diane k. willimack u.s. census bureau
DESCRIPTION
Introductory Overview Lecture Current Practices in Questionnaire Development, Evaluation, and Testing for Establishment Surveys. Diane K. Willimack U.S. Census Bureau. The Response Process in Establishment Surveys and Implications for Pretesting. Diane K. Willimack U.S. Census Bureau. - PowerPoint PPT PresentationTRANSCRIPT
1
Introductory Overview Lecture
Current Practices in Current Practices in Questionnaire Development, Questionnaire Development, Evaluation, and Testing for Evaluation, and Testing for
Establishment SurveysEstablishment Surveys
Diane K. WillimackU.S. Census Bureau
2
The Response Process in The Response Process in Establishment Surveys and Establishment Surveys and Implications for PretestingImplications for Pretesting
Diane K. WillimackU.S. Census Bureau
Survey
3
Outline
Establishments vs. households
The response process in establishment surveys
Questionnaire development, evaluation and testing methods for establishment surveys
Conclusions and challenges
4
Establishments vs. Households
What are establishments?– Economic units– Locations where business is conducted or
industrial operations performed (FCSM, 1988)
How are establishment surveys different?– Population, data, and survey design– Response process
5
Establishment Surveys Are Different!
Household Household SurveysSurveys
Establishment Establishment SurveysSurveys
Qs are about Self Business
Types of dataAutobiographical, Attitudes, Behaviors
Financial, Quantities
Data sources MemoryRecords & R’s knowledge of them
Respondent Identity Self or Proxy Informant
6
Establishment Surveys Are Different! Continued
Household Household SurveysSurveys
Establishment Establishment SurveysSurveys
R selection; R rule
Single R or Selected proxy
Multiple informants
ModeInterviewer administered
Self-administered
AuthorityVoluntary or Mandatory
Mandatory
Response Process
CognitiveCognitive and Organizational
7
The Response Process in Establishment Surveys
Cognitive response model
PLUS
Organizational processes
8
Tourangeau’s (1984)
Cognitive Response Model
1. Comprehension
2. Retrieval
3. Judgment
4. Communication Survey
9
Response Process Model for Establishment Surveys
Sudman et al., ICES-2
Organizational Organizational AspectsAspects
4. Comprehension
5. Retrieval
6. Judgment
7. Communication
1.1.
2.2.
3.3.
8.8.
Business Survey
10
Response Process Model for Establishment Surveys
Encoding in Memory / Record FormationEncoding in Memory / Record Formation
Selection / Identification of Respondent(s)Selection / Identification of Respondent(s)
Assessment of Priorities (Motivation)Assessment of Priorities (Motivation)
4. Comprehension
5. Retrieval
6. Judgment
7. Communication8.8.
1.1.
2.2.
3.3.
Release of the DataRelease of the Data
Business Survey
from Memory and / or Recordsfrom Memory and / or Records
11
The Response Process in The Response Process in Establishment Surveys Establishment Surveys
1.1. Encoding / record formationEncoding / record formation2. Respondent selection / identification
3. Assessment of priorities
4. Comprehension of the data request
5. Retrieval of data
6. Judging the adequacy of the response
7. Reporting the response
8. Release of the data
13
1. Encoding / Record Formation, continued
Needs of company records– Management– Regulations– Standards –
• Generally Accepted Accounting Practices (GAAP)
14
1. Encoding / Record Formation, continued
Encoding includes – R’s knowledge of records encoded in memory
15
1. Encoding / Record Formation, continued
If R’s knowledge of records is – – Flawed– Incomplete– Nonexistent
Response errorsResponse errors
16
1. Encoding / Record Formation, continued
In establishment surveys, encodingencoding must include – Knowledge
of RecordsPersonal Knowledge
Business Survey
17
The Response Process in Establishment Surveys
1. Encoding / record formation
2.2. Respondent selection / identificationRespondent selection / identification3. Assessment of priorities
4. Comprehension of the data request
5. Retrieval of data
6. Judging the adequacy of the response
7. Reporting the response
8. Release of the data
19
2. Respondent Selection / Identification, continued
Distributed knowledge– Multiple respondents
Organizational hierarchies– Authority
Respondent selection is under the Respondent selection is under the control of the business / establishment.control of the business / establishment.
