The first post-Soviet censuses were taken between February 1999 in Belarus and October 2002 in the Russian Federation. They raise interesting issues on the complex links existing between politics and administration, statistical continuity and international recommendations, the adjustment to new situations and the territorial stakes involved. The census questions, particularly those regarding language, ethnicity and migration status, reflect the confrontation of these competing claims and the necessary political compromises that were reached. Here, Dominique Arel examines the census procedures and analyses in detail how the phrasing of the questions was elaborated by the statistical offices in Russia, Ukraine, Belarus, Kazhakstan and the Baltic states. This approach throws light on the political construction of the categories that were finally adopted.

As Table 1 indicates, the official population of all post-Soviet countries under review declined in absolute terms during the last decade. Russia is no exception. Preliminary figures, released as soon as the October 2002 Russian census was completed, placed the population of Russia at 145.1 million (Sedov, 2002), a decrease of 2.3 million or 1.6% since the 1989 census. All post-Soviet countries have experienced negative natural population increase since the end of the Soviet Union, the excess of deaths over births increasing steadily throughout the 1990s (Silver, 2003). The Russian case is the most notable one. Since 1994, deaths have exceeded births by at least 700,000 annually (Avdeev, 2002). In 2001, the crude death rate in Russia (15.6 per thousand) was 71% higher than the crude birth rate (9.1 per thousand) (Population and Labour Market, 2002). Since, as we will see, Russia is the only country whose natural population decline is partly offset by positive net migration, this suggests that the official population figures for Belarus are not credible. With an industrial, urban, and ethnic (east Slavic) profile comparable to that of Russia and Ukraine, there is no plausible reason why the population decrease in Belarus would be so much smaller (once Russia’s migration balance is accounted for) than that of the neighbouring Slavic countries.

Census results, expected in 2003-2004, are unlikely to change the parameters of the debate. This is so because, like the yearly official statistics, they will be based more on official registration, than on actual residence. Most western censuses aim at counting the de facto population, i.e. the population actually residing in the enumerated districts at the time of the census. Others, notably in France and German-speaking territories, have traditionally opted for the de jure population, the population legally residing in the enumerated districts (Eggerickx and Bégeot, 1993). The Soviet Union officially calculated both kinds of population figures, which it called “permanent” (postoiannoe) and “present” (nalichnoe) (Anderson and Silver, 1985). The “permanent” population of a territory was technically the population whose usual residence (obychnoe zhitel’stvo) was located in that territory at the time of the census. In theory, the determination of “permanence” was made on the basis of an oral declaration by the respondent, since official documents were not supposed to be requested. In practice, a clear link existed between a census declaration of residence and the official Soviet residential registration system, the propiska.

Since the end of the Soviet Union, at least 6.4 million people immigrated officially to Russia (mostly from the CIS), and at least 3 million emigrated officially (mostly to countries outside the CIS) [8]. The official statistics, however, do not include unregistered migrants, or migrants deemed “temporary,” that is, with a “permanent” address in another country. Observers estimate that there could be as many as five million uncounted immigrants in Russia (Latsis, 2001). Since Russia’s population decreased officially by 3.6 million between 1992 and 2002 (from 148.7 million to 145.1 million), this means that the number of people actually living in Russia, registered or not, may in fact have increased slightly in the past decade. The 2002 Russian census, in its resolve to count the “permanent” population, will not shed light on the extent of actual migration to Russia [9].

Soviet censuses used to be events of mass mobilization supervised by the party apparatus. There was no question that people had to let the enumerators in. For a post-Soviet citizenry highly suspicious of state intrusion, the mandatory nature of the census is no longer taken for granted. Public protests, muffled by the state-controlled press in Kazakhstan and Belarus, were more widely aired in the Ukrainian media during the December 2001 Ukrainian census, and reached a crescendo during the October 2002 Russian census. The media constantly reported about people refusing to be counted or threatening to boycott the census if some social grievance was not addressed.

A number of western countries, particularly in Europe, have been greatly concerned with issues of confidentiality in censuses. Germany did not conduct a census for fifteen years, between 1987 and 2002, following a 1987 ruling by the Constitutional Court that the 1983 census had violated rights of privacy. The issue is not whether the state has a right to collect census data per se, but whether it can guarantee that census data will be processed confidentially. The heart of the matter is the handling of individual names in census databases. The 1987 ruling in Germany found that confidentiality procedures had been violated when names and addresses from census forms had been passed along to municipalities for comparison with their registers (Rötzer, 2001). A similar case occurred in Estonia in 2000 when the Data Inspectorate sued the Estonian Statistical Office for having improperly passed along private information to the population registration system (Silver, 2003).

