Content uploaded by Leah Melani Christian
Author content
All content in this area was uploaded by Leah Melani Christian
Content may be subject to copyright.
Questionnaire Design Guidelines
for Establishment Surveys
Rebecca L. Morrison
U.S. Census Bureau
Don A. Dillman
Washington State University
Leah M. Christian
Pew Research Center
2
This presentation
Washington Statistical Society seminar,
17 January 2008
Based on a paper (same name) currently
under review by Journal of Official
Statistics
Available from Rebecca Morrison
(contact info on last slide)
3
Outline
Introduction
Background
Guidelines
Conclusions
Introduction
5
Introduction
Other national statistical organizations
have guidelines for questionnaire
construction
U.S. Census Bureau: guidelines for
Decennial Census only
Purpose for today: propose specific
guidelines for establishment surveys
within statistical agencies
Background
7
Background
Guidelines are recommendations
Focus on general guidelines that can be
applied across many estab surveys, typically
conducted on a wide variety of business
issues
Must take into account:
– Agency context
– Visual design research
– Respondent perspectives
8
Agency Context
Paper forms, web survey printouts often
used during data collection process
U.S. Census Bureau designs estab
surveys using multiple methods
Guidelines need to be general enough to
take these issues into account
9
Visual Design Research
Words are primary means of
communication with Rs
Rs also draw information from graphical
features of visual layout
Recent research shows how/why visual
layout and design affects interpretation
of survey questions
10
Visual Design Research
Recent theoretical developments in visual
processing
Gestalt psychology, especially these principles:
– Proximity
– Similarity
– Pragnanz
Many experimental findings relevant to
questionnaire design guidelines (too many to
mention here)
11
Respondent Perspectives
Respondents to estab surveys tend to be
answering questions as representatives
of their businesses, not as individuals
Many estab survey Rs have accounting
backgrounds; generally comfortable
with tables, matrices
12
Respondent Perspectives
Evaluation of the process of completing
questionnaires should be considered in
developing guidelines
Cognitive interviews as a technique for
improving survey design
– Qualitative meetings with respondents
– Focus on 4-step cognitive response
process model
13
Background: Summary
These guidelines link theory and
research on wording and visual layout
to results from cognitive interviews
Affected in turn by agency context and
consideration of use of multiple modes
Guidelines
Guidelines
Address issues of wording and visual
design
Guideline 1:
Phrase data requests as
questions or imperative
statements, not sentence
fragments or keywords.
17
1: Data requests as questions,
imperative statements
Requests for data can be framed as:
– Questions: question word (when, how
many, which) and a question mark
– Imperative statements: subject (“you”) is
implied, command or request is expressed
– Sentence fragment/keywords: no verb, no
punctuation
18
1: Data requests as questions,
imperative statements
Supporting theory:
– Complete sentences stand alone, reduce
need for referring to other sources of info
– Initial words of sentence lay foundation for
understanding remainder. Question word
implies that response is expected.
Cognitive interview findings: Rs prefer
questions over imperative statements
19
1: Data requests as questions,
imperative statements
Converting fragments to questions can
be relatively easy
2002 Economic Census: “Type of
municipality where this establishment is
physically located”
2007 Economic Census: “In what type of
municipality is this establishment
physically located?”
Guideline 2:
Ask additional, simple questions,
rather than fewer, more
complicated ones.
21
2: Ask additional, simple
questions
Supporting theory for asking additional, simple
questions:
– Makes the task easier and less time-consuming,
more manageable
– Complex questions overload working memory.
Result: Rs pay more attention to some words than
others
Possible solutions:
– Split complex question into pieces
– Add a filter/screener question
– Simple diagrams
22
2: Ask additional, simple
questions
Example of simple diagrams
23
2: Ask additional, simple
questions
Example: 2002 Survey of Business
Owners
24
2: Ask additional,
simple questions
2007 Survey of
Business Owners
25
Guidelines
Moving on to visual design and layout
guidelines…
Some guidelines linked under larger
themes
Establish a clear navigational path
(Guidelines 3 – 6)
27
Clear navigational path
Helps ensure that Rs complete the
questions in the intended order
Especially necessary in self-administered
surveys, where there are no
interviewers to help guide Rs
28
Complicated
navigational path
Example:
BEA’s former
quarterly foreign
direct investment
questionnaire
29
Revised
Guideline 3:
Use a consistent page or screen
layout.
