‘Inside Out’: The Politics of Enumerating the Nation by Ethnicity

Abstract

Since the 1990s, state practices of counting and classifying populations by ethnicity have come under increased scrutiny within the social sciences (Arel 2002; Kertzer and Arel 2002; Nobles 2000; Perlmann and Waters 2002; Petersen 1997; Statistics Canada and U.S. Census Bureau 1993). A number of excellent case studies have provided critical insights into how and why ethnic enumeration is pursued in particular times and places. However, with some notable exceptions (Morning 2008; Rallu et al. 2006), little attention has been given to theorizing or empirically testing a global model of ethnic classification and counting. Consequently, there is a limited understanding about the general conditions that impede or encourage state recognition of ethnicity in the national census and the forms that such recognition takes.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Since the 1990s, state practices of counting and classifying populations by ethnicity have come under increased scrutiny within the social sciences (Arel 2002; Kertzer and Arel 2002; Nobles 2000; Perlmann and Waters 2002; Petersen 1997; Statistics Canada and U.S. Census Bureau 1993). A number of excellent case studies have provided critical insights into how and why ethnic enumeration is pursued in particular times and places.¹ However, with some notable exceptions (Morning 2008; Rallu et al. 2006), little attention has been given to theorizing or empirically testing a global model of ethnic classification and counting. Consequently, there is a limited understanding about the general conditions that impede or encourage state recognition of ethnicity in the national census and the forms that such recognition takes.

This chapter represents an exploratory attempt to develop and test a general theoretical model of ethnic enumeration. It is underpinned by two assumptions. The first is that the recognition of ethnic differences in forums such as the census is influenced by factors that have similar effects across states. This assumption marks a significant departure from the prevailing view that ethnic enumeration is best understood as the unique product of a country’s historical relations and contemporary conditions. The second premise is that ethnic classification and counting is influenced as much by factors exogenous to states, as by domestic conditions. Research in the world society tradition has persuasively shown that states are not disconnected islands, but are enmeshed in global networks through trade, participation in International Governmental and Nongovernmental Organizations (IGOs and INGOS), and the endorsement of global human rights instruments (Cole 2006; Tsutsui and Wotipka 2004). We propose that the extent to which states suppress or embrace the recognition of ethnic diversity within their boundaries is likely to be influenced by their level of integration into global civil society (Meyer et al. 1997). This theoretical approach is novel, because it considers how integration into world society ⁴ influences the recognition of ethnic differences ‘at home’.

We examine these proposals through empirical analyses of census and population registration forms for 151 countries for the period 1995–2004. Our comparative approach enables us to empirically weigh the relative importance of internal and external state factors, and it raises a set of fascinating questions. What are the domestic conditions or features that systematically encourage or suppress state recognition of ethnic differences? Does state support for ethnic equality on the world stage translate into the recognition of ethnic differences at home? Why do some states recognise ethnicity as a dimension of difference, but stop short of acknowledging specific group identities?

We begin by considering critical perspectives on census-taking and ethnic enumeration. From the literature we identify a set of factors most likely to predict whether states engage ethnic counting and classification and what form it is likely to take. After showing the relative distribution of how states enumerate, we use maximum likelihood ordered logistic regression models to examine two related outcomes: if states enumerate by ethnicity and if specific collective identities are recognised. Our findings support the prevailing view that national strategies of ethnic enumeration are shaped by dynamics internal to states, but we also find support for our hypothesis that ties to global civil society matter. The findings confirm our general argument that the effects of state-level factors, whether internal or external, can be generalised across vastly different geographic contexts. This latter finding offers much promise for broadening the theoretical understanding of ethnic accounting, particularly if it can be extended in the future to include a greater set of variables and additional time periods. Before discussing the empirical analysis in more detail we first briefly review theories of census-taking and ethnic enumeration below.

2 The Politics of Classifying and Counting by Ethnicity

Given the considerable resources involved in producing a census and the varied ways in which census data are used, it is unsurprising that governments promote census-taking as a universal and efficient model of objective, scientific inquiry (Ventresca 1995; 2002). Among social scientists, however, counting and classifying people is seen first and foremost as a political endeavour (Anderson 1991). As a political process, the census is influenced by pressures exerted from the ‘top down,’ as well as the ‘bottom up’ (Arel 2002; Bonnet and Carrington 2000; Kertzer and Arel 2002; Morning and Sabbagh 2005; Nobles 2000; Prewitt 2000; Rallu et al. 2006). From the former vantage point, ethnic classifications and categories are seen as an extension of hierarchical arrangements and dominant group interests. Early US censuses are an often-cited exemplar of these impositions. Early censuses divided the populace into ‘free’ whites, slaves and Indians, taxed or untaxed. Depending on which group an individual was assigned to, he or she would be counted as a ‘whole’ person, ‘three-fifths’ of a person, or not at all.² This symbolic positioning not only justified and reinforced dominant racial logics and race-based inequalities, but also influenced the balance of power in a way that ensured Southern state interests would have a strong presence in US politics (Ellis 2000). The federal government’s ability to influence race is most apparent in its institution of blood quantum rules to determine who could identify as American Indian, the effect of which was to limit the size of the American population and the state’s obligations to them (Snipp 1989).

