Scottish Government

Sexual Orientation Research Phase 1: A Review of Methodological Approaches

4. Administrative and large-scale data collection

4.1 Use of administrative data in sexual orientation research

While tailored primary research can allow for specific research questions to be explored directly, administrative data collected for other purposes may present a useful source for secondary analysis. The Census and local and central government statistics and databases provide general population level indicators for factors such as poverty, social exclusion, and family composition. There is also a range of large-scale general population social surveys, some of which are conducted on a continuous or longitudinal bias. At present very few of these routine data sources include any measure of sexual orientation, identity or behaviour. There is also relatively little exploratory or methodological work underway assessing the viability of recording same-sex cohabitation or of asking sexual orientation of a general population sample. This Chapter explores the debate over whether a measure of sexual orientation should be included in general population data collection exercises; considers what methodological work has been conducted on the validity of asking sexual orientation and same-sex cohabitation in such contexts; and assesses a number of existing and potential sources of such data.

While official agencies in Scotland and the rest of Britain that produce statistics tend to collect data on gender, age, region of residence, ethnicity, working status and even income, very few ask questions about sexuality. Reynolds argues that by not accounting for sexual identity in demographic research,

"characterisations of minority, abnormality and pathology in non-heterosexual people are left unchallenged. The official statistical 'silence' on sexuality allows the perpetuation of the idea that sexual diversity, and prejudice and discrimination on the basis of sexuality, is a private trouble with no public issues or consequences… Heteronormativity represents that if 'they' do not appear in the statistics that describe the demographic contours of British society, 'we' do not have to acknowledge them or respond to their claims for the equality, rights and justice."

Reynolds, 2001

An absence of official data that can be disaggregated by sexual orientation could have concrete negative impacts when official data forms the basis for decisions about levels of public policy provision and recognition of legitimacy in areas such as pension entitlements, benefits eligibility, and partnership rights. However, it has also been argued that asking sexual orientation in administrative or other general population data collection exercises could provide results that would be subject to political abuse. Barry, for example, argues that if a question on sexual identity were included in a national survey or census, "in the likely circumstance in which a significant or majority proportion of the lesbian and gay community were not in a position to disclose their sexual orientation, such inaccurate quantitative data could then be used to reduce the level of recognition, funding of or service provision to that community."(Barry, 2000) Other related concerns raised against the inclusion of sexual orientation questions in administrative or general population data collection include issues of acceptability to respondents, perceived relevance, and privacy and intrusion.

4.2 Census data on sexual identity

A number of sexual orientation researchers and campaigners, whilst acknowledging the possible drawbacks, have called for the inclusion of a sexual orientation question on national census surveys. Some years ago Smailes argued that "Lesbian existence is generally ignored or made invisible by heterosexual society, so there is no way of knowing how many women identify as lesbian; perhaps the next census could include a question about sexuality as a basis to work from?"(1994) Likewise, Stonewall suggests that "the time has come" for the UK Census to at least collect (and report on) data on same-sex households.(1996) This section considers the most recent two Censuses carried out in the UK (1991 and 2001); the 1990 and 2000 US Census; and the 2001 Canadian and Irish Censuses. All of these now allow for same sex cohabitation to be recorded, but none include a measure of self-perceived sexual orientation or of transgender or transsexual status to be recorded, and not all of them actually analyse or make available data on same-sex cohabitation. The 2001 UK Gay and Lesbian Census, a large survey rather than a Census, will also be considered.

The UK Census 1991 and 2001

The User's Guide to the 1991 UK Census states that the Census edit programme changed the data given by respondents on the Census form, so that an entry of 'cohabiting as a couple' was actively forced to the 'unrelated' code if the two household members were of the same sex:

"The edit programme for the 1991 Census allowed only couples of opposite sexes to be categorised as 'cohabiting'; where same sex couples were recorded as 'living together as a couple' in the relationship question, this was changed to 'unrelated'."

The 1991 Census User's Guide, NS and GRO, 1991 ¹⁹

The Guide states that despite a campaign by some gay and lesbian organisations in the months preceding the 1991 Census for the (then) OPCS and GRO(S) to recognise same sex couples in the Census output, this was not done because "it was decided that a count of same sex couples would provide unreliable results." ²⁰ It is interesting to note that reasons of non-acceptance by the general population were not cited. In deciding whether to include a variable on the Census, various issues need to be considered including respondent burden, cost, usefulness, comprehension, acceptability, and whether alternative data sources exist. The very difficulty of obtaining a reliably representative sample of LGBT respondents for a survey sample is one persuasive reason why it would be beneficial to ask this on the Census.

It is important not to overestimate what Census data is able to provide, it can only give us basic percentages for a relatively small number of variables. However there is clearly plenty of scope for more to be done. Despite the enormous resistance to the inclusion of an ethnicity question on the Census in the 1991, once this was done a whole range of disadvantages and discriminations as experienced by minority ethnic groups could be identified and relevant policies designed in an attempt to tackle them.

