Question development history for 2018
Census: Gender Identity
Overview
Summary
Key driver for question development
New Content
Quality priority level:
TBD
Outcome from question development
Cognitive testing
1 – Purpose
The purpose of this document is to capture the question development process for the Gender
Identity variable, including findings from waves of testing conducted in 2015-2017.
The 2018 variable specification provided by the Customer Needs and Data (content) Team provides
the background and scope of this variable.
This document is intended for use within Statistics New Zealand.
2 – Background
Definition from statistical standard:
Gender identity is an individual’s internal sense of being wholly female, wholly male, or having
aspects of female and/or male. Gender identity is understood to refer to each person’s deeply felt
internal and individual experience of gender, which may or may not correspond with the sex
recorded at birth (adapted from International Commission of Jurists, 2007, p6). A person’s gender
identity can change over their lifetime, and can be expressed in a number of ways and forms. This
expression includes outward social markers, such as name, clothing, hairstyles, mannerisms, voice,
and other behaviours.
3 – Design differences between paper and internet forms
n/a
4 – Findings from testing (or review) and rationale for revision
These tables summarise in chronological order the versions of this question set that were tested (or
reviewed), along with brief findings, and rationale for revision.
Reasons for variables being omitted from a sprint may include: the content need or question design
is not ready, or the variable is not a focus for that sprint (eg it is not suited to the target
respondents), or the sprint is not a test of content.
Summary of sprints this variable has been tested in, plus testing type and mode type:
Sprint 4; cognitive testing and mass completions of paper forms
Sprint 5; cognitive testing and mass completions of paper forms
SPRINT 4; COGNITIVE TESTING AND MASS COMPLETIONS OF PAPER FORMS
March 2016
Christchurch and Wellington
Aim: General
The primary objective of the testing is to provide recommendations to inform a Go/No Go
decision on future development and testing of proposed 2018 Census content.
Targeted LGBTQI+ Testing
Gain insights into acceptability, especially within a census context.
Respondents The cognitive test participants included members of the public, students, and people with step
family. The mass completion test participants included rural fire fighters, secondary students, and
tertiary students (including young parents’ college and ESOL students).
The question design tested in sprint 4 was new:
SPRINT 5, COGNITIVE TESTING AND MASS COMPLETIONS OF PAPER FORMS
March 2015
Wellington and Christchurch
Aim: General
The primary objective of the testing is to provide recommendations to inform a Go/No Go
decision on future development and testing of proposed 2018 Census content.
Targeted LGBTQI+ Testing
Gain insights into acceptability, especially within a census context.
Respondents As with the previous sprint, cognitive test participants included members of the public, students,
and people with step family. The mass completion test participants included secondary and
tertiary students, Age Concern, retirement village residents, and a private workplace.
The question design tested in sprint 5 was:
SPRINT 5 – COGNITIVE TESTING AND MASS COMPLETIONS OF PAPER FORMS – FINDINGS
General Population
Many respondents had a sense of deja vu’ when they got to the gender question. In
many cases the sex-based routing given in the ‘babies born alive’ question immediately
preceding gender in Sprint 5 alerted respondents to there being two similar looking
questions in the form. In Sprint 4, many more respondents didn’t have a sense of Deja-vu
due to the questions being spaced far apart on the form and the absence of sex based
routing prior to the gender question.
Some respondents realised there was a difference between the sex question and the
gender question, but others remained confused.
Some respondents understood the difference between sex (biological) and gender
(identity), but still had difficulty understanding that the categories were not
dichotomous.
Many respondents felt that the distance between the two questions (sex and gender)
was odd and wondered why the gender question was not placed together with the sex
question on the form.
LGBTQI+ testing
Respondents in the LGBTQI community liked the inclusion of this question and several
felt the question was clearer than the sex question.
Some trans respondents selected ‘gender diverse’, while others selected either male or
female.
Some trans respondents selected either male or female for sex and the same for gender,
so it
was not possible to identify them as trans from looking at the form.
Some respondents queried if they could select multiple responses.
SPRINT 5 – RECOMMENDATIONS
Recommend inclusion of Sprint 5 question version for Volume Test with help
information available.
While testing to date shows general acceptance of the gender question, we understand
that the testing done has likely gained the perspectives of the most compliant
respondents and true feelings may be masked due to the presence of the interviewer.
Inclusion in the Volume Test will allow us to understand reaction to the gender question
more fully in a larger scale and non-observed test environment. Examples may include
volume of calls made to call centre, access to help information and for paper forms,
multiple response and form annotations.
Recommend development of some kind of note to explain and affirm that for many
people ‘sex’ and ‘gender identity’ will be the same thing as a means to reassure
respondents we are not in fact asking the same question twice.
Question as to whether the information we can collect will meet information need
expressed in topic specifications is still an issue and one we need to take advice on from
Customer Needs and Data.
Recommend removal of the ‘babies born alive’ question in the context of the sex and
gender questions.
Both ‘babies born alive’ and ‘gender identity’ are questions that must come after the
routing to identify the NZ Resident Adult population as neither questions are appropriate
to ask of children. Further, if asked, we recommend ‘babies born alive’ question precede
the ‘gender identity’ question so ‘male’ respondents are routing on the basis of their sex,
rather than have the complication of which basis to route. However, this increases the
sense of ‘deja vu’ for the majority of respondents for whom biological sex and gender
identity are the same.
