Counting for what purpose? The paradox of including ethnic and cultural questions in the censuses of Croatia, Bosnia-Herzegovina, and Macedonia

Introduction

Collecting sensitive ethnic and cultural data, such as ethnicity, language, and religion, raises an unresolved paradox. On the one hand, if population groups are not defined by category, they do not formally exist and are consequently not given their due rights. On the other hand, the definition of these groups and the linking of political benefits to population numbers may lead to negative discrimination or may even challenge concepts of national identity and state boundaries. Hence the inclusion of ethnic and cultural questions and categories in a population census is highly problematic.^[1] The difficulties that arise can influence the whole census operation and even lead to its failure. This article does not aim to provide answers on whether to record ethnic and cultural characteristics or how this should be done, but it points out the paradox involved and the consequences that follow.

Nowadays census data is used not only for national purposes such as taxation, but also as a requirement by the EU. The demand for reliable and comparative population data through census-taking falls under Chapter 18 of the acquis communautaire, which covers the necessary statistics.^[2] Although census-taking would appear to be a rather small aspect of the process, given the size of the acquis, it is, in fact, an enormous undertaking.

On average, population censuses are taken every ten years. In the Western Balkans, this average has not been attained by all the post-Yugoslav countries, however. For example, due to the difficulties created by war after the dissolution of Yugoslavia (1992-1995), Bosnia and Herzegovina (hereafter Bosnia) did not conduct its first census till 2013. Other countries, such as Croatia, held censuses in 2001 and 2011, and the Former Yugoslav Republic of Macedonia (hereafter Macedonia) conducted censuses in 1994 and 2002, but aborted one planned for 2011.

Although the EU does not actually require data on ethnic and cultural characteristics from its potential member states, all the Western Balkan countries have included questions about these aspects in their censuses, and this was also a practice in the former Yugoslavia. In the post-conflict environment in most of these countries, these questions are now connected with discrimination. Positive discrimination can ensure the granting of minority rights.^[3] However, the labelling of people which puts them into different categories can also lead to exclusion and negative discrimination. Especially when privileges and rights are specifically linked to numerical thresholds, census processes can become problematic. This has been shown in ethnically divided societies, where the process of counting and categorizing a population can lead to negative discrimination against minority groups.^[4] As Croatia, Bosnia, and Macedonia were enlargement countries at the time of their recent censuses, we would expect the censuses to have been carried out in the same way; however, the countries have shown considerable variation in their census-taking procedures. Croatia carried out a complete census in 2011; Bosnia collected census data in 2013 but only published the results, after serious delays, in June 2016; and Macedonia aborted its 2001 census altogether.

The argument of this article is that, though taking a census has been made part of the EU conditions for joining, it is not possible to speak of any ‘Europeanization’ dictating the collection of ethnic and cultural data, since there is no coherent European approach on how to accomplish this. The effects of including questions to elicit ethnic and cultural data in the population census can all too easily politicize the census process, as will be shown in the examples of Croatia, Bosnia, and Macedonia, where population numbers are linked to rights and/or proportional representation.

As there are no EU rules on collecting ethnic and cultural data, this article uses the additional coverage model mixed methods approach.^[5] This combines survey data from census experts who provide an international perspective on census-taking with case analyses of the ethnic and cultural characteristics present in the census questionnaires for Croatia, Bosnia, and Macedonia. By these means, the paper demonstrates the paradox of collecting ethnic and cultural data. The approach is innovative, not only because it combines new methods and data, but because existing research has not yet looked at the censuses in the Western Balkans within the general framework of Europeanization.^[6]

The Sensitivities of Census-Taking

In the Western Balkans, there has been a long tradition of census-taking. Censuses were already being made when these territories were under Ottoman and Habsburg rule.^[7] After the First World War—in 1921—the Kingdom of Serbs, Croats and Slovenes held a census, and, in socialist Yugoslavia after the Second World War, population counts became relatively regular events. Their most important question, closely related to ethnic and cultural characteristics in socialist Yugoslavia, concerned ‘nationality’, though initially without a category for ‘Yugoslav’. This was only added in the census of 1961;^[8] and it was only in 1971 that ‘Muslim’ was first allowed as a national category for Bosniaks.^[9] This remained the practice until the last Yugoslav census, which took place in 1991, after which ‘Bosniak’ was adopted as an official category.^[10] Until the dissolution of the country in 1991, identity questions in Yugoslav censuses were based on self-identification and included open-answer questions for national identity, language, and religion.^[11]

