Biopolitical bordering: Enacting populations as intelligible objects of government

Abstract

Since Foucault introduced the notion of biopolitics, it has been fiercely debated—usually in highly generalized terms—how to interpret and use this concept. This article argues that these discussions need to be situated, as biopolitics have features that do not travel from one site to the next. This becomes apparent if we attend to an aspect of biopolitics that has only received scant attention so far: the knowledge practices required to constitute populations as intelligible objects of government. To illustrate this point, the article focuses on processes of biopolitical bordering: the delineation of the target population that is to be known via statistical practices. Drawing on the example of Estonia I show that methodological decisions involved in this work have important biopolitical implications as they affect the size and composition of the population, thus shaping the design of programmes of government aiming at its regulation.

If spectres, ghosts and other spooks have the habit of travelling light, scoffing the walls and deriding borders – realities are known for just the opposite inclinations. We call them ‘tough’ or ‘hard’. Realities do not move easily; they resist push and pull, claim solid foundations and surround themselves with thick walls and closely guarded borders.

– Bauman (2003)

Populations are usually considered as ‘hard’ realities. They exist and we have various methods to establish their size and composition in terms of wealth, health, sex–age distribution and so forth. These widely shared convictions are supported by an ‘avalanche of printed numbers’ (Hacking, 2015) produced by national statistical institutes (NSIs). However, a closer look at population statistics casts doubts on these widely shared convictions. The Census Hub hosted by Eurostat, the statistical agency of the European Union (EU), gathers figures on the populations of all member states in one place. It thus enacts a European population as the sum of national populations. Following the census hub, Estonia’s usually resident population is 1,294,455 people, according to the 2011 population and housing census (PHC). ¹ This figure was submitted by Statistics Estonia (SE) to Eurostat in autumn 2012 ² and published on the Census hub as Estonia’s usually resident population on 31 January 2011. ³ Yet SE’s web page provides a different number, namely 1,325,217 people. ⁴ This number is supposed to reflect Estonia’s population on 1 January 2012.

Does this mean that the size of Estonia’s population has increased by nearly 30,000 people within 24 h? No, this would amount to a demographic impossibility. The reason for the divergence is that the figure published on SE’s home page is an ‘adjusted figure’ that is meant to account for a phenomenon known as undercoverage, that is, an undercount of people that are, by definition, part of the usual resident population of Estonia, but have not been enumerated in the PHC 2011. This example shows that different methods enact a population in different ways. Importantly, these differences have biopolitical implications as they affect who is (not) considered as an element of the population, and thus also calculations of government regulating this population. Rather than as solid, stable realities, populations emerge as what Bauman calls ghosts: as spectres that tend to be more than one and less than many as they haunt, deride and escape the borders that statisticians try to establish in order to demarcate (and enact) them as coherent objects of government.

The crucial point is that populations do not exist independently of the statistical practices that are used to know and quantify them. This is in line with material-semiotic approaches from Science and Technology Studies (STS) which provide the conceptual starting point of this article. Material-semiotic approaches conceive of methods not as mere tools for the extraction of knowledge from a reality ‘out there’ but as performative devices that help to enact the realities they study and describe (Law, 2004; Law & Lien, 2012; Mol, 2002). In the context of population statistics, different methods do not just provide different, more or less accurate counts of a population ‘out there’. They produce different accounts, that is, multiple versions of a population. This multiplicity has to be negotiated by statisticians in order to enact a population as a singular, intelligible object of government. We are confronted with a politics of method that are ‘ontological politics’ (Law, 2008) in the sense that methodological choices of how to quantify and categorize a population will shape the size and composition of this population. These ontological effects have important biopolitical implications, not only for individuals and groups of people who may be included or excluded from the population depending on the chosen method but also for programmes of government that are, based on this statistical knowledge, devised to optimize this population in terms of its health, fertility, economic productivity and so on.

Hence, the conception of statistical methods as performative yields an important lesson for Michel Foucault’s notion of biopolitics. ⁵ In brief, it suggests that debates about biopolitics, which are usually led on a highly abstract level in generalized terms, need to be grounded in a situated analysis that accounts for how biopolitics operate in particular sites. This becomes apparent if we attend to the question how populations are constituted as intelligible objects of biopolitical regulation in the first place. This question has, however, ‘escaped sustained critical attention in the Foucaultian literature’ (Curtis, 2002, p. 506). Drawing on material-semiotic approaches, this article addresses this question to show that debates on biopolitics need to be situated because biopolitics feature aspects that do not travel from one site to the next.

To this end, the article attends to the first step of enacting populations through statistical practices. I call this step biopolitical bordering: the delineation of the target population that is to be known. Importantly, biopolitical bordering features a micro-politics of method which have biopolitical implications as allegedly minor methodological decisions affect the size and composition of the population to be known. Drawing on a multi-sited ethnography that was conducted in context of a collaborative research project on population statistics in Europe, ⁶ these arguments will be developed through an exemplary analysis of statisticians’ attempts to demarcate the population of Estonia through various methods after the last census.

