Effect of Appearance of Correspondence on Response Rate to a Mail Questionnaire Survey

Probably more than any other single criticism, the criticism of low response rate can deliver the coup de grace to an otherwise well-designed, well-executed survey research project. Various techniques have been employed to increase response rate, e.g., reducing the questionnaire length; making the questionnaire easy to complete and return; using alert letters, follow-up letters, threats, pleas, coercion, rewards, logical arguments as to the importance of a project, illogical arguments, etc. Depending upon the population, the reasonable expectation may vary, say, from 50% to 80% (3). One technique often used to induce a higher rate is that of optimizing the appearance of the correspondence associated with the survey. The purpose of this study was to investigate the effect of the appearance of the correspondence on response rate to a mail questionnaire survey. The data were collected in a survey conducted with all (N = 132) licensed physical therapists in the state of Georgia, chca Summer 1970 (1). Group 1 (n = 66) was termed the "appearance" group and Group 2 (n = 66) the "other" group. There were three mailings: an alert letter, a letter of instruction accompanying the questionnaire, and a follow-up letter. Correspondence to Group 1 was beautifully auto-typed, appeared personal, and was signed by the director of the project (a former president of the state association). Correspondence to Group 2 was mimeographed, including the stylus-signed signature of the director of the project. The content of correspondence was constant; only appearance varied. Before mailing, each questionnaire was subtly identified as to treatment group. Returns were otherwise anonymous. The over-all response rate for the questionnaire survey (the number of questionnaires returned divided by the number of members of that group) was high, i.e., 80%. The response rates for Groups 1 and 2 were 85% and 76% respectively. The obtained t of 1.23 was nor significant (p > .05; 2, p. 138). While response rate for Group 1 was numerically higher than that for Group 2, the difference could be accounted by chance. The appearance of the correspondence apparently had no real effect on response rate for this sample. Two limimtions ought to be mentioned: a larger N would be desirable; and Ss were members of a relatively intimate professional group, but in a typical survey respondents do not know each other or the director of the project. While replication with a larger and a less intimate subject population would be desirable, the data do indicate the correspondence appearance made essentially no difference in response rate for this sample.