The notion that state practices of ethnic enumeration are also shaped from the ‘bottom up’ is a comparatively recent idea that follows transformations in ethnic relations, notably growing diversity and the diffusion of minority rights (Petersen 1997; Rodríguez 2000). Evidence of bottom up politics may be seen in the shifting purpose of ethnic data collection, from a tool for maintaining minorities’ sub-ordinance, to one that helps ensure compliance with anti-discrimination legislation (Morning and Sabbagh 2005; Simon 2005). In many developed, multicultural countries ethnic minorities have successfully lobbied to have ethnic distinctions recognised in official data collections, and sometimes to have their group identities explicitly listed on official forms.³ The US again provides an interesting example with the multiracial lobby. In the lead up to the 2000 census, multiracial activists were instrumental in pushing through changes that allowed people to tick more than one racial category, but failed to institute a specific mixed race category (Perlmann and Waters 2002).

Despite a number of groundbreaking studies of ethnic politics and censuses within states, few have studied these processes in an international context, and attempts to advance theoretical arguments about global enumeration practices are rare. The closest to a general theory comes from Rallu et al. (2006) typology of ethnic enumeration. Their model identifies four dominant paradigms of ethnic counting, characterised by different political goals: (1) for political control (e.g., colonial censuses); (2) to support a discourse of national hybridity (e.g., Latin America); (3) for anti-discrimination policies (e.g., US); and, (4) non-enumeration in the name of national integration (e.g., France). Implicitly, the typology frames enumeration as a top down process, influenced by internal conditions such as migration and inter-ethnic relations. It provides a valuable heuristic framework within which to situate states with broadly similar political motivations but, as a typology, is best positioned to describe ethnic enumeration rather than explain it.

Part of the difficulty of developing a global theory of ethnic enumeration is the lack of a common understanding about the dependent variable – what counts as ethnic counting? In the absence of a global standard of identity, states have at their disposal a wide range of concepts with which to define difference. In the US human difference has historically been filtered through the biological frame of phenotype or race (Omi and Winant 1994). Elsewhere, and at other times, language, origins and culture have served as the salient boundaries distinguishing socially defined groups. Contextual diversity in how difference is understood is complicated by the multiple meanings attributed to equivalent terms. Nationality, for example, is interpreted in France as a civic, legal identity akin to citizenship, but in Eastern Europe and the former Soviet states is more closely aligned with cultural identity derived from ethnic origins (Kertzer and Arel 2002). Even seemingly unambiguous markers such as language may be subjectively rendered. Arel’s (2002) analysis of census-taking in the post-Soviet states demonstrates how the distinction between mother tongue and everyday language became infused with political meaning, bound up with claims to ethnic nationality and territory. Morning (2008) notes these parochial perturbations, but argues that the diversity of ethnic indicators should not dissuade from the identification and analysis of cross-national similarities. Her innovative comparative research has shown that much of the diverse nomenclature used to describe collective identities (e.g., ethnicity, race, ancestry, and indigeneity) is underpinned by the common concept of descent.

Morning’s work provides a valuable starting point for efforts to develop empirically grounded theoretical arguments about ethnic enumeration processes. By integrating census and population registration data with information about state characteristics and conditions, we are able to empirically test whether countries with similar profiles adopt similar strategies for enumerating their populations by ethnicity. We also examine whether states go beyond the acknowledgement of ethnic differences, to legitimise collective identities by listing them on the census or registration form. Such recognition concedes the presence of pre-existing collective identities and may even nominate new ones into existence (Abu-Laban and Stasiulus 2000).⁴ The intent of identification need not be supportive of groups’ rights – indeed, such strategies might be pursued to mark out groups for discriminatory treatment, or to facilitate ‘statistical fragmentation’ (Arel 2002). Our approach is to treat state motivations as unmeasured and focus on establishing whether systematic associations exist between state-level factors and strategies of ethnic enumeration. These relationships are elaborated below.

3 Factors Internal to States That Affect Ethnic Enumeration

In the absence of a general explanatory model of ethnic counting and classification, we look to national and regional studies for clues about the factors most likely to influence state approaches to ethnic enumeration. Though couched in parochial terms specific to time or place, there are common themes that can be rendered in more abstract and thus generalisable, terms. We identified four sets of factors endogenous to states that are critical to understanding how and why states enumerated by ethnicity. They are: ethnic group relations, immigration, post-colonial sovereignty and resources.

