their objectives. At the end of 1942, German forces in the city of Stalingrad were cut off and surrounded by the Soviets, forcing the Germans to retreat from the Caucuses, and resulting in the greatest German defeat of the war, when in early 1943 an entire army was forced to surrender in Stalingrad. Both sides spend the next several months rebuilding their forces, before in summer 1943 the Germans launched their final major offensive of the war in the east, aimed at surrounding Soviet forces concentrated near the city of Kursk. However, the Soviets were well prepared for the offensive, and were able to weather the German assault before themselves going on the offensive.
From this point on, the Soviets dictated the course of the conflict. In a series of advances, they were able to push the Germans back hundreds of miles to the Dnieper River, over which they gained some bridgeheads by late 1943. In the early months of 1944, the Soviets were able to use these to advance over the Dnieper, and in a series of offensives retook most of the Ukraine. Now outnumbering the Germans more than 2:1, the Soviets launched two major offensives in the summer of 1944. The first, operation Bagration, completely destroyed the German Army Group Centre in Belarus, and brought the Soviets up to the Polish border. The second was an offensive south into the Balkans, which led to Romania and Bulgaria to switching sides, and forced the Germans to evacuate from Greece and much of Yugoslavia. By December 1944, the Germans had been forced back to the border of the Reich across almost the entire front. The Oder and Vienna offensives of early 1945 saw the Soviets smash through the German defences in Eastern Germany and Austria, bringing the Soviet armies within reach of the heart of Germany. The end finally came in April and May of 1945, when the Soviet offensives against Berlin and Prague destroyed what remained of the German forces, at last bringing the four-year conflict to a conclusion.
Principles of historical statistics
We now turn our attention to the question of assessing the number of casualties incurred by the combatants on the Eastern Front. This question is still controversial, as it has implications for narratives about the effectiveness of the German and Soviet armed forces, which played an important role in the self-image of both Western and Soviet militaries throughout the Cold War and into the present day. In attempting to arrive at as accurate answer to this question as possible, I will outline three key principles of historical statistics, which inform both the estimates I use and the manner I collect and process information.
The first principle, comprehensiveness, relates to the importance of including all relevant data which covers the entirety of the domain in question. In the case of casualties, this means including not just battlefield dead and missing, but also those who were killed in rear areas, paramilitary organisations, and succumbed to their wounds. Comprehensiveness is vital for ensuring a fuller picture of the military situation, and is essential for the simulation I will present later in the video.
The second principle, comparability, relates primarily to ensuring that statistics gathered for both combatants use as similar definitions and standards of inclusion as possible. One of the primary purposes of gathering statistical data is for purposes of comparison across time and between countries, and equivalence of scope, definitions, and methodology is critical for the validity of such comparisons.
The third principle, contextualisation, involves providing the relevant background details and context to appropriately interpret statistics in their historical context. This includes include economic, logistic, political, geographic, and technical factors. In this short video it will not be possible to provide the ideal amount of background information, however I will endeavour to provide sufficient context to enable statistical data to be meaningfully interpreted. Context also incorporates knowledge of how the statistics were compiled and what they refer to. Absent this contextual background, historical statistics become mere meaningless numbers.
Issues with casualty data
Given these principles, I will now mention some specific issues that must be borne in mind when examining casualty data from the Second World War:
- First, it is vital to distinguish casualties from deaths, POWs, and wounded. Casualty is typically used as an overarching term to include all losses to a military force, including dead, missing, prisoners of war, and wounded. This is usually an overly broad term since it often includes wounded and sick who later return to service. Deaths, on the other hand, is typically too narrow a term, since it often excludes those who died from wounds (though this depends on the data source), as well as those captured by the enemy. Missing is a particularly difficult category, as it essentially refers to a lack of knowledge rather than an actual final state for a given person. Depending on the context, most ‘missing’ are typically either dead whose bodies are never identified, or prisoners of war. POWs are often listed under ‘missing’ since there is never a body, nor any way for a nation to know how many of its soldiers were captured. POWs can also subsequently die of disease or maltreatment, which makes it important not to double count when aggregating different sources of data. ‘Wounded’ is a particularly complex category since different combatants and different sources include wounds of different severities. Often the front-line units would report wounded as the number of soldiers who needed to be sent back from the front to hospitals in the rear. Some of these soldiers would, after a period of weeks to months, be able to return to duty, while others would be permanently disabled and have to be discharged from the military. Unfortunately, data do not usually distinguish between these cases.
- Second, it is always important to clarify which services are included in a given data source. In particular, ‘army’ is not the same as ‘armed forces’, and it is the latter that is most relevant since it provides the most comprehensive picture of the military performance of a combatant. As such, it is critical to ensure that both German Heer (army) and Luftwaffe (air force) are included in casualty data. This is less of an issue of the Soviets since their Army Airforce was not a separate military branch. A further complication is the existence of units that were technically outside the conventional military hierarchy. The German Waffen-SS was technically a separate branch of elite troops outside the Wehrmacht (the armed forces). Often its figures are included with the Wehrmacht, but this is not always the case, and so it is important to always be aware of whether Waffen-SS is included or not. A similar issue, though less severe because of its smaller relative size, arises with NKVD security units and personal in the Soviet Union, which also were technically outside of the Red Army.
- Third, a further complication arises because of the existence of a wide range of reserve and paramilitary forces, casualties of which are traditionally not included alongside the frontline casualty reports collated by the military high command. These services include militia (such as the Volksturm), border troops (particularly important in the Soviet Union), construction troops, police, anti-aircraft defence forces (of which there were large numbers in both Germany and the Soviet Union), and partisans.
- Fourth, it is vital to understand the method by which casualty data were obtained. As I mentioned, official casualty reports for Germany and the Soviet Union, as for most combatant nations, are constructed by compiling reports sent in periodically by frontline military units, who report how many soldiers were dead, missing, and wounded during some period of time. There are many problems with this method of counting casualties. In addition to ignoring casualties from paramilitary and reserve organisations, this process often leads to undercounting of deaths behind the lines due to wounds and disease. The number of ‘missing’ soldiers is also often underreported, particularly during times of military crises where accurate records are not always kept. Some units also at times only reported the difference between available personal at the start and end of a week or month, which will naturally lead to underreporting of casualties if some personal arrive and then are lost in between those dates. Finally, any records that do not reach the central command or are lost will not be incorporated into totals. Thus, these methods based on frontline reports will nearly always undercount the true casualties.
An alternative method to assess casualties is to use the personnel service records of individual soldiers. This process is typically only possible after the war is over, as it requires a great deal of time and work by archivists. However the advantage is that it enables a far more comprehensive picture of casualties, as it enables losses from all causes and in all military branches to be included. Data from these sources form the basis for my preferred casualty estimates, as I will discuss below.
Yet another method to estimate casualties is to use demographic methods. This usually involves calculating the demographic loss due to the war, and estimating the proportion of these losses due to military action. A crude way to do this is to examine the difference in demographic losses between men and women, since military losses are overwhelmingly male. This method is a subtractive process, and so will tend to overestimate the number of losses. Ideally, the demographic estimate should be similar to, or slightly higher than, the calculation based on personnel service records.
- A fifth and final point to consider in the case of German data is whether all fronts are included, or only the Eastern Front. For the purposes of this investigation only the Eastern Front is of interest, so all data must be adjusted or disaggregated to include only the east. In some cases only crude estimates are available, in particular by 1945 when the fronts increasingly converged on the German homeland. For my purposes, the eastern front includes German action in Finland down to the Black Sea, but excludes operations against partisans in Yugoslavia or Greece. Soviet operations in Manchuria in August 1945 are outside the scope of this investigation, and although technically these losses are included in the total wartime loss figures for the Soviet Union, the numbers are so small as to be negligible.
Soviet military losses
The Krivosheev data
We now begin to consider the issue of determining the military losses of the Soviet Union. Some viewers may have heard the figure of 8.7 million Soviet deaths on the eastern front in World War Two. This figure, as well as virtually all casualty data for the USSR, derives from a comprehensive analysis of Soviet wartime military reports completed in the late 1980s to early 1990s headed by Russian military historian Grigori F. Krivosheev. The source I predominantly use is a slightly updated version of these data published in 2010 titled (in English translation) The Great Patriotic War: Book of Losses. The numbers cited differ slightly between sources depending on whether these revised numbers or the original numbers are used, though the differences are minor compared to the issues we will discuss here.
