The dimensions and correlates of respondent burden in personal interviews
This research focused on respondent burden, defined as the negative attitudes experienced by persons participating in voluntary personal interviews. The topic is of interest to education and other researchers who rely upon personal interviews as a data gathering technique. But the use of this technique in federally-sponsored research is subject to severe restrictions, based on federal assumptions that long interviews and repeated interviewing (such as that experienced by members of survey panels) are burdensome. However, survey researchers argue that the relationship between interview length, panel participation, and respondent burden is far from clear, and that there may be other mediating factors which affect reactions to the interview. These researchers also contend that respondent burden is poorly defined, and that, before further research about the correlates of burden is carried out, the field must explore the meaning or dimensions of this phenomenon.
To address these concerns, a reanalysis of a data base developed by the author through a previous study was carried out. In this earlier study, respondents, who were suburban Philadelphia residents, participated in one of four experimental treatments: one short (25- minute) interview, one long (75-minute) interview, or two short or long interviews, with the second administered ten months after the first using a similar instrument. All versions of this instrument involved topics of moderate salience, such as housing and neighborhood conditions. Respondent reactions to the interview were assessed via a self-administered reaction form, completed at the end of the treatment interview. Responses to this form, as well as background information about respondents, constituted the data base for the dissertation.
The reanalysis of this data base addressed two major questions:
l. What are the underlying dimensions of respondent burden as measured through the reaction form?
- In explaining these dimensions, what are the relative contributions of interview length, panel participation, respondent demographic characteristics, and past participation by the respondent in other, unrelated surveys?
The first question was explored by a factor analysis, and the second by a regression analysis. Three major components of respondent burden were identified:
Perceived Uselessness, which loaded on variables relating to the interest, importance, and general benefits of survey participation, and views about the ability of survey participants to affect government decisions. This factor is the most important, accounting for 48 percent of the common factor variance.
Time Concerns, which loaded on three variables related to ''time'': willingness to continue the interview, views about the length of the just-completed interview, and attitudes about the use of short questionnaires as a survey improvement. This factor accounted for 14 percent of the common factor variance.
Privacy Concerns, which loaded on three variables representing agreement/disagreement with statements about the privacy invading nature of surveys. This factor accounted for 10 percent of the variance.
The regression analysis showed no relationship between Perceived Uselessness or Privacy Concerns and any of the independent variables. Time Concerns were significantly related (p.<.01) to interview length, in the direction postulated by federal assumptions, and to (p.<.05) sex, with females more burdened than males, and employment status, with employed persons more burdened than others.
Based on these findings, several recommendations were made, chief among them that researchers use public relations mechanisms to convince potential respondents of the usefulness of survey participation.