3.1 Ethnic Group Relations

Most studies agree that inter-ethnic relations within a country play a key role in shaping the ethnic enumeration approach taken in its national census. Depending on the specific nature of ethnic groups and their concomitant rights claims, relations between them could influence ethnic classification and enumeration in various ways. In seeking to articulate and measure the effects of ethnic relations on ethnic counting and classification we focus specifically on ethnic contenders. Contender groups are those whose collective identities are founded on claims of territorial or political independence, or who occupy distinct social and economic niches as a result of unequal historical arrangements such as slavery. Ethnic contenders include indigenous peoples whose aspirations often include some form of self-government, as well as regionally concentrated ethno-nationalist groups with a history of organised political autonomy (e.g., Turkish Cypriots in Cyprus). From the perspective of governments, the claims of such groups are often seen as more contentious than those of ethnic immigrants. The latter may seek to retain rights of ethnic customs and associations, but they rarely challenge the legitimacy of the political status quo in their host countries (Koopmans and Statham 1999). The claims of indigenous and ethno-nationalist minorities are made especially potent by the growth of international support for indigenous rights in legal and political forums, making it difficult for states to avoid some form of ethnic enumeration. Yet states also have an inherent interest in building national cohesion and limiting claims that might arise from the politics of recognition. This tension suggests a solution that acknowledges the existence of ethnic difference, but minimises the leverage that such groups might gain through the explicit legitimisation of collective identities.

3.2 Immigration

Increasing transnational flows of migrant workers and refugees have heightened awareness of ethnic differences in states hitherto secure in the myth of ethnic homogeneity and further diversified states with visible multicultural populations. Much has been written about how traditional immigrant countries in North America, Europe and Australasia have liberalised their exclusionary immigration policies to meet pluralistic models of entry and settlement (Dumont and Lemaître 2005; Pearson 2002). Just how immigration shapes ethnic enumeration practices is likely to depend on whether immigration is driven by temporary labour market demands, or oriented towards more permanent settlement. In the former instance, we expect governments will be reluctant to emphasise ethnic differences or legitimate collective identities. Rather the impetus is more likely to be geared differentiating the native- from the foreign-born through enumeration strategies that focus on civic-legal status. Even if the census excludes migrants on short-term permits of one or a couple of years, the issue of how to enumerate the rest of the foreign worker population remains.

Where immigration is linked to more permanent settlement patterns, we expect governments will adopt an ethnically cognisant approach. One reason is the increased potential for discrimination and ethnic conflict that arises from significant migrant inflows. Whereas traditional ‘host’ societies tend to contain white majorities, most of the ‘source’ countries comprise persons who constitute visible minorities in their new settings. This disjuncture suggests a growing incentive for ethnic enumeration because governments require ethnic data to monitor discrimination and institute ameliorative policies (Simon 2005). Their motives may, of course, be less benevolent. The collection of ethnic information may also assist governments to monitor and control the dispersion of migrant communities whose ethnic traits and patterns of association appear to defy integration into the existing social and economic order.

3.3 Post-colonial Sovereignty

Historical and political factors feature prominently in research documenting the evolution of ethnic enumeration in a specific country or region (see, for example, Nobles’ research on racial enumeration in the United States and Brazil). A core feature of a state’s political trajectory is its history of independence. Whether a state has an established history of sovereignty, or emerged from the bonds of colonialism or other political struggles, may bear upon the government’s willingness to give expression to ethnic differences. States that gained sovereignty after 1965 emerged in a fundamentally altered world system – one marked by the Cold War, civil rights, ethnic revivalism and the growth of human rights regimes. In 1965 a slew of international organisations emerged, setting an international agenda whose primary goals reflected the post-colonial and post-war civil rights demands for equality. These included the International Convention on the Elimination of all forms of Racial Discrimination (ICERD); International Covenant on Civil and Political Rights (ICCPR) and International Covenant on Economic, Social and Cultural Rights (ICESCR). The subsequent expansion of international human rights organizations under the auspices of the United Nations institutionalised the global recognition of minority rights and formalized the expansion of human rights to the masses. States born into this new world system may be more inclined to adopt ethnic enumeration strategies that reflect these political sentiments.

In addition to coming of age during a period of transformative change, newly independent states are less rooted in historical path dependencies and freer to incorporate new ideas and discourses trumpeting ethnic equality and the rights of groups and individuals to self-define. Processes of ethnic recognition might simply entail the continuation of a colonial legacy of drawing ethnic distinctions, but with the goal of legitimating ethnic differences, rather than for the purpose of domination and exclusion (Rallu et al. 2006). By comparison, established states – especially ones that have been sovereign for many centuries – are much more likely to be vested in an approach that stresses national unity above ethnic differences. The most glaring example of this is France’s resistance to the inclusion of ethnic and cultural distinctions in its own census in the name of secularism and French identity.