In terms of what the 2001 Census covers: 'what is your sex?' excludes, as discussed in Chapter 2, a small but significant minority who are 'intersex', transgendered or transsexual. Sexuality itself is not referred to, but same sex cohabitation will be identifiable. National Statistics and GRO have confirmed that information on same sex couples will be available from the 2001 Census, although they do not have precise dates for the publication of this information yet.

The US Census 1990 and 2000

In 1990 the category of 'unmarried partner' was included in the US Census, allowing for 150,000 same sex partnerships to be recorded (an underestimation given the disqualification of same sex partners who described their relationship as 'spouse'). As for the UK Census, the likelihood of undercounting the LGBT population is a clear concern in the US regarding collecting data by this variable in the Census. Prior to the 2000 US Census the Institute for Gay and Lesbian Strategic Studies (IGLSS) and the Policy Institute of the National Gay and Lesbian Task Force (NGLTF), the two largest LGBT policy organisations in the US, launched a national 'Make your Family Count' campaign in March 2000 to encourage same sex couples to register their partnership. Paula Ettelbrick of the NGLTF argued "all public policy flows from the US Census… If we are not counted we lose out on federal funding for research, funding for community services and passage and implementation of laws that benefit our community."(2000) Smith and Gates published Gay and Lesbian Families in the United States: Same-sex Unmarried Partner Households, a preliminary analysis of 2000 United States census data, in 2001. While they note that the Census found that lesbian and gay families lived in 99.3% of all counties in the US, they also compared the data with surveys to illustrate the extent of the Census's undercount (as much as by 62%, they estimate).

Qualitative and cognitive Census development work exploring what are LGBT people's concerns about completing the Census, and how the format and layout of the Census might be improved to allay any fears about confidentiality, privacy or intrusion would be extremely valuable.

The 2001 Canadian Census

After successful Census tests were carried out in 1996 and 1998, the Canadian Census included a question on living in a 'same-sex common-law relationship' for the first time in 2001. This has become a particularly important issue in Canada since government legislation there extended common-law marital status to homosexual relationships. The 2001 Census Consultation Report describes the discussions that took place and the test studies that were conducted in designing how to collect data on same-sex partners. ²¹ The issue of enumeration of same-sex partners raised more comments than any other during the consultation period:

"It was continually pointed out that gays and lesbians must use a residual category to identify themselves as same-sex partners in the question on the relationship to the household reference person. Some gays and lesbians believe that this way of identifying themselves is not reliable. Changes made to the definitions used in the programs and legislation are the main justification given for enumerating same-sex partners clearly and precisely. With these changes, same-sex partners are now included in the definition of couples, beneficiaries, dependants, and so on."

2001 Census Consultation Report, StatsCan, 2001

Other issues discussed during the consultation included adding a question on sexual orientation and having additional response codes at the question on sex regarding transsexualism. Both these propositions were decided against but may well be discussed again when the consultation begins regarding the content of the next Canadian Census.

The Irish Census also now allows for people of the same sex and living together in a household to indicate that they are partners.

The Gay and Lesbian UK Census 2001

Administered by I.D. Research, a lesbian and gay market and social research company, the UK Gay and Lesbian Census with its achieved sample of 10,500 is described as "the largest survey ever of any lesbian and gay population in the world." ²² The first market and commercial report was published in 2002 and more socially orientated reports will be published throughout 2003 covering: values and beliefs, including voting behaviour; the impact of homophobia; family relationships and history; relationships; geographic mobility and migration patterns; coming out issues; social mixing patterns; lifestyle and spending patterns and health issues.

Lead researcher David Pinson claims that the UK GL Census will "enhance social and political power of the lesbian and gay population, as it will provide robust data that currently does not exist." However it is worth noting that this 'census' will not provide a random sample or comprehensive or representative cover of LGB respondents in the UK. It differs from the national Census in not being compulsory and having a self-selected sample of those identifying as lesbian or gay and aware that the survey is happening. In addition to the discussion in Chapter 2 about the definitions of sexual orientation used, it is worth noting here that individuals who have sex with people of the same sex, but do not self-perceive themselves as one of the sexual orientation categories given are not accounted for.

4.3 Other administrative or routine data sources

There are various national administrative databases available for the UK. These include the General Practice Research Database; Child Support Agency; Disabled Person's Tax Credit; Joint Unemployment and Vacancies Operating System (JUVOS); School Census; NHS Central register; New earnings Survey (Panel Data-set); Homeless Household case returns; Inter-Departmental Business Register; and Scottish Transport Statistics. ²³ While some of these databases include details of sex, country of birth, languages spoken, ethnicity, region of residence, and other characteristics which can make them valuable for research purposes, none include any measure of sexual orientation or same-sex cohabitation. Likewise, Barry's review of equality data in Ireland did not cite a single source of official data disagreggated by sexual orientation. However, several key data sources in Ireland are undergoing reviews of what classificatory data should be collected, and change in the situation in Ireland is likely to be forthcoming.

