Developing survey questions on sexual identity:

UK experiences of administering survey

questions on sexual identity/orientation

Peter Betts

Data Collection Methodology – Social Surveys

Census and Social Methodology Division
Office for National Statistics
January 2008
[Blank page]
Contents
1 Executive Summary ...............................................................................................2
2 Introduction............................................................................................................6
2.1 Scope of the review........................................................................................7
2.2 Defining sexual identity or orientation ..........................................................7
2.3 Methodology ..................................................................................................8
2.4 Surveys reviewed and organisations involved...............................................8
3 Survey designs .......................................................................................................9
3.1 Sample designs and population coverage ....................................................10
3.2 Modes and methods of administration.........................................................10
4 Analysis................................................................................................................11
4.1 Drawing conclusions and confounding factors............................................11
4.2 Estimates from the surveys ..........................................................................12
4.3 Conceptual basis, purpose, context and location of question ......................13
4.3.1 Conceptual basis ..................................................................................13
4.3.2 Purpose and context .............................................................................14
4.3.3 Location in questionnaire.....................................................................14
4.4 Preambles to and explanations of the question ............................................15
4.5 Question wording and format ......................................................................16
4.5.1 Question Stem......................................................................................18
4.5.2 Response categories .............................................................................20
4.6 Missing data .................................................................................................24
4.6.1 Options not to answer the question......................................................25
4.6.2 ‘Don’t know’ responses .......................................................................26
4.6.3 Non-response and ethnic group ...........................................................26
4.7 Confidentiality, privacy and acceptability ...................................................27
4.7.1 Confidentiality pledges ........................................................................27
4.7.2 Providing privacy in the survey environment......................................28
4.7.3 Acceptability of the question and survey/item response......................29
4.8 Accuracy and completeness: honesty and social desirability ......................30
4.9 Interviewer attitudes to asking about sexual identity...................................32
4.9.1 Non-professional interviewers in communal establishments ..............32
4.9.2 Professional interviewers .....................................................................32
4.10 Longitudinal data collection ........................................................................33
4.11 Issues about asking sexual identity not addressed by the review ................33
5 References............................................................................................................34
6 Appendices...........................................................................................................34
Appendix A: Summary table – mode/sample design/sexual identity question
designs and estimates...............................................................................................36
Appendix B: Information about each survey included in the review ......................38

Tables
Table 1 Surveys reviewed and suppliers of information ...............................................9
Table 2 Minimum and maximum percentage rates for each substantive sexual identity
category and type of missing data, by survey mode ............................................12
Table 3 Question stems and response categories .........................................................17

1
1 Executive Summary

Introduction: ONS initiated a project, in 2006, to develop, test and evaluate a

question, or suite of questions, on self-perceived sexual identity for use on
government social surveys. As part of the exploratory phase a review was conducted
of the experiences of other organisations in the UK which have recently administered
survey questions about sexual ‘identity’ or ‘orientation’. The former term is used
generically in the review unless otherwise specified. It is difficult to determine
commonly agreed definitions of either term and there has been much research into the
concepts. The working premise of the project is to develop a question about
respondents’ self-perceived sexual identity (for example, as heterosexual, gay, lesbian
or bisexual), allowing them to take into consideration whichever dimensions of sexual
orientation (e.g. behaviour, attraction, fantasy, emotional and social preferences,
identification and lifestyle) they so choose in order to determine their sexual identity.
The feasibility of operationalising this intention in the social survey environment is
one of the main focal points of the project. The purpose of the review was to obtain
insight to inform subsequent stages of the project.

Methodology: Ten public sector surveys were reviewed1. Their target populations
covered people of all sexual identities, though in some cases were limited
geographically or to people with certain characteristics. Information was obtained by
semi-structured telephone interviews with researchers at sponsoring organisations or
research and fieldwork contractors, and by further correspondence and reference to
survey reports and other materials. Information collected included survey and sexual
identity question designs, estimates produced, and feedback from respondents and
interviewers. A content analysis of the information obtained was conducted. A
summary of information collected (sample designs, modes of administration, question
designs and estimates) is at Appendix A.

Modes of administration: Sexual identity questions were either administered by

interviewers, using CAPI or PAPI, or were self-administered by booklets placed
during face-to-face or postal surveys.

Analysis & drawing conclusions: The review provided useful insight, generated
hypotheses and identified areas for research through the remainder of the project.
However confounding factors, such as the small number of surveys reviewed,
differing survey and question designs, small sample sizes, the uncertainty as to the
robustness of estimates and the lack of a reliable benchmark of the proportion of the
population which is lesbian, gay or bisexual (LGB), sometimes made comparisons
difficult. In this review terms such as ‘the LGB population’ and ‘heterosexual

1
Northern Ireland Life and Times Survey (2005); Policing For London Survey (2000); Employees’
Awareness, Knowledge and Exercise of Employment Rights Survey (2005); Fair Treatment at Work
Survey (2005); The National Mental Health and Ethnicity Census 2005 Service User Survey (2005);
The National Mental Health and Learning Disability Ethnicity Census (2006 ); British Social Attitudes
Survey (2005); Newham Household Panel Survey (2002-6); Scottish Census Small Scale Test (2005);
Civil Service Diversity Survey (2001). While this review was conducted the Adult Psychiatric
Morbidity Survey (APMS, 2006-7) and Citizenship Survey (2007-8) introduced questions on sexual
identity into the field – limited commentary on these is included.

2
respondents’ are used as shorthand for people who selected particular response
categories.

Survey estimates: Estimates were obtained for most of the surveys. Rates of the
proportion of respondents self-identifying as LGB ranged from 0.3% to 3.0%, lower
than the government estimate of LGB people constituting 5% to 7% of the population.

Concept being measured: All surveys asked a single question to capture sexual
identity, rather than asking separate questions about its different dimensions. Little
information was provided about the conceptual basis of the questions, or what
respondents understood them to be asking, and on what basis their answers were
given. A respondent’s answer might vary according to which dimension(s) they are
considering. ONS’s project will include qualitative research with members of the
public into the concept to be measured. It will address the feasibility of capturing
sexual identity in a single question, in the context of government social surveys with
varying purposes.

Purpose, location and explanation of questions: Generally the purpose of the

questions was less to measure prevalence of sexual identities than for use in analysis
relating to domains where sexual identity might have a bearing (e.g. discrimination).
Questions were generally located among socio-demographic questions, usually toward
the end of interviewer-administered questionnaires but sometimes followed by other
modules in self-administered instruments. Explanations about the purpose of
questions, or about confidentiality, were generally not provided. Interviewer feedback
suggested attention should not be drawn to the question. The purpose and location of
the questioning, and what explanation should be provided to respondents, are topics
which will be covered in the ONS project.

Question wording and format:

Question stems: Question stems varied in whether the term ‘sexual identity’ (or
similar) was mentioned, or left the subject to be inferred from the response categories
by asking respondents which category ‘(best) described’ them. Definitions were
generally not included in question wording. Limited feedback on comprehension from
respondents suggested the need for some definition to be given of what is being
measured. This topic will be addressed in ONS’s project.

Response categories: The questions tended to provide ‘discrete’ or ‘absolute’

response categories, clearly separating heterosexual, homosexual and bisexual, rather
than a set which reflected the possibility that sexual identity might exist on a
continuum (e.g. ‘completely heterosexual, mainly heterosexual, bisexual, mainly
homosexual, completely homosexual’). The different approaches might have
implications for measuring the prevalence of the LGB population: for example,
whether those responding ‘mainly heterosexual’, implying some degree of LGB
orientation, should be included. However, identity might be more discrete than other
dimensions such as behaviour and attraction.

Only a few terms were used in the response categories, but there was great variety as
to which formal and colloquial alternatives were used and in which combinations: for
example, whether ‘heterosexual’ was used with or without ‘straight’, or ‘gay’ and
‘lesbian’ with or without ‘homosexual’. Definitions of categories were generally not

3
provided. Some respondent comprehension problems were apparent, particularly
regarding ‘heterosexual’. No effects on estimates resulting from the terms used or the
order of categories were obvious. The ONS qualitative work will investigate
understanding and use of terms, what are the appropriate response categories and the
need for definitions to be provided. The need for additional response categories, such
as ‘other’, ‘none of the above’ and alternatives to the categories commonly provided
will also be explored.

Missing data: All surveys recorded missing data, from item non-response (‘don’t
know’, refusal to an interviewer, or leaving a self-completion question unanswered),
or from response categories such as ‘do not wish to answer’, or from both. Higher
proportions of missing data tended to be found on self-administered surveys (totals
ranging from 1.4% to 25.0%) compared with interviewer-administered surveys (0.2%
to 9.0%). A possible issue was identified regarding potentially higher non-response
among minority ethnic groups due to cultural/religious beliefs which will be looked at
in the ONS development work.

Confidentiality: Surveys did not mention confidentiality at the point the question was
asked, though all gave general assurances elsewhere, for example in advance or
covering letters. Little information was gathered about respondents’ views on
confidentiality and fears of disclosure of their data to other parties; this will be
covered in the ONS research.

Privacy in the survey environment: There was little evidence of respondent concern
about privacy, that is, the interviewer or any other person present knowing which
answer had been given. On interviewer-administered surveys the common approach to
providing privacy and reducing embarrassment was use of concealed response show
cards, where respondents answered by giving a letter or number rather than stating
their answer in words. However no surveys prevented interviewers from knowing the
response category given by the respondent. Views on within-household privacy,
concurrent interviewing, and preferences for interviewer- or self-administered modes
will be explored in the ONS research.

Acceptability of the question:

Survey response
There was no evidence to suggest that survey response rates had been affected by the
inclusion of a question on sexual identity. Indeed one survey tested its questionnaire
with and without a sexual identity question to confirm this.

Item response
Higher rates of missing data were obtained when questions were administered by self-
completion compared with interviewer-administered mode. This suggested that
perhaps there was an issue of invasion of privacy or objection to being asked about
sexual identity. Feedback from one postal survey supported this assumption, although
it found that objecting to being asked and objecting to giving an answer did not
necessarily correspond. Such concerns were less evident when the question was
interviewer-administered, where missing data was lower probably due to the fact that
respondents tend to be more compliant in interviewer-administered surveys.

4
Accuracy and completeness of data: When producing benchmark data, it is
important to minimise missing data particularly when the characteristic of interest is
found in a small proportion of the population. However it is also important that the
accuracy of substantive responses (that is, to the sexual identity categories including
other, or similar) is not compromised. This could occur if respondents feel pressured
into providing an answer which is incorrect but which may be considered more
socially acceptable. Mode of administration is an important consideration in relation
to this.

LGB rates across the modes appeared to be more consistent than those for
heterosexual respondents. So it might be that LGB respondents tended to give correct
answers in either mode. However, it is not known whether a proportion of the LGB
population gave incorrect, socially desirable answers in interviewer-administered
mode, or avoided giving substantive answers in self-completion mode.

Estimates of the heterosexual population tended to be higher when the question was
interviewer-administered compared with self-administered questions. Respondents
would have been able to ask interviewers for help in understanding the question or
response categories, which might have reduced the amount of missing data. Another
possible reason is that a proportion of heterosexual respondents avoided giving
substantive answers to self-administered questions, but answered truthfully when
‘forced’ to answer on interviewer-administered surveys.

The questions raised above will be covered in the remaining research. Issues which
will be considered include:
• whether respondents should be allowed to avoid giving substantive answers, in
different modes, and, if so, how;
• whether respondents might give socially desirable answers, and what
characteristics they have;
• and the optimal balance between minimising missing data and maximising data
accuracy.

Interviewer attitudes: Experiences on some surveys led to the conclusion that

interviewer instruction should be given sufficient attention to avoid problems. Such
problems included interviewers on one survey being unwilling to ask the question,
and lay interviewers, in an establishment context, determining sexual identity by
observation, without asking the respondent. However, as far as the researchers spoken
to were aware, such problems were rare.

Issues not addressed: Some issues of relevance to the development of a question for
ONS and other surveys were not covered by the review. These included CASI or
telephone methods; administering the question face-to-face to more than one
household member; and proxy data. One survey had a longitudinal design, but little
was discovered about effects of asking the question repeatedly over time. These issues
will be addressed by the ONS project.

5
2 Introduction

There is an increasing requirement for data on sexual orientation or identity in order

to meet current and future legislative requirements. Sexual orientation is the term used
in the legislation but sexual identity is the preferred term used by the ONS. The key
users of this data will include central government departments; local government;
public service providers (e.g. police & health authorities); lesbian, gay and bisexual
(LGB) service providers; the LGB population; academia and other research
organisations.

A project was initiated that aimed to provide advice on best practice with regard to
data collection in this field, and also examine the feasibility of providing benchmark
data. The primary outputs from this project will be a question, or suite of questions,
asking people to self-identify to a particular sexual orientation, along with advice on
administration. Alongside the question(s), a user guide will be produced discussing
the conceptual issues as well as the methodological issues, such as context and mode
effects. General information on the project is available at:
http://www.statistics.gov.uk/about/data/measuring-equality/sexual-identity/default.asp
The methodology is described in Wilmot (2007).

The impact of asking a question on sexual identity on an ONS multi-purpose social

survey such as the new Integrated Household Survey (IHS) is unknown. It is possible
that some respondents may be unhappy about being asked for what could be
considered very personal and sensitive information. This may well impact on the
response to the survey overall or to the individual question. Moreover it may impact
on willingness to be contacted for further research which would have implications
were the question to be included on a longitudinal survey.

There may be issues relating to the administration of this question amongst particular
groups in society. For example, people’s cultural, religious or political beliefs may be
offended. The age of the respondent may also relate to their ability to answer or their
attitudes towards such questioning. Examining personal barriers to response would
enable us to suggest how the methodology could be improved. Respondents’ ability to
comprehend the questions and answers has clear implications for the quality of the
data. Furthermore, the comprehension and attitudes of the interviewers is important
since they are often called upon to justify the inclusion of a particular topic in a
survey.

Data Collection Methodology (DCM) branch was commissioned to develop, test and
evaluate question(s) on sexual identity for implementation in a government social
survey context, with particular reference to the IHS. This report forms part of the
initial exploratory phase of the project.

As part of the exploratory phase of the project, ONS conducted a review of the
experiences of other organisations in the UK which have administered survey
questions about ‘sexual identity’ or ‘sexual orientation’, reported in this document.
The purpose of the review was to obtain insight which might be useful in informing

6
subsequent stages of the project, including qualitative exploration of the issues of
interest with the general public and the design and testing of a question, or questions,
for use on social surveys.

2.1 Scope of the review

The review included UK organisations in central and local government and the private
sector which have recently administered questions on respondents’ sexual identity, to
survey populations inclusive of people of all sexual identities. In some cases coverage
was limited geographically or to people with particular characteristics.

The review excluded the following areas.

• Questions specifically about sexual behaviour and attraction, as the focus of the
development project is on self-perceived sexual identity. (However, questions
were included where the precise concept was not specified in the question
stem).
• Surveys conducted with only the lesbian, gay and bisexual (LGB) community
were excluded, as the project is aiming to develop a question for use with the
general population.
• Surveys prior to 2000, because information collected more than a few years ago
might have become out of date, with regard to the general public’s
conceptualisation of sexual identity, language used and understood and views
on the acceptability of such questions.