3.4 Resources

Finally, a state’s level of resources is also likely to bear upon the kind of ethnic enumeration strategy it adopts. The cost of census-taking is well documented. In the US alone the 2000 census cost $6.6 billion, double that of the previous census. The cost of the 2010 census is expected to be twice as high again, at close to $12 billion (United States General Accounting Office 2004). Rising costs create a strong inducement to rationalise census taking so that only items that yield information vital to governance are included. Some countries have dispensed with regular censuses altogether, opting either to construct a virtual censuses from population registration data (e.g., the Netherlands), or to administer a nominal household census form that is supplemented with other administrative data (e.g., Norway and Spain).⁵ A more pointed argument is that the politics of diversity, of which ethnic enumeration is part and parcel, is a distinctly first world practice that can only be afforded by countries with a reasonable standard of living. Although developing countries may contain considerable ethnic diversity, resource constraints may mean that the census is seen primarily as a tool to document the basics of fertility, mortality, literacy and employment, rather than to track the expression of identities.

4 Factors External to States That Affect Ethnic Enumeration

Historical and structural properties of states might account for the key endogamous influences on ethnic enumeration practices but alternative theoretical perspectives suggest external forces also shape processes of ethnic recognition. Research in comparative politics and sociology, especially the world society literature, emphasise the responsiveness of nation-states to global politics and the impacts on domestic policy-making (Boli and Thomas 1997, 1999; Meyer et al. 1997). From this perspective integration into global civil society ought to lead to isomorphism in ethnic enumeration norms and/or practices by drawing countries into a common global culture and providing political activists with the forum within which to advocate for minority recognition. There are at least two channels through which exogenous factors might influence processes of state recognition. The first is through membership in INGOs; the second is through support for specific international human rights instruments (Cole 2006; Tsutsui and Wotipka 2004). INGOs and IGOs are distinguished in the world society literature by their relationship to the state and civil society. While IGOs are often seen as empowering the state and state interests (Olzak and Tsutsui 1998), INGOs are characterised by their unique position to diffuse global norms about civil society and human rights (Boli and Thomas 1997, 1999; Tsutsui 2004). States with strong ties to INGOs open up political opportunities for ethnic groups to pressure states to adopt policies that compliment international agreements about human rights (McAdam and Rucht 1993).

Since the 1960s, the United Nations has established a slew of international treaties enshrining the rights of minorities with the express goal of forcing governments to act even-handedly towards them. As one of the UN’s oldest human rights instruments ICERD expressly prohibits ‘any distinction, exclusion, restriction or preference based on race, colour, descent and national or ethnic origin’, and it allows for the provision of special measures to ensure the ‘adequate development and protection of certain racial groups or individuals belonging to them’. In theory, states that signal their commitment to ethnic equality on the world stage ought to pursue ethnic enumeration because it legitimates the expression of diversity within their borders and provides the bases for the collection of data with which to evaluate and ameliorate group-level disparities (Morning and Sabbagah 2005). In reality, several factors militate against this. One is that ratification of a convention need not engender a genuine commitment to its goals (Cole 2006; Neumayer 2005). The stringency with which conventions are monitored and enforced can vary widely, making it relatively easy for nominally committed members to evade their responsibilities. In some cases, provisions for monitoring, compliance and enforcement are ostensibly nonexistent or weak, with powerful states often loath to use coercive strategies to pressure states into addressing their poor human rights records (Neumayer 2005). To some extent, we see the disconnection between membership and commitment as a problem of measurement. Rather than treat ratification as an expression of a state’s commitment, we prefer to measure commitment directly. This allows for a more careful evaluation of how expressed dedication to ethnic equality in the international arena translates into ethnic enumeration practices at home.

A trickier problem to resolve is the fact that although most signatories to ICERD condemn ethnic discrimination, there is no consensus about whether ethnic data collection is the salve. Some states view the collection of ethnic data as necessary to combat discrimination; others, as an act of discrimination in itself (Arel 2002). Indeed several European countries have argued that their commitment to eliminating discrimination is precisely why they do not enumerate by ethnicity. However, arguments invoking the historical misuse of ethnic data (e.g., to identify Jewish individuals during WWII) and constitutional prohibitions have been found wanting. Investigations have found several of the countries claiming non-enumeration on those grounds nevertheless collect ethnic data ‘under the radar’, especially on visible minorities of interest (e.g., Roma in the Czech Republic, and Romanians and Algerians in France, see Goldston 2004). Moreover, upon closer examination, it has been found most constitutions do not explicitly prohibit ethnic data collection but rather impose restrictions that make its collection subject to specific privacy and protection safeguards (Ramsay 2006). Our view is that concerns about enumeration as a form of discrimination testify to the ongoing salience of ethno-racial distinctions within states’ boundaries. Without ethnic data, strategies to eliminate discrimination by ethnicity are impossible – a genuine commitment to eliminating ethnic discrimination requires ethnic data.