How people are categorized in the census according to criteria of identity has long been a politically charged issue. All countries inquire about the citizenship of their respondents. While the citizenship policy of a country is inherently subjective and a result of political considerations, the fact of citizenship is, in principle, easily measurable, as it is based on official documents [15]. Some countries also inquire about the country of origin of their residents, as in the French census. Like citizenship, country of origin is an objective category.

All Soviet censuses had a question regarding the “nationality” of respondents. “Nationality,” in this context, refers to what western scholars would call “ethnicity,” i.e. the sense of belonging to a community of presumed descent based on the subjectively-determined saliency of such cultural markers as language, religion, and custom. From the early part of the nineteenth century, when German philosophers popularized the idea, “nationality” has acquired this ethnic connotation in eastern Europe, and eventually Eurasia, while the term “ethnic” is seldom, if ever, used in the public discourse of those countries. Western countries (that is, west from Germany and industrialized countries overseas), meanwhile, have used “nationality” to refer to citizenship. The concept of “(ethnic) nationality” has therefore been absent from western censuses and peculiar to those originating in eastern Europe [16].

While there are no conventions on whether or how the census should enquire into ethnic nationality, a clear standard has evolved in the past decade regarding the labelling of national identities. National identity, according to this standard, is a matter of self-definition and cannot be imposed by the state. This had led the European Union to view the use of ethnic or religious categories on identity cards as incompatible with human rights [17]. “Internal passports” in the Soviet Union had an entry for nationality. They no longer do, except in Kazakhstan. The census question on nationality, on the other hand, was based on self-declaration, but within clear bounds. The most politically consequential of all restrictions is that not all national categories volunteered by respondents are recognized as valid by enumerators. As we will see, all Soviet censuses used an exclusive list of recognized nationalities and detailed instructions on how to recode unrecognized categories into official ones. For instance, “Cossacks” were counted as Russians if they claimed Russian as their “native language,” or as Ukrainians if they claimed Ukrainian, but never as “Cossacks” since the term was not accepted as a nationality. A “dictionary of nationalities,” listing all recognized and unrecognized categories, was prepared and updated prior to each census. All countries in this study have maintained the practice, for the simple and fundamental reason that nationality, in the post-Soviet context, continues to be linked to territorial sovereignty.

The claim of ethnic entitlement to a territory entails a need to produce ethnic majorities. This is arguably the main reason why post-Soviet countries have kept a nationality category in their census. National elites understand the power of official statistics very well. Even though the claim of prior settlement is conceptually distinct from contemporary demographics, the ability to construct statistical majorities is a critical tool for strengthening a state’s hold on territories. The case of Kazakhstan illustrates this point strikingly well. Between the censuses of 1926 and 1959, the proportion of ethnic Kazakhs in Kazakhstan shrank by almost half, from 58.5% to 30% (Alekseenko, 2001b). This severe decrease was caused by the collectivization-induced famine of the early 1930s, which cost the lives of over a million nomadic Kazakhs, by the deportation to Kazakhstan of hundreds of thousands of people belonging to the nationalities of bordering countries after 1937 (primarily Germans), and by the mass settlement of Slavs in the so-called agricultural Virgin Lands of northern Kazakhstan in the 1950s. At the last Soviet census, in 1989, the proportion of ethnic Kazakhs had grown to 39.7%, still far from a majority. In preparing for their first independent census in 1999, it became imperative for Kazakh officials to produce a titular majority (Kolsto, 1998). The task was greatly facilitated by the mass out-migration of Slavs, mainly to Russia, and of ethnic Germans, to Germany. Census results from 1999 indicate that the combined population of “Europeans” (Slavs and Germans) declined by nearly three million since 1989, while the number of Kazakhs grew by almost 1.5 million. Officially, the ethnic Kazakhs were now said to comprise 53.4% of the entire population, a huge increase over 1989 (Dave and Sinnott, 2002).