A less detailed review of sexual identity questions asked in other countries was
conducted at the same time – this is reported separately (Taylor [a]).

2.2 Defining sexual identity or orientation

Note that in this review the term ‘sexual identity’ is used, and generally also
encompasses ‘sexual orientation’. However, it is difficult to determine a commonly
agreed definition of either term. The surveys reviewed here referred to, variously,
‘sexual orientation’, ‘sexuality’ or were not explicit about which concept they were
measuring. There has been much research into the concepts of sexual identity and
orientation, with regard to different dimensions such as sexual attraction, behaviour,
fantasies, emotional and social preferences, identification and lifestyle, and whether
they can/should be measured individually or if it is possible to capture sexual
identity/orientation with a single question (summarised in McManus). The working
premise of the project is that a question is developed which should ask about
respondents’ self-perceived sexual identity, allowing them to take into consideration
whichever dimensions they so choose in order to determine their sexual identity. The
intention is that sexual identity will not be defined for them as only being about a
particular dimension (e.g. behaviour), or prescribe a formula to derive the overall
identity. This multi-dimensional approach is informed by models of sexual orientation
such as the Klein Grid and the Sell Assessment2. The feasibility of operationalising
this intention in the social survey environment is one of the main focal points of the
project.

2
See http://www.gaydata.org/ms001_Index.html

7
2.3 Methodology
Data for the review was collected primarily by means of semi-structured telephone
interviews conducted with project managers or others with relevant knowledge (at
either the sponsoring organisation or the survey design/fieldwork contractor). Further
information was provided by email. In addition, research and technical reports,
questionnaires and other field materials, and data estimates were all referred to.

In Appendix B detailed information is provided about each survey, obtained from the
interviews and documents. General information, to provide context for comparison
and assessment of the sexual identity question, includes when it was conducted, who
the sponsor was, the sample design and mode of administration and survey response.
Specific information about the sexual identity question includes the question wording
and response categories; method of administration; how refusals and other missing
information were recorded; estimates (where publicly available); information on
instructions and training given to interviewers; and feedback from respondents and
interviewers on the experience of administering a sexual identity question (e.g.
comprehension and acceptability).

Much of the information, including estimates from the questions where available, is
summarised in a table at Appendix A, to enable easier comparison between surveys
and modes.

Information for some surveys is incomplete, because it was not available, or because
consultation could not always be with researchers involved day to day in the projects,
or because not all the topics could be covered with each organisation. Estimates were
not available for every survey reviewed. Thanks go to the suppliers of the information
at the organisations acknowledged in Table 1 below.

A content analysis of the information was then conducted.

2.4 Surveys reviewed and organisations involved

Information was collected about all but one of the surveys falling in the scope of the
review of which ONS was aware. Ten surveys were included in the review, listed in
the table below. The year of reference for each survey reviewed is given, either one-
off or the most recent full year.

8
Table 1 Surveys reviewed and suppliers of information

Survey or data collection (year) Information/feedback suppliers

Northern Ireland Life and Times Survey Queen’s University, Belfast - survey
(2005) sponsor
Policing For London Survey (2000) National Centre for Social Research
(NatCen) - survey contractor;
Prof. Mike Hough (formerly South Bank
University) – research contractor
Employees’ Awareness, Knowledge and Department for Trade and Industry (DTI)
Exercise of Employment Rights Survey - survey sponsor
(2005)
Fair Treatment at Work Survey (2005) Department for Trade and Industry (DTI)
- survey sponsor
The National Mental Health and NatCen – survey design consultant
Ethnicity Census 2005 Service User
Survey (2005)
The National Mental Health and Learning The Mental Health Act Commission –
Disability Ethnicity Census (2006 ) survey sponsor
British Social Attitudes Survey (2005) NatCen – survey contractor
Newham Household Panel Survey (2002- London Borough of Newham - survey
6) sponsor
Scottish Census Small Scale Test (2005) Information obtained from ‘Sexual
Orientation in the Census’ report by GRO
Scotland
Civil Service Diversity Survey (2001) ORC International - survey contractor

Note that during the period when this review was conducted the Adult Psychiatric
Morbidity Survey (APMS, 2006-7) and Citizenship Survey (2007-8) introduced
questions on sexual identity into the field. Some information was provided by,
respectively, NatCen (the survey contractor) and Communities and Local Government
(the survey sponsor), though no information was available regarding estimates and
little feedback from the field. Generally they are not included in the survey
descriptions and analysis below, except when referred to explicitly.

As part of the development project, in 2006 ONS conducted experiments asking about
sexual identity on the National Statistics Omnibus Survey. A separate report on this
work has been written (Taylor [b]).

3 Survey designs

This section provides some contextual information on the surveys’ sample designs
and modes of administering the sexual identity question, both of which were varied.

9
3.1 Sample designs and population coverage
Seven surveys were conducted in private households, while three were in institutions
or establishments.

Eight of the surveys used random probability sampling. The exceptions were the Fair
Treatment at Work Survey which was included on an omnibus survey employing
quota sampling, and the National Mental Health and Learning Disability Ethnicity
Census (2006) which was a census of inpatients in mental health and learning
disability establishments.

Five of the surveys sampled the general population. They varied geographically as to
whether they covered Great Britain, particular countries within the UK, London, or a
local authority area. Only one general population survey, British Social Attitudes
Survey, covered Great Britain as a whole, and it should be noted that its sexual
identity question was not intended to measure the prevalence of the LGB population.

The two DTI surveys were limited to people of working age who were currently
working as employees or had worked within the last two years. The Civil Service
Diversity Survey sampled civil servants in all departments. The National Mental
Health and Ethnicity Census 2005 Service User Survey (2005) and the National
Mental Health and Learning Disability Ethnicity Census (2006) covered in-patients in
NHS and private/voluntary mental health and learning disability establishments.

The ages of respondents included by the different surveys varied from 15 and over, to
16 and over, and 18 and over. All household surveys selected one person to respond,
except the Newham Household Panel Survey which interviewed all household adults.
Among the institutional/establishment surveys, two sampled individuals while the
third was a census.

3.2 Modes and methods of administration

Predominantly, two modes of administration were applied for the sexual identity
questions, with different methods used within each mode, as follows:
• interviewer-administered mode:
o CAPI;
o PAPI;
• self-administered mode:
o self completion booklet placed during a face-to-face interview;
o postal survey.

Computer assisted self-completion (CASI or ACASI) was not generally employed.

The single exception was the Fair Treatment at Work Survey pilot, which
experimented with self-completion using a ‘CAPI pen’ in some cases, though the
main mode was interviewer-administered.

No surveys employed telephone methods, so no information was obtained which

might inform the potential administration on ONS surveys conducted this way.

Concurrent personal interviewing was not conducted on any of the surveys for the
sexual identity question (the Newham Household Panel Survey utilised paper self-
completion). Thus no experience of the data collection environment used on the

10
majority of ONS’s social surveys could be drawn on to inform the design of ONS’s
questioning.

4 Analysis

4.1 Drawing conclusions and confounding factors

In this section various aspects of the design of the questions are described and
compared across the surveys reviewed. The effect of the design on the estimates is
explored.

Analysis of the information collected provided insight which will be used to inform
the design of the questioning for use on UK government social surveys, and the topics
to be covered in the qualitative research being conducted as part of the development
process. It also generated hypotheses for further testing both qualitatively and
quantitatively. Areas for further research are identified throughout the review.

Comparisons were sometimes difficult to make between the surveys, and conclusions
were hard to draw about the effect of mode, question design, location or context on
the estimates. A number of confounding factors needed to be taken into account. The
number of surveys included in the review was small, overall and for each mode of
administration. The sample designs and survey populations differed. The question
designs also varied, such as in the ways in which respondents could avoid providing
their sexual identity, or answer ‘other’, ‘don’t know’ or ‘do not wish to
answer’/‘prefer to not say’. The robustness of the estimates obtained was in some
cases perhaps doubtful, due to small sample sizes and/or low response rates (both unit
and item), and therefore subject potentially to sample variation and/or non-response
bias.

Furthermore, lack of a reliable benchmark made analysis of estimates difficult. For

comparative purposes the working assumption used was that nationally lesbian, gay
and bisexual people comprise 5% to 7% of the population (the figure used by
government and thought reasonable by Stonewall)3. However this figure should
perhaps be treated with caution due to it encompassing, variously, dimensions of
sexual identity/orientation – self-identification, behaviour and desire - and having
been derived from several sources and countries.

In this review terms such as ‘the LGB population’ and ‘heterosexual respondents’ are
used as shorthand for people who selected particular response categories.

3
The report Final Regulatory Impact Assessment: Civil Partnership Act 2004 (available at
http://www.dti.gov.uk/files/file23829.pdf) states: “Whilst no specific data is available, a wide range
of research suggests that lesbian, gay and bisexual people constitute 5-7% of the total adult
population…” The figure was based on the findings in a number of studies, from various countries,
conducted among differing sample populations and measuring different dimensions of sexual identity.
A footnote acknowledges that there is very little reliable data about the size of the LGB population.

11
4.2 Estimates from the surveys
The estimates from the sexual identity question were obtained for most of the surveys.
Weighted data was supplied in most cases. The percentages are shown in detail in the
table at Appendix A. Estimates were not available for two of the ten surveys;
however, these were the two with mental health inpatients, which were the least
typical surveys in terms of sample and design, so estimates from them would,
perhaps, be misleading anyway.

In this section, the estimates are discussed at a general level, while in following
sections more detailed consideration is given to the relationship of estimates to
particular features of the question design.

For those surveys where estimates were available, the table below shows the
minimum and maximum percentage rates for each substantive sexual identity
category and type of missing value. For the purpose of this review, ‘substantive’
categories means those covering the different sexual identities (including ‘other’ or
similar), while ‘missing data’ refers to responses from categories presented to
respondents such as ‘do not wish to answer’, as well as don’t knows and refusals.

As will be described in Sections 4.5 and 4.6 the category breakdowns and labels
varied across surveys – in the table below they are consolidated. And the types of
missing values did not all apply to each survey.

The category percentages are based on total cases, not just valid cases. The ranges are
shown for all surveys, and separately for interviewer-administered and self-
administered.

Table 2 Minimum and maximum percentage rates for each substantive sexual identity
category and type of missing data, by survey mode

Substantive categories Missing data

Survey Hetero- Lesbian/ Bisexual LGB Other/ All Don’t Do not Ref Blank All
sexual Gay total none subst. know wish to missing
mode
answer
All Min 70.4 0.0 0.2 0.3 0.0 75.0 1.6 1.3 0.2 1.4 0.2
Max 98.9 2.0 2.0 3.0 4.7 99.8 1.6 15.0 9.0 10.0 25.0
Int. - Min 87.8 0.5 0.2 0.9 0.0 90.9 1.6 1.3 0.2 n/a 0.2
admin Max 98.9 2.0 1.2 2.4 0.7 99.8 1.6 1.3 9.0 9.0
Self- Min 70.4 0.0 0.3 0.3 0.9 75.0 n/a 8.5 n/a 1.4 1.4
admin Max 95.3 1.0 2.0 3.0 4.7 98.6 15.0 10.0 25.0

The estimates had to be treated with caution, since most of the surveys were limited in
their population or geographical coverage, and might be subject to sample design
variation. The sexual identity question on the only GB-wide general population
survey was not intended to measure the prevalence of LGB people.

That said, the main observation is that rates of lesbian, gay and bisexual respondents,
combined, ranged from 0.3% to 3%. This is lower than the 5% to 7% government
estimate.

The surveys which recorded the highest rates of LGB respondents were not of the
general population. They might reflect different compositions, with regard to sexual

12
identity, among the specific populations sampled (civil servants and people currently
or recently employed) compared with the general population.

The highest estimate of LGB respondents among any of the general population
surveys was 2.1%.

The estimates varied as to whether prevalence of lesbian & gay (combined) was
higher than bisexual, or vice versa, though lesbian & gay tended to be higher.

Not all the researchers contacted commented on their estimates. But among them were
some who acknowledged that their own figures were unreliable or underestimated the
LGB population – in the words of one, “by any reasonable standards the results are
pretty hopeless…baldly unsatisfactory”. The report on the Scottish Census Small
Scale test questioned the “accuracy” and “utility” of the data, due to high unit and
item non-response.

A hypothesis put forward by researchers on one survey, comparing their estimate with
the benchmark, was that the clustered sample design might have led to under-
representation of the LGB population, if they were concentrated in particular
geographical areas. Whilst factors including survey coverage, sample design and non-
response might explain the low estimates of LGB respondents to some degree, albeit
that the benchmark against which they are being compared is of uncertain reliability,
the findings of the review do provide some useful insight into the effects of the
question design and mode of administration.

The analysis posed a question about the balance to be achieved between, on one hand,
reducing missing values and obtaining as much substantive data as possible, and, on
the other hand, the accuracy of the substantive data. A key question is whether
accuracy is compromised when the question is interviewer-administered, by, for
example, respondents giving more socially desirable answers. This subject is
discussed in section 4.8.

4.3 Conceptual basis, purpose, context and location of question

4.3.1 Conceptual basis

Little information was ascertained about the conceptual basis of the questions, that is,
what precisely they were intended to measure. The surveys apparently all took a
‘lumper’ approach to collecting information on sexual identity (McManus). That is,
they only asked one question, rather than taking a ‘splitter’ approach, asking separate
questions about different dimensions. The questions were generally non-specific
about which dimension(s) of sexual identity/orientation was or were being measured.
Little information was available as to what respondents understood the question to be
asking (see later sections). (Note that the Adult Psychiatric Morbidity Survey, at time
of writing still in the field, asks separate questions about orientation/identity and
behaviour).

Further research
ONS’s programme of development work will be addressing the concept which is
intended to be measured and whether it can be captured in a single question or

13
requires more than one (see also sections 4.4. and 4.5). The work will consider the
points of view of both members of the public and the analysts and users of the data. If
more than one question is ideally required, the feasibility and appropriateness of
asking more than one question in the context of government social surveys with
different purposes will be considered.

4.3.2 Purpose and context

The purpose of the sexual identity question on the surveys reviewed was primarily for
use as an analysis variable, not to measure prevalence of different sexual identities.
The question was usually considered to be a socio-demographic question, so that
experiences and attitudes measured at other questions could be compared between
sexual identities.

All surveys had at least one question about a subject where a person’s sexual identity
could have a bearing, e.g. experience of discrimination, service provision, diversity
issues or social attitudes. The British Social Attitudes Survey question was used to
identify respondents with particular identities for subsequent questions.

Questions tended not to include any mention of the purpose of asking (see section 4.4
on preambles and explanations below). Little was discovered about the public’s
perceptions of the purpose of asking about sexual identity. Focus group research
conducted during the development of one survey indicated that people accepted the
use of the question in monitoring diversity and unfairness in the workplace.

A different purpose applied to the Scottish Census Small Scale Test, which was to
assess the feasibility of asking such a question on the 2011 Census in Scotland.