Problems with Krivosheev data
The Krivosheev data, despite being widely used, are fundamentally flawed. This table here illustrates some of the major issues in the 8.7 million total he provides. First, Krivosheev excludes half a million reservists who died during 1941 before formally being registered with their units. While Krivosheev maintains these should be included with civilian casualties, virtually every commentator disagrees with this and adds these to military losses. Second, Krivosheev provides a drastically lower count for the number of Soviet POWs compared to the Germans, 4.4 million compared to the German figure of 5.8 million. Krivosheev argues that the difference is accounted for by civilians that the Germans captured as part of their huge encirclements during operation Barbarossa. Again, many scholars find this explanation dubious, and Krivosheev seems to have little evidence to support this. It may be the case that large numbers of non-combatants and rear-guard personal were captured by the Germans, but by the principle of comprehensiveness these should be included anyway so long as they were uniformed military personnel. Furthermore, it is implausible that the Germans would have taken civilians without any armed forces role into captivity, as they made few preparations to accommodate the millions of Soviet prisoners, whom they regarded as a burden on their resources. It is therefore unclear why they would take civilians as prisoners. A third issue is that some of the categories that Krivosheev categorises as discharged from the armed services include large numbers of deaths. Those include those who died in prison, were sentenced to death, or died in penal battalions. Krivosheev does not include these deaths in this data, but they should be included in Soviet war dead. I estimate that about 760,000 additional deaths should be recorded in these categories. Making these three adjustments yields a total figure of 11.5 million deaths, far in excess of the 8.7 million Krivosheev reports.
It should be noted that Krivosheev must make the choices he does in excluding so many categories of deaths, because the underlying casualty data he is working with only includes deaths of frontline soldiers engaged in military operations. Aggregating these reports from individual operations, he arrives at a figure of 8.7 million, which he therefore feels obliged to match using aggregate personnel data. However as we have discussed, this approach is fundamentally flawed, military casualties will always be much greater than reported by frontline units.
These various inconsistencies and omissions in Krivosheev’s figures have been noted by numerous other scholars, including Lev Lopukhovsky, Boris Sokolov, Sergei Mikhalev, David Glantz, and Anvar Ismailov. The table shown below provides an alternative way of considering the data, in which Krivosheev attempts to account for all Soviet personal by their status at the end of the war. In doing so, however, Krivosheev makes to major errors. The first is that he counts 9.7 million personnel being discharged during the war, even though he also reports that 1.3 million were later reconscripted. These 1.3 million should therefore not be included as discharged during the war, since that was not their final status. The second major issue is that Krivosheev includes as a ‘final status’ 940,000 MIA who were later ‘found in recovered territories’ and reconscripted. It is important to understand that these are not 940,000 specific persons that are accounted differently in the data and so not included in any of the other totals. Rather, this is just the total number of people who were listed as MIA and then later reconscripted from liberated territories as the Soviet army advanced. As such, by the end of the war these persons will appear as dead, discharged, or still in the armed services just like anyone else. The only reason I can think for their inclusion as a separate ‘final state’ is to ensure that the aggregate numbers balance, because as Lopukhovsky points out, being recorded MIA and then later reconscripted is not a final status. The only final statuses that should be included in this table are dead, released during the war, returned POW, or still in the military at the end of the war. Making the adjustments to remove this inconsistent counting reveals that Krivosheev’s data fail to account for some 2 million Soviet personal.
In combination with the omissions discussed previously, I believe this constitutes strong evidence that Krivosheev’s data grossly underestimate the number of Soviet casualties. In particular, casualties outside frontline military forces do not appear to be included in Krivosheev’s data. This point is repeated by Michael Ellman and S. Maksudov in their article ‘Soviet deaths in the Great Patriotic War’:
“It is important to note that this figure of 8.7 million only includes the regular armed forces and the frontier troops and internal troops of the NKVD. It does not include non-conscripted fighters (partisans, resistance fighters and the underground in territories occupied by the Germans). Nor does it include railwaymen fighting in their own militarised detachments, local anti-aircraft defence, the militarised fire service, police in frontier areas who fought against the invaders, etc.”
My preferred estimate
To arrive at a more comprehensive figure for Soviet losses, I believe we should turn to data that has recently become available from an analysis of Soviet servicemen records. Anvar Ismailov reports on this:
“Immediately after the war and until 1949, military commissars made so-called ‘house rounds’ [dvorovoi obkhod], that is, they made the rounds of homes with personal inquiries of relatives of those front ﬁghters who did not return from the war for the purpose of exposing persons missing in action. Servicemen recorded in this way comprised 60 percent of the card index of losses.”
The totals found from an analysis of these records is reported as follows:
“To sum up: losses among the rank-and-ﬁle and NCOS (12,434,398) naval ofﬁcers (1,100,000) and personnel (154,771), internal troops (97,700) and border troops (61,400) total 13,850,000. These data refute individual researchers, who cite overall losses as 11,944,000, of whom 6,885,000 died and 4,559,000 were captured and/or missing in action.”
Those second two figures are references to the Krivosheev numbers we have just been discussing.
Another approach can be used to corroborate this much higher estimate of Soviet losses. Using data from post-war Soviet population records, we find (highlighted in yellow) the population deficit of working aged men and women from 1946 compared to 1941. The difference between these two values can reasonably be taken as a rough estimate of military losses, and amounts to about 13.2 million. As a result of these considerations, I estimate the total number of Soviet military losses during WWII as approximately 13 million. I believe the much lower number of around 9 million given by Krivosheev is not credible, as it relies on an additive method of frontline reports that is known to systematically underreport casualties, especially during the chaotic period of 1941. Interested viewers can consult further references provided in the method notes linked in this video description for further discussion of these issues.
German military losses
Turning now to the estimates of the number of German casualties, the matter is thankfully somewhat simpler, as a comprehensive series of monthly estimates is available thanks to the work of German historian Rudiger Overmans. His estimates are based on a representative sample of personal files taken from the Wehrmacht records. He estimates that 5.3 million German military personnel died during the war, of whom about 3.7 were killed on the Eastern front. This is significantly higher than the official OKW estimates derived from reports of frontline units, with the discrepancy being due to many of the issues discussed already. In particular, Overmans argues that the Wehrmacht casualty reporting system progressively broke down during the chaotic retreats of late 1944, and thus casualties from the final year of the war were particularly underreported.
Total casualties by nation
Putting together the German and Soviet data, we arrive at an overall picture of the military casualties of the major combatants on the Eastern Front. In this table I provide a comparison of the official (OKW/Krivosheev) numbers against my preferred estimates. From these numbers it is clear that the overall ratio of Soviet to German deaths and casualties is quite similar in both datasets: slightly more than 3:1 in the official data, and slightly less than 3:1 in my preferred data. I want to emphasise this point because I am concerned that some will accuse me of attempting to inflate Soviet casualties by the adjustments that I made to Krivosheev’s figures. As I have argued, however, these adjustments are well motivated by clear methodological problems in Krivosheev’s numbers, and serve to make the Soviet figures comparable to the all-inclusive figures provided by Overmans. This latter point must be also strongly emphasised, namely it is inappropriate to compare Overmans data to Krivosheev’s data, because doing so violates the principle of comparability. Overmans data is derived using analysis of personnel records, and thus serves as an inclusive measure of all deaths including rearguard and paramilitary units. By contrast, Krivosheev’s numbers are based on reports of frontline units only, and exclude all these additional deaths. Overmans figures are thus most appropriately compared to the data found by examining the Soviet personal records, which is what I based my estimate of 13 million on. Ideally a statistical analysis of the Soviet personal data would be undertaken using the same methodology as Overmans used for the German data, thereby providing a more accurate Soviet series and more precise data for comparative purposes. Unfortunately, eighty years on this scholarly work remains to be done.
Looking now at the national totals for all combatants, we see that after the 13 million Soviet deaths and 3.7 million German deaths, Romania, and Hungary each suffered about 400,000 deaths. Italy experienced about 90,000 deaths, Finland 63,000, Poland 40,000, and Bulgaria, Slovakia, and Spain around 5-10 thousand each. This amounts to a grisly total of 17.7 million, or 18 million rounding for the degree of imprecision of Soviet deaths, which is 375,000 every month of fighting, or over 12,000 every single day for almost four years. The human mind is simply not equipped to fully comprehend such numbers. To provide a modern comparison to aid in grasping the sheer scale of such losses, in twenty years of fighting in Afghanistan, the total Coalition dead (including contractors, but excluding Afghan allies) amounted to just under 8,000. Thus, each day on the eastern front the Soviet Union suffered more deaths than did the entire Coalition in two decades fighting in Afghanistan.