When it includes ethnic and cultural questions, a census does not only provide an overview of the demographic, social, and economic characteristics of a country’s population.^[12] It also creates the opportunity for individuals to identify themselves as being members of this population.^[13] With the creation of identity categories through questions on ethnicity, language, and religion, individuals and peoples start identifying with specific groupings which thereby become legitimized. In the end, they are impelled to recognize themselves as representatives of these census categories—as members of societal groups, and eventually of the state.^[14] The relationship between a census and sense of identity is highly political as it connects with claims over territory and the assumption of power over people with other identities or ethnic and cultural characteristics.^[15]

After the dissolution of Yugoslavia, issues of ethnicity and geographic belonging developed into tools for ethnic bargaining over borders, rights and political representation.^[16] A population census is an essential lever by which ethnic groups can (re-)affirm their legitimacy and supremacy.^[17] Although it was recognized that questions on ethnicity, language and religion were sensitive matters,^[18] in the census round of 2010 all the Western Balkan countries asked for these data.^[19] Mapping identity in the region seems still to be a major objective and the key reason for conducting censuses. Visoka and Gjevori have shown how census processes in the Western Balkans have given ethnic groups the opportunity to fight identity battles so as to (re-)claim political ownership.^[20] This is possible because rights are linked to numerical thresholds. For example: in Croatia, minority rights are granted according to the census results.^[21] In Bosnia, they are linked to the Dayton Peace Agreement, which stopped the war of 1992-1995, and which divided the political system among the three constituent ethnic groups—Bosniak, Croat, and Serb—as judged on the population distribution shown in the 1991 census.^[22] In Macedonia, the census is linked to the Ohrid Framework Agreement, which stopped the armed conflict between ethnic Albanians and Macedonians in 2001. By this agreement, minorities making up more than 20% of the population received language rights and proportional representation in the police and bureaucracy.^[23]

Although there is much discussion about these census questions and categories, there is no coherent agreement as to what exactly constitutes ethnicity.^[24] Nonetheless, ethnicity is regarded as being closely related to other cultural characteristics, such as language and religion.^[25] In Central and Eastern Europe, religion and language are taken not only as markers of individual identity, but as indicators of national or ethnic identity: they are part of an ‘ethnic package’.^[26] In the recommendations of the Conference of European Statisticians (CES) on the population and housing census formulated by the United Nations Economic Commission for Europe (UNECE) and Eurostat,^[27] these aspects are taken into account under the heading of ‘ethno-cultural characteristics’.^[28] Language, in particular, is seen as an indicator of national identity, since it cannot be disconnected from the state (unlike religion, through secularization) and tends to fluctuate, since, in comparison with other identity markers like religion, linguistic assimilation can be achieved relatively easily.^[29] There has been much discussion on how to convert exploration of these categories into census questions; and the formulation of the census questions and categories show the construction of ethnic discourses.^[30] These discourses follow (ethnic) group preferences, which are used to define public policy.^[31] In the cases of Croatia, Bosnia, and Macedonia, they are linked to rights and political representation through numerical thresholds. This linking of population numbers to rights and thresholds ultimately leads to discrimination: either positive (in the granting of minority rights) or negative (in enforcing segregation).^[32] This is what constitutes the paradox of collecting ethnic and cultural data. If population groups are not defined in terms of categories, they do not formally exist and consequently have no rights. However, if political benefits are linked to population numbers, this can lead to negative discrimination or even create a challenge to national identity and state boundaries. Since census processes give ethnic groups the opportunity to (re-) claim political ownership, census-taking is necessarily political.^[33] Therefore it is important to look at the dynamics that exist between censuses and the ethnic and cultural categories they may impose. Since the census is also part of the EU acquis and the countries at the time of the last census were all enlargement countries, the next section will look at the potential influence of Europeanization on census-taking.