The first, more conceptual section introduces the notions of enactment and biopolitical bordering. The second section returns to the example of Estonia’s divergent population sizes after the PHC 2011. The third section discusses the biopolitical implications of methodological decisions involved in biopolitical bordering. It shows how particular groups of people are rendered absent or present by specific methods. Based on this analysis, I conclude that discussions on biopolitics need to be grounded in empirical inquiries that begin with an investigation of how a population is enacted in a particular site in order to elucidate, in a second, empirically informed way, what kind of biopolitics operate in these situations.

Moving beyond the lacuna of biopolitics

For Foucault (1979, 2003, 2007, 2009, 2010), biopolitics designates the emergence of a new form of power that is concerned with governing populations in a way that maximizes their life potentials and economic productivity. Since Foucault introduced the notion of biopolitics as a form of power that complements and partly supersedes the sovereign power to ‘let live and make die’ the concept has certainly attained the status of a ‘buzzword’ (Lemke, 2011, p. 1). ⁷ However, as with all concepts that gain currency, there is no agreement on the meaning of biopolitics, as prominent positions demonstrate.

These include certainly Giorgio Agamben’s (1998) reinterpretation of biopolitics as thanatopolitics through its coupling with Carl Schmitt’s conception of sovereign power, culminating in the claim that the camp resembles ‘the hidden matrix of the political space in which we are living’ (p. 166). Likewise, Archille Mbembe (2003) highlights the destructive dimension of biopolitics with his concept of necropolitics. Drawing on Foucault’s notion of state racism, he claims that racism establishes a ‘biopolitical caesura’ between those who must live and those who must die. Hardt and Negri’s (2000) more hopeful reading of biopolitics as the last stage of an all-encompassing regime of capital accumulation that creates the conditions for the emergence of the revolutionary force of the multitude is, in contrast, regarded as a counter point to these gloomy diagnoses. Roberto Esposito’s (2008) position is situated between these poles. Esposito argues that the relationship between life and politics (i.e. biopolitics) is pervaded by a logic of immunization that seeks to preserve life but, through these attempts, reduces life to nothing but biological existence. In a similar vein, Jasbir Puar (2017) highlights the entanglements of the life-enhancing and destructive dimensions of biopolitics. She argues that the biopolitical embracing of disability through discourses of equal rights and empowerment obscures and simultaneously facilitates the proliferation of multiple processes of debilitation that actively decapacitate individual bodies and entire populations. Hence, Puar’s interpretation of biopolitics as a politics of debilitation moves biopolitical thinking beyond the binary of life and death by pointing to technologies of government that aim at ‘slow[ly] wearing down of populations’ (2017, p. xiv).

There are, of course, numerous other scholars who have reinterpreted and developed Foucault’s notion of biopolitics in various ways. Hence, we are confronted with different bodies of literature that each attend to specific issues. This is what characterizes the ‘messiness’ of biopolitical thought (Bird & Lynch, 2019). It is beyond the scope of this article to provide a comprehensive overview of these literature (for such overviews, see Bird & Lynch, 2019; Lemke, 2011; Prozorov & Rentea, 2017). What the necessarily limited selection of works discussed above illustrates is, however, that debates on biopolitics operate on a highly abstract level. ⁸ The question how populations are enacted as intelligible objects of government has, in contrast, rarely been considered so far (Curtis, 2002). Rather, populations are mostly treated as readily available targets of biopolitical regulation.

While Bruce Curtis (2002) argues that this lacuna in debates on biopolitics is already inscribed in Foucault’s own writings, I maintain that there are also passages in his oeuvre in which Foucault clearly frames populations as products of knowledge practices. In his earlier elaborations on biopolitics, Foucault points out that the ‘remarkable entrance’ (Foucault, 2009, p. 67) of the population into considerations of government in the eighteenth century was tied to the emergence of new forms of knowledge, most notably biology, medicine, demography and statistics (Foucault, 2003, pp. 243–245). In the ‘Bahia lecture’ from 1976, Foucault introduces the notion of biopolitics for instance as follows:

[…] the 18th century discovered this principal thing: that power is not simply exercised over subjects; this was the fundamental thesis of monarchy […]. We discover that that on which power is exercised is the population. And what does population mean? It does not simply mean to say a numerous groups of humans, but living beings, traversed, commanded, ruled by processes and biological laws. A population has a birth rate, a rate of mortality, a population has an age curve, a generation pyramid, a life-expectancy, a state of health, a population can perish or, on the contrary, grow. (Foucault, 2007, p. 161)

In this account, the population is construed as an abstraction, or more precisely, a ‘statistical artefact’, in which individual variation disappears in means, distribution and large numbers (Curtis, 2002, p. 507). This makes it possible to identify patterns such as an ‘age curve’ and objects of intervention such as ‘life expectancy’ (Desrosières, 1991, 1998; Hacking, 1990). Hence, population counts become biopolitical when they inform programmes and practices of government that seek to affect the size, composition, life opportunities and so on of a population. ⁹ A concrete example is offered by Jemima Repo (2016) who shows how concerns for low fertility rates – a key indicator of population statistics – have prompted EU policymakers to turn gender equality policies into a technology of government that seeks to increase both women’s fertility and their labour market participation.

