I. National Coming Out

In Britain, tales of the closet in the nation-state are receiving a great deal of coverage as we research and write this chapter. On November 8, 1998, Agriculture Minister Nick Brown publicly declares his homosexuality after being threatened with exposure by a former lover (Morris, 1998; Travis, 1998). This story comes on the heels of the resignation of Ron Davies, Minister of Welsh Affairs, who reported being mugged on Clappham Common late one evening. The mysterious resignation could only be explained, the media claimed, as pre-empting an outing since a likely reason for his presence on the common at night would be to cruise the ‘beat’ for gay sex. Still earlier in the month Cabinet member Peter Mandelson was outed on a national news program. With all these closet doors swinging open, The Sun provocatively asks in a front-page headline if the country was being run by "a gay mafia of politicians, lawyers, palace courtiers and TV bigwigs" (Pallister and Gibbs, 1998 p. 3). And the issue continues to receive press attention, despite the fact that a majority of Britains feel that being gay is morally acceptable.

These news stories raise the possibility and importance of the closet being framed in national space. In each case the closet is situated at the national scale, specifically in the state. For each man, the closet allowed a gay presence in the government. Yet it also immediately conjoined the space of the state with the space of the nation, as debate ensued over whether state officials should conceal their homosexuality. Here, of course, we could recognize national government as part of "the Trinity of the closet" (Signorile, 1993, see chapter 1). In what other ways does the closet operate through the nation-state as a material location? To explore this question, I began to discuss it with a visiting colleague, Paul Boyle, who has expertise with the British census. At first glance the combination may seem odd, a gay qualitative cultural geographer and a straight quantitative population geographer appear to have little common ground in the discipline these days (Johnston, 1997). Yet we have found the collaboration to be enormously instructive, productive, and at least one result is this chapter. Given our overlapping areas of expertise we decided to focus on the ways that national censuses might be a closet space. Specifically we examine both the British and U.S. censuses. We take this empirical tack partly because of our training and interests, but also because censuses are typically argued to provide the most reliable picture of the population of a nation-state; the choice of questions in some sense must be a reflection of what’s deemed to be important about a nation’s inhabitants.

The empirical analysis that follows draws on and speaks to Foucault’s (1991) notion of governmentality. Briefly stated, the concept signals a decidedly modern form of state power/knowledge. It highlights the interactive power between the knowing (in this case, state bureaucracy) and the known (a nation’s inhabitants). Governmentality calls attention to the simple point that the state’s administrative apparatuses play a key role in the way nationals come to know themselves as a coherent nation. State power is no longer simply the power to wage war or pass laws, it also lays in very ordinary, mundane bureaucratic practices. Specifically: a state’s own knowledge of its population powerfully frames the conditions and terms through which its citizens can see themselves as a nation. In this way, they come to ‘govern’ themselves through the state’s ‘mentality’. Such a network of both descending and ascending exercises of power is clearly Foucaultian. And we want to suggest here that national censuses are a clear material example of this practice. More specifically, we want to argue that the census produces the closet in national space through governmentality.

In what follows we chronicle our attempts at seeing the closet in the census. We note in Section II that the census has yet to be considered at the national scale within queer studies. Next, we review the literature on governmentality, arguing that the census is a prime example of such a power/knowledge practice. This literature, however, has failed to consider sexuality as part of national governmentality. We therefore use the remainder of the chapter to address both literatures in two empirical moves. First in Section III we consider macro-level data with the American census, showing how difficult it is to see any queer presence within tract-level data. Turning to individual-level data in Section IV, we extend this point by running an experiment using both US and UK census micro-data that theoretically should reveal single-sex couples in the nation. In the US the number of same-sex couples is surprisingly low, and these data are laden with assumptions we identify as part of governmentality. In the British context, an interesting story is revealed about the way governmentality emerges through bureaucratic standards of reliability and validity. To our surprise same-sex couples were actually coded out by the Britain’s ONS (Office of National Statistics), and we explore the underlying reasons for that closeting move. Both examples identify the governmentality of the closet at the national scale. Gays and lesbians in both countries remain difficult if not impossible to see through this national image. As Foucault would expect, these instances of national closeting arise not so much from some explicit top-down homophobia. Rather, as is always true of any discourse, the devil is in the details. The "causes" of this closet are rarefied and multiple: societal heteronormativity, categorical tyrannies, concerns over personal privacy, and statistical rigor.

Saturating this chapter, of course, is the ethical question of whether or not we ought to be able to see lesbians and gays in the closet with the census. At the onset, however, there are two points we would like to underscore. First, we are using publicly available census data in this experiment, which have been released according to strict confidentiality limitations. In other words, these governmentalities already inhabit the public sphere. In this way we want to call attention to the extent of the assumptions one would have to make to see lesbian or gay people through the census. And in this way we agree with the paradoxical knowing/not knowing emphasized by Sedgwick (1993) and Fuss (1991), but underscore how this paradoxical knowledge is more than just a metaphor for something else: it is materialized in national geographies sanctioned by the state. And here we call attention to the spatiality of this closet: its location at the nation-state. Second, we’d like to stress that this chapter is not an exercise in "gay-spotting", outing, or locating gay people. We are interested in whether sexuality and the census conjoin in the mechanisms of governance, and trying to specify how the closet operates through those mechanisms.