Comparative losses by month
Absolute monthly losses
We now turn our attention to examining the losses series in more detail. This graph shows losses by month, from the start of the campaign until the end of the war. Interested viewers should consult the linked document for more information on how the monthly series was constructed. Losses are disaggregated to show German, Soviet, and minor powers separately. Readily evident are the tremendous losses borne by the Red Army during 1941. Casualties fell during 1942, but remained very high throughout the remainder of the war. German losses were already substantial during Barbarossa, and aside from a huge spike as a result of the defeat at Stalingrad, remained reasonably constant until the middle of 1944, when they ballooned to catastrophic levels. The losses of minor nations were relatively modest by comparison, with the exception of tremendous Italian, Romanian, and Hungarian losses at Stalingrad, and a spike in mid-1944 due to Soviet offensives which knocked Finland, Romania, and Bulgaria out of the war.
Monthly strength and loss ratios
It is now time to consider the relative strengths of the German and Soviet forces on the eastern front over the course of the conflict. The figures I present here are estimates of the total number of regular military personal at the front, including both combat and non-combat troops, and regardless of whether they are immediately fit for duty. Irregular and paramilitary forces, and wounded sent to rear hospitals, are not included. The primary source is Germany and the Second World War Vol V.II for the German figures, and Krivosheev for the Soviet figures. David Glantz also provides a set of figures which is broadly comparable with those I use. Sufficient information will likely never be available to ensure that these figures are entirely complete or consistent between both combatants, however as far as I can tell the criteria used are broadly consistent, and I believe the margin of error for most months is on the order of 5% or so.
Immediately evident is the substantial increase in the Soviet strength from 3 to 6 million over the course of late 1941 to late 1942. Thereafter Soviet strength held approximately steady in the range of 6-7 million, though augmented in the final year of the war by about one million soldiers from Romania, Poland, and Bulgaria. Germany began the campaign with over 3 million soldiers and nearly a million more from its allies. By late 1941 this had fallen to slightly below 3 million, a level maintained until just after the Kursk offensive, when increasing demands in other theatres of war forced Germany to progressively reduce its strength in the east, which fell to a mere 1.5 million by the end of hostilities. The contribution of Germany’s allies was highly inconsistent, with peaks during the initial phases of Barbarossa owing to substantial Romanian and Finnish contributions, during late 1942 owing to the substantial contributions from Romania, Hungary, and Italy to Case Blue, and in mid-1944 due to the mobilisation of Hungarian and Romanian forces in defence of their homelands from Soviet invasion. The contributions from Germany’s minor Axis partners, however, depended largely on the political willingness of their governments to commit forces, which the Germans continually pressured them to do from 1942 onwards.
Comparing this graph to the previous one, you may notice that there is a close association between losses and strengths. As German strength on the eastern front fell over the course of the war, the loss ratio becomes more favourable to the Soviet Union. We can visualise this on a graph showing the ratio of losses as bars, against the ratio of strengths as a line. Apart from some outlier months, in particular those corresponding to the defeat at Stalingrad, the inverse relationship between these two values is quite visually compelling. We get an even more accurate view of the performance of the German military specifically if we exclude the minor Axis nations and just plot the ratio for Germany and the Soviet Union. We see the overall trend is similar, but with less of a sharp decline for Stalingrad (though January 1943 still stands out clearly), and also a more gradual buildup of Soviet numerical superiority near the end of the war. From this graph we also see that (aside from January 1943) Soviet losses do not fall noticeably below twice the level of German losses until June 1944, which corresponds to the launch of Operation Bagration. For the remainder of the war, Soviet losses were about on par with the Germans. However to achieve parity in overall losses, the Soviets required numerical superiority of between 3 and 4 to one in their favour. By comparison, when strength ratios had been close to 1:1, the Soviets suffered roughly ten times the casualties of the Germans. These data strongly suggest a major discrepancy in the overall capability of the Red Army and the Wehrmacht to inflict causalities on the enemy, and the huge importance of to the Soviets of maintaining significant numerical superiority.
To illustrate the importance of Soviet numerical superiority more clearly, we can plot the logarithm of the loss ratio against the logarithm of the strength ratio. Here we see that taking the data from April 1942 onwards, which clusters distinctly from the data for earlier months, 73% of the variation in the loss ratio is explained by variation in the strength ratio. This is even including the substantial outlier of Stalingrad, and is also despite the fact that errors in measurement will actually tend to suppress R-squared values by what is called attenuation bias. Some viewers may object that it is unreasonable to reduce all facets of military performance, including tactics, operational strategy, training, equipment, logistics, planning, terrain, and morale, into just two numbers. Yet the purpose of this exercise is not to say that such factors do not matter, but rather that fixing all those factors as they were in the actual circumstances at each point of the war, and considering only variation in the ratio of frontline strengths, about three quarters of the variation in the loss ratio can be explained. By itself this does not prove that increased numerical superiority was the cause of the increased overall effectiveness of the soviet military over the course of the war, since other factors like improved Soviet operational and planning effectiveness correlate with higher Soviet numerical superiority, and thus could be a confounding variable. However I regard this evidence as highly suggestive, and I would contend that it is reasonable to believe that the large majority of the improved effectiveness of the Soviet military (as measured by loss ratios) over the course of the latter three years of the war was attributable to their growing numerical superiority, since it is unclear why other confounding factors would correlate so well with both loss ratios and strength ratios.
Comparative equipment losses
Soviet numerical superiority was not limited to personnel, but also extended to most classes of equipment. Here we see a comparison of annual losses of single-engine fighters, all aircraft, armoured vehicles, and personnel. Note that Soviet figures in particular are not especially reliable (they are based on calculations by Krivosheev), but the general picture is so clear that it is not sensitive to even substantial inaccuracies in the numbers. It is clear that the Soviets consistently lost far more tanks and aircraft than the Germans, on the order of 5:1 aircraft and 4:1 tanks averaged across the whole war. While the ratios were generally worst in 1941, they remained poor for the Soviets right until the end of the war. For tanks and personnel, there is enough data to calculate effectiveness ratios, which I define as the casualties inflicted per Soviet soldier in a calendar year of combat, divided by the casualties inflicted per German soldier in the same period. For roughly evenly matched forces, the value should be around 1. However as we see here, the value for personnel was less than 0.1 during 1941 and 1942, and rose to only 0.4 in 1945. This means that even in the catastrophic final months of the war, each German soldier was inflicting more than twice as many casualties as each Soviet soldier, despite being outnumbered about 4:1 by this stage. As a point of comparison, the US-led coalition in the Gulf War inflicted about 100 times as many casualties per soldier as the Iraqis. On average over the course of the entire conflict on the eastern front, Germany inflicted about 6 times as many casualties per soldier then did the Soviets. The ratio is even more extreme in the case of tanks, where the overall ratio for the war is more than 10 times.
The poor performance of Soviet tanks is further illustrated in this figure, which shows tank loss ratios as bars and strength ratios as a blue line. This shows that, despite outnumbering the Germans on average of 3:1 from 1942 onwards, monthly loss ratios rarely fell below 2:1, even up to early 1945.
Summary and conclusions
The overall conclusion I draw from these data is that despite the undoubted improvements in Soviet operational proficiency over the course of the war, they never even came close to matching the proficiency of the German military on a man-to-man, tank-to-tank, or plane-to-plane basis. The tremendous successes of the Red Army were predominantly due to overwhelming numerical superiority, especially in tanks and artillery. This is illustrated by the two figures shown here, both taken from Germany and the Second World War Volume VIII. On the left is a depiction of the relative force strengths at the Battle of Kursk in July 1943, and on the right is a depiction of the situation at the beginning of Operation Bagration in June 1944. In both cases the enormous superiority of Soviet men and materiel is clear, with ratios of 3-5:1 in 1943 and 4-10:1 in the 1944 battle.
Of Course, operational examples such of this must be interpreted with particular caution, as they only refer to troops initially available at that region of the front. Also, deployment of troops and equipment in the right place and at the right time so as to be able to attack where the enemy is most vulnerable is a critical aspect of military operations, one that the Soviets repeatedly proved themselves to be adept at in the cases of the Moscow counteroffensive in 1941, the Stalingrad counteroffensive in 1942, the Battle of Kursk in 1943, and Operation Bagration in 1944. Nor do I in any way wish to demean or impugn the fighting spirit or tenacity of the Soviet soldier. However what I am attempting to illustrate with the data I have presented is that at the grand strategic level, abstracting from the details of specific operations, the evidence points to the fact that Soviet success was continent on substantial numerical superiority of men and equipment. Even with such superiority, the Soviets continued to suffer greater casualties than the Germans until the last ten months of the war, and experienced greater losses of equipment until the very end of the conflict.
