Post on 12-Jan-2016
transcript
Rethinking Leverage-Salience Theory and Causes of Survey Nonresponse:
Integrating Emotion, Mood, and Affect into Theory of Nonresponse
Matt JansUS Census Bureau, Statistical Research Division
Rachel LevensteinMichigan Program in Survey Methodology
University of Michigan
Scope
• What makes up leverage?– Affect v. Reasoned judgment
• Application of random effects w/in LST – Affect– Design features
Leverage-Salience Theory (LST)(Groves, Singer, & Corning, 2000)
• Addresses correlates and causes of unit refusal nonresponse
• Probability of response for the individual is a combination of– Leverage of a survey attribute– Salience of the same survey attribute
Visual Model of LSTGroves, Singer, & Corning (2000), p. 300
Definition and Examples of Leverage
• “The sample person’s assessment of a particular attribute of a survey” (Groves, et al., 2000)
– Incentive: Economic need or social exchange– Survey Topic: Interest in or commitment to issue– Survey Mode: Enjoyment of or aversion to
interacting w/ a live person (e.g., temperament or social isolation)
Definition and Examples of Salience
• Awareness of the sampled person to the survey feature
– Explanation of topic, mode, incentive in cover letter or interviewer recruitment script
– Obvious presence of survey feature (e.g., $5 bill clipped to survey; Interviewer on doorstep)
Decomposing Leverage
Leverage = Valence + Distance
Valence is positive or negative– Positively valenced features
dispose the sampled person toward participation
– Negatively valenced features dispose the sample person toward refusal
Focus of the TalkLeverage Salience Theory
Leverage Salience
Valence DistanceAnother Paper
Affect ReasonAnother Paper
Another Paper
Another Paper
Another Paper
Another Paper
Another Paper
Another Paper
Integral Incidental
Decomposing Valence– No explicit distinction between causes of valence
• All cognitive, emotional, judgmental psychological actions are pooled
Positive Negative
Affective Interacting with people makes me feel good
This interviewer seems pleasant
Interacting with people makes me feel uncomfortable
This interviewer’s voice irritates me
That organization makes my blood boil
Reasoned I think this topic is important
Given the current economic situation, the government would benefit from my opinion on this issue
The size of the incentive relative to the time the survey will take and my income is unreasonable
Definitions of Affect
• Affect– Mood
• Non-directed, lower-intensity, longer-lasting
– Emotion• Directed at an object/cause, higher-intensity, acute
Affect and Decision Making
• “Feelings as Information” perspective• Affect impacts/operates in information
processing, judgment and memory(Schwarz & Clore, 2007)
• We often make judgments and decisions on affective or emotional information (Schwarz & Clore, 2007; Schwarz, 2000)
• We perceive the world emotionally first (Zajonc, 1980)
Major Findings on Affect and Decision Making
• Depressed v. positive mood
• Strong arguments more effective for people in sad or negative moods, than positive moods (Schwarz, 2000)
• Mood will influence responses unless there is an attributable cause– Weather & well-being (Schwarz & Clore, 1983)
Incorporating Affect into LST
0 1 2 3 4 5 6 7ln / (1 )i i ij ij ij ij ij ij ij ij ij ij ij ij ip p C S A C S C A S A C S A
Incorporating Affect into LST
• Affect related to design features (integral affect)
– Like or dislike interacting with another person or not (SAQ v. Iwr Admin Modes)?
– Feel that the incentive is a “fair trade” or manipulative?
– Does R have strong feelings (positive or negative) about the survey topic or sponsor?