Queer theory has conceptualized the nation-state, and hence the closet inside it, in a variety of ways, but none have considered the role of state censuses play in the framing of the nation. Foremost there has been a clear recognition of the state’s power to conceal, erase or deny homosexuality through a wide variety of its legal apparatuses (e.g. Adam, 1995; Marcus, 1992; Mohr, 1988; Kinsman, 1987). These works identify just how powerful and pervasive the closet is in national legislation. As a top-down exercise of power, "law" is an efficacious, visible and direct cause of the national closet. For example, several authors have explored the implications of the famous American Supreme Court case Bowers v. Hardwick, which upheld Georgia’s sodomy law, specifically exercised against gay men (Sedgwick, 1990; Halley, 1994; Hunter, 1995). The state supreme court invalidated particular law in 1998, sparking even further discussion of its closeting effects (Sack, 1998). Others have explored the Clinton administration’s infamous "don’t ask; don’t tell" policy vis-à-vis gays in the military (Butler, 1997). In the British context, the ramifications of Clause 28 (which forbade local governments to "promote" homosexuality in any way) and debates over age-of-consent laws have certainly demonstrated the power of the state to closet gays and lesbians nationally (e.g. Wallis, 1989; Carter, 1992; Tatchell, 1992). Still other examples of the national closet are struck in studies of immigration law and practice. Countries like Australia and New Zealand allow same-sex couples to immigrate, for instance. Likewise Binnie’s (1997b) recent work on gay international immigration in Europe also highlights the power of the state to forge the closet. The common point here, of course, is that through its rules and regulations, states sanction the presence of gays and lesbians in their national territories. Moreover, they do so in rather complex and often contradictory and inconsistent ways.

The closet is also identified through a reading of literature, film, and cultural practices that treats such texts as markers of national culture and consciousness (e.g. Parker, Russo, Sommer, and Yaeger, 1992; Gevisser and Cameron, 1995; Gearing, 1997; Edelmen, 1994). Acts of cultural production in place, then, are recognized as revealing the closet in national space. Debate over the American National Endowment for the Arts (NEA) funding homoerotic material, for example, frames the nation as a cultural space of the closet (Crimp, 1994). In Britain similarly Sinfield (1989) has argued there were tight links between Thatcherism and the closet in national culture by its endorsements of certain kinds of national arts. The issue of state censorship also manifests the closet at the intersection of national culture and state power. Several authors, for instance, have described how Canada Customs routinely bans so-called objectionable material at the border on the basis of its queer focus (Fuller and Blackley, 1995; Forbidden Passages, 1995). Ironically much of this same erotic material is allowed past national borders when it is bound for "straight" bookstores, specifying the rather acute heteronormativity of state power. Canada Customs in this case is a powerful closeting force against even NAFTA! On a more activist slant, Queer Nation has invoked the closet as escapable only by direct retaliation against the structurally heteronormative national "society" (Berland and Freeman, 1992). By coding the subject position of queer as a nation itself authorizes modes of opposition to homophobia and violence more or less sanctioned by nation-states like the UK and the US. Finally, Duggan (1995) along with a host of other political theorists argue we must queer the state. As activists, queers must infiltrate and reactivate state power with a queer-positive affirmation. It also means recognizing that queers already do inhabit the state, and this can (though not always) be a powerful strategic position to augur social change (Brown, 1997; Cooper, 1995). As theorists, it means we all must be attuned to the often subtle ways the state is already queered by stressing the progressive possibilities of siting queers within the state apparatus.

From the discussion above, queer studies show us two related themes. First, the nation sees the closet through a wide variety of media and contexts within itself, though certainly a key node is that of the state itself. Second, this encourages us to push on with a search for the variety of ways the closet operates at the national scale. As Bhabba (1990) reminds us, however, there can never be a single authoritative way of knowing the nation. With this point in mind, we want to suggest that census data might be another important way the nation is narrated—a narration that manifests the closet at the national scale. Moreover, given Foucault’s influence on queer theory’s origins (Jagose, 1996; Halperin, 1995), it is surprising that his work on government has not been drawn on already.

Foucault (1991, p. 90) defines governmentality as "the art of government", while Gordon refers to it as (1991, p. 1) ‘governmental rationality", referring to a way that government thinks about what it does, and the way scholars ought to think about government’s power. Both point to a re-thinking of state power away from the typical top-down, coercive use of force model. Foucault himself used the term governmentality to signal a shift (c. 16^th Century) in governmental power from sovereign monarchy to a modern administrative, bureaucratic and liberal-democratic state. Here he notices a relative change in praxis about how a state exercises its power. The standard, top-down practice of state power (the coercive use of force and edict) becomes juxtaposed—if not superseded-- with a rather different, more subtle form of power that is more interactive between government and governed. This way of thinking about state power is akin to Giddens’ (1987) account of the increasing relevance of surveillance as a form of state authority over its territory. By contrast, however, Foucault is not so much interested in state surveillance as a deliberate, self-conscious and policy-directed reification, rather in the center-lessness and capillaries of the ways normalcy is produced. Nonetheless, both authors agree that politics in modernity is hearkened in part by power through governmental knowledge and not force alone.

With the rise of modernity, a governmental rationality emerges through policing and disciplining that blurs distinctions between state and society. Here, of course, the very rise of "society" is noted, as subjects are transformed into democratic citizens and rational economic agents. Likewise, the functions of the state increase and multiply into concerns like public health, medicine, and social work. Concomitantly there is a rise of expert knowledges about the nation (e.g. demography, sociology, and criminology). Foucault links the broad definition of government (as the practice of the state) with the narrower definition of "the conduct of conduct: that is to say, a form of activity aiming to shape, guide or affect the conduct of some person or persons" (Gordon, 1991, p. 2). The rising relative importance of governance over sovereignty and its manifestation through a series of state apparatuses and a whole complex of saviours, or ways of knowing.

Governmentality defines the array and articulation of institutions, procedures, analyses, reflections, calculations and tactics vis-à-vis the population. For Foucault, a population has become governmentalized to the extent that it becomes a datum, a field of intervention, an objective of governmental techniques to be known intimately yet still "at a distance" (Miller and Rose, 1990). This distance is achieved through the rise of statistical methods and calculus. Governmentality is a way of knowing the members of a nation both reliably and confidently through the epistemology of science underwriting quantitative measures.

