Face-to-Face Interviews and Mailed Questionnaires: The Net Difference in Response Rate

Abstract

Traditionally, reflected methodology textbook reviews of the survey nonresponse literature, interviews have been thought to achieve substantially higher response rates than mailed questionnaires. The paper describes a correlational design for assessing the response difference between these two forms of data collection, after controls for other factors known to affect survey response. Attention is given to change over time the response difference between interviews and questionnaires. In the primary data set, the net response due to data collection amounts to 7.5 percentage points, the middle (30-70) of survey response. The analysis quantifies the decline over time interview response, and shows mailed questionnaires to be free from decline. A second data set reproduces the essential findings. John Goyder is Associate Professor, Department of Sociology, University of Waterloo, Ontario, Canada N2L 3G1. The work reported herein was supported by a Canada Council Leave Fellowship held by the author during a stay at Cambridge University 1978, and by a UW-SSHRC grant awarded 1981. The author is greatly indebted to Cathie Marsh and others at the Social and Political Sciences Committee, Cambridge, for hospitality and research resources. Research Assistants were Norleen Heyzer and Kerry Davies at Cambridge and Jean Leiper and Olorunfe Taylor-Cole at Waterloo. Public Opinion Quarterly Vol. 49:234-252 ?) by the Trustees of Columbia University Published by Elsevier Science Publishing Co., Inc 0033-362X/85/0049-234/$2.50 This content downloaded from 157.55.39.181 on Thu, 29 Sep 2016 05:51:04 UTC All use subject to http://about.jstor.org/terms INTERVIEWS AND MAILED QUESTIONNAIRES 235 (1981:150), the 50 ceiling holds true even allowing for follow-ups. Others, however, set the upper limit markedly higher. Fitzgerald and Cox (1975:53), for example, choose 60 while, for True (1983: 218), estimates of likely return rates to 70 percent, a maximum also reported Goode and Hatt (1952:173) and Cole (1980:62). Sanders and Pinhey (1983:152-53) cite 60 an average and note that some researchers have reported returns of 80 and more. Consistent 70 returns from general populations are not unrealistic, argue Weisberg and Bowen (1977:58), given energetic follow-ups (also Miller, 1977:79-81). Estimated completion rate on interview surveys is 70-80 for Kidder (1981:150), 80-85 for Babbie (1973:171), 80-90 for Orenstein and Phillips (1978:229), over 90 according to Parten (1950:351), about 95 percent for Nachmias and Nachmias (1976:107), and as much 85 percent for Sanders and Pinhey (1983: 154). Weisberg and Bowen (1977:36), introducing a factor to be emphasized herein, quote 90 interview response typical during the 1950s, with rates in the 80 range common today. The greater response due to data collection by interview rather than by questionnaire could, from the above estimates, amount to anywhere from zero to 60 percentage points. In noting the variability of the textbook estimates, we do not deny that the accumulated wisdom on survey methodology presents a broad consensus that interview response is generally greater than mailed questionnaire returns, nor that each of the response figures cited by different authors accurately reflects at least aspect of reality. The review of textbooks is effective, however, emphasizing the variability survey response according to both form of data collection and other factors. A typical survey response rate exists only insofar there is a salience level of topic, sponsor, target population, and set of field procedures. In the unsystematic comparisons of questionnaire and interview response presented methodology textbooks, these various factors have varied haphazardly. The argument this paper is that the survey methodology literature is now a position to establish a net effect for form of data collection, controlling for other determinants of survey response by means of multivariate analysis. We develop this methodology herein, and establish the net margin between interviews and questionnaires, both averaged over the history of survey research and mediated over time.