Geographic micro-clustering of homosexual men: implications for research and social policy.

• Author: Hughes, Anthony

Abstract

There is increasing demand internationally for better-quality information on people with a non-heterosexual orientation. Information requirements include both basic demographic characteristics as well as evidence of disparities in outcomes or differences in needs compared to the general population. The availability and therefore collection of such data are essential if social policies are to be responsive to all groups protected under the Human Rights Act 1993, and if the impact of interventions targeted at sexual orientation minorities is to be properly evaluated. In light of ongoing difficulties obtaining accurate data on basic demographic variables for this population, we consider whether the census can provide accurate geographic micro-clustering data on homosexual males by comparing census data with a nation-wide survey of homosexual men. Place of residence information was targeted due to the importance of this variable in guiding future survey sampling and the provision of social and health services. The geographic micro-clustering profile of homosexual men in both data sets was congruent, and considerably different to the general male population: 12-13% of the national population of homosexual men resided in an inner-city Auckland area compared to 1.3% of all males aged over 15.

INTRODUCTION

A persistent problem when identifying the needs of homosexual populations has been obtaining representative samples, since homosexuality is defined by low prevalence indicators that are difficult to measure, private and usually stigmatised. Furthermore, without accurate basic demographic information on gay communities to guide research, it is difficult to fully evaluate the effects of targeted health promotion programmes, or rigorously assess the impact of general health, social and economic policies on this group (Gates and Ost 2004, McManus 2003, Sell and Becker 2001).

This has led to what Plumb (2001) has described as a "catch-22" situation. Unconventional survey methods and opportunistic research have predominated because of difficulties associated with conventional probability sampling, but this has inevitably compromised the credibility of empirical findings due to potential biases. Data quality concerns have in turn made it more difficult to advocate for funding specific programmes and further research, thereby hindering public health interventions for this population at every step (Plumb 2001). As a result, it is still uncertain whether the accumulated findings from studies surveying homosexual men and women provide an accurate estimate of their basic demographic and behavioural parameters, or whether our current understanding is limited by serious conceptual and methodological problems (Blair 1999).

Consequently, the collection of more accurate data on sexual orientation has become an urgent priority internationally (Dean et al. 2000, Saxton and Hughes 2003). Efforts to this end are proceeding in the United States (Gates and Ost 2004, Meyer 2001, Plumb 2001), Canada (Statistics Canada 2004), New Zealand (Ministry of Health 2004) and Scotland (McLean and O'Connor 2003), with cited tasks including the standardisation of sexual orientation measures and the inclusion of such measures in regular surveillance (Sell and Becker 2001). Canada and New Zealand have both begun to explore the feasibility of including a direct question on sexual orientation in their national census in the future (Statistics New Zealand 2003b, Turcotte et al. 2003). Current legal and social policy debates surrounding homosexuality--such as same-sex partnerships, families headed by same-sex parents, and fair access to social services--broaden this project beyond health and increase its urgency (Black et al. 2000, Phua and Kaufman 1999).

This paper focuses specifically on geographic micro-clustering and the role of this basic demographic variable in planning future research, interpreting survey findings and implementing social policies. This paper also concentrates on homosexual men, due to the greater availability of data for which to make comparisons. Findings on the New Zealand Census and lesbian women have been published elsewhere (Byrne 1998, Hyman 2003).

DIFFICULTIES COLLECTING DATA ON HOMOSEXUAL POPULATIONS

Clarity over the dimensions and definitions of homosexuality is fundamental to the estimation of geographic distribution and all other behavioural outcomes. Sandfort (1997) has demonstrated that the notion of homosexuality as a singular, uncomplicated characteristic is problematic, as shown by findings that identify different aspects of homosexuality in men. By differentiating between lifetime and current same-sex attraction, sexual contact, sexual identity and partnership, one can identify interrelated but often not fully overlapping experiences over the lifetime of survey participants. Several large-scale probability surveys have recognised this multifaceted nature of current and lifetime sexuality with questions on sexual attraction and/or sexual identity in addition to homosexual behaviour (Laumann et al. 1994, Sell et al. 1995, Smith et al. 2003, Wellings et al. 1994).