Due to the confounding factors mentioned above it was difficult to assess what effect,
if any, the perceived purpose and context of the questions had on the data quality (e.g.
how carefully respondents considered the question), or on non-response (whether unit
or item, due, for example, to sensitivity over subject matter or data confidentiality).

4.3.3 Location in questionnaire

No conclusions could be drawn about the effect on item non-response or estimates of
the sexual identity question’s location in the questionnaire. The question was
generally included among socio-demographic questions (along with, for example, age,
marital status, ethnicity, religion, employment status, income). In interviewer-
administered surveys it was the final or penultimate question, suggesting caution on
the part of researchers with regard to the possible effect on item or unit non-response,
as was expressed by one researcher: “…we put the question last so that it wouldn’t put
people off. We feared that they might worry what was coming next”. In self-
administered surveys the question was sometimes followed by other topics, perhaps
suggesting less concern due to the fact that respondents could skip the question.

Further research
The purpose, context and location of a sexual identity question will be discussed with
participants in ONS’s planned qualitative research.

14
4.4 Preambles to and explanations of the question
Explanations specifically relating to the sexual identity question – whether about its
purpose or to provide confidentiality assurances - were generally not provided. There
was no conclusive evidence as to the relative benefits and drawbacks of a prescribed
preamble, or an optional explanation, to the sexual identity question. The limited
feedback available was from interviewers’ perspectives, not respondents’.

Only one organisation included a preamble or explanation immediately before the

question, in their current or most recently used instruments. The Citizenship Survey,
in the field at the time of writing, included this preamble:
“The next question is about sexual identity. We are asking this question
because the government department funding this study, Communities and
Local Government, is responsible for helping to reduce all forms of prejudice
and discrimination in society”

A further clause on confidentiality (“As with all our questions, your answers will be
kept strictly confidential”) was included in the question which was cognitively tested
but dropped prior to the survey going live. It was dropped because interviewers felt it
was unnecessary and made the question more of a “big deal”.

Separate explanation, was provided to respondents on the National Mental Health

...Census (2006), in a booklet to respondents called ‘Your questions answered’. It
included the following question:
“Why are you asking me about sexual orientation? This is really very
personal.”

The response explained that:

• LGB people sometimes don’t receive the same standard of care and have more
mental health problems than other people, so the Mental Health Act Commission
wants to know how to improve services. But it doesn’t know how many LGB
patients there are.
• The question was voluntary.

Two examples were given of a preamble being dropped after initial use on
interviewer-administered surveys. The preambles addressed confidentiality and/or
attempted to explain terminology.

One preamble had been used on a pilot of the survey:

“The last question in this section is about sexual orientation - that is whether
someone is gay or lesbian, heterosexual or bisexual. As with all questions on
the survey you do not have to answer this if you do not want to, but all
answers will be treated in the strictest confidence.”

It was dropped in response to interviewer feedback, that it drew attention to the

question and embarrassed respondents.

Another preamble was dropped after being used in the field for some time. This
preamble explained that
“Some people have a partner of the same sex as themselves (homosexual)
while others have a partner of the opposite sex (heterosexual).”

15
The reason for it being dropped was not ascertained. This preamble, being behaviour-
related, would not be appropriate for a question about self-perceived sexual identity.

In an alternative approach to a standard preamble, the two DTI surveys provided an

interviewer prompt for use if respondents objected to or queried the purpose of the
question:
“We’re collecting this information to find out more about the discrimination
people may face for different reasons. The information is kept completely
anonymous.”

It is not known how often it was used by interviewers or its reception by respondents.

Confidentiality and anonymity regarding data respondents provided more generally

was addressed in various ways across the surveys: in advance letters, cover letters, on
the front of self completion booklets and at the beginning of demographic questions.

Further research
The necessity for a preamble or explanation, and the dimensions which should be
covered, is an issue for exploration in the exploratory and main stages of ONS’s
research. That is, whether respondents require explanation of the question’s meaning
and language to aid comprehension and improve accuracy, or of its purpose and
assurances of confidentiality to maximise unit response. Or, on the other hand,
whether to mention these things draws respondents’ attention to its potential
sensitivity which would otherwise not occur to them. Omnibus Survey interviewer
feedback will also be pertinent.

4.5 Question wording and format

The table below lists the question stems and response categories for the data
collections reviewed. It also includes those used on the Adult Psychiatric Morbidity
Survey (two versions were used in a split sample comparison) and the Citizenship
Survey.

16
Table 3 Question stems and response categories

Survey Question stem Response categories

Northern Ireland Can you tell me which of I am 'gay' or 'lesbian' (homosexual)
Life And Times these best describes you? I am heterosexual or 'straight'
Survey Please just give me the I am bi-sexual
number on the card. Other
I do not wish to answer this question
Policing For Please choose a letter from K - completely heterosexual
London Survey this card which best describes D - mainly heterosexual
how you would think of I - bisexual
yourself? R - mainly gay or lesbian
M - completely gay or lesbian
Employee’s Which of these best describes 1. Straight/heterosexual
Awareness...Survey you? - please just give me the 2. Gay/lesbian/homosexual
number on the card. 3. Bisexual
4. Other
Fair Treatment At Please look at this card. R. Straight or heterosexual
Work Survey Which of the answers on the D. Gay or lesbian or homosexual
card best describes you? P. Bisexual
Please just give me the letter H. Other
alongside the appropriate
answer.
National Mental Which of the categories on M. Heterosexual (like people of the opposite
Health...Survey this card would you say sex)
(2005) describes your sexual Z. Gay male or lesbian female (like people of
orientation? Please give me the same sex)
the letter. P. Bisexual (like people of both sexes)
C. Transgender
National Mental Which of the following terms 2. Lesbian/Gay/Homosexual
Health...Census would you use to describe 1. Straight/Heterosexual
(2006) your sexual orientation? 3. Bisexual
4. Other
5. Do not wish to answer
British Social Which of the following best 1. Heterosexual (‘straight’)
Attitudes Survey describes how you think of 2. Gay
yourself? PLEASE TICK 3. Lesbian
ONE BOX ONLY 4. Bisexual
5. Transsexual
8. Can’t choose
Newham Please tell us what best I am heterosexual or 'straight'
Household Panel describes you. I am 'gay' or 'lesbian' (homosexual)
Survey I am bisexual
If none of the above applies. (PLEASE
WRITE):
I am______________________
I do not wish to answer this question
Scottish Census Which of the following best Heterosexual
Small Scale Test describes your sexual Lesbian/gay
orientation? Bi-sexual
Other
Prefer not to answer
Civil Service If you have no objections to Bisexual
Diversity Survey stating your sexuality, please Gay man
could you state it here? Heterosexual
Lesbian

17
Adult Psychiatric Which statement best Entirely heterosexual (attracted to persons of
Morbidity Survey describes your sexual the opposite sex)
(Version A) orientation? Mostly heterosexual, some homosexual
This means sexual feelings, feelings
whether or not you have had Bisexual (equally attracted to men and women)
any sexual partners. Mostly homosexual, some heterosexual
feelings)
Entirely homosexual (attracted to persons of
the same sex)
Other
Adult Psychiatric Please choose the answer Completely heterosexual
Morbidity Survey below that best describes how Mainly heterosexual
(Version B) you Bisexual
Think of yourself...: Mainly gay or lesbian
Completely gay or lesbian
Other

Citizenship Survey Looking at this card, which of W …Heterosexual or straight

the options best describes P … Gay or lesbian
your sexual identity? Please H … Bisexual
just read out the letter next to S …Other
the description. G … [or would you] Prefer not to say?

4.5.1 Question Stem

Little information was obtained on the provenance of the questions, or their
conceptual development. Instances were cited of questions having been taken from or
adapted from other surveys, being provided by the client, and being designed by the
survey consultant/contractor.

It was difficult to assess the effect of different question stem wordings on estimates –
even between those which did and did not mention ‘orientation’ or ‘sexuality’ - since
there were too many confounding factors.

Mention of the concept being measured

The general approach to the question stem wording was one which perhaps avoided
addressing the inherent difficulties and subtle nuances related to defining sexual
identity. The intended concept being asked about - a definition or specific dimension
of sexual identity or orientation - was not usually explicit in the question stem
wording and generally not known to the researchers contacted, beyond it being
respondents’ self perception.

The majority of questions did not refer explicitly to the subject in the stem, leaving it
to the response categories to convey it. Of those which did mention the subject, the
reference was either to ‘sexual orientation’ or ‘sexuality’. (The Citizenship Survey
referred to ‘sexual identity’).

Definition provided
Only the Adult Psychiatric Morbidity Survey provided any definition of the
question’s subject, one of the two variants explaining that ‘sexual orientation’ meant
“sexual feelings, whether or not you have had any sexual partners”. Such a definition

18
of sexual orientation was more limited than that which the ONS is intending to
measure (see section 2.2).

Acknowledgement of the potential complexity in determining sexual identity

The questions generally used variants on a theme, asking respondents to choose a
category which ‘described’ or ‘best described’ themselves. Use of ‘best’ suggests an
acknowledgement by the designers of the complexity of the subject, and the
possibility of sexual identity being fluid (e.g. because it is changeable over time or in
different contexts, or because of ‘inconsistency’ between different dimensions) or not
easy to determine.

One exception asked respondents to ‘state’ their sexual orientation; the designer of
this question perhaps expected that answers would be more ready-made or concrete.
The latter approach is more similar to that often used in other countries (and on the
second NS Omnibus trial). This approach asks ‘Do you consider yourself to be…’
which perhaps anticipates a more self-assertive assessment compared to the passivity
of a category describing them.

Perspective
As for whose point of view should be provided, questions varied as to whether they
made it clear that it was respondents’ own perception (e.g.‘…how you think of
yourself’) or left it more open to interpretation (e.g. ‘…best describes you’).

Respondent comprehension
What respondents understood the question to be asking about is uncertain. The
evidence, from a limited amount of feedback from cognitive testing and written
comments by respondents to one postal survey, suggests the need for some definition
of what is being measured to be given.

Comments from respondents to the Scottish Census Small Scale Test included those
that "suggested they had possibly misunderstood or misinterpreted the question".
These included an elderly woman who answered the question with a sexual
orientation but did not think the question applied to her, and another who had to think
what the question meant but answered as best she could.

In the cognitive testing of the Citizenship Survey question it was found that
respondents understanding of it included “who you like as a partner” and “it’s what
you are or your sexual preference is” . These comments suggest respondents were
considering behavioural and/or identity-related dimensions, and perhaps not
consistently.

Further research
The conceptualisation of sexual identity and the need or otherwise to make reference
to it in the question stem will be explored in the ONS research. The need to provide a
definition, and if so what it should be, will also be addressed. So will the need to
make clear the perspective to be taken – that is, the respondent’s self-perception rather
than it being left more ambivalent. Whether it is better to ask if respondents consider
themselves to have a particular sexual identity, implying they have considered it
previously, or to choose a category which best describes them, which allows for them
not having given it much thought, will also be assessed.

19
4.5.2 Response categories

Discrete identities v scale or continuum

All but two of questions listed in Table 3 used categories which provided ‘discrete’ or
‘absolute’ sexual identities - that is, there was clear separation between heterosexual,
homosexual and bisexual - rather than a scaled set which reflected the possibility that
sexual identity might exist on a continuum.

One of the exceptions, the Policing for London Survey, included categories from
‘completely heterosexual’, through ‘mainly heterosexual’, ‘bisexual’, ‘mainly gay or
lesbian’ to ‘completely gay or lesbian’. While self-perceived sexual identity might
indeed be located on such a continuum, such categories might imply that behaviour or
desire was being asked about, without the question stem being specific. Indeed on the
other exception, the Adult Psychiatric Morbidity Survey, the question stem did
include a definition of sexual identity, as ‘sexual feelings’. It also used slightly
different vague quantifiers, substituting ‘entirely’ for ‘completely’ and ‘mostly’ for
‘mainly’, though which, if either, is better is probably unimportant.

Again the confounding factors make it difficult to compare the estimates from discrete
or ‘continuous’ [sic] categories and draw any conclusions. From an analytical
perspective, though, some interesting questions arise, assuming that it is possible to
capture sexual identity with a single question.

Firstly, would the 0.4% of respondents choosing the ‘mainly’ gay/lesbian category
have placed themselves in the gay/lesbian category, the bisexual category or another
category, if given absolute choices? The importance of this question perhaps is
dependent on the extent to which, analytically, bisexual respondents need to be
distinguishable from gay and lesbian respondents.

Secondly, using less absolute categories would have implications for measuring the
prevalence of LGB people in the population. For example, should the total LGB
population include respondents in the ‘mainly heterosexual’ category, which implies
some degree of LGB orientation? On the Policing for London Survey, the LGB rate
was 1% when including the bisexual and ‘completely’ or ‘mainly’ gay/lesbian
categories. If the ‘mainly heterosexual’ category, chosen by 2.3% of respondents, was
included, the total LGB proportion thus increased to 3.3%. However, it is not known
on which dimension(s) of sexual identity respondents were basing their answers.
Identity might be more discrete than other dimensions such as behaviour and
attraction. Making the question wording more specific would be likely to have an
affect.

Furthermore, all surveys allowed only one response option to be chosen, rather than
the possibility of respondents considering themselves to have more than one sexual
identity simultaneously (which might, in theory, be the case in different contexts – for
example, domestic, work, social – and depending on which dimensions of sexual
identity – behaviour, attraction, lifestyle etc - they wished to consider).

20
Further research
The issue of whether people consider their sexual identity to be absolute or more
qualified, and how they would respond if given absolute categories, will be covered in
the qualitative research exploring the conceptualisation of sexual identity. Whether
people consider themselves to have a single sexual identity (consistent across all
dimensions, or considered to be the ‘overall’ identity when there are differences
between particular dimensions) or more multiple, fluid identities from which they
cannot derive an overall identity, will be another topic.

Terminology used in categories

Regarding the terms used for different sexual identities, the same few ‘formal’ or
scientific/medical terms (e.g. ‘heterosexual’, ‘homosexual’) and ‘colloquial’ terms
(e.g. ‘straight’, ‘gay’, ‘lesbian’) were used across all the surveys. But there was great
variety as to which alternatives were used and in which combinations.

‘Heterosexual’ was used both on its own, and in combination with ‘straight’, but the
latter was never used on its own. No other term for heterosexual was used.

‘Gay’ and ‘lesbian’ were used both with and without ‘homosexual’. No other terms
for homosexual were used. ‘Homosexual’ was used alone only in the Adult
Psychiatric Morbidity Survey.

No alternative term to bisexual was used.

When both formal and colloquial terms were used in the same category, the order
varied for ‘heterosexual’ and ‘straight’, but ‘gay’ and ‘lesbian’ always preceded
‘homosexual’. There were instances where the second term used in a category was in
parentheses, e.g. ‘heterosexual (straight)’. The colloquial terms were sometimes given
in inverted commas, e.g. ‘straight’, perhaps to convey that they were less formal.

There were instances where a colloquial term was included for some categories in the
set but not for others; for example, a category for ‘heterosexual’ (without reference to
straight) and one for ‘gay/lesbian’ (without reference to homosexual).