Incorporating Affect into LST
• Random effect of design features– Design feature is one realization of similar
features (e.g. levels of incentive, personalization of letter)
– Repeated measures from different levels-features across same R’s
ln[pij/(1-pij)]=B0 + B1ijAffDesij + B2iSij + u1iAffDesij + eij
i=Respondent , j=Design Feature
Incorporating Affect into LST
• Affect unrelated to design features(incidental affect)
– The mind state we happen upon when requesting survey participation
• Daily/weekly variation in mood• Individual variation in mood• Societal variation in mood (e.g., anxious mood due to
economic situation; saturation with polling)
Incorporating Affect into LST
• Random effect of respondent– Respondent’s propensity may change over time
irrespective of design feature– Repeated measures from same respondent
receiving same design features
ln[pij/(1-pij)]=B0 + B1ijAffRespij + B2iSi + u1iAffRespij + eij
i=Respondent , j=Recruitment Attempt
Collection of Measures
• Voice and speech indicators of mood• Respondent speech (spoken words)
– Coded for affective content• “I don’t feel comfortable answering questions about
my sex life”
• Interviewer or observer rating of affect• Respondent rating of own affector design
featares
Collection of Measures
• Complexity of measures will depend on definition of affect
– Integral: Affect related to survey design features• Measures collected from initial contact with R
– Incidental: Affect unrelated to survey • Voice at contact• Measures need to be taken outside of the interaction
with the survey• Non-contact v. Refusal• Panel data helpful
Links to Other Error Sources
• Measurement Error– Response by individuals with only positive
affective states would bias measures of affect or wellbeing
– Identifying current mood & assigning it to a cause (e.g. weather) can change satisfaction reports
• Item Nonresponse– Similar mechanisms & opportunities for tailoring
apply
International Component
• Cultural differences in social cognition & emotion (Markus & Kitayama, 1991)
ReferencesBachorowski, J. A. (1999). Vocal expression and perception of emotion. Current Directions in Psychological
Science, 8(2), 53-57. Groves, R. M., Couper, M. P., Presser, S., Singer, E., Tourangeau, R., Acosta, G. P., & Nelson, L. (2006).
Experiments in producing nonresponse bias. Public Opinion Quarterly, 70(5), 720-736. Groves, R. M., Presser, S., & Dipko, S. (2004). The role of topic interest in survey participation decisions. Public
Opinion Quarterly, 68(1), 2-31. Groves, R. M., Singer, E., & Corning, A. (2000). Leverage-saliency theory of survey participation: Description and
an illustration. Public Opinion Quarterly, 64(3), 299-308. Markus HR, Kitayama S. Culture and the self: Implications for cognition, emotion, and motivation. Psychological
Review. 1991;98(2):224-253. Available at: http://doi.apa.org/getdoi.cfm?doi=10.1037/0033-295X.98.2.224.Roose, H., Lievens, J., & Waege, H. (2007). The joint effect of topic interest and follow-up procedures on the
response in a mail questionnaire: An empirical test of the leverage-saliency theory in audience research. Sociological Methods & Research, 35(3), 410.
Schwarz, N. (2000). Emotion, cognition, and decision making. Cognition and Emotion, 14(4), 433-440.Schwarz N, Clore GL. Mood, misattribution, and judgments of well-being: Informative and directive functions of
affective states. Journal of Personality and Social Psychology. 1983;45(3):513-523. Available at: http://content.apa.org/journals/psp/45/3/513.
Schwarz, N., & Clore, G. L. (2007). Feelings and emotional experiences. In A. W. Kruglanski, & E. T. Higgins, Social Psychology: Handbook of Basic Principles, (pp. 385-407). Guilford Press.
Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35(2), 151-175.
Thank You
• matthew.e.jans@census.gov• mlev@isr.umich.edu
LST Findings
• “Peripheral” aspects of surveys (e.g., incentive, follow-up protocol) have larger impact in absence of personal relevance of topic (Groves, et al., 2000; Roose, Lievens, & Waege, 2007)
• Personal relevance does not always lead to increased response (Groves, Couper, Presser, et al, 2006; Groves, Presser, & Dipko, 2004)
Understanding Leverage
• Hard to measure– Internal, subjective– Group membership
• Decompose leverage into components to facilitate measurement
• Affective v. reasoned perceptions and judgments of survey request
Integrating Affect into Survey Practice
• Tailoring to affect states – “I’m sure you’ve had a busy day” if calling in
evening– Listen for vocal cues indicating unease and tailor– Different information/arguments required for
different moods– Moods may be changeable