"Whereas statistics had previously worked within the administrative frame and thus in terms of the functioning of sovereignty, it now gradually reveals that population has its own regularities, it own rate of deaths and diseases, its cycles of scarcity, etc.; statistics show that the domain of population involves a range of intrinsic, aggregate effects, phenomena that are irreducible to those of the family, such as epidemics, endemic levels of mortality…" (Hacking 1991, p. 99)

This knowledge can be used to address social problems by modern state bureaucracies, but it can also lead to a self-disciplining of the population. Foucault referred to this interaction between individual and nation as "bio-politics", a way of rendering certain bodies "normal" while others abnormal, deviant, and still others epistemologically (and thus ontologically) impossible. Consider, for instance, the centrality of the "average" statistic as a way of representing characteristics of the entire population using a Gaussian or normal distribution. As the nation comes to know itself as a coherent entity through the statistics derived from individuals, governmentality not only signals the extent and form of knowledge about national society, but also its role in disciplining orders of normality. This way of thinking about the nation will make, in Gordon’s words (1991, p.3):

"…some form of activity thinkable and practicable both to its practitioners and to those upon whom it was practices."

Hacking (1991, p.181) argues furthermore that the rise of statistical epistemologies are crucial to governmentality:

"Statistics has helped determine the form of laws about society and the character of social facts. It has engendered concepts and classifications within the human sciences. Moreover, the collection of statistics has created, at the least, a great bureaucratic machine. It may think of itself as providing only information, but is itself part of the technology of power in a modern state."

In the national state a population is created statistically as the valid form of knowledge comprising highly reliable facts and figures that reveal problems for the state to address but also the means by which they are to be tackled, and the confidence of knowledge to propose solutions. Populations become data that can be statistically described, measured, and hence diagnosed, treated, and bettered. Through the complex web of state data on its nation, reliable categorization provides a way of representing and knowing that channels state power with state knowledge of its citizens (and non-citizens!). Citizens, and their problems, come to be known, defined, and recognized along certain descriptive lines, as in Foucault’s sense of savoir.

Foucault’s ideas on governmentality have been deployed in a wide variety of empirical areas of state activity from economic management (O’Malley, Weir and Shearing, 1997) to urban and rural planning (Tang, 1997; Murdoch, 1997), to mapping technologies (Elmer, 1996), to public health (Petersen, 1997; Osborne, 1997), to migration (Abram, Murdoch, and Marsden, 1998), and even to garbage recycling (Darier, 1996). Such a breadth of following is indicative of the subtlety of its power/knowledge. Governmentality is an ordinary, mundane practice of state bureaucracies. As yet, however, it has not been used to explore sexuality per se, never mind the closet. That gap is readily filled by this chapter since the census is one of the most obvious and important ways governmentality works and can be identified in the nation-state.

We argue that the census is a key apparatus bearing on the closet, for it is the state’s way of seeing its nation "at a distance". To substantiate our claim, we explore the dimensions of the closet through an experiment: we try to "see" sexuality through the British and American censuses. Neither nation directly represents gay people through its census, demonstrating how the closet operates at the national scale specifically through governmentality. This governmentality, however, has a subtlety that we also demonstrate. Strictly speaking, only certain lesbian and gay people can be seen in the censuses, however only if gross assumptions are made about them. Moreover, there are particular logistic and practical difficulties in the British census that we discuss. The overall point is to show the simultaneous simplicity and complexities of the national closet. On the one hand it is a simple issue: there is no question on the census about sexual orientation. Given the thorough heteronormativity in society, this is hardly surprising. On the other, it is an example of a population’s self-disciplining: citizens fill out the census form and researchers and government agents interpret the data. After all, gays and lesbians obviously are part of the British and American nations.

According to the United Nations, national censuses are:

‘the total process of collecting, compiling and publishing demographic, economic, and social data pertaining at a specified time or times, to all persons in a country or delimited territory’ (quoted in Brown, 1976).

While sexual orientation is clearly not a requirement according to this definition, neither is it precluded. At a practical level censuses serve a number of functions. Most obviously, they act as the benchmark for identifying and understanding changes in national-level or small-area populations. They are used as the baseline for population estimates and projections, and local government revenue-transfers as well as for representational equity in legislatures. They are also used for identifying areas of need, clusters of deprived people, targeting population sub-groups (e.g. ethnic minorities in health campaigns, etc.). In a very practical sense, the census is undoubtedly part of the art of government in the modern era.

Developing this point, we would highlight the fact that a basic source of a census’ governmental power lies in the power of its scientific epistemology. Its power-as-knowledge derives from at least two scientific principles it embodies. Its reliability is guaranteed by its periodicity (that it happens at regular intervals), while its validity is achieved through its universality (comprehensive coverage) because, unlike other national surveys, it is compulsory (you can be fined or imprisoned for ignoring the census in both countries—and of course the state has the legal means to enforce compliance). The census therefore has enormous epistemological purchase because the state can know the population with such authority. What is more, it is not simply a top-down exercise of power. Since the census relies on self-reported information, nationals take an active, albeit already disciplined, role in their governance. The census, then, can be understood as a tool in the savoir of governmentality, a statistical linchpin between the state’s power and confident knowledge of the nation. Moreover because the data are more or less publicly available, census data isn’t simply the state’s knowledge: its everyone’s. People see themselves through the census. This point is often only acute when state savoir does not harmonize with individual nationals’ self-identities. Take for example, the category "Hispanic" in the American census. It is designed to identify ethnic affiliation based on the commonality of speaking Spanish. Yet it corrals such ethnically and culturally diverse identities as Mexicans, Caribbeans, Latin and South Americans (Gonzales, 1997).

Few scholars, however, have gone further to identify specific government statistical techniques of data collection and dissemination as clear examples of governmentality (Miller and Rose, 1991). Perhaps the best example is Owen’s (1996) historical analysis of the Egyptian census of the early 20^th Century to illustrate how its changing categories of families and individuals reflected and reinforced shifts in the agrarian economy. Another involves looking at the use of intra-national migration statistics in local planning and decisionmaking, Abram et al. (1998, p. 238) underscore the importance of government statistics to the exercise of governmentality when they argue:

"The collection of statistics and the proliferation of inscriptions, with their technologies for classifying and enumerating, thus become effective techniques of governmentality, allowing civil domains to be rendered visible, calculable and, therefore, governable. Furthermore, they herald the advent of subjects who remain ‘free’ but come to calculate themselves in terms derived from the tools and techniques of governmentality."

