The Armenian Weekly
April 2011 Magazine
“My central argument is that there is no major contradiction not only between different Ottoman materials, but also between Ottoman and foreign archival materials. So, it is erroneous to assume that the Ottoman documents (referring here mostly to the documents from the Prime Ministry Archive) were created solely in order to obscure the actions of the Ottoman government…. Ottoman archival materials support and corroborate the narrative of Armenian Genocide as shown in the western Archival sources.” (Emphasis mine)
—Taner Akcam in “The Ottoman Documents and the Genocidal Policies of the Committee for Union and Progress (Ittihat ve Terakki) toward the Armenians in 1915,” Genocide Studies and Prevention, 1:2, (September 2006): 127–148.
After reading the above by Historian Taner Akcam, it occurred to me that similar assumptions are reflected in the study of pre-World War I populations within the Ottoman Empire. This is particularly true of the various estimates of the Armenian population prior to the Armenian Genocide.
To date, those studying the Armenian population of the Ottoman Empire have either accepted Ottoman registration records as the sole source for analysis while dismissing the records of the Armenian Patriarchate, or vice versa. Occasionally, the “suspect” records are critiqued prior to dismissal, but more often than not they are dismissed superficially or ignored altogether.
Using the Diyarbekir province as an example, I plan to analyze under what scenarios Ottoman government and Armenian Patriarchate records are consistent and thus complimentary.
There existed within the Ottoman Empire a long tradition of tax registers. Throughout the 19th century, a more ambitious registration system developed. At first, adult males were the primary objective for tax and military objectives. Later efforts can be viewed as the foundation for demographic analysis and governmental policy decisions. However, even with gradual improvements in enumeration, the Ottoman registration system never approached full coverage of the population.
While not exhaustive, the following are some of the weaknesses in the data gleaned from Ottoman records:
Women and children were undercounted;
Registers containing non-Muslims have never been analyzed (only summary data have come to light thus far);
Registration systems are inherently inferior to a census;
The sparseness of data complicates evaluation;
There is some evidence of manipulation;
Borders between districts and provinces frequently changed and thus complicate comparisons.
While detailed records do not exist, summary information has appeared in a number of sources, primarily in Ottoman provincial yearbooks and government documents.
During this same period and for many of the same reasons, the Armenian Patriarchate began an effort at enumerating the Armenian population. Similarly, there are inherent weaknesses in the patriarchate data that include, but are not limited to, the following:
Population estimates for Muslims were often included even though the patriarchate had no way of gathering such data;
The patriarchate censuses were often timed with political objectives;
The sparseness of data makes it difficult or impossible to develop a population timeline;
Detailed records are lacking and there is little hope further data will come to light;
There is evidence of undercounting children and other gaps in data.
The primary source for patriarchate data for 1913-14 can be found in two sources: Raymond H. Kevorkian and Paul B. Paboudjian’s “Les Armeniens dans l’Empire Ottoman a la veille du genocide” (Paris: Les Editions d’Art et d’Histoire ARHIS, 1992) and Teotik, “Goghgota Hai Hogevorakanutian” ed. Ara Kalaydjian (New York: St. Vartan’s Press, 1985).
While most scholars have used the Ottoman statistics unadjusted or made simple aggregate level adjustments, historian Justin McCarthy utilized stable population theory in an attempt to compensate for the known deficiencies. McCarthy’s work is often cited with frequent praise and occasional criticisms, but rarely from a mathematical perspective.
McCarthy utilizes age-specific data from the early 1890’s to calculate an adjustment factor that corrects the aggregate population for the undercounting of women and children. He does so by fitting the known data for males over the age of 15 to standard life tables he deems representative of the population at the time and then doubles the corrected male population to arrive at the total population. Once the adjustment factor is calculated, McCarthy applies this to data from 1914 and then utilizes population growth rates to extrapolate back and forth in time. The graph displays his adjustments for the Diyarbekir province.