Europeanization and Census-Taking

Within the acquis communautaire, the requirement to conduct census-taking appears in the chapter on statistics, and is a condition countries need to comply with before accession to the EU. Unlike the Copenhagen Criteria, which are regarded as being part of the political conditionality for EU accession (focusing on political transformation towards democratization and EU member state-building),^[34] the acquis conditions are regarded as more technical.^[35] The census is part of the acquis for the EU: members and aspiring members have to be able to present ‘sufficiently reliable, detailed and comparable data on the population and housing’ every ten years.^[36] The methodology, definitions and programme of the statistical data and metadata need to be aligned, both within the Union and in its enlargement countries. Officially the enlargement countries do not yet need to adhere to these regulations, but it is strongly recommended that they do so.^[37] Such institutionalization of EU rules, procedures, and policy paradigms is denoted as Europeanization.^[38] In this present article, Europeanization will be defined as the domestic influence of the EU on the censuses conducted in Croatia, Bosnia, and Macedonia.^[39]

The EU conditions for census-taking include EU regulations on, for example, the technical specifications, definitions and data sources of censuses.^[40] They also rely on CES recommendations.^[41] The latter provide guidance and assistance to countries conducting a census, and also have the same purpose as the EU census regulations: to ensure the harmonization of statistical data.^[42] Sensitive aspects around ethnic and cultural characteristics, ethnicity, language, and religion, are not covered by the EU regulations. In the CES recommendations, they do indeed appear, but only as non-core topics. The collection of ethnic data is excluded from the population censuses in most West European countries, with the notable exceptions of Ireland and the United Kingdom.^[43] The reasons for not including such data in the western part of Europe are, firstly, adherence to data protection laws, and, secondly, because the categories are seen as imprecise and inaccurate, with the potential to challenge the self-perceptions of ethnically homogeneous societies.^[44] This last is the case in France.^[45] Most European countries, therefore, use ‘country of birth’ and ‘citizenship’ as proximate indicators of ethnicity.^[46]

In order to go deeper into the issue of collecting ethnic and cultural characteristics, this article uses an additional coverage model mixed method approach.^[47] This enables integration of qualitative and quantitative data. The article combines discourse analysis and descriptive statistics. Since there is no coherent EU approach to collecting data on ethnicity, a survey has been conducted among census experts from all over Europe and beyond, to get a reference frame and to elicit expert opinions on current practices regarding this aspect. The experts came from the National Statistical Offices of the different countries (with the exception of Russian-speaking ones) and were present at the United Nations Economic Commission for Europe (UNECE) census meetings in Geneva. In addition, my research draws on contacts from Eurostat and further census experts whose names were provided by Eurostat—the list was checked by Pieter Everaers, Director at Eurostat, and Eric Schulte Nordholt, Senior Researcher and project leader of the census at Statistics Netherlands. Those consulted included census experts from all the EU member states as well as experts from the Balkans, Switzerland, Iceland, Canada, Turkey, and Israel. Thus there is a wide geographical scope, but as the census community is rather small, it can be assumed that these experts were familiar with the issues at stake.

In total, I contacted seventy-five census experts. The response rate for the first round was 69% (n=52).^[48] The second round had a response rate of 50% (n=38). The survey was conducted following the Delphi method.^[49] It took place in two rounds in June, and then November/December 2015, as this allowed for interaction with the respondents and for reflection and elaboration on the answers the first round returned. In the first round, specific aspects were identified, such as what might be perceived as sensitive issues in a census. These aspects were then taken up in the second round and follow-up questions were asked about them. In addition, the survey included control questions; and the material was filtered, as the respondents were supposed to indicate their knowledge of a particular topic (such as the EU census regulations or the Western Balkans). For the analysis of some questions, a deliberate division was made between experts who declared that they had knowledge about the Western Balkans and those who declared that they did not know much about the region. This was to test whether respondents familiar with the historical situation and ethnic cleavages in the Western Balkans would have a greater tendency to wish to exclude ethnic and cultural aspects from a census than those with less background awareness.

In order to analyse the practices of enquiring into ethnic and cultural categories, this article looks at census methodologies and questionnaires from Croatia, Bosnia, and Macedonia. I have selected the censuses from the 2010 census round so as to compare the three countries.^[50] (Bosnia had no census before 2013.) The analysis is based on documents in the English language. In some instances, interview data and newspaper articles have been used to contextualize the findings. The three West Balkan countries I follow were chosen because, at the time of the census, they were all (potential) candidates for EU enlargement, and because of the variations they showed with regard to census-taking. Croatia—today an EU member—had its census enumeration in 2011 and presented its results in November 2013.^[51] Macedonia aborted its census enumeration of 2011, a few days into the process. Bosnia completed its census enumeration in October 2013, but the census data was only published, after serious delay, in June 2016.

Expert Perspectives on Counting Ethnic and Cultural Characteristics

As already stated, there are no clear EU conditions on collecting data on ethnic and cultural categories in the EU census regulations.^[52] This section of my article therefore analyses the results of the census expert survey conducted, so as to provide a broad perspective.

