Response Rate

Reasons for Nonresponse

There are many possible reasons for nonresponse. People tend to refuse to participate in surveys or interviews, for example, due to lack of time, lack of interest in the topic, competing demands from many requests for their attention, suspicion that the survey request is actually a marketing pitch, or the sensitive nature of the questions asked.

Each mode of collecting survey data has its own challenges. For in-person surveys, interviewers are often unable to reach respondents who live in gated communities or high security apartment buildings. The major challenge for telephone surveys is the difficulty of finding people at home and devices like answering machines, caller ID, and cell phones. Mail surveys, which have some of the lowest response rates, face several obstacles, including requests being ignored due to an increasing volume of junk mail, the lack of personal contact with an interviewer, and the length of the survey on paper. Web surveys are easy to decline, and most types of web surveys (with the exception of web panels) do not have probability-based samples, which limit their usefulness for research.

Calculating Response Rates

The prevailing standards for response rate calculations are brought out and updated by the American Association of Public Opinion Research (AAPOR). AAPOR recognizes six different methods of calculating response rates. Before discussing the details, it is helpful to understand the basic terminology used.

• Completed interviews are cases in which the sample unit (e.g., household, person, business) was contacted and the interview was completed.
• Partial interviews are interviews terminated by respondents and left incomplete.
• Noninterviews occur when a respondent was located but did not complete the interview. This category includes refusals, noncontacts, and other types of noninterviews. Refusals happen when the eligible respondent refuses to participate. Noncontacts include other situations in which contact cannot be made, such as the respondent never being available, answering machines, inability to gain access to a building, [Page 1421]or completed questionnaire not returned (for mail surveys). Other includes other kinds of noninterviews whereby contact was made and there was no refusal, but the survey could not be administered, for example, if there is a deceased, mentally or physically challenged respondent, or language problems.
• Unknown cases are those in which the researcher is not certain whether the sample element is eligible for the survey. Unknown cases estimated to be eligible are the sum of two categories—unknown households and unknown other. Unknown households include cases in which the researcher does not know if the sample element is a housing unit, for instance, when the phone was always busy, there was no answer or an answering machine, or if the address was not locatable or unsafe to reach. Unknown other includes various categories of returned mail from the postal service (for mail surveys) and in screening studies (i.e., when there is a screener before the actual interview is conducted), the screener was not completed.

The AAPOR response rates are as

...