There are many issues with such a methodology. First and foremost, applying corrections based on the recorded population 20 years prior is highly questionable and McCarthy fails to fully appreciate the implications. The methodology is further hampered by the existence of only one source for the reporting of population by age groups.
In the specific example of the Diyarbekir province, McCarthy notes that the growth in recorded population from 1892 to 1914 is unrealistically high. He speculates that the reason is due to improved enumeration of the population. Yet, he still applies the same correction factor calculated from earlier data without consideration that some of the improved counting could have originated in the groups that the factor is meant to correct (i.e., women and children).
In addition, as can be seen from the graph, McCarthy smoothed a dip in the recorded male population aged 35-39. However, this is the age group that would have been affected by the Russo-Turkish War of 1877-78. Stable population theory must be utilized cautiously so as not to remove the very real demographic impact of historical events. The issue becomes more acute when it is understood that the factor thus derived is applied unadjusted to the 1914 population. In essence, the recorded males aged 35-39 in 1914 are being adjusted by a factor derived from the population of males who fought in the 1877-78 war when quite reasonably they should not have been adjusted at all.
While population by age is only available in the 1892-93 data, the breakdown by gender is available for other time periods and the ratio of males to females varies by ethnicity and year of enumeration. The adjustment, which McCarthy applied to all ethnicities equally, should be viewed with caution. In fact, while the data limits the ability to reflect ethnic differences, it is a mistake to assume no such differences exist.
While the ratio of recorded males to females for Muslims in the Diyarbekir province was traditionally around 1.20, by 1911 the ratio had dropped to 1.04. Conversely, the ratio for Armenians was traditionally around 1.05 but had jumped to over 1.17. What can we make of this dramatic change and what are the implications when estimating the Armenian population? The interpretation is complicated by the expectation that the ratio of Armenian men to women should have dropped dramatically following the Hamidian Massacres of 1894-96, which targeted almost exclusively men. However, this could have partially been offset by the forced conversion of Armenian women to Islam. In addition, there is the emigration of Armenian males to consider.
Another way to state the problem is to refine McCarthy’s methodology for the differences in male to female ratios. Based on the life table McCarthy employed, he arrived at a factor of 1.1313 to adjust the male population for the undercounting of young boys. The overall factor, then, for any time period and ethnicity would equal (2 * 1.1313) / (1 + females / males). McCarthy’s resulting adjustment factor based on 1893 data and that ignores ethnicity is 1.2142 (through an error in McCarthy’s calculations, he uses 1.2172). If instead one were to use the 1911 data, the adjustment for Muslims would be 1.1525, while 1.2146 for Armenians.
There is the additional issue of the extraordinary growth in the recorded Muslim population while not quite to the same extent in the Armenian population. McCarthy attributes this to improved enumeration and assumes the improvement is equivalent for all ethnicities. That was not the case and in particular the areas with the greatest concentration of Armenians exhibited the least amount of growth. Not surprisingly, these are also the areas with the greatest differences between the Armenian population indicated by the patriarchate with that of the Ottoman records.
As can be seen from the table above, prior to the Hamidian Massacres Armenians accounted for almost 20 percent of the population in the regions of Chermik, Palu, and Siverek. On the eve of World War I, according to Ottoman records this proportion had dwindled to 10 percent. When compared to the Armenian Patriarchate figures, these three areas account for ~25K of the ~33K difference, even though only one-third of the Armenian population resided in those districts.
Even prior to the Hamidian Massacres, Ottoman records indicated a decline in the number of Armenians within the Diyarbekir province. It was not until 1900 that the Armenian male population recovered, either due to improved enumeration or as part of the post-massacre demographic rebirth.
The central question is under what assumptions do we account for the difference between an Armenian population of 72,124 as stated within Ottoman records to the 105,528 stated by the Armenian Patriarchate?
