,
26 tweets,
4 min read
Read on Twitter
A few months ago, I promised I would post another thread on the contributions of economists and statisticians to the Allied victory in world war 2, but that I would need to double check some things first. I will tell the stories of 2 successes and 1 failure.
(First of all: I got all these stories from Mark Guglielmo, The contributions of economists to military intelligence during world war 2, The Journal of Economic History, Vol 68, N° 1, March 2008)
OK, let's get started. For those of you who are not World War 2 nerds, the first major US operation on the European theater was the landing in North Africa in November 1942, which was then held by the French Vichy Government, which collaborated with the Germans. /
The objectives of the campaign were 2 fold. First, it was hoped (correctly, so it turned out) that, after some token resistance, the Vichy troops would switch sides and that Morocco, Algeria and Tunisia to the Western Allies could serve as a forward logistic base against Italy/
The second objective was that, almost one year after entering the war, it was important to show that the US "was doing something", even if it was not invading France directly. Anyway, I am deviating /
In our world that is accustomed to GPS and Google Streetview, it is difficult to imagine that the US planners were facing a major issue: how to obtain detailed and reliable maps, especially of the build-up areas, of a country that had been completely off your strategic radar? /
No, seriously, where would you get your information? If the military has not build up stock of maps before the war, how could the Allies obtain detailed maps of, let's say, Oran if reconnaissance flights were excluded?
So it was a bit of a conundrum until someone pointed out that US insurers were massively involved in the re-insurance of properties in North Africa, and has thus extremely detailed maps, up to the level of individual buildings. Success n° 1.
Story n°2. How to estimate German casualties? The Germans would be so stupid as to publish casualties in the newspapers, would they? Except that they did, for one key category of their personnel - and that they unknowingly provided all the info needed for extrapolation. /
For some reason, there was a German tradition to publish obituaries of officers who were killed in action - including rank, unit and the place where they fell. The only drawback is that this information was published in local newspapers and was not centralised/
So what the Americans did, was to set up a sampling of newspapers published in neutral countries and aggregate manually all the information contained in those papers. In a world before automated text reading, this was a daunting task /
Anyway, this was just the first step. After all, while it's interesting to know how many officers die in battle, you also want to know total casualties. How do you extrapolate from one variable to the other?
You really need two additional sources of information. First, you need to know the ratio officers/(NCO + enlisted men). That wasn't too difficult: the German organisation tables were available. Second, you need to know that ratio of casualty rates among officers compared to the /
NCOs and enlisted men. The Americans used the data on this subject from world war 1, (If you are interested: it's 21 to 1, confirming that casualty rates are higher among officers) /
So how accurate were these predictions? It depends. they were rather close for the Western Front, but estimated casualties on the Eastern Front were only half so high as estimated. Thus, for some reasons, officers were killed in action more quickly on the eastern front. /
Thus, while this story illustrates the power of extrapolating from sparse public data if one is willing to make some assumptions, it also shows the limitations of the method. This really shows why it is important to use several approaches to obtain an estimate, whenever possible/
Third story, and the one with the most radical implications: how to estimate German war production? Here, German thoroughness would be very helpful. Indeed, every single piece of German equipment contained information on: name and location of the manufacturer, the date of /
manufacture and the serial number. The cave of Ali Baba for applied statisticians, if the Allies could capture a sample of it - which, with the Germans mostly in the retreating mode, was indeed the case /
Kind of information that could be gathered from it: the level of concentration in each industry, the total production per month and the delay between the production of intermediate production and its use on the frontline. /
This gave an idea of the level of vulnerability of each individual market and, also, with some delays, of the effectiveness of the bombing campaign. For instance, the Americans could establish that gearboxes for tanks were manufactured in just two plants /
This implied that the tank industry was highly vulnerable to targeted bombing campaigns. On the other hand, the production of ammunition was thinly spread over the territory of Germany, and would thus not be a suitable target for strategic bombing. /
This is an area where extrapolation based on sound scientific statistical analysis was hugely successful. Economists estimated that around 186,100 tires were produced per month - the actual number was 175,500 /
Tank production was estimated at 327 units per months - the actual figure was 342 units. This should be compared with the competing estimate (based on interrogations of prisoners and prewar production) of 1,550 units /
There are numerous other examples of studies performed during the war, ranging from estimates of the share of the Soviet industry destroyed during the war (and its capacity of recuperation after the war was over), labour supply in Japan, /
how long it would take before mass famine would set it in Japan as the result of the destruction of its merchant fleet, etc etc. All of them illustrate how military intelligence is not just a "James Bond type" profession, but depends heavily on very wonkish work.