So as an instance of governmentality the census provides a means of citizens’ self definition, categorization that derives from and feeds into state authority. Since individuals in the population actively participate in the census process by completing the form, the census is not simply a top-down exercise of state authority. Questions and categories are offered, but so too are they (re)interpreted by respondents. They are ways of seeing a nation’s population done by and for the state to solve practical problems. Where better than the census to look for what the American or British populations "are like", and where better to acknowledge Foucault’s point? The dimensions of likeness, of course, are contingent upon the questions that the state decides to include on the form, the range of values that any such variable could take, as well as the way in which the enumerated interpret and understand what is asked of them.

So how does the census produce the closet at the national scale through governmentality? As we noted above, the simple question, ‘Can we see sexuality in the census?’ immediately prompts an equally simple reply: no. Neither British nor American census forms include an explicit question on sexuality. Table 4-1 illustrates this point well. It compares the categories derived from questions on sex, marriage, family type and household composition in both the U.S. and Britain, questions where one’s sexuality would most likely be at issue. Consequently, the issue of sexuality in any governmental knowledge or planning derived throughout the census would be non-decided. It could not be a significant variable, because it is not a variable in the first place. So there is a fundamental presumption of heterosexuality in both censuses that would be reproduced through any use of the data, and heterosexist presumption, of course, is a clear example of the closet at work. This is surprising if we think that gay individuals (and their living choices) are different to the general population choosing different types of accommodation in different places for different reasons (e.g. Bouthillette, 1997). With information on sexual orientation, wouldn’t planners be able to plan better? Yet in other ways it is all too unsurprising. Gay people are thought to be a small percentage of the overall population, and therefore do not need to be specifically addressed in the census. As well, several social characteristics besides sexuality are not asked on the census. In other words, sexuality is part of a constellation of identities that go unknowable through census governmentality. For example, in 1991 the British census does not ask questions on religion, ancestry or income, which are included on the US census. Conversely, details about housing tenure seem to be a rather difficult identity to isolate from American census data, compared to the British census. Important as these variables are, there are two major considerations that are debated decennially within the context of the modern bureaucratic state: the financial cost of adding new questions and the worry that the question would be unreliably answered or affect the answering of other questions. Despite the potential usefulness of income information in decisions and a vocal lobby in favor of adding it to the British 2001 census, there is still debate about whether it should be added. Clearly then, if the census is conceptualized—and evidence suggests that it surely is—as a mirror by which the national represents itself, lesbians and gays go unseen. They remain hidden, concealed, in the national closet. They do not belong to the nation, even though we know (through other epistemologies) that they are here. This point affirms the paradox that is signalled by Sedgwick and Fuss. The closet is a knowing by not knowing. But here we would suggest the reverse: for the census it is a not-knowing by knowing too.

Rather than making this the end of the question, however, we choose to see it as the start. For if Foucault’s arguments energize our thinking about the closet, then we are spurred to draw a distinction between accuracy and precision. It is accurate to say that sexuality is stealth in the census, but is that precise? From this sort of starting point, we can begin to exercise the subtle ways the closet operates. Notice, for instance, the variable "Respondent’s Relationship to Householder (Unrelated)" (item #4) in the American Census in Table 4-1. Presumably the value "unmarried partner" or "other non-relative" could capture gay men or lesbians who were in relationships with householders of the same sex. Indeed, the American Census allows for such a coding. From its own handbook:

"An unmarried-partner household is a household other than a "married-couple household" that includes a house-holder and an "unmarried partner". An "unmarried partner" can be of the same sex or of the opposite sex of the householder." An "unmarried partner" in an "unmarried partner household" is an adult who is unrelated to the householder, but shares living quarters and has a close personal relationship with the householder." (U.S. Census, 1990, p. b-15; emphasis added)

Here, it would seem, the closet breaks down, at least for the American census—and at least for certain lesbians and gays (see below). The data can be reported and coded in such a way as to reveal same-sex couples. When we turn to the actual data, however, the closet seems to manifest itself again. To test their revelatory power, demographic data were collected on census tracts from a well-recognized gay neighborhood in Seattle Washington, Capitol Hill. These data were accessed from the census’ homepage, which is the most readily available source of census data for the public (http://venus.census.gov/cdrom/lookup/). It is also free data, available at the public library, for instance. These attributes are an important part of the exercise, since the whole idea of governmentality is that power lies in the capacity to represent the population to itself. If governmentality is predicated on a bio-politic of self-discipline, part of those politics also lay in the fact that the population can, in turn, see itself through these data.

Tables 4-2 through 4-6 give a brief statistical summary of Capitol Hill, Seattle in comparison to the rest of the city and the nation overall, for individuals and households respectively. We can see that it has rather odd or extreme values on several variables that mark it as a decidedly "different" place from the city of Seattle and the United States overall. The area is more male (54% of residents) than the city or country percentages (roughly 49%), as illustrated by Table 4-2. Such a fact might be relevant given that gays and lesbians desire people of the same sex, so single-sex concentrations in space might suggest a gay neighborhood. In addition, the ways individuals arrange themselves in household units may provide some clues. Table 4-3 demonstrates that the majority of people living on Capitol Hill live in "non-family households (almost 80%, in fact). Specifically, around a quarter of both men and women on the Hill live by themselves (25% and 23% respectively). This statistic stands in marked contrast with Seattle overall and the United States. In those other places, the majority of people live in traditional, family households (65% and 84% respectively). On Capitol Hill, only 23% of residents live in such a conventional arrangement. Since lesbians and gays cannot marry in Washington State, we might also use data on marital status to infer a queer presence in the neighborhood. Table 4-4 tells us that Capitol Hill is a space of single people. Thirty-seven percent of men and 25% of women (aged 15 or older) in the neighborhood have never been married. These are rather higher than the statistics for the city or the nation. Less than 10% of either women or men on Capitol Hill are married and living with their spouses there. If we move away from data on individuals in neighborhood, and look at its households, similar stories are told. The percentage of households that the Census would define as "family household" (i.e. the presence of marriage arrangement with or without children) is 14% for Capitol Hill, according to Table 4-5, much lower than the city or the country overall. Nearly half of the households in Seattle are "family households" (48%), and well over half—nearly three-quarters—of American households are described that way (71%). Conversely most of the households on Capitol Hill are one-person households (Table 4-6).

