class 5_non response Flashcards
nonresponse
after the sampling step of the survey
reflects the total failure to obtain survey data
unit nonresponse
failure to obtain survey measures on a sample unit
partials
break offs
drop outs
BLS Consumer Expenditure survey and NR/Refusal
Total non response and refusal rates increasing over time for BLS
de Leeuw & de Heer (2002)
Multi Country Studies of Cooperation
16 countries
10 ongoing surveys- varied in topics
on average, 3 percentage decline per year in cooperation rate
Costs and Non response
Costs have risen as surveys take counter measures
- advanced letters
- incentives
- call backs
Forms of non response
Non contact
non cooperation
inability- e.g.,unable to p due to health
Non contact: Face to Face Surveys
Gated communities , locked buildings
gate keepers
Non contact: Telephone surveys
Rise of answering machines
caller ID
Inability to provide data
physical/ mental limitations
Immigrant populations
Theories of non cooperation
too busy
too self absorbed; not integrated in community
-Unwanted intrusions; privacy
Groves et al (2000)
those with higher community involvement had higher response rates
Abraham et al (2006)
little support for busyness as predictor and more for low social integration as the reason for low response rates
Leverage-Salience Theory
people make decisions to cooperate or refuse on diff bases
salience of design attributes or focus of int behavior
Implications for non response bias
could vary by subgroup
happens quickly
advanced letters may help
Benefit-Cost Theory of Survey participation (Singer,2011)
will participate if perceived benefits outweigh the costs
incorrect info- by int can play an important role- benefits need to be made salient to the respondent
Groves and Peytcheva (2008)
Meta analysis of non response errors
59 studies with multiple est.
weak link between nonresponse error and response rates
error not well predicted by response rates
Kreuter et al 2010
admin data available for entire sample
additional recruitment efforts reduced non response bias
slight inc in measurement error associated with increased contact effort offset by reductions in non response bias