In general, census experts are aware that there are census topics which can be very sensitive. When these experts were faced with the question whether sensitive questions should be part of a population and housing census questionnaire in general (n=49), most of them (49%) said they would include sensitive questions that could be answered on a voluntary basis: the questions would be included, but there would be no obligation to answer them.^[53] Some 14% would include sensitive aspects in a census, and 20% would exclude sensitive aspects altogether. Further, 12% of the experts stated that the issue was not applicable, as in their countries the collection of census data was not made in the traditional way, but through registers or other sources. Also, 4% would consider including sensitive questions if they were important for the distribution of subsidies. The question addressed to these respondents was formulated very broadly and it was left open to them to make their own choice of what was meant by ‘sensitive’. However, when asked directly what kind of census topics they thought might cause difficulties, the vast majority of the experts indicated ethnicity (94%) and religion (94%) as (highly) sensitive (Table 1), while language (61%), nationality (61%), citizenship (30%), and place of birth (20%) were identified as sensitive by a smaller proportion of the experts (Table 1 again). Even though nationality is high on the list, the focus of this article will remain on the aspects of ethnicity, religion and language, as these are identified as the key ethnic and cultural characteristics in the CES recommendations and are seen as identity markers for the ‘ethnic package’.

The census experts were asked whether they agreed or disagreed with the inclusion of citizenship, ethnicity, language and religion in census questionnaires for all countries in the UNECE region.^[54] In their replies, there was a clear trend—most experts (strongly) disagreeing with including ethnicity (57%) or religion (61%) in the census questionnaire (Table 2). However, with regard to language (32%) and citizenship (10%), a lower proportion was against the inclusion of questions. This result is not completely surprising, since citizenship is a required census topic by the EU census regulations and a core topic in the CES recommendations, and language is an indicator often used in policymaking.^[55] As mentioned above, ethnicity and religion are not mentioned in the EU regulations and are non-core in the CES recommendations. That most of the experts would not include these aspects probably explains why there are no coherent rules and why practices are so divergent. This theme will be developed at a later stage.

The survey included another question that was very similar, but instead of asking about the UNECE region, this one focused on the EU census regulations: did the census experts think the EU population and housing census regulations should contain more precise rules about census questions that enquired into citizenship, ethnicity, language and religion (Table 3)? The results show answers for citizenship rather spread out and indecisive. But for ethnicity, language, and religion, the results were relatively similar: about two experts in every three (strongly) disagreed with any need for more concrete EU rules. This may reflect the diverging practices in Europe on how to collect ethnic data; but also it may be seen as an indicator of the degree of sensitivity that surrounds these issues.

Table 4 shows that there was a divide in responses between those experts claiming they had good knowledge of the Western Balkan region and those stating that they did not. Although they came from a small sample of respondents, the results were significant (at the 0.05 level) for ethnicity and language.^[56] While experts who indicated that they were knowledgeable about the Western Balkan region preferred to exclude the aspect of ethnicity from censuses or would only include it as a voluntary category, the answers from experts who indicated that they did not know the region were more spread. This shows that the sensitivity of the question is known to census experts especially if they know the Balkan region. Most Balkan experts would also only include enquiry into language as a voluntary category in the census, while the majority of non-Balkan experts would most definitely include this aspect in a census.

In the main, census experts—and especially those with a good knowledge of the Western Balkans—do not agree with the inclusion of questions on ethnicity and religion in a census. Language and especially citizenship are characteristics often used for policy-making and mapping migration, and so questions on these can more often be justified. The survey reflects what was established earlier: that there are no EU conditions requiring the collection of data on ethnic and cultural categories. Even the census experts’ opinions diverge on how to collect such details. Though, in the main, the experts were against including ethnicity and religion in a census, the national censuses under investigation did include these questions.

Ethnic and Cultural Categories in Croatia, Bosnia, and Macedonia

The Western Balkan countries are on their way to EU membership, and, as such, present an interesting case for the analysis of census-taking. On the one hand, the census is part of the EU acquis—though, as the survey results have shown, there are no coherent European guidelines on how to ask for ethnicity, language or religion and the EU does not require countries to enquire into these aspects. On the other hand, since the dissolution of Yugoslavia, ethnicity has been institutionalized.^[57] It is intricately intertwined with positive and negative discrimination. In part, this is a legacy from the old Yugoslav system, which had elaborate mechanisms for ethnic representation and minority protection; but after the wars following the dissolution of Yugoslavia, ethnic representation increased.^[58] Particularly in ethnically heterogeneous societies like Bosnia and Macedonia, this can lead to difficulties in the census process as will be outlined in the following sections.