The view that I present here runs somewhat counter to the trends in recent western scholarship, which has tended to emphasise Soviet military successes and German military failures, and express significant scepticism to the view (commonly found in memoirs of German generals after the conflict was over) that German defeat in the east was due primarily to German numerical inferiority combined with Hitler’s meddling. My claim is that these accounts, while serving as an important counter to the overly simplistic and self-serving claims of many of the German memoirs, nevertheless fail to appreciate the immense importance that Soviet numerical superiority did have, because they do not sufficiently emphasise comprehensiveness and comparability in their accounts, and do not sufficiently contextualise figures when they are given. In particular, the fact that loss ratios show such a strong correlation with strength ratios is, I believe, a novel finding, and illustrates the value of taking a statistical approach to address historical questions. Too often, historians will provide statistics for individual operations or army groups without any clear indication of how these fit into the broader picture, or how the numbers compare between the combatants. It is not enough, for instance, to discuss the severe losses suffered by the Germans, unless one also considers the losses suffered by the Soviets. Authors also point to improved Soviet operational performance in the second half of the war, without making sufficient adjustments for the substantial increase in numerical superiority enjoyed by the Soviet Union in (say) 1944 compared to 1941. By constructing a comprehensive dataset for the entire eastern front, I hope to have shed greater light on these questions than was previously possible using less complete or comparable datasets.
This concludes the first video in this series. In the next video, I will continue my analysis by presenting a novel computational simulation of the eastern front, and attempt to answer the question as to whether Germany could realistically have defeated the Soviet Union even after the failure of Barbarossa. I also use them model to draw conclusions about which factors were most critical in the German defeat in the east. I will reserve my final conclusions and summary to the end of that video. Until then, thanks for watching, and feel free to like and subscribe if you enjoyed the video.
Part 2: Eastern Front Simulation
Introduction to the model
In this video I consider the question: was the German defeat on the eastern front in WWII inevitable? It is not uncommon to find western scholars claiming that, following the failure of Operation Barbarossa and the transition of the eastern front into a war of attribution, Germany had little or no chance of winning in the east. This claim is usually based on the fact that by the beginning of 1942, the men and material resources of the Allies greatly surpassed those of Germany. In its failure to knock the Soviet Union out of the war quickly, Germany had placed itself into a two-front war of attrition that it simply could not win. Here I assess this claim using a novel approach. Instead of relying on isolated datapoints and intuitive judgements, I analyse the fighting on the eastern front using a computational model, which is parameterised using the strength and losses datasets that I introduced in the previous video. I strongly recommend watching that video before this one, as it will provide essential background on my assumptions and general methodology. Here I provide a brief outline of the key aspects of my model. Viewers desiring more detail should consult the document linked in the video description.
Summary of the model
The simulation uses monthly historical estimates for manpower, strengths, reinforcements, losses, and frontline strength ratios. These are loaded into the model, which then initialises the corresponding variables to their values at the desired initial month (by default this is April 1942, but can be changed as desired). The simulation then computes the course of the conflict stepping forward one month at a time, until one of the surrender conditions is met. The monthly loop at the heart of the simulation works as follows. First, the strength ratio is calculated as the ratio of Soviet to German forces deployed at the front. Second, the loss ratio for the month is calculated based on the strength ratio, using parameters found by regressing the loss ratio against the strength ratio. The idea of this is that the ratio of frontline forces determines the ratio of losses experienced by the combatants, which is observed in the historical data I presented in the previous video. Third, the strength ratio is also used to calculate the advance rate (how far the frontline moves in a month), with the parameters again based on regressing the two variables. This part of the simulation assumes that when the Germans have (relatively) greater strength they advance, whereas when the Soviets have greater strength then they advance. Fourth, the total losses (German plus Soviet casualties) are calculated from the advance distance, with parameters estimated by regressing total losses against advance distance. The logic behind this is that months of more intense fighting are also typically those in which one side is able to launch a major offensive. This is not always the case, meaning this estimate is especially noisy. However this does not much matter for the simulation as a whole so long as total casualties match historical trends on average. Fifth and finally, the losses are computed for each side based on total losses and the loss ratio, and their new frontline strengths are estimated based on the initial strength plus reinforcements, less casualties.
As I have already discussed the strength and loss ratio data in the previous video, I will not repeat that analysis here. I will, however, show the data used for estimating the advance rate. This graph here shows the correlation between the strength ratio and advance distance, with positive numbers representing German advances and negative numbers representing Soviet advances. While the correlation is only about 0.36, partly because many months saw neither side advancing and hence no change in territory, there is still a clear relationship between these two variables, with the point of balance occurring at a ratio of around 2:1. As I mentioned, the low correlation between these variables does not present a problem for the simulation, since the main effect is to introduce significant trial-to-trial variation in outcomes, not to bias the average results over many trials.
Clarifications about the model
At this point it is necessary to clarify that the simulation is only a strategic-level simulation. Its purpose is to analyse whether, and under what situations, Germany would have been able to win a war of attrition against the Soviet Union. As such, the simulation does not directly incorporate factors like logistics, equipment quality, leadership, operational decisions, morale, or other such factors. Such aspects are indirectly incorporated into the simulation through the use of historical data to estimate the relevant parameters. This means that the simulation can be regarded as a model of the eastern front at the strategic level, over the range of force levels, territorial expansion, technology levels, and logistical constraints that the combatants actually experienced in the war. The simulation should not, therefore, be used to estimate the outcome of a conflict that took place five years earlier or five years later, since these factors would likely be substantially different, and hence the parameters used in the model would be inappropriate. Likewise, the model should not be used to predict the outcome of an eastern front conflict drawn out far longer than was historically the case. As such, the simulation is hard coded to end in June 1947 (six years after Barbarossa). However the simulation can, in my view, be used to estimate the outcome of the sorts of historical counterfactuals that I will discuss in the next section of this video.
A second important point to understand is that the model is stochastic. This means that values of key variables such as loss ratio and advance distance are drawn from a probability distribution, the parameters of which are estimated from the historical data. Because the historical data are quite noisy, especially for the total losses and advance rates, there is a great deal of variation from one run of the simulation to the next. This is entirely expected and appropriate, given the uncertainty in the data. Trial-to-trial variation should be interpreted as representing uncertainty in the parameter values, as well as the effects of non-modelled factors, including strategic command decisions, operational decisions of generals, the effects of weather, and other idiosyncratic unpredictable aspects of war. The model does not assume that such factors do not matter – indeed I will argue that such factors were likely pivotal in shaping the outcome of the war – but the simulation does not directly model them, and instead seeks to address the question of what impact overall force levels and the availability of reserves had on the outcome of the conflict.
A third point to make about the simulation is that it does not incorporate the strength or losses of minor Axis or Soviet allies. Such an inclusion is not possible because the strength of these minor combatants was primarily determined by political factors, which are outside the scope of the simulation. For example, sizeable Finnish forces participated in the opening phases of Barbarossa, but for most of the remainder of the conflict they remained largely inactive, owing to the reluctance of the Finnish government to commit their forces to offensive operations. Likewise, both Hungary and Romania attempted to limit their involvement on the eastern front in 1943 and 1944, with their contributions only increasing from early 1944 as a result of mounting German pressure, and their need to defend their national borders against encroaching Soviet forces. In August 1944, Romania experienced a coup and switched sides, thereafter fighting alongside the Soviets against the Germans. Occurrences such as this are impossible to incorporate into the simulation, and any decision I made about incorporation of minor combatants would be ad hoc. Furthermore, I do not believe that the role of minor nations was ever particularly important at the strategic level. Germany’s minor axis partners contributed a maximum of about one quarter of the total Axis forces in the east (a proportion reached in mid-late 1942 and briefly again in mid-1944). However owing to the much lower quality of Italian, Romanian, and Hungarian troops, equipment, and leadership, I believe that as a proportion of overall fighting strength, the maximum proportion would have been closer to around 10-15%, and the typical proportion would have been more like 5-10%. These are within the margins of error for German and Soviet force strengths and also Soviet losses (German losses are somewhat more accurately known), and hence make little difference to the overall simulation.
