How many Italian prisoners, really?

The Australian War Memorial holds the document AWM 54, Written records [1939-1945], 780/1/6, Report on the Directorate of Prisoners of War and internees at Army Headquarters, Melbourne, 1939-1951.

Volume 1, part 2 is about Enemy prisoners of war, and pages 101 to 106 list all the arrivals in Australia of Italian POWs.  On page 106, you will find a summary table of which I reproduce here below the part about Italian POWs:

 This figure of 18,432 is what everybody quotes, including the most authoritative authors, like  Gianfranco Cresciani in "Captivity in Australia: The case of the Italian prisoners of war, 1940-1947". Studi Emigrazione / Etudes Migrations, 26(94), 1989, 195–220, on page 204, and Desmond O’Connor in "From Tobruk to Clare : the experiences of the Italian prisoner of war Luigi Bortolotti 1941-1946". Fulgor, 1, 2003, 69–85. , on page 69.  Also Alan Fitzgerald, in his widely quoted book, The Italian farming soldiers: prisoners of war in Australia, 1941-1947, Melbourne: Melbourne University Press, on page 191, reports the magic figure of 18,432.

The problem with that figure is that if you read the text of the original document, rather than just looking at the total, you will see that only 256 seamen were first interned and then held as POWs, not 268 as reported in the summary table.  As a result, it is reasonable to conclude that the Italian POWs were in fact 18,420.

My study of the list of 44,513 POWs and internees of all nationalities made available by the Australian National Archives as series MP1103-1 seems to confirm the lower number.  I eliminated as carefully as possible all entries of non-Italian POWs and all entries of Italian internees and ended up with 18,415 names.

For a while I had in the list exactly 18,420 names!

This series MP1103-1 includes the Service and Casualty Forms of all POWs and internees.  The military authorities filled in those forms when the POWs arrived in Australia.  If it turned out that there are indeed 18,420 forms for Italian POWs, the universally accepted figure of 18,432 might have to be reconsidered.  In any case, it can no longer be accepted at face value!

In case you are wondering, I'm still working on the PhD.  But the focus of the PhD has somewhat shifted from telling the stories of POWs to how to tell those stories.  I'll try to write more often.  The huge void before this article is shameful...

Most returning POWs came from the South

I started adding to the 554 records of Italian POWs who returned to Australia some of the information contained in their Service and Casualty Form (the green forms of which you see an example in italianpow.info/2013/02/passengers-arrivals.html).

In particular, I’m adding the military rank (e.g. Cpl for Corporal), the date of capture (e.g., 1941-4-11), when they were captured (e.g., Amba Alagi), the place of birth (e.g., Palermo), the profession (e.g., Engineer driver tractor), the marital status at the time of capture (e.g., single), the service (e.g., Army), the ship boarded for repatriation (e.g. Otranto), the date of repatriation (e.g., 1947-01-10), and a flag to indicate whether they were “parked” in India before being sent to Australia.

So far, I have processed 116 records, and discovered that three of them were of internees. This is easily explained: with 18,550 records of POWs extracted from MP1103/1 and only 18,420 POWs counted in AWM54 780/1/6, I expect my list of POWs to contain at least 130 internees. Three of them must have travelled to Italy and I discovered them when they returned back to Australia.

After processing about 21% of the 554 records, I couldn’t resist the temptation to start making some statistics.

The first thing I did was counting how many of the 113 processed records referred to singles. It turns out that 63.7% were singles and 36.3% were married. These figures obviously can change once all 554 records will have been processed, but they already give a good indication of the final result.

Somehow, it doesn’t surprise that singles were more adventurous.

Next, I checked where they came from. For two of them, the place of birth was not clearly legible, but the remaining 111 provided an interesting result. First of all, here are the counts for all Italian regions (listed from North to South):

Piemonte	1
Valle d'Aosta
Lombardia	3
Trentino-Alto Adige
Veneto	5
Friuli-Venezia Giulia
Liguria
Emilia-Romagna	3
Toscana	3
Umbria
Marche	1
Lazio	2
Abruzzo	2
Molise	4
Campania	23
Puglia	9
Basilicata	3
Calabria	24
Sicilia	28
Sardegna
Total	111

Here is a map of the regions downloaded from www.mapsofworld.com/italy/regions.html to help you visualise them.

Clearly, many more of the returnees were from southern regions. Without counting the two people from Lazio, which is the region used as a demarcation between North and South, 14.7% came from the North, and 85.3% from the South. A ratio of almost six Southerners for each Northerner.

Again, I will repeat the calculation once I will have processed all 554, but I see no reason for expecting much change.

The figures become even more dramatic when taking into account the fact that the population of Northern Italy was larger than that of the South. Using the results of the 1951 Italian national census (the closest date to the return of the POWs to Australia) as a correcting factor, I calculated that the POWs from the South were close to nine times more likely to return to Australia as migrants.

I see this as a clear indication of the fact that life in the South of Italy was much more difficult than in the North. The South had always been (and, to a certain extent, still is) less developed than the North. Therefore, it is no surprise that more Southerners were prepared to leave their country to seek fortune Down Under.

To confirm this conclusion, I will need to check the ratio of Northerners vs. Southerners among all 18,420 POWs. If it turned out to be heavily biased in favour of the Southerners, that could be a reason for the unbalance in the numbers of those who returned.

To avoid having to obtain the place of birth from 18,420 Service and Casualty Forms, I will select a number of them at random and only check those. I will estimate the statistical error of the resulting ratio, but I am confident that a comparatively small sample (perhaps 1%?) will be sufficient to check whether the Northerners/Southerners ratio of the POWs roughly reflected the ratio of the whole Italian population.

After all, the significant result is that many more Southerners returned, not the precise ratio with which they did so.

Number of Returning POWs

I finally managed to compare the list of 18,420 Italian POWs with the list of 203,813 people who arrived in Australia via Fremantle (WA) or transited there on their way to one of the Eastern ports.

I found 554 entries with matching family and given names.

I had actually expected a higher number of matches. Desmond O’Connor, professor of Italian at Flinders University (Adelaide), in his article From Tobruk to Clare: the experiences of the Italian prisoner of war Luigi Bortolotti 1941-1946, which you can read online or download in PDF format, estimates that 9.4% of Italian POWs held in South Australia returned to Australia as migrants.

If we assume that the same percentage applies to POWs held in all states, we arrive to a figure of about 1,700 returnees (9.4% * 18,420 = 1,731). Further, if we assume, as it seems reasonable to do, that many independent factors influenced the decision to return to Australia, we can apply the Central Limit Theorem (CLT) to estimate the degree of approximation of the mean.

The CLT, for those who are not familiar with it, states that the mean of a sufficiently large number of independent random variables is normally distributed. In practical terms, it means that if we repeatedly measure something that depends on many independent factors and make a histogram of the values we measure, the plot will approximate the familiar bell shape of a normal distribution (thanks to Wikimedia Commons for the following image).

With a normal distribution, we can roughly estimate its standard deviation (indicated with the Greek letter σ in the figure) by taking the square root of its mean. In practical terms, it means that with a mean of 1,731, the standard deviation of the number of returning POWs nationwide is 42. This in turn tells us that, according to O’Connor’s estimate, the actual total number of Italian POWs returning to Australia has a probability of 68% to be between 1,689 and 1,773. And the probability of the actual number being less than 1,605 (3σ below the mean) is less than 1%.

My figure of 554 is so low as to be incompatible with O’Connor’s estimate. Now I have to figure out whether one of us is completely off the mark and why...