Title: A TheoryGuided Interviewer Training Protocol for Survey Participation
1A Theory-Guided Interviewer Training Protocol for
Survey Participation
- Robert M. Groves, University of Michigan and
Joint Program in Survey Methodology - Kate McGonagle, University of Michigan
2Overview of Presentation
- Brief review of theoretical perspective
- Description of results of two experimental trials
- Description of training protocol
- Next steps
3Two Very Different Processes that Produce
Nonresponse
- Noncontact of sample households
- Refusal to cooperate with the survey request
4The Contact Process has a Small Set of Causes
Access Impediments
Social Environmental Attributes
Accessible at-home patterns
Likelihood of Contact
Socio-demographic Attributes
X
Timing of Calls
No. of Calls
5Interviewer Behavior Reducing Noncontacts
- Access impediments
- Call on such cases early to obtain knowledge of
source - Evidence of general ability to overcome
impediment - Callbacks
- Advantage of evening, weekend
- Average 2-3 calls to first contact
6Cooperation is a Complex Phenomenon
Social Environment
Design Features
Household/ Person Predispositions
Interviewer Attributes/ Behavior
Householder-Interviewer Interaction
Decision to Participate
7Three Theoretical Concepts
- Leverage-Salience
- Tailoring
- Maintaining interaction
8Leverage-Salience Theory of Survey Participation
Authority of Sponsor
Authority of Sponsor
Incentive
Topic
Incentive
Burden
Topic
9Effects of Salience Weights on Propensity to
Cooperate
Authority of Sponsor
Burden
Authority of Sponsor
Incentive
Incentive
Topic
Topic
10Tailoring
- I give the introduction and listen to what
they say. I then respond to them on an
individual basis, according to their response.
Almost all responses are a little different, and
you need an ability to intuitively understand
what they are saying.
11Lessons for Interviewer Behavior
- Scripting of interviewers reduces likelihood that
they address concerns of different sample persons - Interviewer training must train them to
- listen for concerns
- identify a way to address them
- deliver answer quickly, in native language
12Goals of a Training Protocol
- Use expert knowledge among best interviewers
- Identify common concerns as expressed in native
language of target population - Train in diagnosis, customized reply, and
delivery speed - Evaluate absorption of training
13Experiment 1
- Current Employment Statistics program of the U.S.
Bureau of Labor Statistics, asks each sample
employer to report six items concerning the
counts and payroll for various types of employees - Request protocol involves three steps
- a sample refinement step
- an advance mailing package sent to the contact
person, - an enrollment call
14Experiment 1 Research Design
- Comparisons of interviewers performances before
and after training workshop - 320 cases from stratified samples of employers in
4 states were contacted in Phase 1 (pre-training)
and 329 in Phase 2 (post-training) - 17 interviewers in 2-day training workshop
15Pre- and Post-Training Cooperation Rates for
Experiment 1
16Random Coefficient Models Predicting Likelihood
of Cooperation in Experiment 1
lnpij/(1 pij) ? 0j ? 1j (POSTij1) ? ?
k (SIZEk 1) ?ij
Two models compared for the random coefficients
1) Base Model ? 0j ?00 u0j ? 1j ?
10 u1j 2) Learning Model ? 0j ? 00
u0j ? 1j ? 10 ? 11 (GRADE)j u1j
17Values of Estimated Coefficients in Experiment 1
Two models compared for the random coefficients
1) Base Model ? 1j 0.43 u1j
(0.21) 2) Learning Model ? 1j
0.40 0.07 (GRADE)j u1j (0.46) (0.36)
18Conclusions from Experiment 1
- Apparent positive training effect (10 percentage
points) - Reduced interviewer variation in cooperation
rates - Potential confounding of effects of experience
and effects of training - Weak predictive power of grade on evaluation
19Design of Experiment 2
- U.S. Census of Agriculture, followup of
nonrespondent farm operators by telephone - Interviewing organized by state (5 experimental
states, 5 control states) - 96 interviewers given experimental training
assigned 11,000 cases - 99 interviewers assigned 12,500 cases in control
states
20Pre-Post Cooperation Rate Differences by
Treatment Group, Experiment 2
Control
Experimental
21Pre- and Post-Training Cooperation Rates by
Experiment 2 Group
22Pre-Post Interviewer Cooperation Rates by
Treatment Group
Pre
Post
Pre
Post
Experimental
Control
23Random Coefficient Models Predicting Likelihood
of Cooperation in Experiment 2
lnpij/(1 pij) ?0j ?1j (POSTij1) ? 2
(FORMij 1) ? ij
Two models compared for the random coefficients
1) Training Model ? 0j ? 00 u0j ?