There did not appear to be any meaningful difference in the estimates obtained from
questions which could be put down to using particular variants of gay, lesbian and
homosexual.

‘Gay’ and ‘lesbian’ were usually combined in a single category. There were two
instances of them as separate categories. There appeared to be little difference in the
estimates of homosexual respondents obtained from separate or combined categories,
though the confounding factors make this difficult to assess. Since respondent’s sex is
generally collected by surveys, there would seem to be no need to distinguish between
homosexual men and women, unless the ONS qualitative research reveals either or
both groups to strongly object to being together in a single category.

It is also difficult to disentangle any possible effect of ordering straight before

heterosexual, and vice versa, from other factors.

21
Explanation/definition of terms provided to respondents
Only two surveys included a definition of each category. In one instance, the National
Mental Health…Survey (2005), the explanation was expressed in terms of having a
‘like’ for people of the same or opposite sex, or both sexes. In the other, the Adult
Psychiatric Morbidity Survey, the definition was in terms of ‘attraction’ to the same
or opposite sex, and ‘feelings’ (as was also defined in the question stem). Such
explanations seem more appropriate to questions about behaviour or attraction
specifically, and would probably not be appropriate for a question about identity.

The need or otherwise for an explanation of each category, or specific categories, will
be explored in the qualitative work.

Comprehension of terms by respondents

Feedback showed that some respondents do not understand terms used in the
questions. An interviewer on the Policing for London Survey commented that:
"Respondents often misunderstood the question. Many laughed and replied
'None of these!', or in several cases 'None of those - just 'Normal'!'. Others
chose an answer, but when probed, it became apparent that they had
misunderstood the meaning of the terms, and had therefore given an answer
which did not describe them accurately....In London, many respondents have a
limited working knowledge of the English language, which does not extend to
polysyllabic technical terms of Greek and Latin origin. The only way to obtain
reliable data from them is to use simpler words."

That survey did not use the word ‘straight’. Such comprehension problems were also
cited in McManus. ‘Heterosexual’ has also been reported to be unfamiliar or
misunderstood, particularly among older or less well educated respondents, from
feedback from interviewers working on the NS Omnibus Survey trials. The cognitive
testing of the Citizenship Survey question found that there was some “confusion”
about the term ‘heterosexual’ – even though the term ‘straight’ was also given – and a
“process of elimination” used to decide the correct category.

A resulting hypothesis is that some heterosexual respondents are less likely to have
considered ‘sexual identity’ and to understand and use the associated terminology,
than LGB or younger or more highly educated heterosexuals.

Regarding the presence or absence of the term ‘straight’ in the heterosexual category,
the two postal data collections did not include it. The high rates of ‘prefer not to say’
or ‘not answered’ they produced might be explained partly by this.

It would be useful to understand how respondents who did not understand terms
answered the question and the implications for the accuracy of data. That is, did they
choose a substantive category, and if so which, or did they choose ‘other/none’?
Alternatively, did they choose ‘do not wish to answer’ or not provide any answer at
all?

Further research
The qualitative work being conducted by ONS will explore the terminology used,
understood and preferred by respondents of all sexual identities, and what are the

22
appropriate response categories. It will also investigate respondents’ answering
strategies when they do not have a good understanding of the terms used.

Category order
The questions generally, but not always, placed the heterosexual/straight category
first, followed by homosexual/gay/lesbian, then bisexual. The rationale seems likely
to have been to reflect the anticipated order of incidence of the main categories
(though while heterosexual is undoubtedly most prevalent, the relative incidence of
homosexual and bisexual is less certain and might vary according to what concept is
measured – e.g. identity, behaviour, attraction). One survey placed the categories in
alphabetical order. It can only be speculated if this was an attempt to avoid implying
any kind of hierarchy.

No obvious effect of the category order, such as primacy, was apparent from the
estimates. It can be observed that the survey which placed ‘bisexual’ first obtained the
highest rate for that category across all surveys. However, other surveys had higher
rates for bisexual than for gay/lesbian, even when it was later in the order.

Further research
Category order will be considered as part of the ONS development work.

‘Other’ category
The surveys varied as to whether or not they included an ‘other’ category (or similar).
Where ‘other’ was not presented to respondents, the research assumption, presumably,
was that the categories offered were comprehensive.

‘Other’ rates ranged between 0% and 0.9%.

No information was available as to why respondents answered other (e.g. if because

of miscomprehension of other categories, or genuine ‘other’ sexual identities, or
satisficing). It is not known what the respondents who answered ‘other’ considered
their sexual identity to be, or if any respondents felt they did not have a sexual
identity.

When questions did not include an ‘other’ option, it is not known how respondents
who did not think any of the substantive categories applied answered (e.g. answer one
of the substantive categories, or don’t know, or not provide an answer).

When questions were administered by self-completion, it varied as to whether

respondents answering ‘other’ were asked to specify further in an open question. No
information about the responses to ‘specify’ questions was available. GRO Scotland
decided against asking in case “facetious” responses were attracted.

‘Other’ was sometimes included when the question was interviewer-administered. It

varied as to whether interviewers probed for a description. On the survey where they
were expected to probe, no ‘other’ responses were recorded, for reasons unknown. It
is possible that respondents were unwilling to answer ‘other’ because of concerns
about how that would be perceived by interviewers. On surveys where no ‘specify’
question was asked, the reason can only be speculated, e.g. that researchers were wary
of causing embarrassment to respondents.

23
The availability of ‘other’, ‘none’ and ‘can’t choose’ categories, and the potential
need for any alternative substantive categories, will be another topic of investigation
during the qualitative stages of the development project.

Alternatives to ‘other’
Alternative categories to ‘other’ were used in two instances. One self-completion
question included: ‘If none of the above applies, (please write) I am_______’, while
another offered ‘Can’t choose’ as a category (with no follow up ‘specify’ question).

There appeared to have been some fluctuation in the rates across the ‘other’, ‘none…’
and ‘can’t choose’ categories. Intuitively, it is possible that each was perceived by
respondents differently. Unfortunately the written responses from the ‘none…’
category were not available for review. Any differences can only be hypothesised:
perhaps ‘none…’ and can’t choose’ were more inclusive for respondents who did not
consider themselves to have any sexual identity. Or perhaps they provided a more
suitable alternative for respondents who didn’t understand the meaning of the sexual
identity categories.

Additional response categories: transsexual and transgender

Only two surveys included response categories additional to heterosexual,
gay/lesbian, bisexual, other and do not wish to answer. The categories were
‘transsexual’ and ‘transgender’ respectively. In one case the question was being used
to identify respondents with certain identities for follow up questions, rather than
measure prevalence of sexual identity specifically, which explains the apparent
mismatch of sexual identity and gender identity concepts. The concept of gender
identity is distinct from sexual identity, and will therefore not be covered in the
question ONS is developing. However, if possible the qualitative research will include
transsexual and transgender people to ensure they are able to answer the sexual
identity question.

Further research
The need to provide an ‘other’ category and a ‘specify question’, and the way in
which they are used, will be addressed (although analysis of ‘other – specify’ data
from the NS Omnibus trial suggests it is a minor issue). Also to be considered are
whether ‘none’ or ‘can’t choose’ should be considered as a category, and provision of
alternative categories at the main question or a follow-up question.

4.6 Missing data

As mentioned above, all surveys recorded some non-substantive data (that is, item
non-response and response categories such as ‘do not wish to answer’). This resulted
because either i) respondents were presented with a category such as ‘do not wish to
answer’ on a show card or in a self-completion question; or ii) interviewers could
record refusal or don’t know; or iii) on paper self-completion forms the answer was
left blank.

On interviewer-administered surveys missing data ranged from 0.2% to 9.0%. The

high end of that range was an outlier; without it, the upper end was 2.3%. On self-

24
administered surveys they were higher, from 1.4% to 25.0%. This time the lower
figure was the outlier; without it the lower end was 8.0%.

4.6.1 Options not to answer the question

For both interviewer- and self-administered modes there were examples both of an
option to not answer the question being presented to respondents and of no such
option being presented.

‘Do not wish to answer’ and spontaneous refusals - interviewer-administered

surveys
Among the interviewer-administered questions, an option such as ‘do not wish to
answer’ was presented as a category on the show card on two surveys. Data for this
category from one of these surveys were not available. On the other survey, the
Northern Ireland Life and Times Survey, the proportion of respondents answering ‘I
do not wish to answer this question’ had reduced from 5% in 2000 to 1.3% in 2005.
The wording of the question and response categories did not change in that period,
and no obvious reason for the reduction was apparent to the researcher providing the
information. We can only hypothesise that respondents in Northern Ireland have
become more willing to answer substantively because they have become more aware
of issues relating to sexual identity, which public bodies have a statutory duty to
monitor as a diversity strand and/or because interviewers have become more
experienced at administering the question over time.

When ‘do not wish to answer’ was not included on the show cards, interviewers were
instead able to code spontaneous refusals. Refusals ranged from 0.2% to 9%, though
see ‘Interviewer issues’ below regarding the latter figure. These surveys also included
an ‘other’ category, or allowed ‘don’t know’ to be recorded (though none of the
estimates provided contained cases in both ‘other’ and ‘don’t know’).

Do not wish to answer & no answer – self-administered surveys

When included on self-administered surveys, higher rates of ‘do not wish to answer’
were recorded than on the interviewer-administered survey which included that
category. The Civil Service Diversity Survey and the British Social Attitudes Survey
did not offer ‘do not wish to answer’, but were subject to the question being left
unanswered. On the British Social Attitudes Survey the rate was much lower than on
the other self-administered surveys, for unknown reasons.

Despite the inclusion of a ‘do not wish to answer’ category the Scottish Census Small
Scale Test and the Newham Household Panel Survey also experienced high levels of
the question being left unanswered.

Clearly, it was easier for respondents not to answer a question on a self-completion

form. The conclusion that might be drawn is to exclude a ‘do not wish to answer’
option on paper self-completion forms, to try to reduce missing answers, since
respondents can and do leave the question unanswered anyway. However, this might
compromise consistency across modes or methods, if it were felt that such an option
had to be offered in CAPI and CASI because respondents would not otherwise have
an equivalent obvious means of not answering (they might not be aware that they
could spontaneously refuse).

25
4.6.2 ‘Don’t know’ responses
Only one survey allowed the recording of ‘don’t know’ answers, if given
spontaneously to the interviewer. At 1.6%, these responses outnumbered refusals
(0.7%). It is worth noting that this survey did not use the term ‘straight’; as previously
mentioned, evidence suggests some people do not know the meaning of
‘heterosexual’, which might explain this rate to some extent.

It is not possible to gauge whether respondents answered ‘don’t know’ due to their
lack of comprehension of the question or knowledge of terms, and/or their inability to
decide what their sexual identity was.

Nor is it possible to say how many respondents would answer ‘don’t know’ if it was
possible to record (other than on the basis of the one survey which recorded it); or
whether they would instead not answer the question at all, would answer ‘other’ or
‘do not wish to answer’, guess, or use a process of elimination to answer with one of
the substantive categories.

4.6.3 Non-response and ethnic group

Potential issues regarding the acceptability and comprehension of sexual identity
questions were highlighted among the surveys.

The Newham Household Panel Survey found that white respondents were more likely
to answer the sexual identity question than respondents from Asian and ‘other’ ethnic
groups:

“In all waves those giving their ethnic origin as 'white' were significantly
more likely to answer the question; whilst those giving their ethnic origin as
'Asian' or 'Other' were significantly less likely to answer the question. All
'Asian' categories displayed this tendency. Around a third of respondents from
Asian origins did not answer the sexuality questions compared with around 1
in 7 of those with a ‘White' ethnic origin.”

Further to not answering the question at all, respondents whose first language was not
English were, on average, slightly more likely to answer ‘none of the above’ and
much more likely to answer ‘I do not wish to answer this question’. It can be
hypothesised that some respondents from minority ethnic groups had cultural or
religious beliefs resulting in objection to being asked the question or unwillingness to
answer it. It is not possible to say whether there were any problems with
comprehension due to English not being respondents’ main language.

The Policing for London Survey carried out an ethnic boost to its sample. The figures
on missing data by ethnicity were not obtained for that or any other survey, so no
comparison with the Newham findings could be made. However, an interviewer
commented that whilst there were problems with comprehension, including among
respondents with limited English, “I don't recall any problems of privacy,
confidentiality or acceptability, and people were quite happy to try to answer the
question.”

26
Further research
Minority ethnic groups’ views of the question among will be an area which ONS
intends to explore in the qualitative research.

4.7 Confidentiality, privacy and acceptability

Following on from the above references to missing data, and respondent beliefs,
embarrassment and objection to the question, the review will now consider more
generally the topics of privacy in the survey environment (i.e. with regard to other
people in the household or an interviewer), confidentiality (i.e. disclosure of
information to other people), and acceptability of asking about sexual identity (i.e.
feelings about the personal nature of the subject and of invasion of privacy), and their
affect on survey response, and the completeness and accuracy of data.

Considered overall, it seems there was little concern expressed by respondents,

particularly to interviewer-administered surveys, about the three aspects outlined
above. The cognitive testing of the Citizenship Survey question found no objections to
the question among respondents, although “several were slightly embarrassed”.
Additional support came from two surveys which asked interviewers to indicate
whether the respondent showed any ‘hesitation’ or ‘reluctance’ – respectively only
2% and 10% answered yes - the latter was ascribed to a few interviewers being
reluctant to ask the question (see ‘Interviewer issues’ below).

Each aspect is considered further below.

4.7.1 Confidentiality pledges

As mentioned earlier, all surveys gave confidentiality assurances in their advance or
cover letters, or within self-completion booklets, or at the start of the section of
questions including the sexual identity question. None of the surveys mentioned
confidentiality at the point when the question was asked, though two provided
interviewers with an optional explanation which mentioned anonymity.

In focus groups conducted as part of the development of the Civil Service Diversity
survey, concerns had been expressed at respondents being identifiable individually.
Such concerns might have been higher in the context of a survey conducted for
respondents’ employers than they would be in surveys conducted for sponsors less
immediately related to them.

No other information was gathered with regard to respondents’ views on

confidentiality of the data they were being asked to provide, such as any concern
about the data being divulged to a third party, deliberately or accidentally.

Further research
People’s views on confidentiality, and fears of the likelihood and consequences of
disclosure to other parties, is another subject to be covered in the qualitative research.

27
4.7.2 Providing privacy in the survey environment
Only limited information was available about respondents’ views on privacy in the
survey environment. Researchers were generally unaware of any widespread concern
fed back by interviewers. There was only a little evidence of respondent concern.

Interviewer-administered questions
The common approach taken to dealing with providing privacy to respondents on the
interviewer-administered surveys was the use of concealed response show cards. That
is, respondents were asked to give a letter or number corresponding to the response
category, rather than state the answer in words. The letters or numbers were usually,
though not always, randomised – i.e. they were not immediately sequential or in
alphabetical/numerical order (e.g. ‘R, D, P, H’ rather than ‘A, B, C, D’). This
approach was intended to reduce the potential for embarrassment. On the Fair
Treatment at Work Survey pilot the show card question was "not met with any
resistance".