Croatia

The latest census in Croatia was conducted, as planned, in April 2011. Even though, due to difficulties in hiring staff, there was a delay in the publication of the census data, no major difficulties were experienced.^[59] In the Croatian census’s methodological guidelines of 2011, ethnicity is grouped with language and cultural identity, being seen as ‘sense of belonging to a society (nation)’.^[60] This includes awareness of integration into the community, and also ‘special qualities in relation to other such communities’.^[61] The description of ‘sense of belonging’ to the society or nation clearly links ethnicity to national identity, in this case the Croatian national identity. The latter part of the description also links it to a possible differentiation from other ‘communities’. The methodological guidelines go on to provide details of how the data on ethnicity should be presented: first the data for Croats should be listed, and then the other twenty-two ethnic minorities in alphabetical order.^[62] This shows Croats being differentiated from other ethnic groups, which are ranked separately in the official presentation. By listing only a set number of ethnicities, the presentation ensures that a selection entailing inclusion/exclusion must be made.^[63] In this case, one might be left wondering why the selection is presented alphabetically and not according to which group is most prevalent.

Thus, in the Croatian census questionnaire, citizenship is grouped with questions about ethnic and cultural identity.^[64] This is not in line with the CES recommendations, according to which the question on citizenship should be clearly distinguished from ethnic and cultural characteristics to avoid confusion or indirect links being made between them.^[65] In addition, there is no option to abstain from declaring citizenship, unless a person does not possess citizenship at all. In the methodology it is stipulated that if a person does not want to declare his/her citizenship, the results will present this person as having unknown citizenship.^[66] By putting citizenship so close to the other ethnic and cultural categories, the census implies that there is a link between citizenship and the ethnic identity package.

With regard to the categories for the questions on ethnicity, language, and religion, instead of offering open categories, as advised in the CES recommendations, the questionnaire first offers the option ‘Croat’ (or in the case of language, ‘Croatian’ and for religion ‘Catholic’ or ‘Atheist’).^[67] Only afterwards is there an open field to declare anything else. In the methodology it is stated, with regard to the categories for ethnicity and religion, that ‘persons could freely declare’ their ethnic or religious affiliation.^[68] This facility is not offered, however, with the question on language. Language in Croatia is taken to be a person’s mother tongue.^[69] Individuals speaking more than one language at home should declare the one they consider to be their mother tongue. This too is out of line with the CES recommendation, according to which the questionnaire should include at least two questions on language and give open-answer boxes.^[70] It is not possible to declare more than one language. According to the enumerators’ instructions, the answers for these categories should not be read out by the enumerators, so as not to influence outcomes.^[71] Fines for possible malpractice by enumerators are described in the census law.^[72] The questionnaire offers a box to tick for undeclared ethnicity and religion; but this option is unavailable when it comes to language.

Though enumerators should not influence responses, when we look at the census questionnaire it seems that ‘Croat’, ‘Croatian’ and ‘Catholic’ are the preferred categories for the questions on ethnicity, language and religion. The questionnaire shows that, amongst the sensitive aspects, language is treated differently from ethnicity and religion. Whereas ethnicity and religion have an option for non-declaration, this is not the case with language. However, when the categories are grouped, they can be seen as an ‘ethnic package’, with language being the most important indicator—and one that may be taken to show national identity.^[73] In addition, in the Croatian questionnaire, citizenship is linked with ethnic and cultural characteristics, which is something that should definitely be avoided according to the CES recommendations.

In Croatia, rights granted to minorities depend on a numerical threshold as outlined in the 2002 Constitutional Law on the Rights of National Minorities. If national minorities make up more than 1.5% of the total population, they receive at least one seat, or a maximum of three seats, for their representatives in the Croatian parliament.^[74] Representation in local government bodies depends on population numbers too: the threshold is 15%.^[75] If a certain minority makes up more than 30% in a local self-governed unit, this language can be used as the official minority language.^[76] The implementation of this law was highly influenced by the EU accession process of Croatia.^[77] According to this law, Cyrillic script could be introduced as the official form of writing in about thirty municipalities where Serbs make up more than 30% of the population. This caused major protest in 2014, as well as a call for a referendum for changing the constitution.^[78] Although, with regard to ethnic and cultural questions, the census in Croatia was not in line with the CES recommendations, the collection of the census data was achieved without major difficulties. The census data is destined for use as a basis in policy-making and can have consequences for ethnic minority groups, as was shown by the protests in 2014. Since there are no EU conditions on how to deal with ethnic and cultural data and Croatia had closed the acquis chapter on statistics before the start of the census, it can be seen that the census did not play a big role in Croatian EU accession.