5 Data and Method

To examine the connections between the foregoing factors and state processes of ethnic recognition and legitimisation, we use data from the Ethnicity Counts? database, which codes national census questionnaires and population registration forms for the period 1985 to 2014.⁶ For the purpose of this study we restrict our analysis to the 2000 census round which spans the decade 1995 to 2004. To define the sample population we consulted the United Nations Statistics Division’s (UNSD) list of countries that existed in June 2005, then referred to a separate UNSD list to determine whether a census was conducted in the 2000 census round (also see Morning, Chap. 2, this volume).⁷ We restricted our analysis to sovereign states for which a 2000 round form could be located. The exclusion of territories and dependencies was necessary to avoid inflating the effects of governing states (e.g., United States), and because membership in ICERD is limited to sovereign nations.⁸ Of the 203 nations in our sample, 175 had conducted a census in the latest round, for which 151 forms were located.⁹ To maximise our sample, we included countries missing from the 2000 round for which forms for preceding or successive rounds could be located. The earliest census form was for 1995; the most recent was for 2004. Because our interest is in the factors underlying state processes of ethnic recognition, we also deviated from a strict focus on the census to include countries that maintained population registers as a substitute.¹⁰ Population registers are prevalent throughout Europe, providing a regularly updated source of information on individuals (Poulain and Herm 2013). The scope of data collection varies across countries, but may include information on births, deaths, marriage and dissolution, family relations, education, employment, taxation, residence and migration status (Legoux and Perrin 1999). For each population registration country, we obtained copies of the appropriate forms directly from the agencies responsible for administering the databases. Once the census questionnaires and population registration forms were assembled and translated, we coded a wide array of ethnic variables. Relaxing the selection criteria to include forms beyond the 2000 census round and population registration countries yielded a final sample size of 151 states. The characteristics of our sample can be seen in Table 3.1.

Table 3.1 Census coverage and types, 1995–2004

5.1 Variables

Our first dependent variable, ethnic cognisance, measures the number of ethnicity items listed on a state’s census questionnaire or, in the case of population registration countries, the total number of unique ethnicity questions asked across all of the constituent instruments. We operationalise ethnicity to include questions that use the following terms: ethnicity, ethnic group, ethnic origin, descent, ancestry, race, indigenous, tribe, language, mother tongue, nationality, national origins and ethnic nationality. The variable is coded on an ordered scale from 0 to 3 where 0 = no questions; 1 = one question; 2 = 2 questions; and 3 = 3 or 4 questions.¹¹ Since the variable only had a range of 0–3 we treat it as an ordered categorical variable in our models. Our second dependent variable, ethnic legitimisation is a dummy variable that indicates whether states explicitly nominate groups into existence by specifying the name of at least one ethnic group on the enumeration form. States were coded as 1 if at least one ethnic group name was listed or appeared as a tick box on the form, and 0 otherwise.¹²

The endogenous variables cover structural properties of states, as well as inter-group dynamics arising from the presence or absence of migrants and ethnic minorities. With respect to immigration we use two separate measures. The first, net migration rate, was derived from the United Nations’ World Population Prospects data (United Nations 2000). It is the net average annual number of migrants from 1995–2000 per 1,000 people. We selected net migration over the number of foreign-born persons because it captures the most recent newcomers to the population, rather than the cumulative effect of immigration.

Unfortunately, the scarcity of complete and internationally comparable data on immigrant workers means we are unable to include a direct measure. Instead, to capture the visibility of immigrant workers in the labour market we use the percentage of international migrants that is male as a proxy.¹³ This variable is taken from the United Nations’ Population, Resources, Environment and Development data bank (PRED Bank 2006). The international spread of migrant workers has historically been characterised as a male dominated phenomena (Houstoun et al. 1984), albeit that women have a growing presence among global migrant communities (Pedraza 1991; Alcala 2006). At the very least, our proxy variable is indicative of a sex imbalance that may translate into differential treatment of immigrants within the borders of host states.