All of the interviewer-administered household surveys only interviewed one person

per household, so there were no issues to deal with relating to concurrent
interviewing. That is, how to maintain confidentiality between household members
hearing or seeing the same question and/or show card.

However, the concealed response show card method was only intended to provide
privacy between the respondent and any people present other than the interviewer. No
surveys made any assurances to respondents that the interviewer would be unaware of
the response category to which the number or letter corresponded. Indeed, it was said
by a respondent to one survey that interviewers knew the answer so it was “clearly
not anonymous”. The issue of social desirability due to the presence of interviewers is
addressed in section 4.8.

In one of the surveys conducted in mental health establishments, the intention was
that the interview be conducted in a private place but that was dependent on
availability. On the National Mental Health … Census (2006), it is unknown what
efforts to provide privacy from other staff or patients were made by the staff
conducting it.

Self-administered questions
Respondent preferences for interviewer-administered or self-administered questions
are little known.

The Fair Treatment at Work Survey pilot, an interviewer-administered survey,

experimented with self-completion using a CAPI pen in a proportion of cases. The
respondent was given the laptop and an electronic pen with which to select their
answer on the screen. This method was reported to have been "slightly preferred" by
respondents as "more anonymous". But it was also reported to have broken up the
interview and been difficult for certain respondents. However, all pilot respondents
were apparently heterosexual, so the views of LGB respondents were unknown. A
question remains as to whether respondents considered self-administered questions to
be adequate in maintaining privacy and confidentiality, despite the lack of an
interviewer.

28
One researcher recollected that a respondent on a self-completion survey had
contacted the organisation to say he had not “felt comfortable” answering the
question, as the answer might be seen by other people (it was not specified whether
this meant other people known to the respondent seeing the form, or the people
collecting or using the data).

The Newham Household Panel Survey was the only survey to collect data from more
than one household member. It is not known whether the high rates of missing data it
experienced were in any way connected to this, for example if there were concerns
that other household members would be able to see answers.

Further research
The subject of within-household privacy and the concurrent interviewing
environment, and views on interviewer- and self-administered modes will be covered
in ONS’s further research.

4.7.3 Acceptability of the question and survey/item response

Survey response
No researcher gave any indication that survey response rates had been affected by the
inclusion of a question on sexual identity.

The Scottish Census Small Scale Test experimented with using four variants of the
questionnaire, to examine the effect on response rates of asking about sexual
orientation. Each variant employed a different combination of two questions: one on
sexual orientation and one on experience of discrimination on the basis of a number of
grounds, including sexual orientation. The combinations were: both questions
(separated by several other questions; with sexual orientation asked second); one
question but not the other; and neither question. Unit and item response rates were
very similar across the variants, suggesting there was no effect, either of asking about
sexual orientation per se, or of asking it with and without the context of a question on
the subject of discrimination.

Item response
The relatively low rates of missing data when the question was interviewer-
administered would seem to indicate that on the surface questions were broadly
acceptable to respondents, in the sense that they did not generally didn’t refuse to give
a substantive answer.

However the rates of refusals or not answering the question were higher on the self-
completion surveys. This suggested that there might after all be an issue of invasion of
privacy or objection to being asked the question, which was not so apparent from the
interviewer-administered surveys.

Indeed, this was borne out in feedback to the Scottish Census Small Scale Test.
Respondents were asked to indicate how happy they felt about each question. The test
report stated that "The question on sexual orientation stands out with around 11% of
respondents selecting very unhappy or unhappy" compared to less than 5% for most
questions. Respondents could also write comments. For the sexual orientation

29
question they "ranged from fully accepting to hostile". Concerns expressed included:
"not understanding the need for the data"; "offence" and "unhappiness at the intrusive
nature of the question"; and concerns about confidentiality (e.g. being asked for their
postcode).

Such concerns did not necessarily "translate into refusal to record an answer". As
reported above, the response rates were similar to the versions of the questionnaire
with and without the sexual orientation question, suggesting that the question did not
cause people not to respond to the survey. Conversely respondents who "accepted the
question without reservation" included some who answered 'prefer not to say'. It
would therefore appear there was a distinction to be made between objecting to being
asked the question, and objecting to give an answer, and that the two positions did not
necessarily correspond.

Non-responders to the Scottish Census Small Scale Test question "generally also
declined to record any comments".

4.8 Accuracy and completeness: honesty and social desirability

An important issue in collecting data on sexual identity with the intention of providing
a benchmark measure is minimising the extent of missing data. Missing data
compromises the quality and usefulness of estimates. This is exacerbated when the
characteristic of most interest – minority sexual identities – is found in a small
proportion of the population: even a relatively small proportion of missing data can
cast serious doubt on the accuracy of estimates, and hence their usability. However,
when considering the potentially sensitive topic of sexual identity, it is also important
that in attempting to reduce the extent of missing data, the accuracy of substantive
responses given is not compromised as a result.

Such compromising of the data could occur if respondents feel pressured into
providing an answer to a question they would rather avoid. If respondents are unaware
of how to avoid answering – the absence of a ‘do not wish to answer’ category and, in
interviewer-administered mode, not knowing or remembering that they can
spontaneously refuse – they might instead give a substantive answer, but one which is
inaccurate.

Reasons respondents would wish to avoid providing an accurate, substantive answer

might include ‘social desirability’ or the urge to conform; an example of which could
be LGB respondents answering heterosexual, being the majority sexual identity,
because of perceived societal disapproval of non-heterosexuality. In the interviewer-
administered context, this might occur because a respondent has concerns that the
interviewer, or any other person present, such as a family member who is unaware of
it, might disapprove of their sexual identity. In any mode of administration, it might
occur because a respondent has concern that the data will not be kept confidential and
will be seen by someone they would prefer did not see it.

As has been shown above, interviewer-administered surveys generally obtained a

higher proportion of substantive answers and lower rates of missing data than self-
administered surveys. The most likely explanations of this observation are the
presence of an interviewer, and in most cases, the absence of an obvious means of

30
avoiding giving a substantive answer. Respondents who wished to avoid giving a
substantive answer for one reason or other were more at liberty to do so when self-
administering – when given the chance, they were more likely to take it.

A number of questions arise. Do those respondents to interviewer-administered

surveys who would avoid answering the question if they could, provide truthful or
untruthful answers? If they provide untruthful answers, why do they do so? Are
respondents of certain sexual identities more likely to do so than others? In what way
are the answers inaccurate?

The generally low rates of homosexual and bisexual obtained, when compared to the
government 5% to 7% estimate, suggest it is possible that some respondents to
interviewer-administered surveys were giving a socially desirable answer, rather than
being willing to respond with one of the ‘sexual minority’ categories, or saying to the
interviewer ‘I don’t want to answer’ or ‘I don’t know’. The latter two answers might
have been avoided due to respondents feeling the interviewer or someone else might
have made an inference about their sexual identity.

However, a number of factors make it difficult to make firm conclusions about any
social desirability effect.

Firstly, the reliability of the benchmark proportion of the population which is LGB.

Secondly, other confounding factors such as population and geographical coverage,

sample designs and clustering of the LGB population, response rates and potential
non-response bias.

Thirdly, the rates of missing data on the self-administered surveys were generally
higher than the estimated proportion of the population which is LGB. And LGB
estimates appeared to be more consistent across modes than those for heterosexuals.
So it might be that LGB respondents tended to give correct answers in either mode.
However, it is not known whether a proportion of the LGB population gave incorrect,
socially desirable answers in interviewer-administered mode, or avoided giving
substantive answers in self-completion mode.

Estimates of the heterosexual population tended to be higher when the question was
interviewer-administered, ranging from 88% to 99%, compared with self-
administered questions, where they were between 70% and 95%. Respondents on
interviewer-administered surveys who did not understand the question or response
categories were able to ask the interviewer for help and so give a substantive answer,
which might have reduced the amount of missing data. On self-administered surveys
they might have been less able to check with someone. Another possible reason is that
a proportion of heterosexual respondents avoided giving substantive answers to self-
administered questions, but answered truthfully when ‘forced’ to answer on
interviewer-administered surveys.

As well as the possibility that some respondents gave deliberately untruthful answers
for social desirability reasons, other forms of satisficing might have been at play.
Respondents might have given any answer just to satisfy the interviewer and/or avoid
having to consider a subject which might be embarrassing or sensitive, rather than

31
deliberately being untruthful. For example, by choosing the first category in the list
presented (the ‘primacy’ effect).

Further research
The questions raised above on missing data, confidentiality, privacy and acceptability
will be covered in the further research. Issues which will be considered include:
• whether respondents should be allowed to avoid giving substantive answers, in
different modes, and, if so, how;
• whether respondents might give socially desirable answers, and what
characteristics they have;
• and the optimal balance between minimising missing data and maximising data
accuracy that can be achieved.

Issues which will be considered include: relating to the presentation, or not, of

‘avoidance’ options (such as ‘do not wish to answer’) in the different modes; and
measurement error effects such as social desirability and satisficing. A key question is
what the optimal balance is that can be achieved between minimising the amount of
missing data on one hand, and maximising the accuracy of the data on the other.

4.9 Interviewer attitudes to asking about sexual identity

Experiences on some surveys, relating to interviewer attitudes to asking about sexual
identity, lead to the conclusion that their instruction should be given sufficient
attention to avoid problems which would affect data quality. However at least from
this sample, and as far as the researchers spoken to were aware, such problems were
not the norm.

4.9.1 Non-professional interviewers in communal establishments

A problem experienced in conducting the surveys was related to the use of non-
professional survey interviewers. The National Mental Health… Census (2006) used
ward and other staff employed in mental health establishments. This is not the
approach which would normally be followed by survey organisations like ONS. The
context and environment of the data collection was non-standard for social surveys.
However the experience is worth noting, for the benefit of any other organisation
which would use non-professional interviewers. Problems were encountered with
many establishments not asking the question at all, or only asking a proportion of the
population. The sponsors learned that views existed among the people administering
or supervising the census that sexual orientation could be determined without asking
the subject, by observation, and that staff could not be expected to ask such an
‘embarrassing’ question. Consistency of administration across and within
organisations, and appropriate training/briefing and supervision of people
administering the question, is clearly required. Indeed, feedback from other mental
health establishments was more positive – the question had been administered to more
people and without reported problems.

4.9.2 Professional interviewers

Even among professional interviewers, who it might be assumed were used to
administering questions on potentially sensitive topics, and to be impartial, some
reluctance to ask about sexual identity was experienced. One of the surveys

32
experienced a high rate of refusal to the question. The researchers attributed this to a
“small number” of interviewers coding a refusal for all or the majority of their
interviews, suggesting that some “do not feel able to ask this question…” The
characteristics of these interviewers (e.g. age, sex, experience) were not discovered
and any correlation with high levels of refusals unknown. It is worth noting that the
interviewers on this survey received written instructions but received no briefing,
whereas on other surveys with lower refusal rates, interviewers had been briefed. It
was concluded that “future attempts to ask a question on sexual identity must pay
particular attention to interviewer training and briefing to ensure interviewers feel
able to ask the question.”

Further research
This recommendation and interviewer-related issues generally will be addressed by
ONS as part of its development work.

4.10 Longitudinal data collection

The Newham Household Panel Survey collected data on sexual identity from
respondents over time – they were interviewed over 4 annual waves (to date). The
sexual identity question was asked at each wave. No information was available on the
stability of responses over time, at the individual person level. However the response
rate to the question overall was stable, from waves 1 to 4. This might indicate that
respondents in general became neither more nor less happy over time to answer the
question (though attrition and sample replenishment has not been considered).

Further research
ONS will be considering the effect of asking about sexual identity on agreement to
take part in further waves on panel surveys. It will also look at whether the question
need only be asked at the first wave or each wave.

4.11 Issues about asking sexual identity not addressed by the review
There are some important issues pertinent to the potential introduction of a sexual
identity question on ONS surveys such as the Integrated Household Survey, which
have not been addressed in this review.

None of the surveys on which the sexual identity question was interviewer-
administered interviewed more than one household member, unlike the majority of
ONS’s surveys. Therefore no issues relating to concurrent administration were
covered. Nor were any issues to do with either telephone interviewing or proxy
interviewing, both of which are conducted on some ONS surveys.

Further research
The three topics above will all be addressed in ONS’s development project.

33
5 References

McManus S (2003), National Centre for Social Research, ‘Sexual Orientation

Research Phase 1: A Review of Methodological Approaches’ on behalf of the Scottish
Executive. Available at
http://www.scotland.gov.uk/Publications/2003/03/16650/19351

Taylor T (2007a), ONS, ‘Developing survey questions on sexual identity:

International experiences of administering questions on sexual identity/orientation’.

Taylor T (2007b), ONS, ‘Developing survey questions on sexual identity: Report on

National Statistics Omnibus survey trials 1 and 2’. Available at

Wilmot A (2007), ONS, ‘In search of a question on sexual identity’ - paper presented
at the 62nd Annual Conference of the American Association of Public Opinion
Research, May 2007.

(The above three papers are available at

http://www.statistics.gov.uk/about/data/measuring-equality/sexual-
identity/default.asp)

6 Appendices

See over page.

34
[blank page]

35
 Appendix A: Summary table – mode/sample design/sexual identity question designs and estimates

Mode/sample design/sexual identity question design/survey response rate Estimates: Substantive categories Estimates: Missing values (inc. Base
KEY: n/a = not applicable; dk = no data (Percentages are weighted; of all cases) presented)

rate %
Survey response

/straight
Heterosexual

gay/lesbian
Homosexual/

Bisexual

Total LGB

similar)
Other

categories
Additional
categories
substantive
Total % in

DK
(presented)
not to answer
Do not wish/Prefer

code)
Ref/no answer (int

SC
Unanswered on

values
Total % missing

(w)=weighted
(u)=Unweighted
Survey Mode of Sample Question stem Response Categories
administ-
ration

(or
Northern Interviewer- Random probability; general Can you tell me which of I am 'gay' or 'lesbian' 61 97.0 0.5 1.2 1.7 0.0 n/a 98.7 n/a 1.3 n/a n/a 1.3 1199 (w)
Ireland Life administered population (NI); 1 these best describes (homosexual);
and Times (CAPI; person/hhold (18>) you? Please just give me I am heterosexual or 'straight';
Survey showcard) the number on the card. I am bi-sexual;
Other;
I do not wish to answer this
question
Policing for Interviewer- Random probability; general Please choose a letter Completely heterosexual 49 completely: completely: 0.2 (1.0 or n/a n/a 97.7 1.6 n/a 0.7 n/a 2.3 2800 (u)
London Survey administered population (London) + from this card which best Mainly heterosexual 94.4 0.4 3.3?)
(CAPI; ethnic boost; 1 person/hhold describes how you would Bisexual mainly: mainly:
showcard) (15>). think of yourself? Mainly gay or lesbian 2.3 0.4
Completely gay or lesbian
DTI Interviewer- Random probability; men Which of these best Straight/heterosexual 58 98.9 0.7 0.2 0.9 0.1 n/a 99.8 n/a n/a 0.2 n/a 0.2 1038 (w)
Employees’ administered 16-64, women 16-59 describes you? - please Gay/lesbian/homosexual
Awareness (CAPI; employees now/in the last 2 just give me the number Bisexual
Knowledge showcard) years; 1 person/hhold. on the card. Other
and Exercise
of Employment
Rights Survey
DTI Fair Interviewer- Quota sampling (part of an Please look at this card. Straight or heterosexual n/a 87.8 2.0 0.4 2.4 0.7 n/a 90.9 n/a n/a 9.0 n/a 9.0 2704 (w)
Treatment At administered omnibus survey); Which of the answers on Gay or lesbian or homosexual
Work Pilot (CAPI; employees now/in last 2 the card best describes Bisexual
Survey showcard) years; 1 person/hhold. you? Please just give Other
me the letter alongside
the appropriate answer.