Yet while Foucault is seemingly aware that the constitution of populations as objects of governments hinges on statistical practices, Curtis (2001, p. 42) is also right when he observes that there are instances in which ‘Foucault tends nonetheless to write naturalistically about population’, treating it ‘as a thing composed of elements and empirical processes’. This tendency is particularly present in Foucault’s argument on the population as the ‘great discovery’ of the eighteenth century (e.g. Foucault, 2009, pp. 140–141). Yet, as Evelyn Ruppert notes, ‘population is not a thing waiting to be discovered, but a particular way states organise social relations’ (2010, p. 159). It is this tendency in Foucault’s work to treat the population as a given reality that explains why the knowledge practices that enact populations as intelligible objects of government have – with the exception of Ian Hacking’s work (2015) – rarely been scrutinized in the literature on biopolitics. To consider these knowledge practices is, however, important because they have implications for how biopolitics operate in particular sites as they affect how the target of biopolitical interventions – namely a population – is enacted in terms of its size and composition.

To address this lacuna, this article draws on material-semiotic approaches from STS to show that populations do not exist as intelligible objects of government independently of the practices mobilized to know them. Material-semiotic approaches suggest that knowledge practices like statistics do not just mirror or quantify a population that already exists ‘out there’. Rather, statistics are performative insofar as they enact the population they are meant to study and describe (Law, 2008; Mol, 2002). Consequently, the practices that are mobilized to know a population become a non-negligible aspect of any analysis or discussion of biopolitics as they bring into being the very reference object of biopolitical regulation, thus informing and shaping interventions of government.

In the following, I speak of enacting to highlight the processual character of these efforts. The notion of enactment avoids connotations carried by terms like ‘constitution’ – which suggests a one-time creational act – or ‘construction’ which construes enactments of populations as accomplishments of wilful human subjects. The notion of enactment underscores, in contrast, the precarious, unstable status of populations which are thought of as ‘always in formation’ (Ruppert, 2011, p. 223). It also highlights the materiality of knowledge practices which are not reducible to actions of human beings but are related to inscription devices and material infrastructures that rely on a range of other sociotechnical networks.

This material-semiotic approach has three important consequences for analysing the enactment of populations. First, if there is no ‘reality out there beyond practice that is independent, definite, singular, coherent and prior to that practice’ (Law, 2012, p. 171), scholars need to investigate the practices through which populations are done. Second, the conception of practices as material and relational requires focusing on the wider method assemblages of which they are part. These method assemblages comprise various artefacts and devices, standards and categories, methods and infrastructures. What distinguishes method assemblages from other assemblages is ‘the emphasis on presence’ (Law, 2004, p. 84). Method assemblages enact realities through a set of ‘relations that make some things (representations, objects, apprehensions) present “in-here”, whilst making others absent “out-there”’ (Law, 2004, p. 14). Hence, scholars have to engage, third, in a situated analysis that accounts for the specificities of the practices involved. The reason is that different practices and sets of relations will enact different populations. This implies refraining from abstracting any general or even universal claims from a particular case. The question whether a feature of an object (like a population) or a related concept (like biopolitics) travel from one site to the next is thus not a question of adequate theorization but a matter for empirical investigation.

Drawing on this framework, this article attends to the first step of the enactment of populations via statistics: the demarcation of the target population that is to be known through statistical methods. The need to delineate the entity that is to be known is not a peculiar feature of populations but concerns all statistical topics. In his book The Rule of Experts, Timothy Mitchell shows, for instance, that the enactment of the economy as ‘a self-contained, internally dynamic, and statistically measurable sphere of social action, scientific analysis and political regulation’ (2002, p. 4) requires processes of boundary-drawing:

[…] the economy depends upon, and helps establish, boundaries between the monetary and the nonmonetary, national and foreign, consumption and investment, public and private, nature and technology, tangible and intangible, owner and nonowner, and many more. (Mitchell, 2002, p. 9)

Likewise, the enactment of populations depends on boundaries between residents and non-residents, citizens and non-citizens, knowns and unknowns, emigrants and immigrants, living and dead and so on. Since these instances of boundary-drawing concern, in the present case, populations – the reference object of biopolitical regulation – I refer to these practices as biopolitical bordering.