Bosnia and Herzegovina

In October 2013, Bosnia and Herzegovina held its first census. Although, according to the International Monitoring Operation led by Eurostat, ‘the census enumeration was carried out smoothly and in accordance with international standards’,^[79] there were reports of minor irregularities.^[80] The results were published after serious delays in June 2016.^[81]

The Bosnian census law and methodology address the ethnic composition of the enumerators, which should follow the ethnic composition revealed by the census results of 1991, and mentions fines for enumerators who try to influence a respondent’s answer with regard to the sensitive questions on ethnicity, language and religion.^[82] This already gives an indication of the sensitivity of these aspects in the region.

The census questionnaire issued by Bosnia and Herzegovina features questions on ethnic and cultural characteristics, which appear in a different order to that adopted by Croatia. Citizenship is not grouped in the same answer section as ethnic and cultural characteristics. However, the Bosnian questionnaire includes the category ‘entity citizenship’.^[83] This has been much debated and it is believed that it might be used by ethnic Serbs to pave the way towards a referendum over independence for Republika Srpska, the entity which has a majority of ethnic Serbs.^[84]

If we look at the ethnic and cultural questions, the Bosnian census questionnaire asks first for ethnicity and religion, and only afterwards for language.^[85] For all these questions, the first answer category is an open-answer field in which to insert the reply the respondent gives. It is only below this open field that there are boxes giving preselected options. For ethnicity, these are ‘Bosniak’, ‘Croat’, and ‘Serb’; for religion they are ‘Islamic’, ‘Catholic’, ‘Orthodox’, ‘Agnostic’, and ‘Atheist’; for language they are ‘Bosnian’, ‘Croatian’, and ‘Serbian’. While the preselected categories for ethnicity seem to be ordered alphabetically, the answer categories for religion appear connected with those for ethnicity in the question above.^[86] The same holds for the answer categories for language. It is made to look as if there is a clear connection between these categories, and that people should stick to the preselected ethnic, religious, and linguistic answers in an ‘ethnic package’ so as to increase the numbers shown for the constituent ethnic groups. Ethnicity in Bosnia and Herzegovina is described as ‘ethnic/national affiliation’ and language is noted as ‘mother tongue’. It is not permissible to tick more than one answer box, and therefore not possible to declare multiethnicity or more than one religion.

Above these answer boxes, it is stated that the questions on ethnicity and religion are not obligatory, but it should be pointed out that an answer does need to be given: ‘[if] the answer cannot be obtained, the enumerator shall write down “Unknown”’.^[87] Here, children whose parents have different ethnic and/or religious backgrounds need to decide which side they are on. From an outsider perspective, this seems strange in a country which contains several ethnic groups with intermarriages between them. This practice also goes against the CES recommendations, according to which respondents should be free to declare more than one ethnic affiliation.^[88] The methodological guidelines allow the declaration of more than one ‘mother tongue’, but only in ‘exceptional cases, when the enumerated person declares that s/he has two mother tongues’.^[89] In such an instance, this has to be put in the open-answer field. Though language can be taken as an indicator of national identity, as described above, Bosnia, it seems, allows more leeway with regard to this aspect than it does with the other ethnic and cultural categories. The language categories offered in the census questionnaire—‘Bosnian’, ‘Croatian’, and ‘Serbian’—are not as different as might appear, since they all developed from the same linguistic origin.^[90] Another aspect taken into account in the Croatian census methodology but not taken up in this one is that, in Croatia, people can declare their ‘regional, religious or any other affiliation’ in the category of ethnicity.^[91] There is no mention of such an opportunity in the Bosnian census law and methodology. This is indeed surprising, since the category ‘Bosniak’ is a somewhat recent invention. Before the break-up of Yugoslavia ‘Muslim’ was the term used.^[92] This may cause confusion among the Bosniak population, who in the last Yugoslav census (1991), had to declare themselves ethnic Muslims, but now find themselves referred to as ‘Bosniak’.