National Interviewer- Simple random sample of Which of the categories Heterosexual (like people of 31 dk dk dk dk n/a trans- dk dk n/a dk n/a 0 <400 (u)
Mental Health administered patients in each of 41 on this card would you the opposite sex) gender:
and Ethnicity (PAPI; NHS/private hospitals say describes your Gay male or lesbian female dk
Census 2005 showcard) sexual orientation? (like people of the same sex)
Service User Please give me the Bisexual (like people of both
Survey letter. sexes)
Transgender
National Interviewer- Census of inpatients in all Which of the following Lesbian/Gay/Homosexual dk dk dk dk dk dk n/a dk dk dk n/a n/a 0 dk
Mental Health administered mental health/learning terms would you use to Straight/Heterosexual
and Learning (PAPI; disability establishments, describe your sexual Bisexual
Disability showcard) England and Wales orientation? Other
Ethnicity Do not wish to answer
Census 2006
British Social Self- Random probability; general Which of the following Heterosexual (‘straight’) dk 95.3 gay: 0.4 0.6 1.4 can't trans- 98.6 n/a n/a n/a 1.4 1.4 1732 (u)
Attitudes administered population; 1 person/hhold best describes how you Gay lesbian: 0.4 choose: sex'l:
Survey (paper) after (16>) think of yourself? Lesbian 1.7 0.2
interviewer- PLEASE TICK ONE Bisexual
administered BOX ONLY Transsexual
survey Can’t choose

36
 rate %
Survey response

/straight
Heterosexual

gay/lesbian
Homosexual/

Bisexual

Total LGB

similar)
Other

categories
Additional
categories
substantive
Total % in

DK
(presented)
not to answer
Do not wish/Prefer

code)
Ref/no answer (int

SC
Unanswered on

values
Total % missing

(w)=weighted
(u)=Unweighted
Survey Mode of Sample Question stem Response Categories
administ-
ration

(or
Newham Self- Random probability; general Please tell us what best I am heterosexual or 'straight' W1: 74 W1: 74.5 W1: 0.0 W1: W1: 0.3 none of n/a W1: n/a W1: n/a W1: 9.5 W1: W1: 1286
Household administered population (Newham); all describes you. I am 'gay' or 'lesbian' W2: 82 W2: 75.2 W2: 0.7 0.3 W2: 1.1 the 76.8 13.7 W2: 8.8 23.2 W2: 1218
Panel Survey (paper) after household members (16>) (homosexual) W3: 78 W3: 70.4 W3: 0.7 W2: W3: 1.3 above: W2: W2: W3: 10.0 W2: W3: 1132
interviewer- I am bisexual W4: 44 W4: 71.7 W4: 1.0 0.4 W4: 2.1 W1: 2.0 79.5 11.7 W4: 9.5 20.5 W4: 636
4
administered If none of the above applies. W3: W2: 3.3 W3: W3: W3: (u)
survey (PLEASE WRITE): I 0.6 W3: 3.4 75.0 15.0 25.0
am_______________ W4: W4: 4.7 W4: W4: W4:
I do not wish to answer this 1.1 78.5 12.0 21.5
question
Scottish Self- Random probability; general Which of the following Heterosexual 31 83.1 0.9 0.4 1.3 0.9 n/a 85.3 n/a 8.5 n/a 6.2 14.7 692 (u)
Census Small administered population (Scotland); 1 best describes your Lesbian/gay (versions
Scale Test (postal) person/hhold. Half sample sexual orientation? Bi-sexual with SO
asked sexual orientation Other Q)
question. Prefer not to answer
Civil Service Self- Simple random sample of If you have no objections Bisexual 48 89.0 gay: 1.0 2.0 3.0 n/a n/a 92.0 n/a n/a n/a 8.0 8.0 7863 (u)
Diversity administered civil servants from every to stating your sexuality, Gay man (rounded) lesbian 0.0
Survey (postal) department, plus boosts. please could you state it Heterosexual
here? Lesbian
Omnibus (July- Self- Random probability; general Which of the following Heterosexual 92.0 1.3 1.2 2.5 0.9 n/a 95.4 n/a 4.6 n/a n/a 4.6
Aug06) (exc administered population; 1 person/hhold best describes your Lesbian/Gay
15% skipped) (CASI) within (16>) sexual identity? Bi-sexual
interviewer- Other
administered Prefer not to say
survey
Omnibus (Nov- ditto ditto Do you consider yourself Heterosexual or Straight 96.8 0.8 0.6 1.4 0.3 n/a 98.5 n/a 1.5 n/a n/a 1.5
Dec06) (exc to be… Gay or Lesbian
14% skipped) Bisexual
Other
Prefer not to say?
Adult Self- Random probability; general Version A: Entirely heterosexual (attracted to persons of the opposite sex) still in
Psychiatric administered population (Eng); 1 Which statement best Mostly heterosexual, some homosexual feelings field
Morbidity (CASI) within person/hhold (16>). Split describes your sexual Bisexual (equally attracted to men and women)
interviewer- sample experiment orientation? This means Mostly homosexual, some heterosexual feelings
administered sexual feelings, whether Entirely homosexual (attracted to persons of the same sex)
survey or not you have had any Other
sexual partners.
Version B: Completely heterosexual still in
Please choose the Mainly heterosexual field
answer below that best Bisexual
describes how you think Mainly gay or lesbian
of yourself... Completely gay or lesbian
Other
Citizenship Interviewer- Random probability; general Looking at this card, Heterosexual or straight still in
Survey administered population; 1 person/hhold which of the options best Gay or lesbian field
(CAPI; describes your sexual Bisexual
showcard) identity? Please just Other
read out the letter next to [or would you] Prefer not to say?
the description.

4
DK survey rate. % of completed individual interviews returning self- completion booklet. Estimates and bases for the sexuality question are for those who returned self-completion booklet only.
37
Appendix B: Information about each survey included in the review

1. INTERVIEWER ADMINISTERED QUESTIONS

1.1 Northern Ireland Life and Times Survey

Sponsor Queen's University Belfast and University of Ulster

Information provided by Queen's University Belfast
Year conducted 2005 (SID has been asked since 2000)
Survey mode CAPI (plus paper self-completion module)
Sexual identity question CAPI (concealed response show card)
mode
Sample design Random probability; general population, Northern Ireland.
One person per household (aged 18 or over) is selected for
interview.
Survey response information 61%
Survey topics Social attitudes (evolved from Northern Ireland Social
Attitudes Survey).
Sexual identity question • Intended to reflect the respondent's 'self-perception', not
purpose/definition anything more specific (e.g. behaviour, attraction).
• In Northern Ireland public bodies have a statutory duty to
monitor 9 diversity strands, including sexual identity.
Location in The final question, in the 'background' section
questionnaire/interview (demographic/classificatory questions) at the end of the
interview. Religion and income are also asked.
Other SID-related questions With reference to equality laws; whether various groups,
including ‘Gays/lesbians/bisexuals’, are generally treated
unfairly.
Sexual Identity question Can you tell me which of these best describes you? Please
just give me the number on the card.
I am 'gay' or 'lesbian' (homosexual)
I am heterosexual or 'straight'
I am bi-sexual
Other
I do not wish to answer this question

Interviewers were expected to probe 'other' and write in the

respondent's description.
Explanations/definitions/assu No preamble to SID question (see below). The background
rances provided to section began with a preamble explaining, among other
respondent things, that the survey was confidential and answers would be
anonymous.
Interviewer instructions -
Estimates (weighted % Gay/lesbian/homosexual: 0.5
unless otherwise indicated) Heterosexual: 97
Bisexual: 1.2
Other: 0
Do not wish to answer this question: 1.3
Non-response / non- • There were no 'don't know' responses.
substantive responses • The rate of 'don't wish to answer' reduced steadily from
5% in 2000 to 1% in 2005.
Pre-testing; piloting; changes • Pilot survey used a preamble: "The last question in this
made over time section is about sexual orientation - that is whether
someone is gay or lesbian, heterosexual or bisexual. As
with all questions on the survey you do not have to
answer this if you do not want to, but all answers will be
treated in the strictest confidence. Can you tell me
whether you would describe yourself as ... READ
OUT...". Interviewers advised that it be dropped, as it

38
 drew attention to the question and made some people
feel embarrassed; thus it was dropped for the main stage
(beginning 2000).
• After piloting they discarded various versions of the
question which did not give the options in both formal
and informal language. Interviewers felt that some
people weren't quite sure what heterosexual meant, and
confused it with homosexual. The options now used
include 'I am heterosexual or straight'.
Feedback from researchers, • The researchers had no feedback from the fieldwork
interviewers, respondents agency on acceptability and comprehension of the
question by respondents.
• No known explanation for there being no 'other'
responses being given in 2005.
• The phrase 'I do not wish to answer this question' was
thought to be clearer for respondents than 'Refused',
having less negative nuances.
• Interviewers thought that it was less embarrassing for
respondents if they were given a show card than if the
responses were read out by the interviewer. Also,
respondents only give the associated numbers, rather
than the phrase on the card, which again cuts down
potential embarrassment.
Links http://www.ark.ac.uk/nilt/
http://www.esds.ac.uk/government/nilts/

39
 1.2 Policing For London Survey

Sponsor Metropolitan Police

Information provided by National Centre for Social research (NatCen) - fieldwork and
research design contractor;
Prof. Mike Hough, formerly South Bank University Criminal
Policy Research Unit
Year conducted Fieldwork conducted July-December 2000
Survey mode CAPI.
Sexual identity question CAPI (concealed response show card).
mode
Sample design Random probability; general population, London; plus ethnic
boost (over-sampling of areas with high density BME and
focused enumeration).One respondent per household
randomly selected from members aged 15 or over.
Survey response information Overall response 49% (though “difficult to calculate when
using focused enumeration”)
Survey topics Experience of living in their locality; contact with the Police;
views of the Police; views on policing priorities; collaboration
with the Police; and experiences of crime.
Sexual identity question -
purpose
Location in In the final section, along with other demographic questions.
questionnaire/interview Positioned between questions on religion and political
affiliation. Further questions followed.
Other SID-related questions -
Sexual Identity question Please choose a letter from this card which best describes
how you would think of yourself?
K - completely heterosexual
D - mainly heterosexual
I - bisexual
R - mainly gay or lesbian
M - completely gay or lesbian
Explanations/definitions No preamble included before the section or this question.
provided to respondent Confidentiality was mentioned in the advance letter sent.
Interviewer instructions "This Q has previously been used in a Sexual Lifestyles
survey – but not in a ‘normal’ survey. It asks respondents to
self-report their sexuality and as such may be a sensitive
question to ask. Please ensure the respondent is able to
answer the question themselves (if a translator is required,
take care to ensure the answer is the respondent’s). […] This
Q is asked to investigate the relationship between sexuality
and any experience of crime and attitudes to the police.
Respondents can refuse to answer. If necessary, remind
respondents that homosexual = ‘gay/lesbian’ and
heterosexual = ‘straight’."
Estimates (weighted % Weighted Unweighted
unless otherwise indicated) all valid all valid
Completely 94.4 96.6 93.4 97.4
heterosexual:
Mainly 2.3 2.3 1.7 1.8
heterosexual:
Bisexual: 0.2 0.2 0.3 0.3
Mainly gay or 0.4 0.4 0.3 0.3
lesbian:
Completely gay or 0.4 0.4 0.2 0.2
lesbian:
Not answered: 0.7 - 1.6 -
Don't know: 1.6 - 2.6 -

40
 Base =8293 =2800
Non-response / non- No 'prefer not to say' category was offered. Interviewers could
substantive responses code ‘don’t know’ and ‘not answered’.
Pre-testing, piloting, changes -
made over time
Feedback from researchers, • Technical report didn't indicate any particular issues
interviewers, respondents about the question on 'sexual orientation'.
• Mike Hough: “by any reasonable standards the results are
pretty hopeless… baldly unsatisfactory.”
• One interviewer, contacted by NatCen for this ONS
review, commented that there were problems with
comprehension, including among people with limited
English, but that there were no problems with privacy,
confidentiality or acceptability: "Respondents often
misunderstood the question. Many laughed and replied
'None of these!', or in several cases 'None of those - just
'Normal'!'. Others chose an answer, but when probed, it
became apparent that they had misunderstood the
meaning of the terms, and had therefore given an
answer which did not describe them accurately....In
London, many respondents have a limited working
knowledge of the English language, which does not
extend to polysyllabic technical terms of Greek and Latin
origin. The only way to obtain reliable data from them is
to use simpler words."
Links Technical report: www.blink.org.uk/docs/policereport.pdf

41
 1.3 Employees’ Awareness Knowledge and Exercise of Employment Rights Survey

Sponsor DTI
Information provided by DTI
Year conducted 2005
Survey mode CAPI
Sexual identity question CAPI (show card).
mode
Sample design Random probability (PAF). Population: individuals of
working age (16-64 for men and 16-59 for women) who
were employees or had been employees in the previous two
years. One person selected per household.
Survey topics Various issues relating to employment rights
Survey response information 58%
Sexual identity question A 'socio-demographic' sexual identity question was included
purpose in order to examine the experiences of LGB employees
compared to heterosexual / straight employees, in respect of
employment rights and fair treatment.
Location in Last substantive question (followed by a recontact question).
questionnaire/interview
Other SID-related questions Relating to awareness of employer’s obligation re. unfair
treatment and whether employee had been unfairly treated
Sexual Identity question "Which of these best describes you? - please just give me
the number on the card." (Showcard categories numbered 1
to 4).
1. Straight/heterosexual
2. Gay/lesbian/homosexual
3. Bisexual
4. Other
Explanations/definitions/assu Prompt for interviewer use if respondents objected to or
rances provided to queried the question:
respondent "We’re collecting this information to find out more about the
discrimination people may face for different reasons. The
information is kept completely anonymous."
Interviewer instructions Briefed (by video) and received written instructions.
Estimates (weighted % All Valid
unless otherwise indicated) Straight/heterosexual 98.9 99.1
Gay/lesbian/homosexual 0.7 0.7
Bisexual 0.2 ?
Other 0.1 ?
Refused 0.2 -
Base =1038
Non-response / non- A 'refusal' category was available to interviewers. It was not
substantive responses shown on the card in order to maximise response to the
question.
Pre-testing, piloting, changes -
made over time
Feedback from researchers, • DTI researchers hypothesised that the clustered design
interviewers, respondents might have led to under-representation of LGB people,
if they are concentrated in certain urban areas.
• For each interview the interviewer was asked "Was
there any reluctance or objection from the respondent
in answering this question?" To which the responses
were Yes: 2%; No: 98%. The fieldwork contractor
reported after piloting that "everyone was happy to
answer the sexual orientation question - the show card
helped lessen any potential embarrassment for
respondents and interviewers asking the question".
Links