So does this census profile frame an epistemology of the closet for that piece of the nation? Facts, of course, do not simply speak for themselves. And this is where we begin to see the complexity of the governmentality in the census. It must work through other epistemologies that we have juxtaposed with this government data. Indeed the very original set-up of this experiment is based on our own experiential epistemology: ‘we know it’s gay because Michael’s gay and he’s been there.^’ The facts only reveal gays and lesbians if we also draw on other epistemologies through which we already know them. To run through the tables again—and here we must suspend political correctness and follow stereotypes, preconceived notions, assumptions, etc (the ‘other epistemologies’ washing across the facts to make them sensible)-- Table 4-2 suggests Capitol Hill is a gay neighborhood because they tend to be concentrated male spaces. Capitol Hill is a queer area according to Tables 4-3 through 4-6 because we know that gays and lesbians cannot marry same-sex partners, so they do not conform to census definitions of family households; they are legally "single". The potential limits of the truth we are trying to tell here, of course, can easily be demarcated because other identities can also be suggested by these facts. In other words, different and perhaps equally plausible conclusions could be drawn from these data. For example, Capitol Hill might be interpreted as a place of young single people who might all be straight. Yet recall that our point is not to use the census to definitely claim Capitol Hill as a gay space. Rather it is to see how the closet operates through the census.

On that tack, we would underscore the immediate conjoining of alternate epistemologies with the census because our reliance on them to confirm "what we already know" shows an interesting complexity to the closet. To put it succinctly, if we bring certain gays and lesbians out of the closet with the census, we do it by simultaneously closeting others with our biases and stereotypes. To know Capitol Hill as a queer space, we must know it as a gay-male space and closet lesbians. To see gays and lesbians because there is a majority of non-family households in the area conversely closets those gays and lesbians who do live in family households. Not all gays and lesbians are single, not all have never been married. Some are children, stepchildren or grandchildren and so on. If we infer that Capitol Hill is a gay area because most households are populated by individuals, we likewise closet those gays and lesbians who are partnered but do not live with their partners in the same household. The point we are trying to make here is that even when we use the census against the governmentality of the closet (in it already because there is no question to capture same-sex situations), we wind up re-instanciating it. We cannot break down the closet without inevitably reconstructing it for somebody else. To our minds this point affirms the "knowing by not knowing" argument put forth by Sedgwick, but in a thoroughly spatial and complex way. In the end, only certain types of gay people are seen through these census data, the rest remain closeted. Moreover, the validity of this epistemology is quite precarious. This is a form of closet governmentality: a knowing by not knowing—or perhaps more accurately, a not-knowing by knowing.

The complexity of the closet in the census can also be understood through the power of scientific and bureaucratic rationalities that reflect and reinforce the power of the census to narrate the nation truthfully. Recall from page 117 that the proper way for an American in a same-sex-cohabiting couple to describe their relationship using the categories in Table 4-1 is as "a unrelated household member" who is an "unmarried partner". Consider how the categories offered in Table 4-1 were translated into the census output chronicled in Tables 4-2 to 4-6. You can see that there is no such category as "unmarried partner" in the output. The one category gays and lesbians should use to describe themselves when filling out the form disappears when they go to the publicly available data see themselves. To our minds, this is another complex manifestation of the closet nationally. We would explain it, moreover, not through some intentionally homophobic machination, but rather through the force of governmentality operating through discourses that sustain the truth and power of the census. This closet could be due to the tendency of the variable to reveal small numbers, which would threaten the anonymity and confidentiality guaranteed by the census. Cell counts are suppressed or aggregated in published data when there is a possibility of identifying an individual. Here, the closet appears as a consequence of the census’ attempts to live up to important principles that help maintain its own authority and power. If Americans could be identified individually through the census, they would be far less likely to participate willingly in its surveillance, or tell it their truths. Another possible reason might be that the Census Bureau simply did not think the data on unmarried partners (some of whom would be same-sex) or the specifics of non-family household composition would be of interest to the general public. The presentation of data in science is always a process of selection and editing. If the census did offer everything it could, no doubt its public accessibility would be compromised by the sheer avalanche of possible information. Either way an upshot of this erasure is that governmentality’s positive emphasis on normalcy is noticeable. The categories "roomer/border/foster child", housemate/roommate", "other non-relative" and "unmarried partner" have all been collapsed into the category "not living alone". Overall, we would argue that a complex closet is operating through the collapse of categories, done not so much to erase identities but to protect their privacy and make the data accessible.

A large part of the problem, of course, is that the data in the tables above are provided at the census tract, rather than the individual, level. Hence any attempt at reading individuals from that spatial scale belies the ecological fallacy (Babbie, 1997). This is the problem of generalizing from aggregate level data (in this case census-tract scale) to individuals inhabiting those numbers. For instance, we can know or calculate the number of men who have never been married, and we also know the number of men who live in non-family households, but we cannot logically infer that they are necessarily the same individuals. It is a recurrent problem for those dealing with readily available census data, because it is provided for areal units in order to protect the anonymity and confidentiality of respondents (a concern which sustains its governmental power through its power to tell the truth). For that reason, we turned to individual, micro-level data, which are taken from a sample of the population, and whose geographic identification is not as precise as the census tract. The benefit of these data are that we can cross-tabulate characteristics of individuals with ease, thus escaping the ecological fallacy. The cost, as it were, is that these data are not necessarily widely available or readily manipulated without a certain degree of familiarity with the census. Nonetheless, since bureaucratic experts (like academics!) are part of the exercise of governmentality, so the validity of the exercise remains largely in tact.