We constructed the variable ethnic contender using Phase IV data from the Minority at Risk (MAR) project, maintained by the Center for International Development and Conflict Management at the University of Maryland. The global dataset identifies 284 communal groups that were politically active in 2003, classified into 6 types of minorities: ethnoclass; ethnonationalist; indigenous; religious; communal contenders; and national minorities.¹⁴ We constructed a dummy variable coded 1 if an indigenous, ethnonationalist, or ethnoclass group existed within a state; and 0 otherwise. We note that MAR data is compiled from multiple sources besides the census, which minimises the risk that the variable ethnic contender is endogenous to the measured outcomes. For example, MAR identifies five separate ‘minorities at risk’ in France, including Muslims, Basques and Corsicans. The absence of official ethnic or racial data in France as the criteria used to define these groups in the MAR dataset was not based on census or population registration forms alone.

To capture a state’s level of national resources we use the 3 year average (1990, 1995 and 2000) of Gross Domestic Product per capita in (GDP), measured in constant US$. A natural log-transformation was performed to correct for a skewed distribution. The historical emergence of state autonomy is measured with a dummy variable new sovereign nation. We coded countries as 1 if they gained sovereignty after 1965; and 0 otherwise.¹⁵

We constructed three variables to capture the effects of exogenous factors on state enumeration practices. Two of them measure a state’s commitment to international treaties associated with the elimination of ethnic and racial discrimination. The first is a dummy variable ICERD signatory coded 1 if the state is a signatory to ICERD; 0 otherwise. The second is a measure of state commitment to ICERD based on the following factors: (1) if ICERD was signed; (2) if Article 14 was enforced, vesting the ICERD committee with the power to hear individual and group grievances against member states; (3) if at least 50 % of reports due were filed within the allotted timeframe; and (4) if countries signed ICERD on or before 1975. We treat commitment as a 4-point interval variable with 0 representing no membership in ICERD and 4 being fully committed. Ties to the international community are measured through a state’s involvement with international nongovernmental organisations. The INGO variable represents the total number of organisations of which a state is a member, either directly or through the presence of member organisations within that country. We take the natural log of INGO to correct for skewness.

Finally, we include several control variables. The Gini index controls for the effect of relative income inequalities on state processes of ethnic recognition. We derived Gini values from the 2007 World Bank development indicators (World Bank 2007) for a range of years over the 2000 census time frame (1994–2005). Where recent data was unavailable we used older data extending back to 1989 (n = 8), or imputed values using multiple imputation techniques (n = 20, see Shaefer 2002). Index values vary from 0 to 100, with 0 representing perfect equality. We also control for regional variation by using a region variable that denotes the broad geographic area in which a state is located. The literature suggests regional variation in ethnic recognition is partially due to historical trajectories such as slavery, colonisation, civil wars and so forth. We do not try to model these processes directly but treat the region variable as a weak proxy for these historical variations in state governing practices. Lastly, for reasons that may include differential resources, lack of infrastructure or newness of states, there are varying lengths of questionnaires in terms of the number of questions asked. We created a dummy variable that was coded as 0 if states had less than the median number of questions on their census form (median = 20) and 1 if states had a greater than median length of questions (see Table 3.2).

Table 3.2 Distribution of Key Variables by Region, 1995–2004 (N = 151)

6 Results

We start by briefly discussing the results of our descriptive analysis by broad geographic region (see Table 3.3). Like Morning (2008), we found significant regional variation in the prevalence of ethnic enumeration and in the concepts used to define difference. In Oceania, for example, ethnic enumeration was near universal but in Africa, Europe and Asia, ethnic enumeration ranged from 59.1 % in Europe to roughly 68 % in African and Asia. Overall, ethnic terminology was most often used to define difference, although regional preferences were also apparent.¹⁶ In South America, concepts of indigeneity and tribe prevailed while references to ethnic nationality and race were largely confined to Europe and North America respectively. In about half of all countries (56.7 %) the recognition of ethnic differences did not extend to the recognition of specific group identities. Only a little more than a third of European and Asian countries explicitly recognized group identities on their forms and only one-fifth in Africa, significantly less than the proportion that counted by ethnicity. North America and Oceania were the only regions for which the listing of ethnic groups on the census was commonplace.

Table 3.3 Maximum likelihood ordered logistic regression models for ethnic cognizance

In terms of the independent variables, a clear distinction is evident between immigrant receiving (Europe, North America, and Oceania) and sending regions (Asia, South America, and Africa). The growing presence of female immigrants in developed regions of the world is also apparent, with North America and Europe having slightly less male than female immigrants. The ethnic contender variable highlights two outliers. Oceania and South America represent two ends of the continuum with 70 % of South American countries having at least 1 ethnic contender group, compared to just one-fifth of Oceanic countries. Regional variation was also apparent with respect to the exogenous variables. Europe, Africa and South America had the highest percentage of state signatories to ICERD, whereas Asia and Oceania had relatively few signatories. Regions with the highest proportion of signatories also tended to be the most committed. Finally, we see regional variation with respect to participation in INGOs. The large number of INGOs in Europe and the sparse involvement of Oceania and Asia most likely reflect regional variation in the level of involvement in global civil society generally and national resource capacities.

