Evolution of Armenian Surname Distribution in France between 1891 and 1990

Abstract

The evolution of the Armenian presence in mainland France from 1891 to 1990 is described on the basis of an inventory of more than 7000 family names of Armenian origin extracted from the INSEE surname database. Several surname samplings are proposed, and parameters such as the number of different Armenian names, the number of births with these names and their proportions are used as descriptors for each of the 320 French arrondissements and the four successive 25-year periods between 1891 and 1990. Before 1915, Armenian surnames and births with these names are infrequent and almost exclusively located in Paris and the arrondissements of Marseille. From 1915 onwards, subsequent to the genocide in Turkey, the number of births and the diversity of Armenian surnames rose sharply until 1940, before stabilizing thereafter. The diaspora remains essentially centred in Paris, Lyon, and Marseille, with little regional extension around these poles.

1. Introduction

The Armenian genocide of 1915–1916 triggered a major wave of emigration. The extermination of the Armenian population which, at the beginning of the 20th century, was mainly established between the Russian Empire to the north, the Ottoman Empire to the west, and Persia to the southeast, had begun in Asia Minor some twenty years earlier with the Hamidian massacres (1894–1897) and the Cilician massacre (1909). The process continued after the genocide with massacres in Iranian Azerbaijan, the Caucasus, Cilicia, and Smyrna between 1918 and 1922 (Kunth 2007; Kaiser 2010). Survivors chose to go into definitive exile in the Middle East, Europe or America, forming a diaspora that has been well documented (Mouradian and Kunth 2010; Hovannisian 2006; Boudjikanian 1982; Ter Minassian 1989, 1994; Ter Minassian 1997).

In France, the presence of Armenians is attested from well before the mass exodus of survivors beginning in 1920. It bears witness to the long-standing commercial exchanges (from the late Middle Ages on) between France and regions of Asia Minor and the Caucasus. In this context, it was above all in ports such as Marseille and in the capital city of Paris that Armenians first settled. While there is an abundant historical documentation in the form of local or family monographs (Temime 2007; Boudjikanian-Keuroghlian 1978; Hovanessian 1988; Huard 2007; Morel-Deledalle et al. 2007; Adjemian 2020) shedding light on the conditions of arrival of these new Armenian communities from the Ottoman or Russian empires and the ways in which they integrated into French society, major surveys of France as a whole are poorly documented and provide scant demographic and geolocalized statistics. This is why we propose to fill these gaps using a statistic that is seldom employed in this context, and which is based on family names. Indeed, the vast majority of Armenian surnames are characteristic and easily distinguishable from surnames of French origin. They can be readily traced in France, across both space and time, using the surname databases currently available, thus enabling a quantitative approach to the Armenian presence in France, which is the subject of this note.

2. Materials and Methods

Our aim is to examine how Armenian surnames are distributed in France and how this distribution has changed both in space and time. To do this, the dataset of surnames from 1891 to 1990 produced by INSEE (1985) was used, as it represents the most complete source of surnames in France. It covers over four periods (P1: 1891–1915; P2: 1916–1940; P3: 1941–1965; P4: 1966–1990). There is, however, a certain bias in these data because according to INSEE, they only record people who were still alive in 1972. Indeed, it has been found to include almost 20% more births for the first two periods (Darlu and Chareille 2020).

The next question is that of identifying which of the 500,000 or so surnames in the INSEE dataset are of Armenian origin. In order to make the list as representative as possible, it was necessary to compile beforehand a list of such names by including, where feasable, the commonest names in Armenia. Two different strategies were applied in this study. The first consists of identifying the commonest names in Armenia, and then locating them in France. The second involves searching directly the INSEE dataset for all Armenian-sounding names, that is, those ending in “–YAN” or “–IAN”, which are known to indicate their essentially Armenian origin.