To clarify the concept of biopolitical bordering, I elaborate on four of its key aspects. First, I speak of biopolitical bordering to stress that the delineation of populations through statistical practices is just as essential for the (re)production of the nation state as the demarcation of a state territory through geopolitical borders. Biopolitical bordering is indispensable for the creation of a legible population which James Scott regards as a ‘central problem of statecraft’ (1998, p. 2). One response to the state’s ‘blindness’ (Scott, 1998, p. 2) was the ‘rise of statistical thinking’ which suggests that regularities, causal relationships and correlations that permit to govern populations are to be found in large numbers (Desrosières, 1998; Hacking, 1990; Porter, 1986). The intimate relationship between the institutionalization of statistics and the consolidation of the nation state is reflected in the etymology of the word ‘statistics’, which means essentially ‘state-istics’, the science of the state (Schmidt, 2005, p. 15). ¹⁰

Secondly, I speak of biopolitical bordering to emphasize that I am referring to a set of practices and not a physical site. While this understanding of biopolitical bordering follows from the conception of enactment as a sternly practice-oriented approach, this clarification is needed to avoid confusion with works on ‘biopolitical borders’. The concept was coined by William Walters (2002, p. 573) to highlight that ‘the border can be regarded as a privileged institutional site where political authorities acquire biopolitical knowledge about populations – their movements, health and wealth’. Since then the concept has been picked up by numerous scholars (e.g. Pallister-Wilkins, 2015; van Baar, 2016; Vaughan-Williams, 2015). While this body of literature has done a great job in facilitating ‘a paradigm shift in border studies from a geopolitical to a biopolitical horizon of analysis’ (Vaughan-Williams, 2015, pp. 5–6), it maintains, nevertheless, a conception of borders as physical sites of control where knowledge on populations is generated as a by-product in the attempts to regulate the circulation, economic transactions, health and wealth of these populations.

If we start from the STS-inspired enactment agenda and take seriously that the population constitutes, from the perspective of the state, a void, a problem of not knowing the people that are supposed to be governed, then the biopolitical border is no longer thought of in spatial terms as a physical site. It becomes a conceptual boundary, an ‘epistemological device’ (Mezzadra & Neilson, 2013, p. 14) that helps to enact that which it tries to delimit which is, in this instance, a population. Thereby we move from biopolitical borders, understood as physical sites for regulating populations, towards biopolitical bordering – a set of knowledge practices and their performative effects. This move underscores the ‘constituent moment’ (Mezzadra & Neilson, 2013, p. 6) of bordering practices, their ‘world-configuring function’ (Mezzadra & Neilson, 2013, p. 6).

This points, third, to the micro-politics of method involved in biopolitical bordering. These entail manifold controversies and decisions that emerge because there exist potentially infinite methodological possibilities to delineate the population to be known. Relevant issues range from ‘big’ methodological decisions about adequate data sources or the definition of the so-called population base ¹¹ to allegedly minor decisions on data cleaning or imputation of incomplete data. Inspired by Mike Savage’s work, I call related decisions and disputes micro-politics of method. In contrast to the ‘big’ politics of methods analysed by Savage (2010) in post-war Britain, these micropolitics of method do not concern the formation of national identities but the ontological effects that statisticians’ methodological decisions have on their reference object. Seemingly minor methodological decisions affect the size and composition of the population to be known. Consequently, these decisions also shape interventions of government mobilized for optimizing the population in terms of its wealth, health, economic productivity and so on. What makes the bordering of populations biopolitical are then the ontopolitical implications of these micropolitics of method. For different methods enact different populations. This is because method assemblages make things absent or present. What is made absent or present in the case of biopolitical bordering are people who are either considered as elements of the population, and thus a concern of biopolitical regulation, or not. Consequently, the biopolitical implications of the micropolitics of method involved in processes of biopolitical bordering concern individual subjects, groups of people and practices of government, as the example of Estonia illustrates.

Haunted by the shadow of a doubt: Biopolitical bordering through signs of life

Generally, a census is regarded as the most reliable method to know a population (Ruppert, 2010). In contrast to a sample survey, a census targets the entire population of a country (Tiit et al., 2012). Yet, due to increased levels of mobility of ever more people census results are believed to increasingly suffer from ‘undercoverage’, that is, the non-enumeration of people who should be counted as part of the usually resident population.

Due to an alleged undercoverage, SE could not simply publish the results of the PHC 2011 – 1,294,455 people – as the population size of Estonia. The reason was media reports in which people claimed they had not been enumerated (Tiit & Vaehi, 2012; Tiit et al., 2012). These reports constituted an embarrassment for SE which had run a campaign under the slogan ‘everyone counts!’ [Igaüks loeb!] for increasing participation in the census (SE, 2011). The slogan alludes to the widespread fear that Estonian language and culture might become extinct due to a declining population (Servinski, 2012; Uibu, 2014). Hence, statisticians created a working group to assess the undercoverage and to develop a method which would enable them to ‘correct’ the PHC 2011 results (Tiit et al., 2012, p. 100). These efforts culminated in the publication of the adjusted figure of 1,325,217 people.