42
 1.4 Fair Treatment at Work Pilot Survey

Sponsor DTI
Information provided by DTI
Year conducted 2005
Survey mode CAPI
Sexual identity question CAPI (concealed response show card).
mode of administration
Sample design Quota sampling. Administered as part of an omnibus survey.
Population: adults in work as employees or who had worked
as employees in the previous two years. One person was
selected per household.
Survey response information Response rate not applicable (quota sample). 2366
interviews achieved in main survey plus 1570 from boosts.
Survey topics Perception of unfair treatment in workplace; reporting of
personal and observed unfair treatment
Sexual identity question A 'socio-demographic' sexual identity question was included
purpose in order to examine the experiences of LGB employees
compared to heterosexual / straight employees, in respect of
employment rights and fair treatment.
Location in Last substantive question (followed by a recontact question).
questionnaire/interview
Other SID-related questions Awareness of/personal experience of unfair treatment due to
sexual orientation.
Sexual Identity question "Please look at this card. Which of the answers on the card
best describes you? Please just give me the letter
alongside the appropriate answer." (Show card categories
labelled R, D, P, H).
R: Straight or heterosexual
D: Gay or lesbian or homosexual
P: Bisexual
H: Other
Explanations/definitions/assu Prompt for interviewers to use if respondents objected to or
rances provided to queried the question:
respondent "We’re collecting this information to find out more about the
discrimination people may face for different reasons. The
information is kept completely anonymous."
Interviewer instructions Interviewers received written instructions. No briefing.
Estimates (weighted % All Valid
unless otherwise indicated) Straight/heterosexual 87.8 96.5
Gay/lesbian/homosexual 2.0 2.2
Bisexual 0.4 ?
Other 0.7 0.7
Refused 9.0 -
Base =2704
Non-response / non- A 'refusal' category was available to interviewers. It was not
substantive responses shown on the card in order to maximise response to the
question.
Pre-testing, piloting, changes On the pilot, self-completion was also tried, using a CAPI
made over time pen, which was reported to have been "slightly preferred" by
respondents as "more anonymous". But it broke up the
interview and was a struggle for certain respondents. The
show card version was "not met with any resistance",
though it was said that interviewers knew the answer so it
was not confidential. However, it appears that all pilot
respondents were straight - different views might be
expressed by LGB respondents.
Feedback from researchers, • The Fair Treatment at Work Survey asked "Was there
interviewers, respondents any reluctance hesitation at all when answering the
earlier question about sexual orientation?" To which the

43
 responses were Yes: 10%; No: 89%; Don't know: 1%.
• The "considerably higher levels" of refusal to answer,
and of interviewer perceptions of reluctance or
objection, have been ascribed by DTI to "a small
number of interviewers" coding a refusal for the
majority or all of their interviews. The refusals did not
occur at random, suggesting "that some interviewers do
not feel able to ask this question and record a refusal,
instead of asking the question and recording the correct
response." It is worth noting that those interviewers
received no briefing. DTI conclude that "Future
attempts to ask a question on sexual identity must pay
particular attention to interviewer training and briefing to
ensure interviewers feel able to ask the question."
Links

44
 1.5 Count Me In: The National Mental Health and Ethnicity Census 2005 Service User Survey

See also 1.6 below regarding experiences of the 2006 NHME Census itself.

Sponsor Mental Health Act Commission

Information provided by NatCen - provided expertise in the questionnaire and sample
design of this survey, conducted by another fieldwork
contractor.
Year conducted 2005
Survey mode PAPI. “Action research” was employed: interviews were
conducted by non-professional interviewers - people with
mental health problems themselves (“trained service user
interviewers from a wide range of ethnic backgrounds”), and
there were problems with administering the sample properly.
The aim was to conduct interviews in a private place in the
hospital but that was dependent on availability.
Sexual identity question PAPI (concealed response show card).
mode
Sample design The survey was of in patients in mental health hospitals - a
sample taken in each of 41 NHS and private/voluntary
hospitals.
Survey topics Patient needs and services received. To compare the
experiences of BME patients with white patients (e.g. how
appropriate were the services received, with reference to the
culture of patients).
Survey response 31%.
Sexual identity question To explore whether sexual orientation was a factor having a
purpose bearing on patient needs or service received. The question
had been included in the survey rather than the ‘main data
capture element’, i.e. the 2005 MH Census, because of
“anticipated difficulties in collection” – not elaborated - (from
Census 2006 ‘Inclusion of data capture field on sexual
orientation: Background and supporting information’). NB In
the 2006 MH Census, a question was included – see below.
Location in Dk
questionnaire/interview
Other SID-related questions -
Sexual Identity question Which of the categories on this card would you say describes
your sexual orientation? Please give me the letter.
M. Heterosexual (like people of the opposite sex)
Z. Gay male or lesbian female (like people of the same
sex)
P. Bisexual (like people of both sexes)
C. Transgender

Categories were given random letters so as not to imply any

hierarchy and to ensure privacy.
Explanations/definitions/assu Dk
rances provided to
respondent
Interviewer instructions Dk
Estimates (weighted % n/a. Achieved sample <400 (white and BME) was too small to
unless otherwise indicated) allow meaningful analysis.
Non-response / non- Don’t Know and refused options available to interviewers.
substantive responses
Pre-testing, piloting, changes -
made over time
Feedback from researchers, None.
interviewers, respondents
Links Report accessible from this web page:
http://www.mhac.org.uk/Pages/publications.html

45
 1.6 Count Me In: The National Mental Health and Learning Disability Ethnicity Census 2006

Sponsor Mental Health Act Commission

Information provided by Mental Health Act Commission
Year conducted 2006
Survey mode IT staff in NHS, private and voluntary sector mental health
and learning disability establishments were compiled the
required information in a spreadsheet – much from
administrative records – see also below.
Sexual identity question PAPI. A member of the staff on each ward, assumed to
mode usually be nursing staff, administered the question. They were
informed by their IT people of what gaps there were in the
information required, and expected to look it up from notes or
ask the patient. Sexual orientation is not routinely collected
administratively.
A “flash card” (show card) was used, to promote privacy and
sensitivity.
Sample design The Count Me In Census of mental health and learning
disability inpatients is conducted annually, of all patients in all
MH and LD establishments in England & Wales. A Census
Day is fixed (31 March). Question intended to be asked of all
patients >16.
Response Dk
Survey topics Age, sex, preferred language, religion/faith and health related
questions. Most of the information required is, in theory,
already collected for administrative or medical reasons, but
this is not always the case in practise.
Sexual identity question The main impetus is that DH has a 5 year programme to
purpose improve the service to BME patients. A sexual orientation
question was introduced in 2006. To provide a baseline so
services can be appropriate.
Location in -
questionnaire/interview
Other SID questions -
Sexual Identity question Which of the following terms would you use to describe your
sexual orientation?
2 Lesbian/Gay/Homosexual
1 Straight/Heterosexual
3 Bisexual
4 Other
5 Do not wish to answer

Question and response categories both shown on flash card.

Explanations/definitions/assu A ‘your questions answered’ booklet was available to patients,
rances provided to including “Why are you asking me about sexual orientation?
respondent This is really very personal.” The response explained that:
• LGB people sometimes don’t receive the same standard
of care and that they have more mental health problems,
so MHAC wants to know how to improve services.
• But it doesn’t know how many there are.
• You do not have to answer the question.
Interviewer instructions A briefing note was sent to all participants, covering the
following points:
• It stressed confidentiality.
• People who hadn’t considered or didn’t know their sexual
orientation, or did not fall within options, and those for
whom it was deemed inappropriate were to be recorded
as ‘other’.
• It said that staff’s discomfort should not be a reason for
not asking and using ‘other’.

46
 • Patients did not have to reply.
• It explained difference between orientation and activity.
Estimates (weighted % Full results not published. 2.1% of all records were coded LG
unless otherwise indicated) or B. But many providers (nearly 50%) did not return any
records coded LGB – they might not have asked the question.
Of those providers who returned any LGB cases, the rate was
3.1%. They ranged between under 1% to over 10%. Highest
proportions were in private and voluntary establishments
rather than NHS (were felt to have a better environment).
Non-response / non- There was no category for ‘not asked’ (will be from 2007) but
substantive responses a value had to be entered somewhere. So ‘other’ might have
included the cases where the question wasn’t asked, as well
as genuine others/don’t knows.
Pre-testing, piloting, changes -
made over time
Feedback from researchers, • No information on feedback from patients.
interviewers, respondents • Problems were encountered with many providers
administering the Census not asking the question.
Providers were given a feedback questionnaire; only a
third responded to a question on the % of patients who
had been asked the SO question; of which 63% said
they had asked 75>% of patients; 14% had asked <25%.
• MHAC conducted road shows with providers to discuss
issues about conducting the Census. There were some
“shocking” findings including views that the question
didn’t need to be asked, you could tell just by looking if
someone was gay (no information on how often it was
answered by providers rather than asked of patients);
and how can staff be expected to ask - it was too
embarrassing. But there was also positive feedback from
the establishments which had administered the question
to more people.
Links www.mhac.org.uk/census

47
2 SELF-ADMINISTERED QUESTIONS IN INTERVIEWER-ADMINISTERED SURVEYS

2.1 British Social Attitudes Survey

Sponsor Various bodies. ESRC for sexual identity/orientation.

Information provided by NatCen (research and fieldwork contractor)
Year conducted 2005
Survey mode CAPI
Sexual identity question Paper self-completion. Booklet given at end of interview.
mode Preferred model is for the interviewer to return to the
household to collect the completed questionnaire.
Alternatively respondent posted it back.
Sample design Random probability, general population, GB. One person
selected in household. The sample was subdivided - different
versions of the questionnaire were given to each randomly
selected subgroup (so more questions could be administered
over the whole sample, for cost efficiency). The SI question
was not on all versions.
Survey response Dk
Survey topics A range of topics from Health to Transport to general attitude
statements. Sexual identity question within social identities
module.
Sexual identity question The question was not primarily intended to measure
purpose prevalence. It was part of a self-completion module on social
identities.
Location in Self completion booklet. Question 38 out of 55.
questionnaire/interview
Other SID questions Question on how much respondents felt they had in common
with others identifying in this way: a lot more, a little more, or
no more than with other people.
Sexual Identity question Which of the following best describes how you think of
yourself?
1. Heterosexual (‘straight’)
2. Gay
3. Lesbian
4. Bisexual
5. Transsexual
6. Can’t choose
Explanations/definitions/assu None (booklet anonymised).
rances provided to resp.
Interviewer instructions n/a
Estimates (weighted % Heterosexual (‘straight’) 95.3
unless otherwise indicated) Gay: 0.4
Lesbian: 0.4
Bisexual: 0.6
Transsexual: 0.2
Can't choose: 1.7
Not answered: 1.4
N=1732
Non-response / non- ‘Can’t choose’ response category provided. Respondents
substantive responses able to skip.
Pre-testing, piloting, changes The pilot study didn't ask a sexual identity question but one
made over time where people chose from shuffle cards some identities which
they related themselves to (inc heterosexual and gay). No
respondents identified as LGBT.
Feedback from researchers, One respondent contacted them to say that he didn't feel
interviewers, respondents comfortable ticking his sexual orientation on a paper
document, which might then be seen by other people.
Links http://www.natcen.ac.uk/natcen/pages/or_socialattitudes.htm
Questionnaires/topics/methodology:
http://qb.soc.surrey.ac.uk/surveys/bsa/bsa05.htm

48
 2.2 Newham Household Panel Survey

Sponsor London Borough of Newham

Information provided by Newham Corporate Research Unit
Year conducted Annual waves since 2002
Survey mode Face to face interview. Interviewers speak a range of
community languages that Newham residents speak and
can carry out interviews in the languages required.
Questionnaires are also translated into community
languages.
Sample design A simple random sample of addresses (every 56th address).
People stay in the panel as long as they wish to - if they
move within Newham they are traced if possible, though
there is a high attrition rate due to this. When respondents
have moved, the new occupants are invited to join. New
addresses are added each wave.
All adults (16>) in the household are interviewed.
Survey topics SC booklet: mental health, satisfaction with various aspects
of life, questions about harassment/discrimination (which
precede the sexual identity question) and questions about
friends.
Survey response information Number of achieved households over waves ranged from
857 to 1051. Number of individuals ranged from 1481 to
1828.
Response rates for self-completion:
W1: 75%
W2: 82%
W3: 78%
W4: 43%
(% of all completed individual interviews)
Sexual identity question To provide context for other questions about
purpose discrimination/victimisation.
Sexual identity question Paper self-completion booklet. This is administered to
mode of administration individuals before or after their individual questionnaire, the
appropriate time being judged by the interviewer. The
booklet can be completed then and given to the interviewer,
or later and posted back.
Location in 11th of 14 questions
questionnaire/interview
Other SID questions Questions about unfair treatment, harassment/abuse or
discrimination by employers because of sexual orientation.
Sexual Identity question In Wave 4:
Please tell us what best describes you:
I am heterosexual or 'straight'
I am 'gay' or 'lesbian' (homosexual)
I am bisexual
If none of the above applies. (PLEASE WRITE): I
am______________________
I do not wish to answer this question.
Explanations/definitions/assu Preamble used up to Wave 3 but dropped – see below. The
rances provided to booklet explains confidentiality, anonymity and data
respondent protection. It contains a serial number to match to the
questionnaire.
Interviewer instructions n/a
Estimates (weighted % See table below
unless otherwise indicated)
Non-response / non- Total % who left question blank or answered ‘I do not wish
substantive responses to answer…’ See estimates below for breakdown.
Wave 1: 23.1
Wave 2: 20.5
Wave 3: 25.0

49
 Wave 4: 21.5
Pre-testing, piloting, changes A preamble was formerly used, but dropped from Wave 4. It
made over time read "Some people have a partner of the same sex as
themselves (homosexual) while others have a partner of the
opposite sex (heterosexual)."
Feedback from researchers, • ‘none of the above’ – verbatim responses not available.
interviewers, respondents • None from interviewer debriefing or from respondents.
No apparent problems.
• In all waves ‘white’ respondents were significantly more
likely to answer the SID question (1/7 did not answer),
‘Asian’ (3/10) and ‘other’ (base too small) ethnic groups
significantly less likely.
• No analysis of individual responses from wave to wave
was available.
Links http://www.newham.info/research/NHPS.htm

Survey Estimates
% of all individual % of all cases % of valid cases
Wave 1
interviews where s/c returned
Heterosexual or straight 55.3 74.5 97.0
Gay or lesbian (homosexual) .0 .0 .0
Bisexual .2 .3 .4
None of the above 1.5 2.0 2.6
Do not wish to answer this question 10.1 13.7 -
Total answering question N=988
Not answered on s/c 7.0 9.5 -
Total returning s/c form 74.2 N=1286
Self completion not returned 25.8
Total individual interviews N=1733