(1) The first strategy consists of referring to the list of Armenian names on the “globalsurnames.com” website, which ranks the 1000 most frequent surnames held by people currently living in Armenia.1 The vast majority of the names on this list (991/1000) end with the suffix “–YAN”. This ending is a marker of filiation (in the sense of “son of”), comparable to such suffixes as “–SON” endings (Johnson = son of John), “–EZ” (Martinez = son of Martin), or the forms “–VICI”, “–WITZ”, “–VITZ”, “–WICZ”, “–VI(T)CH” (Mikhaïlovitch) or “–CHVILI” (Davitachvili) and “–DZE” (Shevardnadze) in Georgia.2 This Armenian list ranges from the name GRIGORYAN (83,517 occurrences) to the 1000th name, XAZARYAN (8 occurrences). Very few of the 1000 names on this list can be of ambiguous, possibly non-Armenian origin, and all of the following rank above 500: PETROV, PETROVA, IVANOV, IVANOVA, SARKISOV. They were therefore eliminated from the list. On the other hand, other names such as BAKUNC, SHALUNC, SHEGUNC and BZNUNI were retained because they seem to be present almost exclusively in Armenia, and are not found in the INSEE database for France.

The localization in France of the surnames on this list in the INSEE dataset cannot be determined without taking into account the francization of names, whether voluntary or imposed, when the immigrants arrived in France. Armenian characters were transliterated into the Latin alphabet in various possible ways: the suffix “–YAN” could be also spelled “–IAN” (e.g., MANUKYAN to MANUKIAN), and U as OU (e.g., MANUCHIAN to MANOUCHIAN), implying sometimes changes in pronunciation. A new list was therefore compiled containing not only the 995 initial surnames ending in “–YAN” (systematically employed in the initial Armenian list), but also all the possible spelling variants of these names. Ultimately, only those surnames for which at least one birth in mainland France was recorded in the INSEE file were retained from this set of names. This resulted in a list of 346 names whose Armenian origin is not open to question, designated as sample I of “ARM” surnames.

(2) The second approach is to list all surnames ending in “–IAN” or “–YAN” found in the INSEE birth records. Clearly, not all such names are necessarily of exclusively Armenian origin, even if this is most probably the case. Therefore, in order to minimize possible errors in assigning origin on the basis of “–IAN” or “–YAN” endings alone, selection was also dependent on two other criteria: (a) firstly, a decision to retain only names with at least eight characters (including the suffixes “–IAN” or “–YAN”), on the basis of the fact that 747 of the 1000 most frequent Armenian names in Armenia contain more than seven characters (the length of the name is therefore a criterion for retention in this selection); (b) the second criterion is to exclude names from this selection when the probability of their being of Armenian origin is low or nil. We know that certain names, particularly in the southwest of France, also possess endings of the “–IAN” type: SAILHIAN, SEBASTIAN, BAUSSIAN, VALENTIAN, DARMAYAN, COURBIAN, CARLHIAN, COUSTURIAN, etc. The suffix “–IAN” is also common in Iranian surnames.3 However, of the 1000 most frequent Iranian names, none of the 35 ending in “-IAN” appear in the “IAN list” of names of Armenian origin. Nor are they found in the INSEE list corresponding to period P1 (the 35 possibly Iranian names ending in “–IAN” correspond, incidentally, to a total of only 39 births in France during periods P2–P4). Attentive, but (as we shall see later) probably not perfectly exhaustive, examination led us to consider that 176 of these names are probably of “non-Armenian” origin. This is supported by the fact that these names are among the most frequent in period P1 (1891–1915), i.e., before the great Armenian migration. Once these names have been excluded a priori, we obtain sample II, the so-called “IAN” list, of 7533 different names.