Tables 3.3 and 3.4 examine the intersection between the foregoing elements using ordered logistic and binominal logistic models. We begin by exploring the factors associated with a state’s level of ethnic cognizance. In light of the strong regional differences apparent in the earlier tables, we expected geographic location to exert an effect on a state’s propensity to recognize ethnic distinctions. This was the case for North and South America and, to a lesser degree, Oceania. Compared to European states, those in North America were about 9 times more likely to enumerate by ethnicity, while states in South America and Oceania were three to four times more likely to do so. The coefficient for questionnaire length was positive and significant, with instruments of above median length more likely to elicit ethnic information. This cannot be attributed to national differences in resource capacities, as the effect remained even when GDP was included. Rather, it suggests governments that are committed to fully documenting their population’s characteristics are more disposed to recognize ethnicity as a necessary component of the national stock-take.

Table 3.4 Maximum likelihood logistic regression models for ethnic legitimization

Results from model 2 in Table 3.3 are consistent with the prevailing view that endogenous factors are important influences of whether states support ethnic enumeration. States with higher levels of net immigration were slightly more likely to recognize ethnic distinctions, whereas the presence of immigrant workers, measured by our proxy variable, male immigrants, had a countervailing effect. Consistent with our expectation, the presence of an ethnic contender enhanced the likelihood of ethnic recognition; although the variable’s magnitude and significance was slightly reduced once exogenous variables were introduced (model 3). Nevertheless, states with at least one indigenous, ethnonationalist, or ethnoclass minority were twice as likely to recognize ethnic distinctions as those without. Taken together, the group variables suggest the benefits of specifying the ethnic terrain within which decisions about ethnic recognition are undertaken. Immigrants, foreign workers and ethnic contenders all appear to influence, in different ways, the willingness of states to enumerate ethnicity, irrespective of their structural properties.

With respect to the latter, we were surprised to find that state resource capacities, measured by GDP, were negatively associated with ethnic cognizance. We can only speculate that countries with high GDP are less likely to have internal pressures on the state due to the relative wealth of its citizens. That is, favourable material conditions may have a unifying effect on the country, reducing the likelihood of ethnic disenfranchisement, or making it difficult for ethnic advocates to gain traction. In turn, such conditions may be less likely to draw the attention of international organizations that might otherwise push for minority differences to be addressed. States that gained independence in the post-1965 period of ethnic and post-colonial transformations were also more likely than the established states to recognize ethnic distinctions.

Turning to the exogenous variables, we find significant support for our hypotheses that international ties influence state processes of ethnic recognition. Consistent with findings from world society research (e.g., Neumayer 2005), we find simply being a signatory of ICERD has little effect on whether or not a state enumerates by ethnicity. Of greater importance is the level of commitment that states exhibit in their participation in these organizations. For each level of state commitment to ICERD, the likelihood of enumerating by ethnicity doubled. These findings challenge the implicit assumption that ethnic enumeration is exclusively or even primarily the product of endogenous state characteristics. It also suggests state expressions of support for eliminating ethnic and racial inequities and ensuring minority rights are not just displays of empty benevolence, but can translate into the active recognition of those differences. Our argument that global civil society matters is buttressed by the positive effect of participation in INGOs, with an increased likelihood of ethnic recognition among those states with more INGO linkages.

The models in Table 3.4 provide insights into the factors that underlie state recognition of ethnic differences generally, but do they extend to the recognition of specific group identities? To answer this question we re-ran the same models, changing the outcome variable to reflect the recognition of specific group identities (see Table 3.4). Interestingly, immigration flows, ethnic contender groups and commitment to ICERD – all variables significantly associated with ethnic cognizance – were inconsequential to state recognition of specific groups. In keeping with the earlier results, the presence of foreign workers decreased the likelihood of group recognition, as did a higher level of state resources. Regional effects were also stronger than in earlier models. North American states were about 17 times more likely than European states to name a group on their enumeration forms, perhaps reflecting the prevalence of identity politics in the former part of the world.¹⁷ States that gained independence after 1965 were also much more disposed to recognize specific group identities. Included among these states were the Czech Republic and Slovakia whose ‘Velvet Revolution’ began a series of events that led to the mostly peaceful dissolution of the former Czechoslovakia; Fiji and East Timor, which broke from the colonial legacies of oppression and control; and states such as Namibia and Lesotho whose independence was secured only after civil war and/or internal disputes. To some extent the effect of recent independence is also likely to capture international influences, given that many of these states were born into a fundamentally altered world system.