The origins of this second part of this experiment came from Paul’s ongoing cross-national research project investigating whether or not women’s labor-market status suffers as a result of family migration (in Britain and the US), which usually stems from a change in male partner’s job circumstances. A requirement of the analysis in this project was the identification of "linked partners" in the respective micro-data samples: the 1991 British Sample of Anonymized Records (SAR) and the 1990 US Public Use Micro-data Set (PUMS). Both the SAR and the PUMS are individual-level data sets taken from a small sample of the entire national population recorded in the census. Each individual’s exact geographic location is withheld, however, so confidentiality and anonymity are retained while the ecological fallacy can be avoided. In both data sets linked partners can only be identified reliably in couples that include a self-declared householder. This is because in both censuses household relationships are structured around the head of household, with individuals identifying their relationship to this person on the census form. The resulting couples extracted in the ‘tied migration’ project were therefore either married or cohabiting couples involving a head of household. Other married or cohabiting couples not involving the householder in multi-couple households could not be reliably identified. In theory, of course, we would expect to be able to identify same-sex couples in both these data sets, so long as gays and lesbians had self-identified themselves as cohabiting with someone of the same sex. So here, it seemed, even though there was no explicit question on sexuality, gays and lesbians who were in household relationships, might be a visible part of the British and American nations, according to the census. And it was at this stage that Paul and I began discussing the possibility that the census might not be so much of a closet after all.

The 1990 American census has more questions than its British equivalent. As a result the PUMS are more comprehensive than the SAR in terms of variable content and are presented in a much less refined format than the SAR data. The benefits of this ‘rawness’ is that the PUMS data can be manipulated to create variables that match the information from the SAR (Boyle, et al., 1999). Although the relationships within households are defined slightly differently in the US census, it is still possible hypothetically to identify same-sex couples. In much the same way as the British census, individuals must state their relationship to the householder.

We utilized a 0.1% random sample of the total population drawn from the 5% sample of the PUMS data. From this, 59,978 individuals or 29,989 linked partners were extracted, one of whom was the householder. Selection criteria included nuclear family households, partnered adults, not in armed forces, not living in institutions, permanent residents, and persons aged 16-60. We drew information on marital status and sex first. Table 4-7 crosstabulates marital status (for opposite sex couples) and sex (for same sex couples) against region, demonstrating that while the majority of individuals were in conventional, heterosexual relationships, 74 men and 64 women self-identified themselves in same-sex partnerships. So, crudely estimated, there would have been 138,000 individuals identifying themselves in same-sex couples conforming to the selection criteria established above in the US, according to the 1990 census. Even if we make the terribly problematic assumption that gay individuals may be less likely to live in partnerships than heterosexual individuals, the extremely low percentage of individuals living together in same-sex couples as a percentage of total persons in the US (0.055%) seems remarkably small and contradicts cultural assumptions about the very presence of gays and lesbians in the country. The oft-quoted statistic (from Kinsey) is that 10% of the population is homosexual. Even if we reject this benchmark for the lower 2% estimate that has been suggested elsewhere (e.g. Smith, 1991; Muir, 1993; cf. Laumann, Gagnon, Michael and Michael, 1994; Sell, Wells and Wypij, 1995), that would put the 1990 lesbian and gay U.S. population at 4,974,197 which in turn (according to the PUMS data) would mean that only 2.77% of gays and lesbians are living in same-sex relationships.

Again, however, we immediately confront the problem of assumptions and alternate epistemologies at work in reading these data, even if they did jive with our less scientific epistemology. To our minds they are also part of the power/knowledge effects of the closet in a national context of governmentality. Consider the assumptions and caveats that had to be made in the above exercise. First, we are relying on a self-reported survey by using census data. Fears over homophobia might very well have cautioned certain gay couples from describing their relationship to the government and ultimately the rest of the nation. This certainly plays into the governmentality of the closet, since people would essentially be closeting themselves on the census form. For argument’s sake, what if a question on sexuality was included on the census form? The closet would still be at work since many lesbians and gays would no doubt be highly reluctant to reveal their "private" sexual orientations to a state authority. That concealment would, of course, be especially apposite to those who would self-identify as already being in some form of the closet (for instance, married men who are bisexual and have sex with other men)! Second, of course, couples who simply appear as two women or two men living together in the same household might be closeted by being subsumed into a population of roommates. Third, we are picking up couples who reside within a single household. With both of these assumptions, a powerful conventional heterosexism is at work. Here gay couples are presumed to ape the arrangements of their straight counterparts. Yet a decided thrust of queer theory has been to question and restructure the hegemony of those arrangements in the first place (Weston, 1991; Benkov, 1994; cf. Bawer, 1996). Do relationships need to be hierarchical? If they do, can the hierarchy shift between partners? To what extent do assumptions about monogamy infect the very category of "couple" and its heterosexual bias? Do members of a couple need to live in the same household, at the same address? The crux of the census is a record of the population and how it is distributed across housing units. In other words, the data structure affects the story the data can inevitably tell. Fourth, and perhaps most obviously, we are only detecting gay and lesbian couples. If there were that hypothetical question on sexuality, this point would not be a problem. However, we have no way of revealing a whole series of gay men and lesbians who help comprise the nation. What about those 97.23% of gays and lesbians—if we believe the controversial 2% estimate-- who are single? We simply have no way of detecting them through these data. They remain closeted relative to those revealed above in the sense that the data that shows them also works to conceal others. Likewise, what about those queers who might have multiple partners or alternative forms of family? Bisexuals also suffer the effects of the closet since as a sort of categorical pharmakon, they are both categories, but neither. The point here is that each set of assumptions above creates a closet in order to reveal a small percentage of the population as gay or lesbian. It is fascinating that in order for them to be revealed, we must closet others. This paradox is at the core of governmentality: to see some lesbians and gay men in the nation we must inevitably closet others because of our partial assumptions. Additionally, it is interesting to note that the deconstruction above reveals how the census might be concealing certain numbers of opposite-sex couples too!