There is an important difference between sample I (ARM) and sample II (IAN). The ARM sample contains only the most frequent surnames in Armenia, whereas the IAN sample, which contains only a selection of these surnames (those of at least eight letters), includes names that are less frequent in Armenia than the 1000 most frequent names in the ARM list. The ARM list therefore does not reflect the totality of Armenian immigrants, but only those with frequently found names. It is nevertheless reasonable to consider that very few Armenian immigrant names have failed to be included in this list, unless either the distribution of such names in France were to differ radically from that of the most frequent names in Armenia, or their geographical localization were to diverge from that of the most frequent Armenian names. From this perspective, it is unlikely that bias has been created. On the other hand, the larger list—sample II (the IAN list)—contains a considerable number of Armenian names, but also possibly (and unfortunately) a few names that are not of Armenian origin. The procedure adopted to establish the IAN list minimizes the number of “false positives” (non-Armenian names ending in “–IAN” and “–YAN”), but can neither guarantee that the names included are true positives (hence names that are markers of Armenian origin) nor ensure that certain rejected names are not, in fact, authentic Armenian names (and therefore wrongly considered to be French).

For this quantitative study, which covers France as a whole, the most relevant geographical unit needed to be larger than the commune, given the size of our surname sample compared to the number of communes in France (around 34,000). Therefore, the intermediate administrative scale of the arrondissement, situated between the commune and the department, was preferred. The initial, commune-level INSEE data were thus aggregated at the arrondissement level. The relevance of this regrouping is reinforced by the fact that from the P3 (1941–1965) and P4 (1966–1990) periods onwards, births are often registered in the localities where maternity units are located, which are generally in the main town of the arrondissement, and no longer in the communes. However, despite this data aggregation, commune-level figures remain available for discussion. The number of arrondissements taken into account here is 320. For the statistical processing of the data, we calculated the following variables for each arrondissement and each period (P1 to P4):

(1)

the number N_ij of births registered in the INSEE dataset under one of the names in the ARM list (depending on the sample selected), for period i in arrondissement j

(2)

the proportion f_ij (×100), expressed as a percentage (%) of these births for period i in arrondissement j compared to the overall number of Armenian births calculated for all arrondissements j in period i

f_ij = 100 × N_ij/∑j N_ij

(3)

The number S_ij of different Armenian surnames in each arrondissement j.

With regard to the cartographic representations, the wide dispersion of the number N_ij of births per arrondissement j, from 0 or a single birth to more than 1300, led us to carry out logarithmic transformations of the values (Ln(1+ N_ij)) and to constitute identical classes for the four periods in order to propose a single identical scale for all the maps (cf. Figure 1 below) and to better understand the variations from one period to another. In order to be able to compare the increase in the number of Armenian births from one period to the next and by arrondissement, we have compared the f_ij values (and not the gross numbers).

3. Results and Discussion

The first result is not really new. Coincidentally, the 1915 break in the INSEE data coincides with the first Armenian arrivals in France subsequent to the genocide. This is clearly reflected by the results in

Table 1. Statistics on the number of births registered under Armenian surnames and the number of different Armenian surnames by period, calculated for the 320 arrondissements (m: mean; sd: standard deviation; max: maximum; Ardt > 0: proportion of arrondissements with at least one Armenian surname; f_Paris, f_Lyon and f_Marseille (×10⁻⁵): proportion of births or number of different surnames in Paris, Lyon and Marseille.

Births		1891–1915	1916–1940	1940–1965	1966–1990	Names		1891–1915	1916–1940	1940–1965	1966–1990
ARM	m	0.225	12.206	15.009	11.291	ARM	m	0.100	3.259	3.759	3.484
	sd	2.277	90.130	95.377	71.858		sd	0.921	13.397	13.200	12.423
	max	29	1380	1428	1055		max	15	165	134	141
	Ardt > 0	3.75%	38.44%	48.44%	55.63%		Ardt > 0	3.75%	38.44%	48.44%	55.63%
	fi,Paris %	40.28	17.77	13.82	12.12		fi,Paris %	46.88	11.03	10.39	9.78
	fi,Lyon %	0.00	10.39	11.89	14.81		fi,Lyon %	0.00	8.44	7.73	9.15
	fi,Marseille %	8.33	35.33	29.73	29.20		fi,Marseille %	18.75	15.82	11.14	12.65
IAN	m	1.638	67.172	79.697	61.209	IAN	m	1.063	29.184	29.538	25.319
	sd	11.683	487.947	495.496	376.615		sd	7.977	164.568	136.285	114.073
	max	181	7271	7409	5530		max	127	2145	1692	1423
	Ardt > 0	23.44%	77.19%	89.06%	86.56%		Ardt > 0	23.44%	77.19%	89.06%	86.56%
	fi,Paris %	34.54	19.62	14.46	12.18		fi,Paris %	37.35%	18.38%	14.58%	13.16%
	fi,Lyon %	1.55	9.23	10.56	14.10		fi,Lyon %	0.88%	8.75%	8.1%	10.33%
	fi,Marseille %	19.27	33.82	29.05	28.23		fi,Marseille %	18.53%	22.97%	17.9%	17.55%