Population pyramid comparison
Because the primary purpose of this simulation is to examine the capabilities of Germany in fighting a war of attrition against the Soviets, it is important to consider the manpower reserves of both combatants. This is a subject rarely discussed in depth in popular accounts, but is one of the single most important factors in determining the outcome of the war in the east. This diagram shows an overlay of the population distributions of Germany (grey) and the Soviet Union (red) for working-age men as of 1941. The number of men eligible for service in the armed forces (traditionally ages 18-50, with ages 18-30 preferred for frontline service) is the ultimate limiting factor of the number of reinforcements available to replenish losses.
In this diagram it is useful to distinguish several important age groups. First there are the veterans of the First World War, who by the outbreak of WWII were in their 40s. Next are the so-called German ‘white ages’, which were the cohorts who came of age between 1919 and 1932, a time when the German army was limited to 100,000 men and conscription was forbidden. This meant that most men in this age group did not receive any military training, hence the appellation ‘white’. By contrast, a large proportion of the men in the corresponding Soviet cohorts served in the armed forces for about two-three years. The third group consists of those born during WWI, who came of age shortly after the Nazi rise to power, and therefore received military training in Germany before the outbreak of the war. Men of demographic, which was significantly smaller owing to the fall in birth rates during the First World War, were in their early to mid-20s at the outbreak of World War Two. The final demographic were men who turned 18 during the crucial period 1940-1946 (the latter included as both Germany and the Soviet Union conscripted large numbers of 17-year-olds by the end of the war). Also note on this graph the considerable decline in Soviet birth rates corresponding to the period of forced industrialisation in the late 1920s to early 1930s.
Several observations should be made here. First, note that while during the First World War the Russian Empire (which had similar borders to the Soviet Union) had larger numbers of available recruits than Germany, the ratio was not enormous, only about 1.5:1. By contrast, by the Second World War the ratio had increased to over 3:1. This may represent one overlooked reason why Germany was able to defeat Russia in the First World War while fighting on two fronts, when in the Second World War they failed to defeat the Soviet Union despite fighting mostly only on a single front for the first two years or so. There were of course many other differences, including the fact that the Soviet Union was far more industrialised and had more effective and ruthless leadership than Imperial Russia, but nevertheless the fact they also had more than twice the available manpower relative to Germany than they did in the previous war, must have also been highly significant. The second observation to be made is that while both the Soviet Union and Germany began the conflict with significantly smaller than usual cohorts of frontline troops due to the reduction in births during World War I, the effect was far greater for Germany, which had a reduction of about 50%, compared to only about 20% for the Soviet Union. It seems therefore that by an odd confluence of fate, Hitler came to power and reinstituted conscription at just the moment when he had the smallest number of young men available to bring into the military. How much this impacted the German war effort is difficult to say, but it is a potentially highly significant fact that is seldom mentioned at all in historical accounts of the conflict.
Cohorts of 18-year-olds
Turning now to the cohorts of 18-year-olds coming of age during the war itself, we see the annual increase in manpower available to the Soviet Union was over three times as great as that available to Germany. The overall population of the Soviet Union was only about twice that of Greater Germany, illustrating how important it is to move beyond crude comparisons of population and examine the actual numbers of combat-age men available. However even this doesn’t tell the whole story, because for reasons that I have not been able to fully uncover, the Soviet Union (despite experiencing severe manpower shortages later in the war) only conscripted about 50% of the male population who turned 18 during the war years, whereas the corresponding figure for Germany was about 80-90%. While I have not found any clear explanation for this discrepancy, I suspect that a greater proportion of German 18-year-olds were conscripted due to a combination of the much lower efficiency of Soviet agriculture requiring more young men to stay in the fields, a diminished ability of the Soviet government to forcibly conscript young men from more peripheral areas like Central Asia and the Caucuses, and nutritional deficiencies rendering a higher proportion of young Soviet males unfit for service.
But whatever the reasons, when adjusting for this difference, we find that the Soviets had about twice as many 18-year-olds added to the manpower pool every year during the war than the Germans. This means that in a war of attrition, casualty rates of more than 2:1 were unsustainable for the Soviet Union, while rates of less than 2:1 were unsustainable for Germany. This is a highly significant result, since as we saw previously, the loss ratio never fell consistently below 2:1 before the final ten months of the war.
This result is also shown in this graph, where every month above the horizontal line represents unsustainable losses for the Soviets, and every month below the line represents unsustainable losses for Germany. As we see, in all the fighting prior to June 1944, the Soviet Union was suffering losses at a rate that was unsustainable, and hence would have depleted their Soviet manpower before German manpower. Of course, this ignored the impact of the opening of a second front by the Western Allies, which we will discuss in a moment.
Comparative reserves during the war
We can see the manpower situation more clearly if we plot the available Soviet manpower reserves over time against those of Germany. Note that the plot begins in January 1942; if it was extended back to June 1941, the plot would be hard to read because of the huge quantity of Soviet reserves available at the outset of the conflict that were lost during 1941. These reserves figures are estimates of all manpower potentially fit for military service, including those currently recovering from wounds or lacking military training. This distinction is critical for operational purposes, since at the beginning of 1942 Germany certainly did not have 2.5 million trained reservists available (the number was less than half a million), and untrained men cannot immediately be deployed into combat. However from a long-term strategic point of view, untrained but potentially fit men can be trained and sent into combat. This was done by both Germany and the Soviet Union with their reserve pools of untrained manpower over the course of the conflict.
In this plot, reserves are added each month in accordance with an estimate of the number of young men turning 18, and deducted in accordance with casualties suffered. In the Soviet case, additional recruits also became available from 1943 onwards due to liberation of Soviet territory from German occupation. In the case of Germany, the plot also shows manpower that was taken from the Wehrmachtsgefolge (‘followers’, or various support units of the Wehrmacht) in the final 10 months of the war and redeployed into the Wehrmacht proper. This practise meant denuding training schools, administration, construction, antiaircraft, and other support branches of their personnel, which was an unsustainable and ultimately counterproductive practise, as it would eventually destroy the ability of the Wehrmacht to function and provide a continual stream of trained reinforcements. Nonetheless, this was a desperate action taken at a time when Germany had already clearly lost the war, and so Hitler felt Germany had nothing left to lose by this point.
It should be noted that overall manpower data series are very difficult to reconstruct, and numerous estimates and assumptions had to be made. Nonetheless, I am confident that those slopes of the Soviet and German lines are fairly accurate, as these are based on casualty data and the actual number of personnel recruited into the armed forces. The position of the lines on the vertical axis is likely less accurate, since this depends on assumptions about the total number of reserves available, potentially including personal who were never actually drafted. In my judgement, the German line could perhaps be shifted up or down by half a million men, while the Soviet line could be shifted by maybe one million men, depending on assumptions.
These uncertainties, however, do not alter the overall message of this plot, which is that for the first two years of the war, both powers were experiencing unsustainable losses, causing their manpower reserves to be slowly depleted. By around the middle of 1943, both Germany and the Soviet Union had effectively exhausted their manpower reserves. In order to provide reinforcements, Germany was forced in September 1943 to alter its mobilisation policies, extending the age of military service from 18-50 to 17-55, and eliminating large numbers of service exemptions for those working in industry. By these measures and even more draconian enforcement of recruitment regulations, Germany was able to increase their reserves by about one million men. This was able to sustain their losses on the eastern front for about another year, until the onset of enormous losses brought on by Operation Bagration. From around July 1944 onwards, Germany was only able to meet its manpower needs by cannibalising the Wehrmachtsgefolge, which as we have seen is a self-defeating policy which effectively represented tacit acknowledgement that manpower reserves had been exhausted. The Soviet Union, by contrast, was able to meet its manpower requirements by inducting millions of men from territories liberated from German occupation. Without this additional source of manpower, the Soviet Union would have had a manpower deficit of 5 million by the end of 1944.
Soviet manpower problems
The importance of the Soviet reliance on manpower from occupied territories is further illustrated by this figure, which shows an estimate of the population of the Soviet territories occupied by Germany over the course of the war (note that this includes territories annexed by the Soviet Union in 1939-1940). This plot shows that the population under German occupation was fairly constant at around 60 million from late 1941 to late 1943. Between August 1943 and August 1944, almost the entire occupied population was recovered, providing the Soviet Union with an estimated 6 million additional military recruits. This additional manpower came at just the time in the war when the Soviet Union had essentially depleted its existing manpower reserves. Interestingly, the end of this period of liberation coincides with another new source of manpower, this time coming in the form of new allies, first Romania in August 1944, followed by Bulgaria in September. Together these two nations contributed about 800,000 troops to the Soviet side in the final six months of the war, which is more than the total ever contributed at any one time to Germany by Romania, Hungary, and Italy (max about 700,000).