What this example illustrates is the micropolitics of method involved in biopolitical bordering as well as their biopolitical implications. In brief, the working group was tasked with determining the degree of undercoverage in the PHC 2011. Since undercoverage expresses the ratio between the number of people who have not been enumerated in the census and the number of people who should have been enumerated according to the UN definition of the usually resident population, statisticians had to find a method which would allow them to establish the target population of the PHC 2011 (Tiit, 2012). Put differently, SE’s statisticians had to develop alternative methods of biopolitical bordering.

While engaging with this task, SE’s statisticians were confronted with three different population sizes: First, the population size that had been calculated by SE with the so-called cohort component method. The latter is used to update the population size on an annual basis by adapting the results of the last census (the PHC 2000) by recorded births, deaths and net migration figures. According to this method, Estonia’s population comprised 132,000 residents in 2011. Second, there was the census result of 1,294,455 people enumerated through online and door-to-door enumeration. Third, there was the figure of 1,365,000 people officially recorded as residents of Estonia in the population register (RR; Tiit, 2014).

What this example illustrates is that, in their quest for accuracy, statisticians generate multiplicity, that is different versions of the Estonian population. The reason for this production of multiplicity is that, contrary to what the influential epistemological register of ‘metrological realism’ (Espeland & Stevens, 2008: 417) suggests, there exists no independent population reality ‘out there’ that could serve statisticians as a yardstick to evaluate the accuracy of their results. Metrological realism frames statistics as measurements of an external reality that can be evaluated in terms of ‘accuracy’ (Desrosières, 2001). It thus confirms the central assumption of a singular, coherent reality ‘out there [that] substantially precedes our actions and attempts to know it’ (Law, 2012, p. 156). The above example illustrates however that populations are abstractions that do not exist independently of the statistical practices that are used to know them. Populations are quantified into being through statistical practices and related method assemblages. This is why statisticians are only able to assess the ‘accuracy’ of any method by comparing its results with those of another method. This implies that over- and undercoverage do not express a relation to an external reality. They express relations between different versions of the Estonian population enacted with different methods of biopolitical bordering.

This can be illustrated through the two versions of the Estonian population with the largest discrepancy. Statisticians could only assess the scale of the alleged undercoverage of the census in relation to another method of biopolitical bordering like the number of recorded residents in the RR. The discrepancy between this number and the number of people enumerated in the PHC 2011 is about 71,000. But this number expresses only the potential undercoverage of the census since the RR entails, according to assessments by statisticians, an unknown overcoverage of emigrants who have left Estonia without notifying authorities (Tiit, 2014). Hence, statisticians were confronted with an unknown undercoverage of up to 71,000 people who were absent from the Estonian population of the PHC 2011 but present in the population enacted by the RR.

Since 71,000 people are ‘enough to populate an average county [of Estonia]’ (Tiit & Maasing, 2016a), SE’s statisticians had to address this conundrum. The fact that they developed a fourth methodology of biopolitical bordering that was subsequently declared the best method shows that over- and undercoverage resemble epistemic devices which permit statisticians to hierarchize between different versions of the same object in order to negotiate the multiplicity that is generated by their quest for accuracy. This hierarchization is needed to comply with the convention of common-sense realism that there can only be one single, coherent reality. For ‘if two objects that go under the same name clash, in practice one of them will be privileged over the other’ (Mol, 2002, p. 47). In case of the multiple versions of the Estonian population, this privileging is achieved by framing the new methodology as a solution for overcoverage of unregistered emigrants in the RR and undercoverage of non-registered residents in the PHC 2011. Hence, an article in SE’s house journal celebrates the new methodology by concluding, in the logic of metrological realism: ‘From the statistics viewpoint the best result is not “the largest possible”, but “the most accurate” one’ (Tiit & Vaehi, 2012, p. 119).

Following statisticians, the new methodology enables them to establish, based on data from 10 different government registers, the residency status of the 71,000 people who have not been enumerated in the PHC 2011 but nevertheless have a record in the RR. This method is based on a relatively simple idea: if a person does actually live in Estonia, it is assumed that this person will engage in more transactions with state institutions than a person that has left the country. These activities leave behind traces – the so called ‘signs of life’ (Tiit & Maasing, 2016b, p. 54) – in various government registers. Hence, statisticians try to delineate Estonia’s population with the help of transactional data that are produced as a by-product of everyday activities like working, seeing a doctor, studying or driving a registered car (Ruppert, 2011, p. 221). What these activities have in common is that they create records in government databases like the tax register, the health register or the education register. Thanks to the unique personal identification number used across all administrative registers, Estonian statisticians can link and analyse data from different registers.

After statisticians had established that the signs of life a person leaves behind across registers varies according to age and gender, they constructed nine sex–age groups to establish whether a person is resident in Estonia or not (cf. Tiit et al., 2012). For each sex–age group, statisticians identified the five most important signs of life in terms of occurrence. Subsequently, they developed an algorithm that analysed, beginning with the most relevant sign of life of a person’s sex–age group, whether this individual should be considered as an element of Estonia’s usual resident population. For small children, the most important sign of life was for instance child benefits, for the working age population the tax register and so forth (interview SE, June 2016). Based on this methodology, statisticians concluded that 30,760 of the 71,000 people were residents of Estonia that had not been enumerated in the census. This number – 30,760 people – is the difference between the population size published on Eurostat’s Census hub and the ‘adjusted figure’ of 1,325,217 people published on SE’ web page.