31
3: Consistent page/screen layout
Rs do not have to reorient themselves to
each new page/screen
Booklet format (paper surveys) good,
since it is familiar to Rs
Generally, one-column format easier,
only have to process information in one
direction (vertically)
32
3: Consistent page/screen layout
Two columns:
– Rare for web surveys
– May be desirable in paper surveys
Survey of Business Owners:
– Categorial info about owners and business
– No numerical information
– No complex instructions
Guideline 4:
Align questions and
answer spaces / response options
34
4: Align questions and response
options
Supporting theory for aligning questions
and response options:
– Gestalt principle of proximity: visual
elements that are closer together are
perceived to be related
– Gestalt principle of good continuation:
visual elements arranged along a straight
line are more likely to be perceived as a
group
35
4: Align questions and response
options
36
4: Align questions and response
options
Arrange response options in a single
column, if possible
Multiple columns:
– Increases space between options,
increasing the risk that some options will
be missed
– Horizontal vs. vertical processing
Guideline 5:
Clearly identify the start of each
section and question
38
5: Clear start
Sections and headings used to help Rs:
– Recognize that some questions are related
– Discern basic organization of survey
– Understand what is being asked of them
39
5: Clear start
Identify questions using numbers, or
some other consistently applied font or
symbol variation
Rs often move back and forth between
paper and web questionnaires;
numbers can help ensure that Rs are
providing their response to each
question
40
5: Clear start
Guideline 6:
When the navigational flow needs
to be interrupted, use strong
visual features.
42
6: Interrupt flow using strong
visual features
Needed for branching/skip instructions
Needed to indicate change in R’s action:
Eliminate visual clutter from the
questionnaire
(Guidelines 7 - 10)
44
Visual clutter
What it is:
– Symbols, graphical information that
competes for attention
– Information that is irrelevant, from a
respondent perspective
– Apparent lack of info organization (BEA’s
old quarterly foreign direct investment
form: lines divided page into small units)
45
Visual clutter: example
46
Visual clutter
Supporting theory for avoiding visual
clutter:
– Gestalt principle of proximity
– “Near means related” heuristic
– Multiple languages (verbal, symbolic,
numeric, graphical)
Guideline 7:
Use blank space to separate
questions and make it easier to
navigate within questionnaires.
48
7: Use blank space: example
Guideline 8:
Avoid unnecessary lines that
break up or separate items that
need to appear as groups.
50
Avoid unnecessary lines: example
Guideline 9:
Use visual cues to achieve
grouping between questions and
answer spaces.
52
9: Group questions and answers:
example 1
53
9: Group questions and answers:
example 2
Guideline 10:
Avoid including images or other
graphics that are not necessary.
55
10: Avoid unnecessary images,
graphics
Symbols can be beneficial
Examples:
56
10: Avoid unnecessary images,
graphics
Symbols can be confusing, unhelpful
Example:
Use visual design to help
respondents process instructions
(Guidelines 11 – 13)
58
Instructions
Need to build common frame of reference
between survey researchers and Rs
Instructions important for conveying
correct specifications, intent of question
Rs tend to believe they understand
exactly what the question is asking, or
know answer without clarification
Guideline 11:
Incorporate instructions into the
question where they are needed.
Avoid placing instructions in a
separate sheet or booklet.
60
11: Incorporate instructions
Supporting theory for incorporating
instructions:
– Rs not likely to expend extra effort to look
for instructions in a separate place
– Likelihood of using instructions increases
when they are located with question
61
11: Incorporate instructions:
example 1
62
11: Incorporate instructions:
example 1
63
11: Incorporate instructions:
example 2
Guideline 12:
Consider reformulating important
instructions as questions.
65
12: Instructions to questions
Rs pay attention to questions more than
instructions
Can help clarify/correct reported data
Guideline 13:
Consider converting narrative
paragraphs into a bulleted list.
67
13: Paragraphs to lists
Supporting theory for using bulleted lists:
– Readers spend more time on initial
sentences of paragraphs
– Density of text is reduced
68
13: Paragraphs to lists: example
69
13: Paragraphs to lists: example
Be consistent in how answer
spaces and/or response options
are displayed
(Guidelines 14 – 15)
71
Display of answer spaces /
response categories
Important because this is where Rs
record their response
Convey type of info or level of detail
expected
Should be easy for Rs to locate, and
should stand out from the question,
instructions, other information
72
Display of answer spaces /
response categories
Supporting theory on the display of
answer spaces:
– Rs quickly decide what is foreground and
background
– Gestalt principle of pragnanz
– Gestalt principle of similarity
Guideline 14:
Use white spaces against a
colored background to emphasize
answer spaces.
74
14: White spaces, colored
background
Lighter color and smaller area mean
answer spaces become objects of
interest and become figure
Visual guide to keep answers inside
provided space
75
14: White spaces, colored
background
Example
Guideline 15:
Use similar answer spaces for the
same task.
