Research on Mail Surveys

Abstract

Five recent mail surveys carried out by the Government Social Survey are examined for their bearing on mail survey technique. Experimental features introduced into these studies allow measurement of non-response bias, of early/late response bias, of response by non-addressees, of the influence of a variety of factors on the response rate, and of response reliability and validity. An attempt is also made to evaluate all published research on each topic, and thus to present a definitive summary of the present state of knowledge about mail survey technique.

Mail Surveys and Response Rates: A Literature Review

“…Bressler [46] Gullahorn and Gullahorn [37] Clausen and Ford [18] Robinson advantage for airmail over first class for the return envelope, while Kephart and Bressler [46] did not for the outgoing mail; however, both Kephart and Bressler and Gullahorn and Gullahorn [35] did find a significant advantage in special delivery over first class, Clausen and Ford [18], using airmail and special delivery combined, obtained a very significant advantage over first class, Robinson and Agisim [66] obtained a significantly larger response using first class stamps rather than a postage meter, while Clausen and Ford [18] did not, Scott [71] obtained a small but significant difference using a stamped (but not addressed) return envelope versus a franked, addressed label attached to the flap of a return envelope. However, his results are somewhat ambiguous because of the presence of two experimental variables, Longworth [53] found no significant difference in a test of multiple small denomination stamps versus one large denomination stamp, while Watson [82] found a definite advantage in using multiple stamps.…”

“…Levine and Gordon [52] achieved a 100% response after three foilowups; an advance letter was also sent, Suchman and McCandless [78] achieved 95,1% after three foilowups; in a second study, they obtained 98,5% after only two foilowups. Eckland [24] reported 95% after four foliowups; in two different studies, Scott [71] achieved 95.6% and 93.2% with the use of two foilowups. Bachrach and Scoble [3] used five foliowups to obtain an 85% response rate; Donald [23] used three foliowups to obtain a 77.3% return.…”

“…Many writers on mail surveys recommend the use of official support of some kind wherever possible, particularly in surveys sent to commercial firms or professional people; however, there is little experimental evidence on the effects of sponsorship (see Table 4), Scott [71] speculated that the high response rates achieved in many of the British Social Survey mail surveys might have been due to government sponsorship, because of the implication that the government had the power to compel response. A split ballot survey using three different letterheads-the British Central Office of Information (a government agency), the London School of Economics, and the British Market Research Bureau (a market research firm)-did not result in a significant difference in response rates.…”

“…Efforts to increase the response rates to mail questionnaires are periodically reported in the literature, yet the most recent literature review appeared in a British journal back in 1961 [71], Various research reports published both before and since that time indicate that, contrary to general belief, some very high mail questionnaire response rates have been achieved.…”

“…Scott [71] and Sletto [74] tested two short questionnaires against one long questionnaire (consisting of the two short questionnaires strung together) and found no significant difference in results. However, their results are not necessarily conclusive because of the possibility of interaction between particular questions and questionnaire length.…”