Regardless of the effects of asking babies born together with gender identity, it is
QM&Ds view that it would be preferable to remove ‘babies born’ in the interests of
reducing overall respondent burden, especially given the continued and consistent
feedback we get from respondents regarding the insensitivity of the question.
Recommend some advice from experts in Māori language/Māori world-view in the
implications for asking in Māori language specifically and of Māori respondents in
general.
5 – Data quality
<Expectations based on testing, known issues, question interactions (suggested edits). To be
completed towards the end of QMD testing>
Appendix 1: testing methodology
Research objectives
The broad research objectives of testing may vary with each sprint, but generally are to:
Understand how well individual questions and key concepts/definitions are understood by
respondents
Understand how well individual questions and the overall form design enables respondents
to answer quickly and accurately
Understand how new and changed questions may impact on other questions in the forms
Understand respondent burden
Understand public attitudes to new and changed questions which may influence their
willingness to answer
Topics or questions may be allocated as a primary or secondary focus or not a focus of testing in a
given sprint. This depends on the priority of the variable itself and how well it has tested previously.
Desktop review (paper and online)
Questions and questionnaires (paper and/or online) are reviewed before they are tested with
respondents. The aim of desktop review is to:
Check whether the forms accurately match content and design specifications;
Identify any usability issues in the online forms (across a range of devices, operating systems
and browsers);
Identify any potential issues that should be subjected to further testing with the public.
Test participants
Testing aims to include people from a wide range of backgrounds, with a mix of age, sex, ethnicity,
income, employment status, etc. However an individual sprint may target respondents with
particular characteristics, for example, students, people who have children or stepfamily, Māori, or
tenure (renting, home owners, etc).
Test participants have been recruited using a variety of methods. These have included flyers posted
in public spaces such as libraries and YMCAs, Twitter and Facebook posts, contacting community
groups eg LGBQTI+, Step Family Network and the Retirement Village Association.
Testing methods
Three testing methods have been used, each with a different focus.
Cognitive testing
This is a qualitative, observational research method that helps identify problems with questionnaire
design. This methodology involves one-to-one interviews where respondents complete a
questionnaire. It uses techniques such as concurrent probing, retrospective probing and think-aloud
to highlight how respondents get to their answers and how they interpret certain terms.
Cognitive tests last around one hour, during which the first 30-40 minutes will involve the researcher
observing the respondent completing their dwelling form and individual form. The remaining 20-30
minutes will take a semi-structured interview approach. This time will be used to probe in-depth on
the focus questions described in this plan, which are relevant for respondents.
Mass completion + group interview
Mass completion tests involve asking a large group of respondents to complete a questionnaire
unobserved, in a supervised environment. Mass completion is a useful diagnostic tool to confirm
suspicions about a particular design or uncover unexpected reactions to questions using a larger
group of respondents.
Mass completion and group interview will last about one hour. In the first half of the session,
respondents will be asked to complete one or both Census forms. The remaining time will be used to
probe in-depth on the focus questions described in this plan, which are relevant for respondents.
The same semi-structured interview protocol can be used for cognitive testing and group interview.
Usability testing (online)
User testing involves one-to-one interviews where respondents complete a set of given tasks (e.g.
complete household set up page, complete Individual/Dwelling form) on a device ie a tablet,
smartphone or desktop. It is a qualitative, observational research method used to identify problems
with a user interface. User testing employs think-aloud, concurrent probing, and retrospective
probing techniques to understand how the design of the user interface impacts on the user
experience.
Analysis
From sprint 7 onwards, findings were coded to approximately 20 codes, which were further
summarised into themes:
Table: Analysis of testing findings – codes and themes used
Codes
Themes
Theme Description
Total nonresponse due to
Relates to how and why respondents
Sensitivity
sensitivity
perceive question content to be sensitive
Protest response
to themselves and other people.
Selection of ‘object to answer’
Sensitivity is often based on the individual
response
person’s personal experiences, worldview
Reluctant response
and personal values and can affect their
willingness to respond.
Sensitivity on behalf of others
Value / Value +
Relates to the explicit or implicit value
Questioning why we ask
judgements that respondents make about
Questioning use of data
a question and whether they perceive it
as having value, or not. Whether
Willingness to answer based on
respondents perceive a question to have
value judgement
value or not will affect both their
Positive comment volunteered
willingness to answer and the quality of
regarding info need
their response should they choose to
answer.
Difficulty in recalling the
Relates to the ease with which
requested information
respondents are able to answer questions
Difficulty in interpreting the
and the extent to which they have a
question
positive respondent experience. There
Difficulty in fitting their answer
Burden
are many aspects of respondent burden
into the response formats/
which respondents may experience when
categories
answering questions. Some of these arise
from ambiguous or unfamiliar terms or
Confusion or difficulties arising
concepts in the questionnaire, while
from interactions between
others may be a direct effect of the
questions
poorly designed question or form.
Effort required to answer
Missed routing instructions
Error
Relates to causes of respondent error
Instructions missed or incorrectly
that can affect data quality and reliability.
followed
Sources of error usually arise from poor
Subjective response
question and form design, but may also
Proxy response error
include contextual factors specific to the
respondent which can’t be controlled for.
Guesses
Poor question construction
Relates to respondent burden and error,
Dissatisfaction with
specifically arising from poor question
question/response options
Defective design and form design. A fundamentally
defective question or set of questions
may negatively impact on data quality
Visual design of form
and/or the user experience.
Testing collects information about people’s willingness and ability to answer. Not all of these
findings will result in alterations to the questionnaire, and any changes that are made may not
necessarily resolve the issues found.