However, also this method was haunted by doubts regarding its ‘accuracy’. The reason was that people had been enumerated in the PHC 2011 but were considered as nonresidents by the new methodology because they did not show any sign of life in any of the 10 registers (Tiit & Vaehi, 2012, p. 119). Following Ene-Margit, Tiit (2015, p. 5), the head of the methodology of the PHC 2011, these ‘people were mainly working-aged men who did not study, did not visit doctors, did not get any social support and probably made some non-registered […] work’. Put differently, the new methodology enacts people as absent that were present in the population enumerated in the PHC 2011. This example demonstrates, once more, that different methods enact different versions of the Estonian population which differ in size and composition. Depending on the methodology used, being a member of the Estonian population means either to not have been recorded as dead or emigrated since the PHC 2000 (the composition cohort method), a completed questionnaire in the PHC2011, a record in the RR or a sign of life in one of the five most relevant registers for a particular sex–age group. These are all different methods of biopolitical bordering that enact different versions of the Estonian population because they make some individuals and groups of people present as elements of the population while enacting others as absent.

The crucial point resides in the biopolitical implications of these different enactments of Estonia’s population. For statisticians’ methodological choices affect who is (not) considered as an element of the population. As the example of the middle-aged men with no signs of life shows, some people are rendered absent by a particular method and thus not considered in the allocation of resources and programs of government aiming at the optimization of the population’s life opportunities and wealth. This raises important political questions since these subjects hold, in most cases, formal citizenship.

Similar political concerns arise – on a larger scale – for groups of people. The different enactments of Estonia’s population demonstrate that ‘differential undercount’ (Anderson & Fienberg, 2000) does not only concern ethnic minorities but can also other population segments in terms of sex, age, economic background and so on. ¹² In the present case, the RR enacts people as present who are, according to the PHC 2011, unregistered emigrants and thus absent. Yet the PHC 2011 also enacts a group of people as absent, namely, the ‘younger and more mobile people who tend to be left out’ (Tiit et al., 2012, p. 99). These people are however present according to the new, register-based methodology. The latter enacts, in turn, another group of people as absent which is present in the population enacted by the PHC 2011: middle-aged men with no signs of life. But how the population is enacted in terms of size and composition shapes interventions of government aiming at its biopolitical regulation. ‘Differential undercount’ is for instance known to affect ‘legislative apportionments and policy based on the numbers’ (Anderson & Fienberg, 2000, p. 88). The population size affects in turn population projections that, while anticipating possible future developments, shape biopolitical interventions in the present like migration policies or social policies seeking to influence the fertility of the ‘native’ population (Schultz, 2018).

In Estonia, where the population has decreased by 44,835 people according to the adjusted population figure since the PHC 2000 (Tiit, 2014, p. 95), the prospect of a shrinking population implicated for instance the introduction of a program that ‘aims, by 2015, to achieve a birth rate that is higher than the death rate’ (Uibu, 2014). It entails measures that allow ‘to better reconcile work and family life’ like the creation of more childcare placements as well as financial incentives like an increase of child benefits (Government of Estonia, 2018). While it is difficult to establish a direct link between Estonia’s population decrease and specific interventions of government, it is reasonable to assume that the perceived urgency for and scope of measures aiming at increasing the birth rate would have been even higher if the population figure had not been adjusted after the PHC 2011 as this would have nearly doubled the intercensal population decrease to 75,595 people.

The residency index: Enacting present absences and absent absences

The methodological challenge of how to demarcate Estonia’s population did not end with publication of the adjusted population figure on SE’s home page. Rather, statisticians were confronted with an even bigger methodological challenge: they had to develop a method of biopolitical bordering that would enable them to determine the size of Estonia’s population without a traditional census. Accomplishing this task was viewed as a major milestone for realizing the ambitious plans of Estonia’s government to conduct a fully register-based census in 2020 (Tiit, 2014). Hence, SE’s statisticians decided to advance the methodology they had developed for adjusting the population figure after the PHC 2011.

One crucial difference was that this methodology – the Residency Index (hereafter: RI model) had to assess the residency status of the 1.5 million people with a record in the RR – the potential usually resident population of Estonia – and not just 71,000 people (Tiit & Maasing, 2016b, p. 55). Moreover, statisticians had to solve two issues raised by the previously used methodology: First, they had to dispense with the sex–age groups which were only useful for determining residency status at one particular point in time. Usage of different models for each sex–age group over a longer time span implicates that some people will ‘get lost’ if they get older and move from one sex-age group to the next (interview SE, March 2016). This would result in highly volatile population figures, as ‘people are counted in [the population] and out all the time’ (interview SE, June 2016). Second, statisticians had to address the issue of the 2% of the population enumerated in the PHC 2011 that were enacted as absent by the register-based methodology because the people concerned did not show any signs of life (Tiit & Vaehi, 2012, p. 119).