and a comparison of the maps in Figure 1a,b for P1 (1891–1915) and P2 (1916–1940). Before 1915, the Armenian presence in France was significant only in Paris and Marseille. The value of f_1,Paris is 40.28% for the ARM list, with 23 different surnames for 29 births, and 34.54% for the IAN list, with 146 different surnames for 220 births. For Marseille, the value of f_1,Marseille is 8.3%, with 5 different surnames for 5 births for the ARM list, and 19.3% for the IAN list, with 63 different names for 101 births.

Although the ARM list contains fewer surnames than the IAN list, the ratio between the number of surnames and the number of births is 1 (5/5) in Marseille and 0.66 (146/220) in Paris during the first period. Each surname in Marseille therefore corresponds to a single birth, whereas in Paris each surname may be borne by a larger number of births. This allows us to conclude that before 1915, the Armenian population in Marseille was more recent than in Paris, allowing it less time to register a large number of births.

In Lyon, Armenian names and births are either not found or particularly rare before 1915: the value of f_1,Lyon is in fact zero with regard to the ARM list, and barely exceeds 1% with regard to the IAN list. Even so, among this 1%, there are names whose Armenian origin is uncertain (e.g., CAZAMIAN, CECILIAN), but, even if they were Armenian, they would only represent a small proportion of the data, because they do not appear in the list of the 1000 most frequent surnames in Armenia.

Between 1891 and 1916, Paris and Marseille alone accounted for 54% of births bearing an Armenian name in mainland France (IAN list), and 48.6% for the ARM list. Other births are distributed according to the number of names. For example, in the IAN list, which does not preclude the inclusion of non-Armenian surnames ending in “–IAN”, we found four different names for a total of eight births in the Doubs department, three of which differ by only one spelling variant, while the fourth was probably non-Armenian for eight births. We further found that in Finistère, there were five names in “–IAN”, but of questionable Armenian origin; in the Aube, there were two probably non-Armenian names; in the Marne, there was one Armenian name and three births; and in the Aisne, there was one Armenian name and one birth, to mention but a few of the statistics for departments outside Paris, Lyon or Marseille. In the Morbihan, where four names ended in “–IAN” (IAN list), two births were registered under the surname ASLANIAN in the town of Hennebont during P1. This name is common in present-day Armenia, ranking 63rd (with 8501 occurrences). However, these ASLANIANS did not settle in the region, as they were not found in Hennebont in later periods. On the other hand, between 1916 and 1990, there were 195 ASLANIAN births, mainly in Paris, Marseille and Lyon.

As illustrated above, many of the names in these departments, which are far from the main Armenian reception centres, are “false positives”, i.e., names retained by our sampling procedure for construction of the IAN list because they resemble Armenian names, but which turn out to be non-Armenian. One example is the Landes department, where the name DARBAYAN is found, exclusively in this department (with ten births between 1891 and 1990). It is difficult to decide a priori whether it is of Occitan or Armenian origin, even though it does not appear in the list of the 1000 most frequent Armenian names. All such surnames ending in “–IAN” or “–YAN” that are rare and generally very localized constitute a sort of “background noise”, which does not alter the information provided by all the clearly Armenian names, whose frequencies are higher.

