non-response

The proportion of people among those invited to participate in a non-compulsory interview survey , or other study, who choose not to take part or are unobtainable for other reasons. Non-response covers all causes of non-participation: refusals; people who are away temporarily, on holiday for example; and non-contacts for other reasons, who may be covert refusers. Those who are found to be outside the scope of the survey are classified as ineligible and excluded altogether. Ineligibles would include people who had died or moved to an area outside the survey area, businesses that had closed down, and demolished addresses. Non-response is a good indicator of response bias: as a general rule the higher the proportion of non-respondents to a survey, the greater the degree of bias among those who chose to participate. Rules of thumb for acceptable levels of survey response vary, but 60 per cent would generally be regarded as the bare minimum, with 75 per cent regarded as very good, and anything above that as excellent. Non-response above 40 per cent would normally be regarded as high enough to vitiate the results obtained from a survey or study, as non-participants roughly equal participants.

Non-response analysis compares the characteristics of respondents and non-respondents; usually this is limited to information from the sampling frame, such as sex and geographical area. Where additional information is available in the sampling frame it may be possible to assess more precisely the extent of response bias. In industrial societies, there are periodic declines in survey response rates, as reflected in all the regular major national surveys which monitor response-rates very closely. Declining response rates are met with renewed efforts to encourage participation, and reassure people on doubts about confidentiality and uses of the data. But they indicate that the survey method could become over-utilized, and that people are becoming increasingly well informed about social research and its uses. See also sampling ; sampling error.