20
The Response Process in Establishment Surveys
1. Encoding / record formation
2. Respondent selection / identification
3.3. Assessment of prioritiesAssessment of priorities4. Comprehension of the data request
5. Retrieval of data
6. Judging the adequacy of the response
7. Reporting the response
8. Release of the data
21
3. Assessment of Priorities
Financial reporting priorities1. Requests from mgmt / investors2. Annual Reports, SEC & IRS filings3. Qtr & monthly financial reports4. Regulatory requirements
(govt & non-govt)5. Other govt data requests
(Census Bureau)6. Everybody else
(e.g., trade associations)
22
3. Assessment of Priorities, continued
Financial reporting priorities Job performance & evaluation criteria Professional standards Pride
R’s MotivationR’s Motivation
23
The Response Process in Establishment Surveys
1. Encoding / record formation
2. Respondent selection / identification
3. Assessment of priorities (Motivation)
4.4. Comprehension of the data requestComprehension of the data request5. Retrieval of data
6. Judging the adequacy of the response
7. Reporting the response
8. Release of the data
24
4. Comprehension
Interpreting the form– E.g., mailing label = reporting unit
Understanding the data request– Types and levels of data requested
Interpreting the meaning of the questions– Language, terminology, jargon
25
The Response Process in Establishment Surveys
1. Encoding / record formation
2. Respondent selection / identification
3. Assessment of priorities (Motivation)
4. Comprehension of the data request
5.5. Retrieval of dataRetrieval of data6. Judging the adequacy of the response
7. Reporting the response
8. Release of the data
27
5. Retrieval of Data, continued
Retrieval incorporates –– Cognitive step – retrieving knowledge of
data sources– Access to records– Physical act of retrieving data
• Extracting information from files• Consulting multiple sources• Compiling information
28
5. Retrieval of Data, continued
Different Rs differ in their – – Knowledge of records– Access to records– Ability to retrieve data
29
The Response Process in Establishment Surveys
1. Encoding / record formation
2. Respondent selection / identification
3. Assessment of priorities (Motivation)
4. Comprehension of the data request
5. Retrieval of data
6.6. Judging the adequacy of the responseJudging the adequacy of the response7. Reporting the response
8. Release of the data
30
6. Judging the Adequacy of the Response
Does R’s answer retrieved from records or memory match the data being requested?
Business
Survey
RecordsMemory
31
6. Judging the Adequacy of the Response , continued
Does R’s answer retrieved from records or memory match the data being requested?
R doesn’t have the requested data in records. How does R estimate that figure?
Business Survey
32
6. Judging the Adequacy of the Response, continued
Does R’s answer retrieved from records or memory match the data being requested?
R doesn’t have the requested data in records. How does R estimate that figure?
R thinks, “My answer is good enough.”
Business Survey
33
The Response Process in Establishment Surveys
1. Encoding / record formation
2. Respondent selection / identification
3. Assessment of priorities (Motivation)
4. Comprehension of the data request
5. Retrieval of data
6. Judging the adequacy of the response
7.7. Reporting the responseReporting the response8. Release of the data
34
7. Reporting the Response
The act of writing the data onto the form or entering the data into an electronic instrument.