The configuration changes radically after 1915. The arrival of Armenian migrants after 1915 and the resulting births over the following 25 years are remarkably high (see averages and standard deviations for the 320 arrondissements, Table 1). It is well known that migrants arriving in a country tend to group together where their relatives or compatriots have already settled. Armenians are no exception to this rule. According to the International Labor Office, there were almost 30,000 Armenian refugees in France in 1925 (Ter Minassian 1994, 1997). They point out that by this date “almost half of the Armenian refugees in France were living in Issy-les-Moulineaux or Alfortville [Parisian suburbs]. The diaspora that emerged from the break-up knew these two islands of refuge.” The INSEE data show that Armenian births were mainly, but not exclusively, located around conurbations such as Marseille, Lyon and Paris, and here and there in the south west (Figure 1). The proportion of arrondissements recording at least one birth with an Armenian surname rose from 3.75% in P1 (ARM list) to 38.4% in P2. This quantitative shift can also be seen in the IAN list: 23% of arrondissements registered at least one birth with an Armenian name in P1, and this rose to 77% in P2 and to 86.6% in P4. The jump between P1 and P2 shows the extent of migration following the Armenian genocide.

The Armenian presence can be assessed over the long term; some names, already present before 1915, are still present in subsequent periods. There are 25 such names (in the IAN list: ABDALIAN, BAGHDASSARIAN, MOUTAFFIAN, PANOSSIAN, TCHAKIRIAN, for example), while 1665 names not found in the first period are present continuously in at least one arrondissement over the last three periods. Thus, they demonstrate the arrival of Armenians after the genocide.

Whether we consider the number of births per arrondissement or the number of different Armenian surnames per arrondissement (Table 1), the conclusions we can draw are very similar. This is because there is a strong correlation between the number of births N and the number of different surnames S, as shown in Figure 2. This log–log correlation is strong (r = 0.99). The greater the variety of surnames, the greater the number of births. This explains why the comments made about the number of births per arrondissement remain valid for the number of different surnames per arrondissement.

The results from the two samples, IAN and ARM, show a high degree of agreement, as can be seen in Table 1 and Figure 1a,b. While the correlation between the logarithmic transformations of the data for the first period P1 (1891–1915) between the two samples IAN and ARM is relatively low (r = 0.486 and ρ = 0.234, respectively, Bravais–Pearson and Spearman correlations), these correlations are much higher for the following three periods, respectively: r = 0.889 and ρ = 0.755, r = 0.865 and ρ = 0.749, and r = 0.858 and ρ = 0.785. The weak correlation observed for P1 stems from the very small number (12) of arrondissements wherein an Armenian name is attested (this number is 75 for the IAN list). It is also explained by the fact that the ARM list contains fewer surnames and neglects possibly infrequent Armenian names, unlike the IAN list, which contains more surnames. This difference reduces in subsequent periods (P2 to P4). Despite this, we can consider that the two sampling strategies used to select Armenian surname data allow fairly congruent conclusions regarding the evolution of the spatio-temporal distribution of Armenians in France between 1890 and 1990.

While the geographical expansion of Armenians outward from Paris, Lyon, and Marseille into new arrondissements exploded just after 1915, it remained modest in subsequent periods. It is remarkable to note that the proportion of arrondissements receiving Armenians increased between P1 and P4, as measured by the proportion of agglomerations with at least one birth (Table 1). Between P1 (23.4%, IAN list) and P2 (77.2%), this increase is a sign of a significant geographical expansion. It then continues between P2 and P3 (89%) before stagnating or falling slightly in P4 (86.6%).