The lesson to take from this analysis is that, had Germany been able to maintain its occupied territories in the Soviet Union for even six months longer, it is plausible that the Soviet Union would have been unable to raise the manpower needed to replenish its forces, which would have led to a reduction in its force strength on the eastern front. By itself this would not have led defeat, but as we have seen, loss ratio and strength ratio are highly correlated, and the Soviet Union required a strength ratio of about 3:1 in order to inflict unsustainable losses on the Germans. A reduction in Soviet frontline strength brought about by a failure to retake occupied territories could have led to a positive feedback effect, in which lower Soviet frontline strength led to greater Soviet losses, which further diminished their strength and hence reduced the chances of capturing the territory needed to recover recruits to replenish manpower. It is these complex patterns of interactions which are often neglected by historians when they rely on their intuitive judgements of cause and effect. We will attempt to test these ideas rigorously in the computational model.
German commitments in the West
One final aspect of the model that is important to understand is that of the German additional losses in the West, and the German frontline ratio. In the simulation, German losses to the Western Allies are hard coded at their historical values for each month. While not ideal, this allows the simulation to focus entirely on the eastern front, and also permits modification of these values to serve as a method of counterfactual intervention, effectively asking what difference would it have made if Germany suffered greater or fewer casualties to the Western Allies at a given point in the war. Similarly, the ratio of total German forces to the strength on the eastern front is also hard coded to historical values, and is thus also available as a point of intervention to determine the impact of Germany devoting a larger or smaller proportion of its total resources to the eastern front. This provides a simple way of gauging how important the contribution of the Western Allies was to the defeat of Germany, which is one of the main questions we seek to address.
The figure here shows the magnitude of German additional losses, as well as the percentage of German personnel and divisions deployed against the Soviet Union. The casualties shown here include all German losses not attributable to the eastern front, including casualties at sea, pilot losses to the defence of the Reich, and losses suffered in North Africa, Italy, and Western Europe. The personnel strength percentage is far lower because this includes all personnel in the Wehrmacht, including the Luftwaffe, Kriegsmarine, and the replacement army. The division percentage is not used in the model, but provides a clearer picture of the proportion of the field army which was deployed in the east. Note that following the decline in personnel strength during Barbarossa, the percent of personnel deployed to the east remained at roughly 30 percent until July 1943, when the allied invasion of Sicily forced the Germans to begin transferring significant forces away from the eastern front. The large spike in losses in May 1943 corresponds to the surrender in Tunisia, where over 100,000 Germans were captured. The tremendous increase in losses following the Normandy landings in June 1944 is also clearly visible.
Before showing the results of the simulation, it is important to understand the conditions we are attempting to simulate. This table describes the scenarios we will consider.
The simplest is called ‘standard’, and simply initialises the simulation to the historical values, and then allows it to run. The idea here is to see how well the simulation approximates the historical outcome. Note however that this simulation is not always be expected to replicate the historical outcome, because this would be to assume that the historical outcome was the most likely outcome given the situation at the starting date. As I have discussed previously and will further consider later however, certain events (most importantly the German defeat at Stalingrad) were statistically anomalous given the comparative force strengths at the time, and therefore we do not expect the historical outcome to be replicated in most runs of the simulation.
The second condition is called ‘no surrender’, and is a minor modification of the standard condition in which the massive surrenders to the Western Allies which occurred in April and May 1945 do not occur. The purpose of this is simply to allow the simulation to proceed more naturally without always being forced to terminate in May 1945 because of the collapse on the Western Front.
The ‘no West’ condition is an extreme one, in which the German non-eastern losses and frontline ratio are set to the average of their values for the months prior to the Torch landings (in November 1942), which marks the rough time when Germany began to deplete their forces in the east to deal with increasing western pressure in North Africa, as well as the time when the combined bomber offensive began. The purpose of this condition is to estimate the effect of the Western Allies on the eastern front.
The ‘late Torch’ condition maintains German non-eastern losses and frontline ratio at pre-Torch levels for an additional year, to roughly simulate the counterfactual of what would have happened if the Germans had an additional year before the Western Allies began to open a second front.
The ‘failed D-Day’ condition is a counterfactual based around what might have happened if Operation Overlord failed, and Germany had managed to push the Allied landing in Normandy back into the sea. The assumption is that, had this occurred, the Allies would not have been able to launch a similar operation for another 9 months. This is a largely arbitrary number, a compromise between 6 months and 12 months. It is very hard to say what the allies would have done if D-Day had failed – I don’t believe any definitive plans were made. I suspect the Americans would have pushed for an even bigger landing in France at a later date, which would almost certainly have required another year’s preparation. Churchill might have pushed for a smaller-scale intervention somewhere like Greece or Norway, which could have caused problems for Germany, but probably would not have had a major impact on the war, as Germany pulled out of Greece in November 1944 anyway, while Germany’s large garrison in Norway could probably have held out for many months in the mountainous terrain. Whatever the case, for purposes of illustration I have assumed that in the event the Normandy landings failed, the Germans would have reverted to the western losses and frontline ratios of nine months previous, which then would gradually return again to their levels in June 1944. This is a way of implementing the idea that Germany would have been able to temporarily transfer some forces away from the West to the east, as the Western Allies would not pose an immediate danger for at least several months as they regrouped.
The final condition called ‘early mobilisation’ differs from the others, as it starts at the standard date of April 1942, and does not alter the frontline ratios or Western casualty numbers. Instead, the purpose of this condition is to simulate what might have happened if the Germans had more fully mobilised their available manpower earlier in the war. Specifically, this condition assumes that Germany mobilised an additional one million men between mid-1940 and mid-1941, and correspondingly fewer in the years 1942 and 1943. The total number of recruits is not changed, just the time of their induction into the Wehrmacht. This reflects an assumption that the Germans did not sufficiently prepare for a lengthy campaign of attrition in the east, and therefore did not have sufficient numbers of trained reserves available to replenish their forces in 1942, after the severe losses of Barbarossa and the Soviet winter offensives. This is reflected in German eastern front strengths we saw previously, which dip noticeably in mid-1942, indicating the lack of sufficient replacement troops.
Wehrmacht recruitment by year
To motivate this condition more fully, it is helpful to consider this graph showing Wehrmacht recruitment by year up until 1944 (only annual data are available). I have divided the recruits into rough cohorts based on their training status: WW1 veterans, men from the ‘white years’ who had little or no military training, cohorts who had received some training before the war, those who had been fully trained prior to the war , and those who turned 18 during the war (marked ‘trained during war’). This diagram shows that by 1941, the Wehrmacht had mostly exhausted its reserves of WW1 veterans and those who had received training before the war. However, there were still sizeable numbers of men from the ‘white years’ period (mostly in their 30s during this time) who were not trained before the war, but who were eventually inducted into the Wehrmacht during the conflict. These recruits are indicated with a yellow box, and number approximately two million.
The significance of this is that these men were available for recruitment during 1940-41 (unlike the top darker grey region, who weren’t yet of age at that time), but simply had not received training yet, even though they would eventually be called up and receive training later in the war. The reasons why these men were not called up earlier are many and varied. A large factor was the reluctance to withdraw so many men from the war economy. This was only overcome gradually, because of a combination of necessity brought about by defeats, increased restrictions of the civilian economy freeing up additional labour, and increased usage of foreign workers. Another factor was the limited training capacity of the replacement army, which had the officers, resources, and capacity for training only so many new recruits at once. This capacity was greatly expanded during the war to meet the needs of training multiple cohorts simultaneously. A final factor was the belief of Hitler and many of his generals that a long campaign against the Soviet Union would not be necessary, and hence there was no need to withdraw so many men from industry. Indeed, planning had already begun for a shift of priorities to the Luftwaffe and Kreigsmarine, which would be the key services needed to defeat the United Kingdom, after the Soviet Union was conquered. This type of air and sea-based warfare would require significant industrial resources, but fewer personal in the armed services.