A further divergence in these census questions from those of Croatia is that language is treated differently from religion and ethnicity. While the questions on ethnicity and religion are not mandatory, the language question must be answered. Here there is no direct option allowing respondents to avoid declaration of their mother tongue. What the CES recommendations insist should not appear is any link between ethnicity and national affiliation.^[93] But this recommendation is ignored: the category for ethnicity is called ‘ethnic/national affiliation’. There seems altogether to be a preference for the set ethnic packages of ‘Bosniak’, ‘Croat’, and ‘Serb’ in the census form.

These ethnic packages clearly represent the three constituent ethnic peoples included in the Dayton Agreement:^[94] If a person ticks the box ‘Bosniak’, it is expected that ‘Islamic’ and speaking ‘Bosnian’ will follow. The same seems to be the case for the combination ‘Croat’, ‘Catholic’, and ‘Croatian’; and, in turn, for ‘Serb’, ‘Orthodox’, and ‘Serbian’. It comes as no surprise that the publication of the census data for ethnicity, language and religion only mentions the categories connected to these three constituent groups.^[95] Initially, there was concern that the data from the census in Bosnia might affect the power-sharing structures in the country, which, following the Dayton Agreement, divide representation proportionally among the ‘constituent [ethnic] peoples’—Bosniaks, Croats, and Serbs—as they appear in the census results of 1991.^[96] However, the population numbers given in the 2013 census do not differ significantly from the 1991 figures.^[97] Though the census was supervised by the EU through an International Monitoring Operation, the census results have been questioned by the ethnic Serbs.^[98] This challenge, indirectly connected with a potential increase in ethnic numbers, is part of a political impasse in Bosnia on how to count the resident population.^[99] The whole issue is still causing debates. Even though the results have been published, Republika Srpska authorities do not recognize them.^[100]

Macedonia

Macedonia conducted censuses in 1994 and 2002. In the 2010 census round, a new census was planned for early October 2011.^[101] However, on the tenth day of the enumeration process, the Macedonian Parliament annulled the Census Law it had adopted earlier and cancelled the census altogether. This was due to political disagreements, especially on whether or not to include the diaspora population of the country in the count.^[102]

Macedonia’s 2011 census questionnaire is the only one of the three studied here that gives the title ‘ethnic characteristics’ to a section.^[103] The order of questions on ethnic and cultural characteristics is the same as in the Croatian questionnaire—a question on ethnicity, followed by questions on language and religion. As in the Bosnian questionnaire, the ethnic and cultural questions are not grouped in the same answer section as the question on citizenship. However, unlike the other questionnaires, the Macedonian one does not give any preselected answer categories for the questions on ethnicity, language, and religion: it offers open-answer fields for all. In addition, it provides three questions about language: ‘mother tongue’, ‘language usually spoken’, and ‘knowledge of foreign languages’. There is one question each for ‘ethnic affiliation’ and ‘religious affiliation’.^[104] Apart from the question on ‘knowledge of foreign languages’, none of these are mandatory. Though there is no direct option offered to avoid declaration in these categories, the census methodology states that the enumerators may fill in that the person did not declare, or put ‘I do not know’.^[105] In short, this questionnaire is the only one making all three ethnic and cultural categories voluntary, and is thus the sole example fully in line with CES recommendations.

Another differentiating feature is that the Macedonian census law allows the collection of census data in a language other than Macedonian and a script other than Cyrillic. The law does not mention the name of the language (Albanian) but refers to it as ‘the official language used by a minimum 20% of the citizens of the Republic of Macedonia’.^[106]

As all the Macedonian census questions only offer open-answer options, the questionnaire potentially allows the declaration of more than one ethnicity. This questionnaire is thus the only one without a prominent ethnic preference. It is remarkable that, although this census form was most in line with CES recommendations, it was the one that did not go through with the enumeration. The Macedonian census form did not suggest any ethnic preference, but the census failed due to lack of trustworthy data.^[107] The census was stopped because there was an inconsistent implementation of the accompanying methodology.^[108] This revolved around the issue of whether to include the diaspora population to boost ethnic (and linguistic) numbers, or whether to exclude this group.^[109] Already one year before the census operation started, Albanian politicians were proclaiming that they would not accept the census results.^[110] Then, throughout the census process, ethnic Albanians wanted to increase their own numbers in the survey, while ethnic Macedonians were just as eager to play these numbers down.^[111] Following the Ohrid Framework Agreement, language rights should be given to communities if the language is spoken by more than 20% of the population. Also, proportional representation in the police and in the bureaucracy depends on recorded population numbers.^[112] Since the census of 2002, the only ethnic group meeting the threshold of 20% is the Albanian group. It was the political struggle between the ethnic Albanians, who wanted to boost population numbers so as to keep their rights and standards, and the ethnic Macedonians, who wanted to decrease Albanian influence, that led to the cancellation of the census operation.