Table 2. Comparison of means m and correlations r between successive series of the logarithm Ln(1 + N) of the number of births N calculated over the 320 arrondissements. Student’s t-test: * = p < 0.0001; NS = not significant. r is the Bravais–Pearson coefficient of correlation and ρ the Spearman correlation between two successive periods.

	m	t	r	ρ
P1: 1891–1915	0.323
P2: 1916–1940	1.692	17.17 *	0.489 *	0.337
P3: 1941–1965	2.164	10.39 *	0.875 *	0.760
P4: 1966–1990	2.182	0.35 NS	0.829 *	0.784

provides some additional results: the number of births increases significantly on average between the P2 and P3 periods, but then stagnates between P3 and P4. The correlation between the numbers of births between periods P1 and P2, although significant, is relatively weak, so it can be considered that the determination of the geographical locations of Armenians before 1915 is only moderately predictive of those observed in subsequent periods.

Another question concerns the increase in the proportion of Armenians in each arrondissement over time. To answer this question, the variation in f_ij from one period to the next is a good indicator. It should be remembered that f_ij represents the proportion of births with an Armenian surname in France occurring in arrondissement j and by period i. The map in Figure 3 shows the distribution of variations in f_ij between period P2 and period P4 (f_P4,j − f_P2,j) for the names of the IAN list. Among the arrondissements showing a very significant increase in this Armenian component is Lyon, where the value of f between P2 and P4 rises from 9.23% to 14.10% (Table 1). This is the case, to a lesser extent, for increases seen in Grenoble (from 1.83% to 2.48%), Versailles (from 0.3% to 1.7%), and Istres (from 0.3% to 1.6%). Among the arrondissements with a decline in figures are Paris (from 34.54% to 12.18%, cf. Table 1), Valence in the Drôme (from 3.96% to 1.40%), and Marseille, Vienne, and Nanterre. These decreases could be explained by the wider spatial distribution of names in P4 than in P2. The situation in Bouches-du-Rhône is more complex. In the Marseille arrondissement, the drop is one of the largest (from 33.83% to 28.23%, cf. Table 1), while in the Istres arrondissement, on the other hand, there is an increase in the proportion (from 0.51% to 1.55%). However, this increase only concerns a very limited number of names (48 in P2, and 126 in P4) compared with the situation in Marseille (respectively 2145 and 1453). These results suggest a transfer of population from the Marseille arrondissement to another (Istres, among others). The regions where there is an increase in the proportion of births with an Armenian surname (in red on the map) are mainly located (cf. Figure 3) in the north east of the Paris Basin, in Languedoc, and in Aquitaine, to the detriment of the areas (in blue on the map) along an axis from the Rhône to the Paris Basin and beyond, including Paris.

4. Conclusions

The surname approach used in this study has demonstrated its ability to provide useful information. The number of surnames selected here, over 7000, is sufficiently high to provide a solid statistical basis for our conclusions, even if possible biases should not be overlooked, such as those generated by variations in the spelling of names, the selection of names according to their length, the inclusion of “false positives” (French names mistaken for Armenian names), and/or the exclusion of “true negatives” (elimination of Armenian names “mimicking” French names). One might question the choice made here of migration indicators such as the number of births and the diversity of Armenian names by arrondissement. However, it would be difficult to deny that these indicators validly reflect migration in the sense that the arrival of new families is always accompanied by the arrival of a descendant generation. The diversity of surnames (Armenian surnames by arrondissement) remains an indicator linked to the number of births; the influx of migrants is accompanied by an increase in the diversity of surnames, even if this diversity is not interpreted here in terms of an origin in Armenia.

All the results confirm the massive arrival of Armenians after 1915, whereas their presence before this period was minimal, and mainly located in large cities such as Paris and Marseille. Lyon, despite being a major city along the Marseille–Paris axis, does not seem to have originally been an Armenian centre of any importance. Subsequently, between 1916 and 1940, with the arrival of migrants, the number of births and the diversity of surnames increased sharply, before slowing between 1940 and 1990. However, the geographical distribution of the Armenian population increased, first gaining ground in the arrondissements near the major towns where Armenian immigrants had first settled, and subsequently, albeit modestly, in more distant arrondissements. Over the years, the diaspora has largely remained within relatively narrow geographical areas, seeing a moderate degree of expansion across the country as a whole, while remaining stable at the local level.