The use of the census to obtain desired majorities also involves the manipulation of ethnic categories. The main identity cleavage in Kazakh society is between Kazakhs and “Europeans” (primarily Russians, Ukrainians, Germans, and Jews). European nationalities are in theory distinguished by language, but with overwhelming assimilation to Russian among the contemporary Europeans, this distinction has long lost its relevance. In daily life, most of them are perceived as “Russians” and would probably think of themselves as such if it were not for the Soviet system of passport nationality, which “froze” nationalities on paper. Under this system, people were assigned the nationality of their parents in their internal passports. They were socialized into associating the census question on nationality with the nationality inscribed on their passports. This had the effect of statistically masking extensive linguistic assimilation which, in the Kazakh context, amounted to ethnic assimilation. What can possibly distinguish a third-generation Ukrainian from a third-generation Russian in Ust-Kamenogorsk (one of northern Kazakhstan’s main industrial cities)?

Soviet census officials used an exclusive list of nationalities. This meant that groups that did not appear on the list were recoded into acceptable ones. In Latvia, for instance, a “Latgalian” (latgalietis) was counted as a “Latvian”. In Tajikistan, where national consciousness is weak, no less than forty unrecognized names were recoded as “Tajiks” [21]. Various arguments in favor of maintaining an exclusive list can be made. One is statistical. Recording every conceivable answer in an open-ended question on nationality could prove costly and cumbersome. The 1989 Soviet census list of nationalities had 128 entries, slightly more than the previous three postwar censuses, but the number of unrecognized answers reached 823. A more complex argument, however, is political. Soviet ethnographers have long worked under the assumption that people may be mistaken in declaring their nationality in the census (Cadiot, 2001). They may volunteer an identity that ethnographers categorize as “regional”, or “subethnic”, a component of a larger “ethnic” nationality, as documented by the work of the ethnographers themselves.

The great majority of these “new” nationalities are obscure groups known only to a narrow circle of specialists, and unlikely to register more than a few hundred or a few thousand members. It is the inclusion of larger groups that has generated stormy debates. Three of these groups are connected to the Tatars. The two most prominent ones during the 2002 census in Russia were the Kriashens and the Cossacks.

All censuses in the former Soviet lands, going back to the first Imperial Russian census of 1897, have included a question on native language (rodnoi yazyk). Instructions to census-takers, however, have never made it clear whether the question should be understood as the first language learned as a child (as later became the standard UN recommendation), or the language that people prefer to speak (as understood in German-speaking lands). Soviet ethnographers often complained about the imprecision of the concept of rodnoi yazyk, but to no avail. As sociological surveys revealed in the 1990s, many people understood the concept in a third way, namely, as the language of their nationality. In Ukraine and Kazakhstan in particular, the proportion of titulars who speak Russian at home, and who, presumably, were raised in that language, far exceeds the proportion who claimed Russian as a native language in the 1989 census. In Ukraine, over one third of Ukrainian titulars speak Russian at home, while only 12% claimed it as a native language (Arel and Khmelko, 1995). In Kazakhstan, up to one third of ethnic Kazakhs are estimated to be Russian speakers at home, compared to a minuscule 1% declaring Russian in the census (Dave and Sinnott, 2002).

Rather unexpectedly and literally at the time when census forms were going to press in April 2002, six months prior to the October census, the government decided to break with a century-old tradition in Russia and dropped the question on native language (rodnoi yazyk) altogether. Census authorities were apparently swayed by the argument offered by the director of the Institute of Ethnology and Anthropology that rodnoi yazyk is widely interpreted as a restatement of one’s nationality, irrespective of actual language preference, and therefore provides little additional information (Tishkov, 2002). The sudden disappearance of rodnoi yazyk from the census provoked an outcry among groups promoting ethnic interests, forcing the government to backpedal [28]. Since millions of census forms were already printed, the question reappeared in the instructions to the enumerators. The census form had an entry for “knowledge of other languages” [than Russian], with three lines. The instructions indicated that the first line should be used for rodnoi yazyk. However, it became clear during the census that the question was not asked systematically. Moreover, as long as a separate question on nationality is used in the census, any question on non-Russian languages is likely to be interpreted as a restatement of one’s nationality. The Russian government underestimated the extent of symbolic attachment to the native language question. Russia’s wavering and vulnerability to lobbying efforts also contrasted remarkably with the rather secretive and top-down methods used to prepare censuses in other post-Soviet states.

Census categories are not neutral. The use and meaning of categories such as “immigrant”, “permanent population”, “nationality”, and “language” are the results of what is essentially a political process involving state agents, census takers, scholars, and social groups. Census decision-making is shaped by inertia (past practices) and by evolving national interests. While Western censuses have grappled with issues of migration and confidentiality for decades, the problems were new for post-Soviet countries, and particularly for Russia. Mistrust of the state and massive migratory flows had not been factors affecting the conduct of Soviet censuses, at least in the postwar era.