Secondly, the private and usually stigmatised nature of sexuality heightens the difficulties surrounding data collection, although substantial progress has been made identifying the range and impact of possible biases across various methodological approaches in sexuality research (Bagley and Tremblay 1998, Bancroft 1997, Catania et al. 1990, Fenton et al. 2001). These range from issues of participation (Dunne 1998, Groves et al. 1992, Johnson and Copas 1997) to biases in operation after the selection of respondents (Catania 1999, Gribble et al. 1999), and include sampling and recruitment strategy, interview mode, respondent motivation, survey topic and question wording.

Thirdly, most studies return a low population prevalence and incidence of homosexuality over its various dimensions (ACSF Investigators 1992, Binson et al. 1995, Butler 2000, Fay et al. 1989, Johnson et al. 2001, Laumann et al. 1994, Paul et al. 1995, Rogers and Turner 1991, Sell et al. 1995, Smith et al. 2003). In a recent national study using random sampling, 5.9% of Australian men reported any lifetime homosexual behaviour, reducing to 1.9% for those reporting same-sex behaviour in the past year (Grulich et al. 2003). This places limits on the reliability of data collected, even when extracted from large general population samples. Furthermore, the small size of this population subgroup makes it difficult to justify the inclusion of a broad range of additional questions in national surveys that may be relevant for this group, such as micro-residential information.

The intersection of social stigma with the low population prevalence of homosexuality has also had implications for the most appropriate research technique for obtaining demographic and behavioural estimates. For small populations such as homosexual men or people living with HIV (Grierson et al. 2004), non-random, opportunistic and self-selected surveys can yield richer information and larger samples than can probability surveys. However, non-random techniques such as these often rely on comparisons with estimates derived from probability surveys in order to assess the generalisability of findings. Probability samples themselves usually look to the most recent national census to design and evaluate sampling strategies and define post-stratification weightings (Catania et al. 2001), and yet most countries do not have this point of reference for homosexual men. The census is without question the benchmark instrument for providing data on demographic profiles and residential sampling lists, particularly for low prevalence populations (Statistics New Zealand 2003a). Yet for many of the reasons mentioned above, the census has rendered most gay, lesbian and bisexual individuals invisible through the absence of any direct questions on sexual orientation as a basic demographic variable.

The upshot of these issues of definition and quantitative technique is that different questions and different methods have been shown to identify different types of homosexual men (Donovan 1992, Prestage 2002, Ross et al. 2000). Men participating in opportunistic studies have been found to differ in several demographic and sexual behaviour characteristics, and are more likely to be homosexually identified, compared to homosexual men recruited in probability studies conducted among the general male population (Harry 1986, Sandfort 1997).

One variable returning consistent findings in national probability studies has in fact been place of residence, with the prevalence of various dimensions of homosexuality being higher in large urban centres (ACSF Investigators 1992, Binson et al. 1995, Laumann et al. 1994, Sandfort 1998, Smith et al. 2003, Wadsworth et al. 1996). Suggested explanations for variations in the geographic distribution of reported homosexuality have included migration to more gay-friendly environments and to places where there is a higher likelihood of meeting a potential sexual or life partner, as well as the greater likelihood of affirmative gay identity formation and higher disclosure of homosexuality in urban as opposed to rural environments (Binson et al. 1995, Laumann et al. 1994, Laumann et al. 2004, Sandfort 1998, Wellings et al. 1994).

Census information distinguishing homosexual from heterosexual males remains limited to same-sex cohabiting couples (SSCC), acquired through indirect, direct or relationship-to-householder questions. These have provided some corroboratory data with the probability studies, at least at the mid-clustering level. Black et al. (2002) determined that 78% of gay couples in the 1990 United States Census lived in a selection of 50 large cities compared to 52% of the total United States population, and Smith and Gates's (2001) consideration of the 2000 United States Census reinforced this...