If we begin with the 1911 Ottoman document, which seems to represent the population as of 1905-06, the Armenian male population is stated as 34,645. The first adjustment is to account for the undercounting of male children. As we have already seen, McCarthy assumed 1.1313 based on data from 1892. If we do not adjust the male population aged 35-39, which assumes the dip is due to higher deaths from the 1877-78 war, then the adjustment is 1.1215. The fundamental problem is that the recorded population is 80 percent Muslim and there is no way to discern whether Armenian children were undercounted to a greater or lesser extent.
In addition, the total population grew by ~26 percent between 1892 and 1906. A more reasonable growth rate would have been 10-11 percent. The additional growth has been assumed to come from better enumeration. So, one could assume that no adjustment need be made for the undercounting of children since improvements in enumeration entirely came from those under the age of 15. While that is probably not a reasonable assumption, it is a possibility that children were counted to a greater or lesser extent in 1906 than in 1892.
In addition, there is the matter of the reasonableness of the life table that McCarthy has chosen. It is beyond the scope of this article to address this issue, but for these reasons
I prefer a range of assumptions. Here I will assume three different adjustments for the undercounting of male children: 10 percent (low), 12.5 percent (mid), and 15 percent (high).
Low Mid High
1906 Recorded Armenian Males 34,645 34,645 34,645
1906 Adjusted Armenian Males 38,110 38,976 39,842
The Muslim population grew by ~14 percent between 1329 Ottoman document and the 1330 Nufus (which is thought to represent the population as of 1914), while the Armenian population grew by ~12 percent. Again, this represents better enumeration plus normal population growth. Either the Armenian population grew at a slower pace or there were greater improvements in registering Muslims than Armenians. For this purpose, let’s assume 10 percent, 12 percent, and 14 percent, respectively.
Low Mid High
1914 Adjusted Armenian Males 41,920 43,653 45,420
Interestingly, this is about 6,000 less than what might be expected based on the growth in the Muslim population. Based on other estimates of the time, this would be an estimate for the number of Armenian deaths during the Hamidian Massacres combined with emigration in the intervening years.
As pointed out earlier, you cannot simply double the male population to arrive at the total population, as Armenian males exhibited deaths and emigration beyond those of females. In addition, conversion to Islam needs to be accounted for. I am going to assume a range of between 0 and 4,000 Armenian women converted to Islam in the years between 1890 and 1914.
Low Mid High
1914 Adjusted Armenian Total Population 85,841 91,305 96,839
This represents a difference from the patriarchate figures of 9-23 percent. From 1890 to 1914, the population of Diyarbekir displayed growth rates that indicate improved registration. Over that period, there was no indication that the trend had leveled or even slowed. Thus, omissions of men over the age of 15 may still have existed.
In addition, there is ample evidence that even in developed countries the undercounting of minorities is greater than the rest of society. For instance, even in the 1990 United States census, African Americans are undercounted almost five times that of whites. Hispanics are undercounted to an even greater extent. Further, the omission rates for African Americans have been estimated to be greater for males aged 15-40 than for ages 5-15.
This is not to say that Armenians within the Ottoman Empire and African Americans within the United States would exhibit the same rates of omission in census enumerations, but it does indicate that differences between ethnicities is a reasonable assumption.
One area that should be looked to for evidence of undercounting of Armenians, whether purposeful or not, is the town of Chungush. Armenian sources indicate a very large Armenian population, yet Ottoman records as late as 1900 indicate only one village containing non-Muslims in the Chermik District where Chungush was located (as well as the towns of Adish and Chermik, which also contained Armenians). The Ottoman records indicate the Armenian population dropped from almost 6,000 in this district to less than 800. The population was well above 10,000 and closer to 15,000. This alone could explain much of the difference.
The analysis above, to a large extent, assumed that the undercounting in the Ottoman registration system was equivalent for Armenians and Muslims. That was most likely not the case. But even with that assumption, the Ottoman records indicate the impact on the Armenian population of policies initiated by the Ottoman government.
Imperfect data is the norm in historical demography. However, even with the flaws in available information, much can be learned from such analysis as that above. The goal is not to arrive at a definitive number of Armenians, but more to understand the issues that must be overcome to fully understand the magnitude of the crime that was committed.