Overall, our models suggest a commonality between the factors that influence processes of ethnic recognition and identity legitimization. That is, these processes may be theoretically distinguished, but are difficult to separate out empirically. We note, however, that the models in Table 3.4 provide a somewhat weak test of identity legitimization processes. A more nuanced distinction between the specific groups listed on census questionnaires and registration forms might have yielded somewhat different results but our goal was to tap general processes, rather than to home in on particular sorts of ethnic minorities.

7 Discussion

This study was motivated by two broad concerns. The first was the dearth of a comprehensive, empirically driven research agenda on the factors underlying state recognition of ethnic differences and group identities in enumeration practices. Notwithstanding the important contributions made by the numerous case studies of ethnic classification regimes, we were concerned by the implicit message that state practices were best understood as parochial products of unique historical, political and social factors. To some extent, our findings support the prevailing view that structural conditions, internal group relations and state histories are important influences on processes of ethnic recognition. However, we go further to show that these influences can be generalized beyond specific national contexts. Consistent with Morning’s finding of an underlying cohesion to apparently divergent indicators of ethnicity; our cross-national comparative approach has revealed systematic patterns in the factors that influence state processes of recognition. Often these factors have similar effects across states that, at first blush, would appear to have very little in common.

Our second concern was the almost exclusive focus on endogenous state factors and the contradiction this posed with broader sociological efforts to nest state-level processes within a global context. By including exogenous influences alongside endogenous ones, we were able to demonstrate that ties to global civil society do have some bearing on national ethnic enumeration practices. First, we showed that the expression of state commitment to human rights instruments on the world stage was strongly associated with a greater willingness to recognize ethnic differences within national borders, even if this did not extend to the recognition of specific group identities. Second, state participation in INGOs enhanced the likelihood of both ethnic recognition and state legitimization.

Our findings may be best understood by conceptualizing the effects in terms of internal and external pressures. Internally, there are two types of pressures on the state: those that emanate from groups and those associated with structural conditions. With respect to groups, we singled out those that were most dominant in the literature – that is, established ethnic minorities with particular rights claims that challenge the state. In our models, higher levels of net immigration and the presence of an ethnic contender group were positively associated with the likelihood of ethnic enumeration. These findings are consistent with our expectation that these groups are sufficiently entrenched in state politics and able to mobilize at least some support for recognition, even if it does not always translate into a formal legitimization of the group on questionnaires and population registries. In addition, the presence of immigrants heightens awareness of difference within the state and the likelihood of recognition of ethnic groups as members of its citizenry. The increased visibility of immigrant workers had a countervailing effect, decreasing the odds of ethnic cognizance. This latter finding suggests states may be more likely to minimize differences in the context of the perceived economic and social threat posed by a large population of immigrant workers. We note too that migration variables are not strictly endogenous as they clearly tap into international migration flows. However, to the extent that inward migration shapes the ethnic terrain of the receiving society, we are able to conceptualize them as primarily endogenous.

Pressures from the ‘inside’ also arise from structural conditions. Our finding that older independent states are less likely to enumerate by ethnicity points to historical path dependencies and a distancing from the post-colonial push for minority and indigenous rights. Finally, states with a high GDP may have less at stake in terms of violating global human rights norms and be less likely to receive opposition from ethnic groups within their states, given their economic position and relative lack of deprivation.

To our knowledge this research represents the first empirical attempt to model the processes underlying state enumeration practices in a systematic fashion. By providing theoretical motivations for our variables and identifying key mechanisms, we have made a start to more general theorizing about ethnic enumeration. In the process, we have contributed to knowledge about the ways racial and ethnic categories work around the globe. As our title suggests, ethnic enumeration appears to be best understood as the product of factors that are internal and external to states, rather than one or the other. Our findings are of growing importance as states are confronted with increased rates of immigration, and the recognitive demands of national minority groups show no signs of abating. Despite the limitations associated with cross-sectional research, we are careful to distinguish between causality and relationship. For the time being, it is difficult to make a causal argument, but given the availability of previous censuses, there is a great deal to learn from exploring repeated cross-sections of census rounds.

Theorizing these processes is at an early stage in its development, but we have been greatly assisted by developments in data collection. Until recently, the collection of census forms was a long and tedious project. Each country had to be contacted separately, which took a great deal of time as attested to by our efforts to collect questionnaires in this manner. The concerted efforts of a few individuals and organizations and the widespread availability of electronic versions of questionnaires from online sources means the collection of enumeration forms is now much more efficient. These collective efforts have made this and future research projects possible. Consequentially, the future of theorizing and empirically mapping global strategies of ethnic enumeration appears to be blossoming.