To address the issue of the sex–age groups, statisticians analysed people’s register activity again. They concluded that, on average, all sex–age groups showed the same number of signs of life (i.e. 4–5) across the 14 different registers used for calculating the residency index. By increasing the number of registers from 10 to 14, statisticians also tried to address the issue of how to capture people with no or very few signs of life. Due to this increase, a person could, theoretically, show up to 21 signs of life in 1 year (Tiit & Maasing, 2016a). Furthermore, statisticians included the register activity of previous years in the analysis by introducing a stability parameter (d). ¹³

Hence, the residency index (R) of a person (j) in a given year (k) is calculated by multiplying the residency index of the previous year (R_j , k−1) with a stability factor (d) of 0.8. The result is added to the weighted sum of the person’s number of signs of life in the previous year (X_j , k−1) which is multiplied with the signs of life parameter (g) of 0.2. The formula for calculating a person’s residency index is thus

$R_{\left(j,k\right)}=d\times R_{\left(j,k-1\right)}+g\times X_{\left(j,k-1\right)}$

The value of a person’s residency index ranges between 0 and 1, depending on the number of the signs of life he or she accumulates across all government registers and his or her residency index of the previous year. The higher the value of the index, the higher is the probability that the person is a permanent resident of Estonia. To be considered a resident of Estonia, a person’s residency index has to be above the threshold (c) of 0.7 (Tiit & Maasing, 2016b).

The values of the signs of life parameter (g), the stability parameter (d) and the threshold (c) have been set in such a way that a person will lose her residency status (and be considered an emigrant) after showing no signs of life in two consecutive years. Conversely, an immigrant who has registered in Estonia will only be considered as a resident in the following year if that person has produced at least two signs of life. While the setting of these values was subject to long discussions and calculations, the values eventually chosen had the twofold objective to make the model stable and to approximate its results to the adjusted population size of the PHC 2011 (SE interview, June 2016).

Ultimately, statisticians were satisfied with the RI model although they conceded that it required constant testing and adaptation, for instance, to accommodate changes in the rules determining the production of records in one of the used registers (SE interview, June 2018). One reason why statisticians were nevertheless satisfied with their model was that it offers a viable methodology for improving SE’s migration statistics. The figures provided had been considered as unreliable and ‘too low’ for a number of years (cf. Scheel & Ustek-Spilda, 2019). In particular, the RI model allows statisticians to address the issue of unregistered emigration as it enables them ‘to detect’ emigrants who have not registered their departure from Estonia. Consequently, SE reported, after introducing the RI model for the production of migration statistics in 2015, an increase of nearly 400 per cent in immigration and 300 per cent in emigration numbers due to this change in methodology (SE, 2016). Put differently, the change in methodology made a group of people present as absent that had hitherto been enacted as present, that is, as an element of Estonia’s population.

However, while the RI model enacts unregistered emigrants as a present absence, the new methodology continues to enact some people as absent who are present according to the PHC 2011: middle-aged men with no signs of life in the registers. Enactment of this absence persists despite of statisticians’ efforts to reduce the number of people concerned by using additional registers and introducing a stability factor. In interviews, statisticians stress that they estimate this group of people to be rather small, while framing their absent absence from future population counts as negligible: ‘It is not a problem if there are a few thousand people missing. It becomes a problem for us if 10,000 or 20,000 people are missing. Our aim is to give numbers in ten-thousands…Every day I am looking for these people…for people who are not in the registers and for ways to find them’, emphasizes a statistician in an interview. He cites the example of a man he identified, based on register data, as a resident of Estonia because the man had seen a doctor (without having a health insurance) and renewed his driving license to conclude: ‘I think the group of people we are talking about here are less than 1000 people’ (interview SE, December 2015). Yet the number of the people concerned remains unknown. It could actually be much higher if one considers that this group of people accounted for 2 per cent of the population according to the PHC 2011 (Tiit & Vaehi, 2012, p. 119). Moreover, the people concerned are enacted as emigrants by the RI model and may thus be part of the increased emigration figures resulting from the change in methodology. What is rendered as an absent absence by the RI model is thus the very problematic of a group of residents of Estonia which does not leave behind any signs of life in government registers.