% of all individual % of all cases % of valid cases

Wave 2
interviews where s/c returned
Heterosexual or straight 61.8 75.2 94.5
Gay or lesbian (homosexual) .6 .7 .8
Bisexual .4 .4 .5
None of the above 2.7 3.3 4.1
Do not wish to answer this question 9.6 11.7 -
Total answering question N=969
Not answered on s/c 7.2 8.8 -
Total returning s/c form 82.2 N=1218
Self completion not returned 17.8
Total individual interviews N=1482

% of all individual % of all cases % of valid cases

Wave 3
interviews where s/c returned
Heterosexual or straight 55.0 70.4 93.9
Gay or lesbian (homosexual) .5 .7 .9
Bisexual .5 .6 .8
None of the above 2.6 3.4 4.5
Do not wish to answer this question 11.7 15.0 -
Total answering question N=849

Not answered on s/c 7.8 10.0 -

Total returning s/c form 78.2 N=1132
Self completion not returned 21.8
Total individual interviews N=1448

50
 % of all individual % of all cases % of valid cases
Wave 4
interviews where s/c returned
Heterosexual or straight 31.4 71.7 91.2
Gay or lesbian (homosexual) .4 1.0 1.2
Bisexual .5 1.1 1.4
None of the above 2.1 4.7 6.0
Do not wish to answer this question 5.3 12.0 -
Total answering question N=500

Not answered on s/c 4.2 9.5 -

Total returning s/c form 43.9 N=636
Self completion not returned 56.1
Total individual interviews N=1451

51
3 SELF-ADMINISTERED - POSTAL SURVEYS

3.1 Scottish Census Small Scale Test

Sponsor General Register Office Scotland

Research and/or fieldwork General Register Office Scotland
contractor
Information provided by Obtained from report ‘Sexual Orientation in the Census’, on
the small scale postal test
Year conducted 2005
Survey mode Postal
Sexual identity question Paper self-completion.
mode of administration
Sample design • Representative random sample of households. One
person in the household, nominated by the
householder, was asked to fill in the form, for him or
herself. So issues relating to privacy within the
household, which might apply in the real Census, were
not addressed.
• Four variants of the form were used, two of which
included a 'sexual orientation' question, and two of
which included a question on discrimination
experienced on the basis of gender, ethnicity, sexual
orientation and seven other grounds. The
discrimination question was included to see if it had any
effect on answers to the sexual orientation question.
The following combination of these two questions was
each sent to a quarter of households:
o sexual orientation/discrimination;
o sexual orientation/no discrimination;
o no sexual orientation/discrimination;
o no sexual orientation/no discrimination.

Response information 31% (biased towards respondents aged over 30, females,
and those in less deprived areas, suggesting response
cannot be generalised to the whole population). Response
rates to the variants were very similar.

Survey topics Socio-demographic

Sexual identity question To assess feasibility of asking in 2011 Scottish Census (to
purpose provide a baseline)
Location in The question was number 17 of 29, after religion, ethnicity
questionnaire/interview and confidence about going out in the evening in your area,
and before caring and employment status.

Other SID questions Experience of discrimination – see above

Sexual Identity question Which of the following best describes your sexual
orientation?
Heterosexual
Lesbian/Gay
Bi-sexual
Other
Prefer not to answer

Explanations/definitions/assu Instruction between question stem and response categories:

rances provided to ‘Only answer this question if you are 16 years old or more.’
respondent Confidentiality was mentioned in the cover letter. Nothing
was mentioned about the voluntary nature of questions (but
there was a 'prefer not to answer' category).
Interviewer instructions n/a

52
Estimates (weighted % Unweighted Total Both No
unless otherwise indicated) data % Qs discrimination
Q
Heterosexual 83.1 82.7 83.5
Lesbian/gay 0.9 1.4 0.3
Bi-sexual 0.4 0 0.9
Other 0.9 0.6 1.2
Prefer not to 8.5 9.2 7.8
answer
Non response 6.2 6.1 6.4
Base 692 346 346
Non-response / non- • 'Prefer not to answer' and item non-response far
substantive responses outweighed response to the LGB categories.
• Item non-response is higher than for religion and other
questions (but lower than for caring, the next question
on the form - a possible order effect?).
Pre-testing, piloting, changes -
made over time
Feedback from researchers, •A 'specify' box for other was not included, considered
interviewers, respondents relevant only to a small number of respondents and a
potential ‘magnet’ for facetious responses.
• The conclusion of the author was that the "accuracy"
and “utility" of the data is questionable.
• The researchers conclude that the "absence of a
difference in response rates between the variants ...
may indicate that the most compliant sections of the
population ...are not put off by the inclusion of a sexual
orientation question."
• At the end of the form respondents were asked to
indicate how happy they felt about each question. "The
question on sexual orientation stands out with around
11% of respondents selecting very unhappy or
unhappy" compared to less than 5% for most
questions.
• Respondents could also write comments. For the
sexual orientation question they "ranged from fully
accepting to hostile".
• People who "accepted the question without
reservation" included people answering 'prefer not to
say'.
• Concerns expressed included:
o "not understanding the need for the data";
o "offence" and "unhappiness at the intrusive
nature of the question"; and concerns about
confidentiality (e.g. being asked for their
postcode). They did not necessarily "translate
into refusal to record an answer" [to the
question].
• Some respondents' comments "suggested they had
possibly misunderstood or misinterpreted the question".
These included an elderly woman who answered the
question with a sexual orientation but did not think the
question applied to her, and another who had to think
what the question meant but answered as best she
could.
• Non-responders to the question "generally also
declined to record any comments".
Links Report: http://www.gro-
scotland.gov.uk/census/censushm2011/question-
development/sexual-orientation-in-the-census.html

53
 3.2 Civil Service Diversity Survey

Sponsor Cabinet Office

Information provided by ORC International (fieldwork contractor)
Year conducted 2001
Survey mode Postal.
Sexual identity question Paper self-completion.
mode
Sample design Simple random sample of civil servants from every
department, plus boosts. Paper questionnaire sent to the
selected sample's work address.
Survey response 48%.
Survey topics Subjects relating to the respondent’s job, including work-life
balance; diversity/fairness/equality;
harassment/discrimination; dealings with members of the
public and career development.
Sexual identity question Equality and diversity monitoring.
purpose
Location in At the end of the questionnaire, in a section on demographics,
questionnaire/interview as the penultimate question. It followed question on gender
and preceded a question asking if the respondent considered
him or herself to be transsexual.
Other SID questions Question about unfair treatment at work as a result of various
characteristics including ‘sexuality’.
Sexual Identity question If you have no objections to stating your sexuality, please
could you state it here?
Bisexual
Gay man
Heterosexual
Lesbian

The question and answer categories were provided by

Cabinet Office.
Explanations/definitions/assu The cover letter mentioned confidentiality e.g. that
rances provided to respondents would not be identifiable in the published results
respondent and that no-one in the Civil Service would see their data. The
questionnaire was to be returned in a prepaid envelope to
ORC, and was anonymised - no serial number/barcode or
name was recorded. The department would be recorded but
nothing about the unit in which the respondent worked.
Interviewer instructions -
Estimates (weighted % Bisexual: 2%
unless otherwise indicated) Gay man:1%
Heterosexual: 89%
Lesbian: 0%
Missing: 8%
Non-response / non- • No 'prefer not to say' category was included; the
substantive responses researchers thought it a waste of a box and that
respondents could just leave the question blank instead.
• No 'other' category was included as it was thought the
categories captured all eventualities.
Pre-testing, piloting, changes In developing the survey focus groups were conducted,
made over time during which people said they accepted the question's use in
monitoring diversity and unfairness. Concerns included being
able to be identified individually, and at the accuracy of any
count based on the question (no further details available).
Feedback from researchers, The researcher had no recollection of any feedback from
interviewers, respondents survey respondents.
Links http://www.civilservice.gov.uk/diversity/publications/index.asp
(scroll down to Diversity Survey).

54
 4. SURVEYS IN FIELD DURING TIME OF REVIEW – NO ESTIMATES AND LIMITED FEEDBACK
AVAILABLE

4.1 Adult Psychiatric Morbidity Survey 2006-7

Sponsor Information Centre for Health and Social Care

Information provided by NatCen - research and fieldwork contractor
Year conducted October 2006- September 2007
Survey mode CAPI
Sexual identity question CASI. May be administered by interviewer in a small
mode proportion of cases where respondents unable to self-
administer – interviewer reads question and respondent keys
answer.
Sample design Random probability; general population (England);
1 adult selected in each household, age 16 and over (no
upper age limit as there was in APMS 2000 [age 74])
Survey response information In progress.
Survey topics Focus on mental health (common mental disorders,
psychosis, personality disorder) with additional questions on
other topics such as general health, use of services,
intellectual functioning, alcohol consumption, key life events,
socio-demographics.

Sexual identity question Designed to compare two versions of the sexual identity
purpose; definition of question (see below). Version assigned at random during
SexID/orientation course of interview (CASI controlled).

Also, evidence from earlier research of variation in rates of

psychiatric morbidity by sexual identity.

Sexual identity may be conceptualised in terms of attraction,

identity and/or behaviour. The two questions used are a basic
way of assessing feelings and behaviour (see questions A1,
A2; B1, B2).

Location in At the end of a CASI module in the middle of the interview.

questionnaire/interview Module takes c15 minutes. Covers general health; use of
health services; drinking; drug use; gambling; personality
disorders; violence; self-harm. In section 'Discrimination and
Sexual Identity', the sexual identity questions are asked after
questions about discrimination.

Other SexID-related There is a question about sexual partners, asked as a follow

questions up to the question about sexual identity. Two version of the
question about sexual partners were used - see questions A2
and B2 in section below.
Sexual Identity question Two versions of the sexual identity question (A1 and B1) and
follow up question on sexual partners (A2 and B2).

VERSION A
A1. Sexori
Which statement best describes your sexual orientation?
This means sexual feelings, whether or not you have had any
sexual partners. "

Entirely heterosexual (attracted to persons of the opposite

sex)
Mostly heterosexual, some homosexual feelings
Bisexual (equally attracted to men and women)
Mostly homosexual, some heterosexual feelings)
Entirely homosexual (attracted to persons of the same sex)

55
 Other

A2. Sexpart
Have your sexual partners been...
/ "Sexual partners"
:
only opposite sex",
mainly opposite sex but some same sex partners",
mainly same sex but some opposite sex partners",
only same sex",
or, I have not had a sexual partner")

VERSION B
B1. Sexdes
Please choose the answer below that best describes how you
Think of yourself...:
completely heterosexual
mainly heterosexual
bisexual
mainly gay or lesbian
completely gay or lesbian
Other

B2. SexPart2
Sexual experience is any kind of contact with another person
that you felt was sexual (it could be just kissing or touching, or
intercourse, or any other form of sex).
Has your sexual experience been..."
:
Only with women/men (or a woman/man), never with a
man/woman,
More often with women/men, and at least once with a
man/woman,
About equally often with women/men and men/women,
More often with men/women, and at least once with a
woman/man,
Only with men/women (or a man/woman), never with a
woman/man,
I have never had any sexual experience with anyone at all"

Explanations/definitions/assu No specific assurances about this question, but all

rances provided to respondents assured of confidentiality of all data provided in
respondent the study.

Interviewer Interviewers shown questions and told about the different

instructions/briefing versions tested in this study, but are not given any special
instructions.

Estimates (weighted % None available

unless otherwise indicated)
Non-response / non- Not yet available. To refuse or answer ‘don’t know’
substantive responses respondents would have to inform the interviewer.
Pre-testing; piloting; changes Cognitive pilot carried out. No feedback on this question
made over time available but no change made.
Feedback from researchers, Not yet available
interviewers, respondents
Links None

56
 4.2 Citizenship Survey

Sponsor Communities and Local Government

Information provided by Communities and Local Government
Year conducted 2007
Survey mode CAPI
Sexual identity question CAPI (concealed response show card)
mode
Sample design Random probability; general population, England and Wales.
One person selected in household.
Survey response information Not yet available
Survey topics Social networks; community; influencing politics; trust;
volunteering; civil renewal; race; and religion.

Sexual identity question Requested by Women and Equality Unit and others: Will
purpose; definition of provide an indicator of the respondents self-reported sexual
SexID/orientation orientation/identity.
Location in In the middle of the second of two demographics sections,
questionnaire/interview towards the end of the questionnaire. Preceded by: media
consumption (newspapers, radio, TV, internet); vehicle
ownership; health; and caring. Followed by: employment
status; main job; qualifications; and income (banded).
The first demographics section, at the start of the
questionnaire, covers: sex; marital status; tenure; ethnicity;
nationality; whether employed.
Other SexID-related A question on personal identities: "Suppose you were
questions describing yourself, which of the things on this card would say
something important about you?"
The 16 response categories include work, family, income and
several identities including ethnicity, religion, social class,
gender and "your sexuality". The question is multicoded,
followed by one asking which is "the most important thing'.
Sexual Identity question (Asked if interview is not being translated by family member
or friend)

Looking at this card, which of the options best describes your

sexual identity? Please just read out the letter next to the
description.

W …Heterosexual or straight
P … Gay or lesbian
H … Bisexual
S …Other
G … [or would you] Prefer not to say?

[NB Abridged question shown on show card: Which of the

following best describes your sexual identity?]

Ask if =Other

How would you describe your sexual identity?

(Open question).

Explanations/definitions/assu Preamble:
rances provided to The next question is about sexual identity. We are asking this
respondent question because the government department funding this
study, Communities and Local Government, is responsible for
helping to reduce all forms of prejudice and discrimination in
society.

Interviewer

57
instructions/briefing
Estimates (weighted % Not yet available
unless otherwise indicated)
Non-response / non- Not yet available
substantive responses
Pre-testing; piloting; changes Cognitive testing conducted by NatCen, with regard to
made over time respondent reactions, objections and comprehension.

Question tested:

“The next question is about sexual identity. We are asking this

question because the Department for Communities and Local
Government is responsible for helping to reduce all forms of
prejudice and discrimination in society. As with all our
questions, your answers will be kept strictly confidential.

Please read out the letter which best describes your sexual
identity?
C. …Heterosexual or straight
B. ... Gay or lesbian
D. ... Bisexual
A. ...Other
E ...[or would you] Prefer not to say?

(If Other) How would you describe your sexual identity?

Feedback from researchers, Cognitive testing: no objections, refusals encountered, though
interviewers, respondents “several were slightly embarrassed”. One LGB respondent
was interviewed (the testing was not specific to this question).

Respondents either said the response option in full or just the

letter.

Comprehension of the question included “who you like as a

partner” and “it’s what you are or your sexual preference is”.
There was some confusion about the term ‘heterosexual’ - a
process of elimination was used or people said they would
have chosen “Straight” if that had been an option on its own.

It was recommended that the question be included in the

dress rehearsal.

NB the confidentiality clause was dropped – interviewers felt

that it was unnecessary and made asking the question more
of a "big deal"
Links

58