1j ? 10 ? 10 (TRAINING1)j u1j 2)
Learning Model ? 0j ? 00 u0j ? 1j ? 10
? 10 (TRAINING1)j ? 12 (GRADE)j u1j
24Estimated Coefficients for Experiment 2
Two models compared for the random coefficients
1) Base Model ? 1j 0.038 0.63
(TRAINING1)j u1j (0.08)
(0.10) 2) Learning Model ? 1j 0.038 -
0.15 (TRAINING1)j 0.051 (GRADE)j
(0.08) (0.51) (0.033) u1j
25Conclusions from Experiment 2
- Training produces positive effects beyond those
expected from experience - Interviewer variation in performance reduced by
improving the low performers - Evaluation tool still needs development
26What are the Steps of the Training Protocol?
- Focus groups of experienced interviewers,
reporting words used to communicate concerns
about the survey request (200-400 examples
identified) - Thematic sets were refined through feedback from
senior interviewers (9 main thematic groups
identified)
27Themes of Sample Person Concerns
- Time Its harvest time
- Burden I want to be paid for my time
- Government Surveys drive the prices down
- Privacy Thats none of your business
- Crisis/Change Im bankrupt. I lost the farm
- Small Operator Go after the big guy
- Policy We dont do surveys anymore
- Burn-out Im bowing out this time
- Pass-off The countys got that stuff
28Steps of the Training Regimen Development (contd.)
- The training workshop in learning to
- understand the themes of concerns
- classify actual wording into those themes
- deliver, in compatible wording, a set of
statements relevant to the concerns - increase the speed of performance on the above
- Structured evaluation of how well interviewers
obtained the skills taught by the workshop
29Classification of Actual Wording into Themes
- Interviewers completed a match-column exercise to
learn to link theme groupings to response
verbatims - Pairs of interviewers role-play
- Listened to unknown concerns
- Articulated an oral diagnosis giving thematic
concern type (Diagnosis) - Selected and delivered a relevant response
30Desirable Behaviors to Address Concerns
- Lecture and discussion on the importance of call
preparation and initial contact protocols - Learning the nature of previous reluctance
- Anticipating respondent concerns
- Sensitivity to respondent tone and nonverbal cues
- Listening exercise using a pre-recorded series of
study-specific introductions - Group critique of content and delivery of
introductions - Suggested improvements brainstormed
31Delivery of Statements Relevant to Respondent
Concerns
- Paired exercise Interviewers used scripted
interactions to practice responding to one them
(government concerns) - Triad exercise Interviewers divided into groups
of 3 with rotating roles - Silent diagnosis of wide variety respondent
concerns - Delivery of rebuttal
- Structured written critique and oral delivery
-
32Increasing the Speed of Performance
- Rapid Fire Exercise 1. Divided interviewers
into small groups of 6-8 members. Exercises
focused on - Interviewers crafting rebuttals to concerns
delivered by facilitator - Eliciting suggestions and comments, giving
alternative approaches from group members - Rapid Fire Exercise 2. Focus on increasing the3
speed of performance by - Elimination of comments between interactions
- Facilitator increasing the pace of role-play
33Structured Evaluation of Interviewer Skills
- Written evaluation Interviewers are given 5
respondent concerns - Listen to concern delivered orally by facilitator
- Identify theme group for respondent concern
- Select a relevant rebuttal for the concern
- Quickly draft a handwritten rebuttal addressing
the concern - Complete handwritten response to each respondent
concern within a time limit of 2 minutes
34Summary
- On multiple surveys, training leads to higher
cooperation - Nature of training effects are to bring up the
low performers to higher levels - We have not yet produced a trainee-level
evaluation instrument with predictive value
35Work in Process
- Development of new post-training assessment
- Telephone call-in by trainee to scripted staff
member - Simulated survey introduction
- Successive concerns raised by staff member
- Grading of interviewer behavior