What this example illustrates is that enacting populations as intelligible, quantifiable entities involves not only knowledge production but also active production of different types of non-knowledge. A population is always known (and thus enacted) in a particular way, depending on the methodology chosen. Hence, each methodology will generate not only knowledge but also different forms of non-knowledge about the population in question by rendering either some of its features or some of its elements as absent, uncertain, not knowable, opaque, subject to speculation and so forth (for a non-exhaustive overview over different types of non-knowledge, see Aradau, 2017). This production of non-knowledge may be ‘strategic’ in the sense that it is the outcome of a conscious strategy of particular actors (McGoey, 2012), or the effect of information loss occurring during the transfer of knowledge from one field of practice or epistemic community to the next (Scheel & Ustek-Spilda, 2019). Yet studying and uncovering the production of non-knowledge raises serious methodological issues, not the least because actors often have an interest to render their production of non-knowledge as not-knowable, as Lindsey McGoey (2012) and Brian Rappert (2012) rightly note. It is nevertheless crucial for the study of biopolitics to interrogate how the production of non-knowledge features in enactments of populations since instances like the manufacture of absent absences exemplify how particular methods of biopolitical bordering affect programs of government aiming at the regulation of a population.

In the present example, residents not engaging in any transactions with institutions of government are enacted as inexistent by the RI model. Rather than being constituted as a potential target for biopolitical interventions, the people concerned are enacted as emigrants by the RI model. Since the enumeration-based method that made this group of people present is about to be discontinued, the problematic of mostly middle-aged male residents with no signs of life will become an absent absence. Instead people concerned are enacted as an element of the present absence that is enacted by the RI model: emigrants who did not inform authorities about their departure. In this instance, the biopolitical caesura between those who are made to live and those who are left to die (Foucault, 2003) is already inscribed in the methodological decision to replace a method that enacts people with no transactions with institutions of government as an absent presence (the traditional census) with a method that enacts another group of people – unregistered emigrants – as a present absence (the RI model).

The biopolitical implications of this methodological decision become apparent if one considers that unregistered emigrants are targeted as a biopolitical treasure by programmes of government aiming at the inversion of Estonia’s population decline. In 2015, the government launched a Compatriots’ Program that seeks to preserve Estonian language skills and sense of Estonian identity among Estonian emigrants and their children. The programme explicitly aims at ‘encouraging the return of expatriates to Estonia’, ¹⁴ providing advice and financial assistance to those following this call. This contrasts with the framing of the mostly male, middle-aged residents with no signs of life as a negligible statistical issue:

Perhaps, despite everything, the list of Estonian permanent residents does not include a man from Supilinn [name of a deprived neighborhood in Tartu, Estonia’s second largest city], who lives off collecting bottles and does not need social assistance, who has not yet attained pensionable age, has managed to avoid visiting a doctor within the last year, and has thus left no trace in any of the registers during the last year. (Tiit & Vaehi, 2012, p. 119)

In this account, people without any transactions with institutions of government emerge as a negligible variable whose ‘undercoverage’ by register-based methodologies is justifiable due to their ephemeral presence that does not yield much biopolitical potential and can therefore be rendered an absent absence, a soon-to-be forgotten methodological issue. Yet, from a theoretical viewpoint, it is precisely this allegedly negligible, soon-to-be forgotten group of people without signs of life which highlights that the ways in which biopolitics are done also depend on how, and with what kind of methods, a population is enacted as an intelligible object of government.

Conclusion

This article has introduced the notion of biopolitical bordering to attend to an important aspect of biopolitics that has not sufficiently been considered so far: knowledge practices required to enact populations as intelligible objects of government. In sum, the article yields two important insights for debates on biopolitics: First, analyses of biopolitics should begin with an inquiry into the methods that are mobilized to enact populations as intelligible objects of government. This becomes apparent if we consider that traditional censuses and the RI model offer just two of many possible methods of biopolitical bordering. While Statistics Sweden trusts its register data, using the ‘Total Population Register’ as a method to delineate Sweden’s population (UNECE, 2016c), Albania’s NSI introduced a migration module in its Labour Force Survey to estimate emigration, the most important factor for the country’s population decline (UNECE, 2016a). Columbia’s NSI has in turn developed another method that draws on data from the country’s border control register (UNECE, 2016b). Each of these methods enacts a different version of the population under consideration, with crucial implications for how biopolitics are done in these sites.

From this follows second that debates on biopolitics need to be situated. Rather than engaging in sweeping generalizations scholars should conduct empirically informed analyses of how biopolitics operate in particular sites. Instead of engaging in abstract theoretical debates about the ‘correct’ interpretation of Foucault’s notion of biopolitics, scholars could illustrate the wide spectrum of interventions of government covered by this term while highlighting that biopolitics carry features that do not travel from one site to the next. Such an approach would certainly better align biopolitics with the impetus of its twin concept of governmentality which was introduced by Foucault (2009, p. 104) to study the specifics of ‘the art of government’ in particular sites (Curtis, 2001, p. 40; Lemke, 2011; Walters, 2012). Despite the methodological difficulties plaguing such efforts, situated analyses of biopolitics should pay particular attention to individuals and groups of people that are rendered absent by particular methods of biopolitical bordering. For it are, ironically, the erased and forgotten that highlight the biopolitical implications of supposedly purely technical decisions as they haunt and derail the boundaries that statisticians try to establish in order to enact populations as hard realities and solid, readily available objects of government.