Practical implications of nonresponse bias in sample surveys : a thesis presented in partial fulfilment of the requirements for the degree of Master of Business Studies at Massey University

Abstract

Researchers world-wide are concerned about a decline in survey response rates. One consequence of such a decline is the potential for increasing nonresponse bias. This research reports the results of an attempt to establish a tentative 'minimum acceptable response rate' at which the interim estimates for two surveys did not differ significantly from final estimates. Data from a mail survey with a sample of 1270 respondents randomly selected from New Zealand electoral rolls, and from a telephone survey with a sample of 183 respondents randomly selected from five telephone directories were used for the research. The results indicate that a tentative 'minimum acceptable response rate' may be close to 50%. The study found that, at a response rate of 48%, demographic and awareness variables were prone to nonresponse bias in the telephone survey, and that altitude and demographic variables had a very low potential for nonresponse bias in the mail survey at a response rate of 51%. Perhaps researchers can now be more confident that a response rate close to 50% is acceptable for many practical purposes. Ultimately, however, the potential for nonresponse bias in a particular survey will depend on the demographic characteristics of respondents and nonrespondents and the strength of the relationship between these characteristics and the key variables of interest.