18-02-2013, 09:34 AM
Design & Planning 8: Non-Response
Design & Planning.pdf (Size: 268.35 KB / Downloads: 31)
Why is Non-Response Important?
¨Potential Bias
¨(Increased Variance)
¨Cost implications
¨Image problem
Almost all surveys suffer from it - even “compulsory” ones!
Some aspects of NR are common across all (or many) types of surveys; others are quite
specific to certain characteristics of a survey.
In many countries, NR is perceived to have increased in recent years. This leads to
worries about the effects of NR on time series.
Non-Response Error in Context
Reminder: Total Survey Error Framework
Errors of non-observation have three components:
· Coverage error is error caused by the omission of some population units from the
sampling frame;
· Sampling error is error caused by the omission of some sampling frame units (usually
most!) from the sample;
· Non-response error is error caused by the omission of some sample units from the
data
Components of Non-Response Error
Non-response errors can be divided into two components:
· Errors due to unit non-response
· Errors due to item non-response
Each of these can in turn be decomposed into sub-sources, for example:
Unit non-response due to:
· Non-contact
· Refusal to respond
· Inability to respond
Item non-response due to:
· Routing (instrument) error
· Routing (interviewer) error
· Refusal to respond
· Inability to respond
Reasons for Unit Non-Response
· Failure of the data collector to locate/identify the sample unit;
· Failure to make contact with the sample unit;
· Refusal of the sample unit to participate;
· Inability of the sample unit to participate (e.g. ill health, absence, etc);
· Inability of the data collector and sample unit to communicate (e.g. language
barriers);
· Accidental loss of the data/ questionnaire.
Unit non-response is defined relative to the eligible sample. In other words, if the
sampling frame contains ineligible units, these do not contribute towards response/
non-response.
Unit non-response is often divided into three components: non-contact, inability to
respond, lack of co-operation (refusal). Response rates can usefully be divided into
these (or other) components (see later).
Refusals: Relevant Survey Design Features
Mode of the initial contact. This affects
· The number of channels of communication between interviewer and respondent
(Groves, 1978).
· The selection of persuasion strategies to employ and the effectiveness of
alternative strategies (Groves, Cialdini and Couper, 1992).
Length of the interview being requested. A basic indicator of the response burden.
Topic of the survey. Helps to determine the respondents’ level of interest and knowledge
in the survey.
Use of respondent incentives. To motivate and to invoke the norm of reciprocity.
Design and content of advance letters / covering letters / questionnaires.