Comparison

Since there are EU and international guidelines for census-taking, one might expect that countries would harmonize their practices in this area. However, when we looked at the ethnic and cultural questions in the census forms of the case countries, we saw diverging approaches. In all the census forms, the questions on ethnicity and religion were not obligatory (Table 5), and this, at least, is in line with the CES recommendations. While Croatia had closed-answer categories presented first, and then an additional open-answer field, Bosnia offered these open and closed fields the other way round. Macedonia, by contrast, offered open-answer fields only. Macedonia also differed on the language question: while in Croatia and Bosnia the question on language was mandatory, the Macedonian census form did not insist on this. It did, however, present several questions about language, and, for a question on ‘foreign languages’ spoken, made an answer obligatory. The only questionnaire fully in line with the CES recommendations was the one Macedonia produced. Ironically, the Macedonian census was also the only one to be aborted. Meanwhile, the census operations in Croatia and Bosnia were completed, despite these countries’ suggesting preferred ethnic packages in their questionnaires. Indeed, Croatia, which produced the questionnaire least in line with the CES recommendations was the most successful of the three states in completing the census operation. Bosnia was more in line with the recommendations than was Croatia, but since this country’s questionnaire suggested categories and included a much-debated question on entity citizenship, it emphasized a linking of the political with ethnic and cultural identities. All this shows that there is far more to a census than a simple population count. As in Macedonia and Bosnia, a census can be highly politicized.

Conclusion

Because political participation and mandate representation quotas have become linked to the political structure of West Balkan countries, ethnic and cultural characteristics need to be collected. But there are no EU conditions on how to manage this, and international recommendations are only partially followed, even though enquiry into ethnic and cultural categories often forms part of the population censuses in the Balkan region. Recognizing this problem, the present article has used survey data to provide an international framework to review these categories. The results, based on the survey data and the census questionnaires of three sample countries, clearly demonstrate that there is no one-size-fits-all approach on how to collect ethnic and cultural data. Some subsidiary points have emerged. 1) There seems to be a difference in the degrees of sensitivity attached to different aspects: ethnicity and religion are highly sensitive and citizenship and language are less so. 2) Since there is no coherent EU approach on how to deal with ethnic and cultural data, it is not possible to speak of Europeanization in this policy area.

In all cases the main paradox surrounding the collection of ethnic and cultural data is present: in order to be recognized, minority groups need to have their presence known, but the collection of data about their existence can cause discrimination or bring about other negative consequences. The negative implications are already visible in the politicization of censuses and census outcomes. This politicization was somewhat weaker in Croatia, where the debate on ethnic and cultural characteristics came up largely after the census, when, following the results, about thirty municipalities were allowed to introduce Cyrillic script for their official language. This led to major protests and even a referendum on whether to change the constitution to increase the numerical threshold.^[113] The Bosnian census categories recognized three constituent ethnic groups, which all had (and have) their own agendas for ethnic national state-building.^[114] Intense political debate on the census definitions was the main reason why the publication of the census results was so seriously delayed. Even now, after the results have been published, the Serb authorities do not recognize them.^[115] In Macedonia, the census was highly politicized, as both the main ethnic groups—Macedonian and Albanian—wanted to maintain or increase their political influence. The linking of language rights to numerical thresholds led to the spread of different methodologies in the census operation, ultimately causing it to fail altogether.

Thus, in all three countries studied, the inclusion of questions on ethnic and cultural characteristics in the census has had political repercussions. Moreover, these have been divergent in character. The present article does not attempt to provide a solution, but highlights the paradox: including ethnic and cultural questions in a census shapes ethnic preferences, which may influence policymaking and the distribution of rights; but these ethnic preferences may also be used for ethnic nation-building.^[116] The influence of ethnic preferences has so far been underestimated by the EU. Lack of attention to how it is addressed in census-taking flaws the current EU approach to state-building and enlargement.^[117] There is no coherent European approach on how to collect ethnic and cultural data, and we cannot speak of Europeanization of census-taking.

As minority rights and political representation in the Western Balkans are connected to the results shown in its censuses, further research is needed on the possible outcomes of this linkage, which so far have been unanticipated. That this is one of the important challenges the EU faces in the Western Balkans is made clear by the ongoing debate on the definition of ‘usual residence’ in Bosnia and the current political crisis in Macedonia.