35
The Response Process in Establishment Surveys
1. Encoding / record formation
2. Respondent selection / identification
3. Assessment of priorities (Motivation)
4. Comprehension of the data request
5. Retrieval of data
6. Judging the adequacy of the response
7. Reporting the response
8.8. Release of the dataRelease of the data
36
8. Releasing the Data
Business relationships with the outside world– Review and verification of data– Reconciliation
37
8. Releasing the Data, continued
Business relationships with the outside world– Review and verification of data– Reconciliation
Consistency and Sensitivity Consistency and Sensitivity of Released Data of Released Data
– Confidentiality– Security
38
The Response Process in Establishment Surveys
1.1. Encoding / record formationEncoding / record formation
2.2. Respondent selection / identificationRespondent selection / identification
3.3. Assessment of priorities (Motivation)Assessment of priorities (Motivation)
4.4. Comprehension of the data requestComprehension of the data request
5.5. Retrieval of dataRetrieval of data
6.6. Judging the adequacy of the responseJudging the adequacy of the response
7.7. Reporting the responseReporting the response
8.8. Release of the dataRelease of the data
39
The Response Process in Establishment Surveys: Personal Cognitive StepsPersonal Cognitive Steps ++
Steps at the Organizational LevelSteps at the Organizational Level
1.1. EncodingEncoding / / record formationrecord formation
2.2. Respondent selection / identificationRespondent selection / identification
3.3. Assessment of priorities (Motivation)Assessment of priorities (Motivation)
4.4. Comprehension of the data requestComprehension of the data request
5.5. Retrieval of data Retrieval of data from memory / recordsfrom memory / records
6.6. Judging the adequacy of the responseJudging the adequacy of the response
7.7. Reporting the responseReporting the response
8.8. Release of the dataRelease of the data
40
Some General Characteristics of Establishment Surveys
Complex response process Request technical, records-based data Skewed target populations Estimate change vs. level Emphasis on timeliness of reporting
41
Emerging Themes That Affect Establishment Survey Pretesting
Nature of requested
data Survey response is
labor-intensive
Respondent burden
Uses of economic
data
42
Questionnaire Development, Evaluation and Testing (QDET)
in Establishment Setting
How QDET for establishment surveys has addressed –1) Nature of the requested data
2) Survey response is labor-intensive
3) Respondent burden
4) Uses of economic data
44
1) Nature of the Requested Data: Issues for QDET
Technical concepts with precise definitions
Data found in business records
45
Technical Concepts with Precise Definitions
Personnel involved in questionnaire development– Subject area experts– Stakeholders and data users – Questionnaire design experts
1) Nature of the Requested Data
46
Technical Concepts with Precise Definitions (cont.)
Consultation methods– Expert / User / Advisory Groups– Iterative rounds of prioritizing data needs– Large-scale content reviews– Exploratory focus groups– Observers during pretesting
1) Nature of the Requested Data
47
Technical Concepts with Precise Definitions (cont.)
Standardization and harmonization of questions and concepts
Instructions
1) Nature of the Requested Data
48
Data in Records
In order to design effective questionnaires, information is needed about –
– Data availability– Degree of comparability– Estimation strategies– Ease of retrieval– Respondent identity
1) Nature of the Requested Data
49
Data in Records(cont.)
Methods – Exploratory / feasibility studies and
site visits• Small, purposive samples • Meetings with company staff• Topic protocol
– Record-Keeping Studies
1) Nature of the Requested Data
51
2) Labor-Intensive Response Process:
Issues for QDET
Tangible cost to businesses
Difficult to observe the response process– Multiple data sources– Multiple respondents– Elapsed response time
52
Tangible Cost to Businesses
Methods to minimize additional workload– Respondent focus groups
• Limitation: Reluctance of business participants to reveal information to competitors
• Solution: Teleconference
– Informal interactions with respondents
2) Labor-Intensive Response Process
53
Difficult to Observe the Response Process
Adapt cognitive research methods– Traditional methods– Hypothetical probes– Vignettes / mock records– Interview protocol
• Flexibility
2) Labor-Intensive Response Process
54
Difficult to Observe the Response Process (cont.)
Logistics for cognitive research– On-site interviews– Time constraints– Personnel
• Questionnaire design experts
• Subject matter observers
2) Labor-Intensive Response Process
55
Difficult to Observe the Response Process (cont.)
Usability testing of electronic instruments– Methods
• Prototypes: paper, non-functioning screen shots, instruments with partial functionality
• Cognitive-like methods• User evaluations
– Logistics• On-site testing
2) Labor-Intensive Response Process
57
3) Respondent Burden:
Issues for QDET
Avoid adding burden due to questionnaire testing
Make effective use of respondent contacts
58
Avoid Adding Burden
Respondent-free methods
3) Respondent Burden
•Expert reviews and cognitive appraisals
•Electronic instruments— Heuristic reviews — Style guide
59
Effective Use of Respondent Contacts
Feedback from survey personnel– Interviewers – Survey staff interactions with respondents
• “Help Desk” calls• Follow-up calls regarding edit failing data
– Methods for collecting feedback• Debriefings / focus groups• Integrated database
3) Respondent Burden
60
Effective Use of Respondent Contacts (cont.)