My contention is that the German general staff were unrealistic about the prospects of a short war against the Soviets. Thus the ‘early mobilisation’ counterfactual is used to estimate how much difference it would have made, if one million additional men been called up and trained during the period between the defeat of France and the invasion of the Soviet Union. While the Heer did expand and reorganise considerably during this period, I believe that more could have been done, and more men could have been trained as available replacements, if the will and commitment to do so had been present. Germany eventually managed to train all these additional men during a time when it was far more engaged on various fronts, and under more pressure from allied bombing attacks. If they could train these extra cohorts under those conditions, I believe they could have done so in the relative peace of mid-1940 to mid-1941.
As for the issue of withdrawing manpower from the economy, this would have imposed additional burdens, but I believe that solutions could have been found, as indeed they were later in the war. The early withdrawal of an additional one million men from the war economy could have partly been achieved by reducing the number of discharges of WWI veterans following the fall of France. I have not been able to get very good quality data on the number of such discharges, but they certainly numbered hundreds of thousands. My own series of estimates shows that during this crucial period, about two million men were discharged from the Wehrmacht – though I suspect a sizeable number of these were later re-inducted when the military situation deteriorated. Additional labour could also have been freed up by slightly depressing the civilian sector (which happened later anyway), bringing more women into the industrial workforce, and making greater use of foreign workers (again, which happened later anyway). A few hundred thousand from each of those sources over the course of a year would have been quite achievable, had the political will been present.
German industrial misallocation
One final concern about the ‘early mobilisation’ counterfactual is whether Germany would have been able to produce the equipment for these additional soldiers. It is important to bear in mind that these one million additional soldiers are not all deployed on the front in the simulation. Only a few hundred thousand (governed by the frontline ratio) are actually deployed on the eastern front. The one million addition manpower is spread over the entire Wehrmacht just like the historical recruits were. Hence the increase to the frontline strength of the Germans in this simulation relative to historical levels is only about 15% (from 2.7 in 1942 to about 3.1 million).
As to whether even this modest increase could have been equipped by the Germans during 1942, I believe the answer is clearly yes. This slide shows some statistics to illustrate how misallocated German industrial resources were during the critical period between the fall of France and the invasion of the Soviet Union. We see that between Q2 1940 and Q3 1941, the share of steel allocation to the Heer, (a useful proxy for overall industrial priority), fell from 40% to 20%, before quickly rising back up again over the course of 1942. This decline in inputs is reflected by a proportional decline in outputs, as shown by this second graph, in which the share of output for the Heer fell from 60% in 1939 to 30% in 1941, before rising back to above 60% by 1943.
This reduction in the priority of the Heer was a deliberate decision made on the assumption that victory over the Soviet Union would be swift, and hence resources should be diverted to the Luftwaffe and Kriegsmarine to prepare for the war against the UK and USA. This made strategic sense if the assumption of a short campaign was correct. As many have argued, however, this assumption was based largely based on faulty intelligence, gross underestimation of the Soviet state, wishful thinking, and racist attitudes towards Slavic peoples. Had proper planning been carried out to accurately assess the manpower and industrial capacity of the Soviet Union (analysis which was only carried out in early 1942), it should have been clear that the Germans that they had no realistic chance of defeating the Soviet Union in 1941 – due to logistical limitations if nothing else. As such the Germans should have planned on a campaign that would last at least two years. Had this been their assumption at the outset of planning, resources would not have been prematurely diverted away from the Heer, and significantly more tanks, artillery, shells, and other equipment could have been available for the eastern army, easily enough to supply the modest increase in personal strength in my counterfactual.
The final graph here shows the mobilisation of Germany, the Soviet Union, and the United Kingdom, measured as a fraction of GDP devoted to the military. Such comparisons are problematic for numerous reasons, probably the most significant of which here is that all three nations received significant foreign military assistance – Germany from its European conquests and satellite nations, and the UK and USSR from American lend-lease. These inflate the percentages, since they represent resources made available for war uses that did not have to come from domestic production. However even in 1940, before these became major factors, we see that the UK was slightly more mobilised than Germany, despite starting from a much lower baseline in 1939. Also note how rapidly the USSR managed to mobilise between 1941 and 1942. By comparison, Germany’s mobilisation was much slower, even if by the end of the war they surpassed both the UK and USSR. My contention is that although the German economy was highly mobilised in 1940, there was still significant slack available, which was only utilised relatively slowly compared to the UK and USSR. Even an increase from 0.40 to say 0.45 in 1940 would have meant a 10% increase in military output. Alternatively, manpower could have been withdrawn from the civilian economy and diverted into the Wehrmacht, while maintaining output at the same level.
The point here is to reinforce my argument from earlier, that in in the period mid-1940 to mid-1941, there was enough additional capacity available in the German economy to train and equip hundreds of thousands of additional replacements for the army in the east, had there been the will and plans to do so. Thus I believe my ‘early mobilisation’ counterfactual, although perhaps implausible in the light of Hitler’s attitude towards the Soviet Union, at least represents a possibility that was within the ability of Germany to implement at the time.
We will now examine the results of the simulation model. First, I will present several illustrative runs, to show how the model works, and help you to visualise the key factors at play. Note that these are only single runs, and there is a lot of variation from one to the next, but I have chosen them to illustrate particular aspects of the model.
Individual simulation runs
The first example we will consider is that of a German victory in the standard simulation, here starting in August 1942. On the screen here you see a table of the field strengths, losses, total armed forces strength, and advance rates (both monthly and total). On the right you can see a map depicting the progress of the simulation. Note that the map is based on the rate of advance as determined by the simulation, however the specific regions that change hands are predetermined and are for illustrative purposes only, as the model has no ability to simulate gain or loss of specific territories.
In this run of the simulation we see some back and forth over the course of 1942 and early 1943, but in the course of this fighting the Soviets experience much higher losses than the Germans. Notice that Germany experiences lower than historical losses in 1942, and hence is able to sustain a larger than historical strength on the eastern front. This in term means a lower strength ratio and higher relative losses for the Soviets. As a result, by late 1943 you can see that the Soviet manpower situation is becoming critical, and they are unable to maintain a large enough field army strength at the front. This is due to a combination of the ongoing heavy Soviet losses, as well as the failure of the Soviets to recover territories in the Ukraine which hold several million potential recruits. While intense fighting continues during 1944, eventually the collapse of the Soviet military leads to German victory in early 1945 .
Now considering the case where the Soviets achieve victory, we note that despite experiencing significant losses, the Soviets manage to recapture the east bank of the Dnieper by the end of 1944, providing them with significant numbers of additional manpower. We see also that the strength ratio is consistently much higher than in the previous scenario, leading to proportionally higher casualties for the Germans and lower casualties for the Soviets. By the end of 1944, Soviet frontline strength has been depleted, but the German strength has been depleted even more, so the strength ratio is still higher than it was in 1942. This allows the Soviets to continue their advance and ultimately achieve victory by mid-1945.
Turning now to the ‘early mobilisation’ condition. This earlier mobilisation of reserves allows Germany to replenish losses more fully, thus allowing them to field a larger army in 1943 than they did historically. This in turn enables them both to hold the eastern Ukraine, and also inflict greater losses on the Soviet Union. The consequences are by now familiar: the Soviet Union experiences higher losses, which depletes their frontline strength, preventing them from recovering recruits from liberated territories, and also leading to even higher losses in the future. This particular simulation depicts a case where the Soviet Union is forced to surrender in early 1945.
Considering now the ‘late Torch’ scenario, where operation Torch is delayed by one year. Here we observe that Germany is able to build up to a force of 3.4 million (the maximum allowed in the simulation), owing to the fact that they do not need to divert forces to respond to the Allied Landings in North Africa. As with the previous simulations, this greater strength leads to much higher losses for the Soviet Union, and enables Germany to maintain occupied territories, thus cutting off the Soviets from those reinforcements. As before, this leads to a gradual grind down of Soviet reserves, leading to German victory by early 1945.
The final scenario to consider is that of the possibility of the failure of the Normandy landings. While this does allow Germany to redeploy some forces to the eastern front, in most cases this is simply too little too late, and the Germans are still defeated. However, in a small fraction of simulations, Germany is able to perform a particularly successful counteroffensive, which inflicts significant losses to the Soviet forces and recaptures some territory. Germany then has to redeploy some of their forces back to the West, but the huge losses sustained as a result of the German counteroffensive have weakened the Soviets even more, leading to a lower strength ratio and hence higher Soviet losses. In this particular simulation, the Soviets are forced to surrender in mid-1946.