77
15: Similar answer spaces, same
task
Rs use all available info to help in
formulating an answer, including info
provided by response categories and
answer spaces
Census Bureau (estab surveys): different
types of answer spaces for dollar info
78
15: Similar answer spaces, same
task
When requesting similar info, use the
same type and size of answer space
Electronic surveys:
– Radio buttons: only one response
– Check box: more than one response
– Some Rs might not know the difference;
augment with written instructions
Reduce the use of matrices; when
they are needed, simplify their
visual presentation
(Guidelines 16 – 17)
80
Matrices
Used often in estab surveys
Efficient use of space, but cognitively
burdensome
Rs must keep multiple pieces of info in
mind at the same time
81
Matrices: example
Guideline 16:
Limit the use of matrices.
Consider the potential
respondent’s level of familiarity
with tables when deciding
whether or not to use them.
83
16: Limit use of matrices
Supporting theory for limiting the use of
matrices:
– Reading tables is a learned skill
– Matrices may be appropriate when survey’s Rs
are likely to have learned the skill
When Rs not likely to be familiar with tables,
minimize their usage, or at least provide
more open space
84
Matrices: example
Guideline 17:
Use lines and spacing to help
respondents process information
vertically and horizontally as
needed to complete the matrix.
86
17: Matrix, using lines and space
Take advantage of Gestalt principles of
proximity and connectedness
Communicate the expected navigational
path through the matrix by
– Connecting items
– Increasing/decreasing space accordingly
87
Matrices: example
88
Matrices: example
Guideline 18:
Use font variations consistently
and for a single purpose within a
questionnaire.
90
18: Use font variations
consistently
Same font/text style for different
purposes can be confusing for Rs
91
18: Use font variations
consistently: example
92
18: Use font variations
consistently
Supporting theory for using font
variations consistently:
– Gestalt principle of similarity
– Improves usability, helps Rs learn things
quickly, focuses attention on relevant
information
93
18: Use font variations
consistently
Can be helpful to establish font variation
rules for a given survey(s)
Census Bureau’s estab surveys:
– Item numbers in reverse-print bubbles,
e.g., , ,
– Font printed in 8-point or larger
– Instructions in italics (for some paper
surveys)
94
18: Use font variations
consistently: example
95
18: Use font variations
consistently
Keycodes:
– Necessary for data entry, but should be de-
emphasized for respondents
Suggestions:
– Reduce size
– Print in a darker shade of the background color (or
gray)
– Place outside the
navigational path and
answer spaces
Conclusions
97
Conclusions
Historically: paper surveys, cram as
much onto the page as possible
Electronic methods have questioned the
wisdom of that philosophy
Also, research indicates that visual layout
and question wording affect cognitive
burden on Rs
98
Conclusions
18 guidelines for constructing
establishment survey questionnaires
Applicable to both paper and electronic
instruments
Grounded in visual design theory,
experimental evidence, research into
information processing
Informed by evidence from cognitive
interviews
99
Conclusions
This is a first step – developed for use by
one agency’s estab surveys
More issues can and should be
addressed
Need for testing to evaluate applicability
across diverse estab populations in
multiple countries
100
Conclusions
Proposed guidelines can be corroborated
by embedding experiments into estab
surveys
Additions/adjustments might be made,
especially when it comes to matrices
101
Conclusions
Ultimately, move from “what looks good
to me” to “what encourages
respondents to process and pay
attention to what is important”
102
Selected References
Dillman, D.A., Gertseva, A., and Majon-Haft, T.
(2005). “Achieving Usability in Establishment
Surveys Through the Application of Visual
Design Principles.” Journal of Official
Statistics, 21: 183-214.
Gernsbacher, M.A. (1990). Language
Comprehension as Structure Building.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Lidwell, W., Holden, K., and Butler, J. (2003).
Universal Principles of Design. Gloucester,
MA: Rockport Publishers.
103
Selected References
Tourangeau, R., Couper, M.P., and Conrad, F.
(2004). “Spacing, Position, and Order:
Intepretive Heuristics for Visual Features of
Survey Questions.” Public Opinion Quarterly,
68: 368-393.
Ware, C. (2004). Information Visualization:
Perception for Design. (2nd ed.) San
Francisco: Morgan Kaufmann.
104
Author Information: Morrison
Rebecca L. Morrison is a Survey
Methodologist in the Office of Statistical
Methods and Research for Economic
Programs at the U.S. Census Bureau.
rebecca.l.morrison@census.gov
105
Author Information: Dillman
Don A. Dillman is a Regents Professor at
Washington State University in Pullman
in the Social and Economic Sciences
Research Center
dillman@wsu.edu
106
Author Information: Christian
Leah Melani Christian is a Research
Associate with the Pew Research
Center for the People & the Press in
Washington, DC
Lchristian@pewresearch.org
107
Requests for the Paper
Rebecca L. Morrison
301-763-7595
Rebecca.L.Morrison@census.gov