If we’re interested in the geography of the closet subnationally, we can identify the regional location of these couples. Table 4-7 problematically locates these individuals regionally. Again, truths seem hard to reconcile with our alternate epistemologies of queers in the U.S. Moreover, statistically there is a problem in inferring any grand truths from these figures. Does it make sense, and whatever could it mean, for instance, that there "are" twice as many partnered lesbians in the Middle Atlantic region of the U.S. than in New England? The very small numbers make it hard to say anything conclusive even at a regional scale. In order to assume these data are the truth, what other truths do we have to suspend? How much closer are we to revealing the lesbian or gay presence in the American nation?

As in the 1990 United States PUMS, it is possible to extract linked partners from the 1991 Great Britain SAR. Selection criteria were the same as in the U.S. sample and a relatively large sample of 164,496 linked partners was drawn (Table 4-8). However, to our initial surprise, none of the couples were the same sex (Table 4-9) begging the question why? Of course, the power closet itself might discourage couples from identifying themselves as gay, despite repeated assurances that census information is confidential. Presumably this would be more important for those living completely in the closet, than for those who are "out", living together as couples. Certainly we would expect some gay partners to acknowledge their relationship with a householder of the same sex. Thus, while we might expect the number of gay couples identified in the census to be smaller than the true figure, we would not expect it to be zero! Our initial assumption, that we had made some kind of error extracting the data, was proven to be unfounded and this prompted us to further investigate this anomaly. Unbeknown to us the issue of same-sex couples had initiated a debate prior to the 1991 census as a decision had been taken at an early stage of the preliminary census planning process to recode individuals who recorded themselves as cohabiting with or married to a householder of the same sex. The exact process is described below.

The results immediately raise the question of whether this was some sort of plot by the Office of Population Censuses and Surveys (OPCS), or the government department under which it served, to ‘hide’ gay individuals. In fact, discussions with census statisticians seemed to reveal a less guileful and more practical reason had influenced the committee meeting discussions that resulted in this decision, which resonate with the subtlety of governmentality. The problem was twofold, according to the census offices. First, the relationship question on the census form provided no guidelines about how gay or lesbian individuals should complete the form. That absence could create a problem of interpretation when lesbians and gay people (at least those in couples) filled out the form. The fear was that while some gay couples would view themselves as ‘cohabiting’ others may not regard this term as relevant to their status, and without additional information clarifying this point, the resulting information would therefore be unreliable. According to the ONS the decision was based less on the sensitive issue of sexuality, and more on the concern that the information that would result from the returns would be incorrect. The second problem was also of a more practical nature. The numerous computer programs that had already been written to handle the census information, according to ONS, would need to have been edited and there was a series of programs that would have needed changing. It was not possible to simply alter a single line or chunk of code to allow same-sex couples who self-identified as married or cohabiting to remain. This intricate re-programming would be expensive and risky, it was argued. Also, the decision to prevent gay couples from being identified in the census output had been agreed by the time that the draft census form was submitted to Parliament for ratification. The story does not end there, however. A key figure in the census offices only became aware of, or realised the implications of, this decision subsequent to this time. He reopened the debate and, while he was unconvinced that the decision was correct, he was persuaded by the argument that the identification of gay couples should not be allowed if the wording on the census form could not be changed. This was impossible in the time frame of the census machine as altering the wording of any part of the census form required that it be resubmitted to Parliament causing impossible delays and prohibitive costs.

The decision stood, although some attempt was made to investigate the potential reliability of such information if it had been allowed to remain in the census output. A great deal of effort is expended prior to the census proper to make sure that the data are collected reliably. The 1989 census test included the same relationship question as used in the 1991 census (Figure 4-1) and a sample of these forms were examined to quantify the number of same-sex couples. Of the 7,500 households that were checked, only one ‘genuine’ case occurred. There were 36 additional cases where the couples were recorded as being of the same sex, but when examined further these all turned out to be errors such as people completing the form incorrectly or incorrect sex imputation by the census offices (see below). Prior to the census, then, it appeared that the information on same-sex couples that could be derived from this question would probably be unreliable.

The processing of the actual census returns is a computerized process, but clerks are required to solve specific problems. Same-sex couples were one example and each time one was identified, a cleric was alerted to revisit the original forms. Various checks were then made to confirm that the apparent same-sex couple was ‘genuine’. Other information on the original forms helped in the identification of errors made by the respondents – the most obvious being that the wrong box had been ticked on the form. Thus, some individuals that were blood related, rather than partners as the form suggested, were identified. However, errors could also have been introduced by the ONS. In certain cases where information was missing from the returned forms the ONS imputed data and, occasionally, an individual’s sex would have been missing from the original form. The sex of some individuals could have been imputed incorrectly, creating fallacious same-sex couples and this would also be checked by the cleric. Of course, after these thorough checks a number of genuine same-sex couples would have remained. The cleric was then instructed to alter the coding such that the householder’s partner was recoded as an ‘unrelated’ member of the household. Returning to Table 4-1, this would mean culling those gay individuals who described their situation as, say "husband/wife" or "living together" and placing them instead into the "unrelated" category. Thus, the exclusion of same-sex couples from the census output resulted from this conscious manipulation of the census data. Allegedly, of course, the rationale for the change was not so much a direct exercise of heterosexism or homophobia per se, but rather a scientific concern with reliability. Here it seems the governmentality of the closet operated through the epistemology of science that underpins the state’s savoir of its population "at a distance". More precisely, the concern that drove the alteration of data was scientifically justifiable in terms of reliability: the criteria that each time the question was posed, it measured the same thing.

The census offices did at least make some attempt to quantify the number of couples that were being recoded in the 1991 returns. During this editing process, the opportunity was taken to count those cases where a clear statement of same-sex cohabiting status had been made. A brief analysis of this information was carried out during the processing period and based on five sixths of the 10% Britain sample only 360 such couples could be identified. This information was published in a short article in Population Trends, which concluded that:

"…the results bear out the earlier indication from the 1989 census test that the 1991 census question on relationship could not properly identify gay or lesbian couples" (Office of Population Censuses and Surveys, 1991, p. 1)

Since the census question was not designed to capture such couples this may not be too surprising. Interestingly, the decision has been made that this minor manipulation of census information will not be carried out in the 2001 census, although the wording of the question on household relationships that has been tested to date makes no mention of same-sex couples at all! While this argument was used at the time, it was probably the difficulty and cost of altering the computer programs that was the more forceful reason.