Pilot tests and pretests– New surveys or major redesigns– “Dress rehearsal” vs. limited scope
3) Respondent Burden
61
Effective Use of Respondent Contacts (cont.)
Development strategies for electronic instruments– Small-scale pilots– Carefully screened survey participants– Respondent evaluation and feedback
3) Respondent Burden
62
Effective Use of Respondent Contacts (cont.)
Split sample experiments– Empirical evaluations of
• Alternative questions• Positioning of instructions• Mode selection
– Pilot studies vs. production data collection
3) Respondent Burden
64
4) Uses of Economic Data:
Issues for QDET Precise definitions Emphasis on timeliness Preservation of time series Recurring surveys that require
longitudinal designs and stability in the measurements (i.e., questions)
Post-Collection Quality EvaluationsPost-Collection Quality Evaluations
65
Post-Collection Quality Evaluations
Empirical analyses– Item nonresponse– Imputation rates– Outliers– Edit-failing data– External consistency checks
4) Uses of Economic Data
66
Post-Collection Quality Evaluations (cont.)
Respondent debriefings or Response Analysis Surveys– Re-contact respondents
• Actual response strategies• Reported data vs. definitions• Use of records, estimation strategies
4) Uses of Economic Data
67
Post-Collection Quality Evaluations (cont.)
Respondent debriefings
– Unstructured interview using a protocol
– Qualitative summary
Response Analysis Surveys
– Structured questionnaire
– Quantitative summary
4) Uses of Economic Data
68
Post-Collection Quality Evaluations (cont.)
Reinterview studies and content evaluations– Empirically assess response error and
bias in summary statistics
4) Uses of Economic Data
69
Post-Collection Quality Evaluations (cont.)
Primary goals – Assess data quality– Correct errors
Feedback into questionnaire design is rare
4) Uses of Economic Data
70
Conclusions and Conclusions and Future Research Challenges Future Research Challenges
for Establishment Survey for Establishment Survey QDETQDET
71
Conclusions
Establishment surveys are different. The establishment survey response
process model appears to be robust. An alternative
QDET cultureculture for Establishment Surveys seems justified.
72
An Alternative QDET Culture in Establishment Surveys
Complex response process
High “loss-function” for incorrect data
Role of stakeholders and data users
Advent of electronic methods
Methods for studying data availability and respondent selection
73
An Alternative QDET Culture in Establishment Surveys (cont.)
Estimation of change in key variables
Respondent burden with tangible reporting costs
Constraints on QDET activities
Adaptation of cognitive methods
Questionnaire stability
Post-collection quality evaluations
74
Future Research Challenges
The respondent– Identifying the “right” respondent– Influencing respondent selection
Facilitate the response process– Task analysis– Multiple respondents– Multiple data sources
Respondent role vs. data quality
75
Future Research Challenges (cont.)
Data needs and question development– Concepts Measurements Records– Records Measurements Concepts
Empirical evaluation of response errors– Edit failures– Item nonresponse– Other evidence?
76
Future QDET Challenges
Make better use of current activities Strengthen feedback loops between
production and development
Questionnaire
development
Post-collection processes and evaluation of data
quality
77
Future QDET Challenges (cont.)
Embed research into production data collection for recurring surveys
Evaluate “value-added” of QDET activities
78
Measuring Value-Added
Following production data collection – – Respondent debriefings– Response analysis surveys – Re-interviews– Validation studies
Intensive follow-up of nonrespondents– Evaluate bias– Nonresponse adjustments
79
Measuring Value-Addedcontinued
Examine post-collection downstream processes before and after redesign– “Plausibility indices”– Edit failure rates– Imputation rates– Item nonresponse– Costs
80
Reference:
Willimack, D.K., L. Lyberg, J. Martin, L. Japec, & P. Whitridge (2004) “Evolution and Adaptation of Questionnaire Development, Evaluation, and Testing Methods for Establishment Surveys,” Ch. 19 in Methods for Testing and Evaluating Survey Questionnaires, S. Presser, et al., ed., New York: Wiley.
Diane K. WillimackDiane K. WillimackU.S. Census BureauU.S. Census Bureau
[email protected]@census.gov