Health and clinic statistics are currently problematic for research, though new monitoring systems are being set-up. For example, while routine STI clinic data can include a measure of sexual orientation, it is problematic to use anonymised case based data for estimates of sexual orientation specific STI trends. Anonymised case based data means that an analyst is unable to distinguish between 10 individuals being diagnosed once from one being diagnosed ten times.

Statistics relating to crimes reported to the police provide both a potentially vital and yet highly problematic source of data for sexual orientation research. As Mason and Palmer point out, there has been no official statistics on homophobic crime or incidents.(1996) A number of secondary analysis studies have drawn on crime statistics from incident reports to the police in order to build up one picture of patterns of anti-gay violence (e.g. Kuehnle and Sullivan, 2001). Gadd also used statistical analysis of incident data recorded by the Scottish Police to explore domestic abuse as experienced by men.(2002) As part of a larger study, this data could be compared with qualitative and survey findings to explore who was least little to report to the police, and what types of incident the police were least likely to record. The use of official statistics in this context was able to enrich the empirical research data whilst at the same time the empirical research was able to expose the biases in reporting to the police and recording by the police.

The GLA London Partnerships Register provides another potential resource that may become more valuable in time: "I hope other cities will follow suit and that other organisations will accept it as proof of a relationship."(Ken Livingston, 2002) ²⁴

4.4 Continuous or large-scale surveys

Some of the key UK survey and longitudinal data sources are listed below (by no means a comprehensive list). ²⁵ Details in relation to each have been gathered either by talking directly with researchers currently working on the survey, through reference to the Question Bank ²⁶, or by going through published technical and substantive reports and other documentation. While each questionnaire as reproduced may indicate that it is technically possible for same sex co-habitations to be identified in the data (particularly where a full household grid is recorded), it is also possible that an 'edit programme' was run on the collected data, changing any answers considered to be contradictory or illogical. This occurred during data processing of the Census 1991, where same sex household members originally coded as 'cohabitees' were deliberately changed to being 'unrelated adults' (see comments on the UK Census above). However what this summary does allow us to gauge is that there appears to already be considerable data on same-sex co-habitation, across a diverse range of policy areas. Of all the studies listed here, only the Metropolitan Police Study and the Northern Ireland Life and Times Survey were identified as having asked respondents' self perceived sexual orientation. ²⁷

In the reporting of these surveys, the derivation of the cohabitation variable (i.e. whether or not same sex couples are actually included) is rarely defined. So, not only are same sex cohabiting couples not presented as a separate group, but it is not clear to the reader whether they are included in the general 'cohabitation' group either.

This table lists just some of the diverse continuous or large scale general population surveys currently being carried out in Britain. While only two have started to ask about sexual orientation, same sex cohabitation measures are potentially included in a majority. Continuous surveys are also able to provide time series trend data, of which there is currently little in the sexual orientation literature. Although there is likely to be an initial undercount of LGBT respondents and of same sex cohabitation partnerships identified, with careful methodological development and use of concealed response show cards or self-completion elements, this can in time be improved. This is likely to be difficult to interpret for some time to come however; Natsal found a substantial increase in reporting of same sex experience amongst women - it is almost impossible to disentangle, however, how much of this increase is a real change in behaviour and how much is an increased likelihood to report.

Although a staff rather than a general population survey, it is worth noting here that the recent large-scale Civil Service-Wide Diversity Study also included a question about self-identified sexual orientation (and transgender identity).(ORC International, 2002) The issues of confidentiality and acceptability, however, are very different for a work place study compared with the other surveys cited in this section, which mostly take place in private households.

4.5 Conclusion

Qualitative and cognitive Census development work exploring what are LGBT people's concerns about completing the Census, and how the format and layout of the Census might be improved to allay any fears about confidentiality, privacy or intrusion would be extremely valuable.

There is scope for detailed secondary analysis of existing large-scale general population surveys. This would focus on patterns of same sex cohabitation as very few large scale studies currently include a sexual orientation question. Amongst those surveys that do, the indication is that acceptability is high (quite low proportion refuse to answer, and these are primarily the oldest respondents). It is harder to assess how many LGBT respondents describe themselves as other than LGBT, however. Like ethnicity, once the message of the value of discrimination monitoring gets across, and that that is the purpose of asking the question, the proportion feeling able to respond honestly is likely to increase. Comprehension may be an issue though, anecdotal feedback from interviewers working on the Metropolitan Police Survey what that the main query made by respondents was 'what's heterosexuality?'.

There is also potential for starting up working groups of survey methodologists currently working on large-scale surveys to raise the issue of collecting data on sexual orientation, debate approaches and share experimental and methodological work being undertaken.