Of course, the examples I have just shown are only individual runs, and as I have noted, there is a great deal of variation from one run to the next. To provide an overall picture of the outcome of a given simulation condition, we can plot a histogram of how often each nation achieves victory, and how long the war lasts in each case. In the historical condition starting in April 1942, we observe that the war typically lasts until the middle of 1945. These results indicate is that absent catastrophic and (according to my model) statistically unlikely defeats such as Stalingrad, the Germans would likely have defeated the Soviet Union, but doing so would have taken about 4 years. Whether Germany would have been able to defend itself against the Western Allies for this long in the counterfactual case of Soviet defeat is another question entirely.
Considering now the histogram beginning in April 1943, we see that the results are similar, but now with a much larger proportion of Soviet victories, given that the starting point is after the German defeat at Stalingrad.
Now examining the histogram for the early mobilisation condition starting in April 1943, we see that the proportion of Soviet victories has shrunk dramatically. This reflects the greater ability of the Germans to replenish their losses after Stalingrad, if more trained reserves had been available. These results indicate that Germany could still have defeated the Soviet Union after their losses at Stalingrad, but insufficient preparation earlier in the war meant the necessary replacements were not available early enough. Also note that the duration of the war is largely unaffected. Again, this leaves open the question of how Germany would have fared against a further intensification of effort by the Western Allies during 1945, which is not incorporated into the simulation.
Finally, we see here the histogram for the failed D-Day condition. Note that the Soviet Union still defeats Germany in the majority of cases, and the times when Germany does win involves them prolonging the war into 1946/47, so that they have sufficient time to wear down Soviet reserves. Again, whether Germany would have been able to survive for an additional 18-24 months against further intensification of Western attacks is a further question. Nonetheless, this shows that had the Germans been able to defend against the invasion of Normandy, there is a chance that they could have successfully counterattacked in the east and worn down the Soviets enough to at least bring about a negotiated peace.
Estimates of German victory probability
It is now time to consolidate the information from all the simulation conditions. This graph shows all simulation conditions that we have considered so far: the historical case, the early German mobilisation case, the late Torch case, and the ‘no West’ condition. Rather than selecting a single initial point for the simulation to begin, this graph plots the proportion of German victories starting at each month from March 1942 up to March 1945. I have also fitted a sigmoid curve to each set of points, which represents a guess at Germany’s overall chances of being able to win the war as of that date, in the relevant simulation condition. Remember that the probabilities here represent aspects not captured in the simulation, including weather, political decisions, operational plans, the vicissitudes of battle, and all such factors.
Starting with the historical condition, we must first discuss the three aberrant points of April, May, and June 1942, which are clearly far off from the others and were not used in plotting the sigmoid curve. The reason for these highly divergent results actually stems from a quirk in the data used to calculate German frontline strength. As shown here, there is an unexplained peak in the Wehrmacht strength in April 1942, and corresponding drop in the months afterwards. I have decided to leave this in the data, as it is present in the source material, though I have no explanation for this anomaly, and I suspect it is a statistical artefact.
Returning to the main plot, we see that Germany’s estimated chances of winning fell rapidly over the course of 1943. Much of this occurred as a result of the defeat at Stalingrad, which alone was responsible for at least a twenty percentage-point reduction in Germany’s chances of winning. By the end of 1943, Germany’s chances of victory had fallen to below ten percent, which is my estimate of how likely it is that they could have successfully fended off the allies at D-Day and then counterattacked against the Soviets. Even this residual chance is gone by September 1944, after which the chance of achieving even a limited victory against the Soviets is effectively zero. It is sobering to note that of Germany’s five million total military deaths during WWII, a full two million, forty percent of the total, were sustained during the final ten months of the war, by which time there was not even the faintest hope of changing the outcome of the conflict.
Considering now the early mobilisation condition, we see that German victory is a near certainty until the second half of 1943. This counterfactual must be interpreted with caution, as the simulation assumes that the war progressed as per the actual historical outcome until the start month of the simulation. Had Germany mobilised greater numbers of reserves during 1940-1941, however, obviously events wouldn’t have proceeded as they did historically all the way until 1943. The purpose of this graph is therefore to provide a way of illustrating how much difference the counterfactual scenario makes to Germany’s chances compared to the historical case. One way to see this is to observe that the early mobilisation condition pushes back the time when Germany’s chances of victory fell below 50% a full year, from the end of 1942 to the end of 1943.
A one-year delay of Torch has an even greater effect, delaying the 50% cut-off a further six months until mid-1944.
Finally, removing the Western Allies altogether practically guarantees that Germany would win the war, pushing the 50% cut-off to early 1945. This interesting result suggests that had the Western army, aerial, and naval attacks against Germany suddenly ceased in January 1945, and had Germany been able to immediately redeploy its forces to the east, the chances of them defeating the Soviet Union by this point would still have been around 50%. Some may well find this implausible, but recall how hard-fought were the campaigns in the east up to the very end, and how heavy Soviet causalities and material losses continued to be. This was even though by this point in the war, Germany was only deploying about half of its army manpower against the Soviets, which meant far less than half of its overall armed forces. Combine this with the fact that Soviet reserves were very low by this point of the war, and the result begins to look a lot less absurd in my view. Also note that the simulation does not factor in the effects of lend-lease, which would have been absent in this counterfactual ‘no west’ scenario, thereby further weakening Soviet forces.
Industrial production comparison
As a final accompaniment to the simulation data, this slide shows my estimates of the total economic output devoted by Germany and the Soviet Union to supporting their military activities on the eastern front. For the USSR this is effectively just GDP times the fraction spent on the military, but for Germany adjustments must be made for the estimated proportion devoted to the eastern front, as well as adjustments for military contributions of its occupied territories. What we observe is perhaps not surprising. After having the lead in 1941 while the Soviets mobilised and rough parity during 1942, from 1943 onwards Germany had to devote more and more of its resources towards other fronts, while increased lend-lease and recovering lost territories helped boost Soviet output. German production also continued to be hampered by allied bombing, allied blockade, lack of critical raw materials (particularly oil and rubber), and lack of industrial manpower. Indeed, the content of this graph is one of the main reasons why Hitler planned on a short war against the Soviet Union, since it was his belief that if they failed to win a decisive early victory, Germany did not have the industrial capacity and manpower reserves to fight a prolonged war of attrition against the Soviet Union and the United Kingdom, both of which backed by the industrial might of the United States. This perspective is reflected in Hitler’s prescient quote on the matter from summer 1942.
Hitler was correct. Germany failed to capture the oil or achieve a decisive victory in the east by the end of 1942. Thereafter, increasing pressure from the Western Allies meant that Germany was unable to deploy the forces necessary to defeat the Soviet Union.
Final summary and conclusions
Let us now summarise the key findings from the past two videos:
- Germany failed to prepare for a prolonged war in the East. Had they done so doing the period 1940-41, they would likely have defeated the Soviet Union in 1943/44.
- Germany’s defeat at Stalingrad was critical. Before this the chances of victory against the Soviet Union were about two thirds, while afterwards this fell to less than one third.
- Soviet manpower reserves were close to exhausted by mid-1943. Only through recruitment from liberated territories were they able to maintain force levels, given their enormous casualties. This feedback effect of success providing more recruits leading to more success was vital, and is underappreciated in conventional historical analysis.
- When adjusted for their numerical superiority, the performance of the Soviet military was poor. On average, the Soviets inflicted about one sixth as many casualties per man as the Germans, and about one tenth as many kills per tank.
- The increasing diversion of German men and material to fight the Western Allies was decisive in enabling Soviet victory in the East. Absent increasing Western pressure from late 1942 onwards, it is unlikely the Soviet Union would have been victorious.
- Following the success of Operation Overlord in June 1944, German victory was essentially an impossibility. About half of all German military deaths, some 2.5 million, occurred after this time.
The outcome of the titanic struggle between Germany and the Soviet Union defined the course of the twentieth century. Soviet Victory meant the end of Germany as an independent military power, and the domination of half of Europe by the Soviet Union for the next fifty years. Every nation in central and eastern Europe was profoundly affected by the conflict, and its consequences still loom large to this day. Hitler and the German high command were fully aware of the monumental task they were undertaking when planning Operation Barbarossa. It is astounding, therefore, that so little effort was made to plan for the possibility of a prolonged conflict. By contrast, Soviet mobilisation following the German invasion was swift and intense. Furthermore, the allies made good use of their material advantages and cooperated effectively to bring the war to a victorious end. Poor German planning and ineffective use of resources lost them the war; good Soviet planning and effective use of resources won them the war.