It appears that the closet was produced through heteronormativity at the irreconcilability of two scientific aims, rather than some direct policy of homophobia. On the one hand, data were recoded in order to preserve the principle of reliability. The explicit changing of the data, however, raises the worrisome and equally important issue of validity: that the measurement device (in this instance the question) actually captures what it is it is supposed to measure (arrangements within the household). Governmentality seems to have been exercised through this impasse, producing a closet that conceals lesbians and gay men in the nation.

In this chapter we have moved from the urban scale to the national scale. It has spatialized the closet at the national scale in the United States and Britain. While gay and lesbians studies have examined the closet through a number of framings of the nation (from the literary to the judicial to the military), Foucault’s concept of governmentality was introduced as a means to expand that operationalization. The term describes the network of power relations between the state and population that operate through the categories through which the former knows the latter. We suggested that the census is an interesting example of the governmentality of the closet. Using the U.S. and British censuses, we tried to "see" gays and lesbians in the two nation-states. Neither country asks a specific question about sexuality or the nature of same-sex households. In this broad sense there is certainly a closet governing the way gay people cannot be counted directly through this portrait of the population. The story is more complex than that, however. In the U.S. the number of same-sex couples that could be identified also suggest a certain closet governmentality. Here we noted the complexities of governmentality. The tyranny of categories also meant, for instance, that only gay people in hierarchically-organized couples could be identified. Likewise, stressing the self-governance aspect of governmentality, respondents’ concerns over privacy and homophobia could certainly explain why more people did not describe their relationships in such a way that we could have allowed their identification.

The point of these exercises certainly was not to cast aspersions or lay blame for the closet at the feet of individual respondents or the census departments. Indeed, the interesting point is just how multifaceted the forces that produced the closet were. In the cases above the closet results from structures of heterosexism that conjoin with issues over personal privacy, administrative efficiency, and scientific rigor.

Here the ethical question arises: should gays and lesbians be counted on the census? If they were, the closet would certainly be dismantled at this spatial scale. Gays and lesbians would arguably increase their visibility in the nation given this additional opportunity to "come out". It would be compelling evidence that they are, in fact, everywhere. The statistical legitimacy of this incontestable fact would be powerful, indeed. Given the legislative importance of the census, the impacts for congressional representatives might be quite direct. On the other hand, arguments could certainly be made against self-declaring one’s sexuality on a government form. Despite the assurances of confidentiality, worries over privacy certainly do still exist. People would, in effect, be forced by law to admit their sexuality, since it is illegal to lie on a census form. What would happen, for instance, if PUMS data revealed an otherwise hidden enclave of lesbians in a politically conservative census track? What kind of politics might result from that knowledge being publicly available? Before one can take an ethical stance on the question, we would press them to consider the variety of interweaving sources of power/knowledge that already place the closet in the census. Simply arguing for categories that better reflect gay and lesbian relationships on the form, for instance, would not necessarily dismantle the self-imposed closet, for instance—though it would certainly be a start!

While we stand behind our arguments about the relationship between the closet, governmentality, and the census, we recognize the dangers of over-extending the point of this chapter too. For example, one might argue that lesbians and gays do not use the census principally as a means of framing themselves nationally. In this way, other modes of representation become "more important" to interrogate vis-à-vis the closet. One commentator, for example, noted the ability of marketing firms or insurance companies to "see" gays in the nation through their own profiling techniques and statistical prowess. We certainly agree that the census is not the only possible framing for the closet. Likewise, we definitely do not want to be interpreted as saying that the census is the only location of the national closet within the state itself. The vignettes that opened this chapter, and the charged debates over "family values", age of consent laws, same-sex marriage, immigration policy, inter alia, obviously demonstrate the census is one—albeit important—part of a broader constellation of state power/knowledge around sexuality. In sum, we see our contribution as rather modest: to point out how the census is a location for the closet in the nation in both simple and complex ways. We see this contribution as relevant because this dimension of the census has been neglected by queer studies population geographers, and students of governmentality to date and is a particularly interesting example of Foucault’s ideas.

This chapter has shown the relevance of seeing the closet as a material force at the national scale. When we think about the presence of the closet in the nation, we would do well to think beyond the likes of Peter Mandelson, Ron Davies, Bowers v. Hardwick or "Don't ask, don't tell". The closet appears in far less sensational national representations. It exists in the census forms and data files through which the populations is defined and governed. Moreover, as an exercise of power/knowledge, its operation is at once simple and complex. There is no question on the form about sexuality, to be sure. Yet gays and lesbians can still be seen if they, and the people looking for them, make a variety of assumptions. Ultimately, the national closet is not a simple space, nor is it simply metaphorical. It is, as queer theory predicts, a knowing by not knowing, owing to the vagaries of heterosexist categories and bureaucratic rationalities. Nonetheless, the closet is also a not-knowing by knowing as well. Closet governmentality is sustained not merely by surveillance techniques, but by its assumptive relations with other epistemologies we have.

------------------------------------------------------------------------------------------------------------------

1. Sex

Male

Female

2. Marital Status

Married

Non Married

Widowed

Divorced

Separated

Never Married

3. Respondent’s Relationship to Householder (Related)

Husband/Wife

Natural born/adopted son or daughter

Brother/Sister

Father/Mother

Grandchild

Other Relative

4. Respondent’s Relationship to Householder (Unrelated)

1. Sex

Male

Female

2. Marital Status

Single (Never Married)

Married (First Marriage)

Remarried

Divorced (Decree Absolute)

Widowed

 

3. Respondent’s Relationship to Head of Household

Source: 1990 U.S Census, 1991 British Census.

Source: US Census, 1990

Source: US Census, 1990

Source: U.S. Census, 1990

Source: US Census, 1990

Source: U.S. Census, 1990

Source: 1990 PUMS

Source: 1991 SAR

Source: 1991 SAR