First go at a graph of British killed in the Second World War

After a fair share of swearing, toil, tears and sweat, I’ve come up with a first attempt at a graph representing the numbers of Britons killed during the Second World War. As I’ll explain, this has involved a fair amount of fudging and compromise, and I would not claim that what I’ve produced is perfect, but I still think the project is worthwhile and I’ll also give some examples of how I think it’s been useful.

A. Sources of imperfection

1) Casualties, deaths and killed.

In other posts in this sequence, you can read about some of the problems I’ve faced. At a fairly early stage, I decided that I would have to graph deaths, not casualties or losses. First, deaths are final, whereas an individual could become a casualty several times: not a problem in itself, but potentially a source of misleading interpretation. Second, the military counts PoWs as losses, and important as this was for their purpose, their loss has a rather different emotional quality to death. In addition, their return at the end of the war played havoc with statistics. Since even wartime death could be defined in very different ways in different services, I tried as the basis for my work to use contemporary definitions. So these figures don’t include those who died ‘of natural causes’ during their military service, but the RAF figures do include those who died from training accidents as well as during ground or flying battles. For the civilian figures, I drew on the reports from Registrars General in Fighting with Figures, which give a higher figure than those calculated by the Ministry of Home Security at the time, and include those killed as passengers at sea as well as those killed by enemy bombing. Even so, looking at the classifications available to Registrars under the international system in use at the time, (a modern version here, which hasn’t changed much) it’s plain that there wasn’t room to stretch the boundaries of ‘war-related’ death too much.

2) A cumulative graph

Wartime deaths were not always reported at the time they occurred. It therefore makes more sense to produce a graph of cumulative casualties rather than month by month deaths. This also helps with point 3)

3) A complete run of results?

For none of the armed forces is there a complete list of month by month casualty figures in the National Archives. For the Army and Air Force, there are such runs for part of the war, but not all of it. For the Navy and Merchant Navy, I had to calculate these myself. The civilian casualty figures produced from air raid reports at the time are complete from March 1940, but omit losses from other sources (you can read Richard Titmuss’ thoughts on civilian casualty figures here). For the Army and RAF, I have therefore sometimes had to rely on estimates and averages. Where I can, I have tried to include extra information here to give better ‘texture’ to the averages. For example, there are no cumulative figures for Army deaths before mid 1941. Rather than average the results over the months in between, however, I have taken information from the contemporary records of casualties by campaign to indicate the losses of April-June 1940 and the desert/Mediterranean campaigns of 1940-41. At other times, averaging has been simply unavoidable, for example with RAF losses between Marcy 1944, May 1945 and the end of the war.

4) Comparable results.

Wartime casualty figures have the advantage of giving a month-by-month account of losses. They have the disadvantage that they are almost always inaccurate, usually in that they underestimate losses. The army’s figure for its cumulative losses produced in the summer of 1945, for example, is significantly below that given in June 1946 in the White Paper Cmd 6832. Partly, that’s because people kept dying after the war’s end, but mostly it’s because it took time for combat accountancy to catch up with actual deaths. Such figures, however, are given as totals, rather than as month by month losses. How to solve the problem of using a set of comparable and accurate figures but retain the ‘feel’ of what was happening during the war? Here is where I’ve done the greatest bit of ‘fudging’ which I am least comfortable with statistically. With the exception of civilian figures, I have used those given in Cmd 6832, and taken those as broadly representative of the ‘reality’ at the end of the war. I realise that they include ‘extra’, post-hostilities deaths, but I also know that compared to the figures produced by a year by year search of the CWGC database, Cmd 6832 still seems to represent a conservative estimate, so those postwar deaths will be matched by those missed whilst battle was going on. I have assumed that wartime under-reporting was consistent as a proportion of casualties, divided up wartime totals by monthly figures, then allocated the post-war figures of Cmd 6832 in these proportions. I suspect this is probably a bogus assumption: it seems logical to suggest that reporting would be more subject to error at times of greater stress on the system, so that my method if anything flattens out losses over time. But I can’t see a way around this whilst maintaining my sanity. Fortunately, it doesn’t seem to have created any really anomalous results in a graph on the scale I’m working at, but it does mean that I would be hesitant to use my findings to make cast-iron assertions on the precise dates at which, for example, military deaths exceeded civilian ones.

B. Why is this still useful?

1) Visualising the war in one go

I think this graph is very useful in thinking about what Britain’s war was like. The spikes of military casualties at the start and end of the war are very apparent, as is the sudden loss of civilian life during the Blitz of 1940-41 (and during the V-weapon offensive of 1944, although the curve of RAF casualties at the same point risks hiding that). The relatively attritional nature of the wars that faced the RN, RAF and Merchant Navy is also apparent from the shape of their graph lines. The levelling out of civilian casualties after spring 1941, and Merchant Navy casualties after May 1943, also tells us about important swings in Britain’s favour. Using this graph, we could repeat what have become cliched statements about ‘more civilians than soldiers dying before… (insert date)’, but we’d also have to point out the huge increase in military deaths in 1944.

We can also super-impose the lines of the three armed services to suggest that, despite the way the war is usually discussed, it is unlikely that the numbers of civilians killed ever exceeded the number of servicemen and women killed. although the two came close in the spring of 1941. This depends on whether one includes merchant seamen as civilians – something I’m not really happy to do, since I think they deserve their own category (’semi-combatants?’).

2) Some comparative suggestions

On this graph, I’ve included two other data points as a way to encourage some thought about how Britain’s war experience compared to an ally and an enemy. One is an estimate of the number of Germans killed in the 1943 bombing of Hamburg – a one off, not a cumulative figure. The other is the estimated number of Russians who were killed during the Stalingrad campaign alone. These were just two figures that were easily to hand: obviously one could add more depending on what point was being made. The problem with cumulative figures for Germany and Russia is not only reliability, but the difficulty of actually showing such data on the same graph as one for Britain – since the British figures would tend to get lost in a blur at the bottom of a larger scaled graph. This could, of course, be exactly the sort of point one might like to make.

3) Further information and research.

The journey of making this graph has been rather more important and intellectually satisfying than the destination. The issues thrown up by my research – particularly on the place of death and mourning in the myth of the war – are really fascinating, and will definitely influence what I right. Just having easy access to lots of casualty statistics is also useful. For example, I’m currently writing about the Battle of Britain, and being able to compare graphically the casualties from Civil Defence regions in July-August, as opposed to September-October 1940, makes a point about how bombing was experienced differently in those periods quickly and clearly.

I’d really welcome comments, suggestions on adaptations or improvements and queries.