Contingency Question

Questions that are limited to a subset of respondents for whom they are relevant are called “contingency questions.” Relevancy is sometimes based on a respondent characteristic such as gender or age. For example, it is typical to ask only women of childbearing age if they are currently pregnant; conversely, only men are asked if they have ever have had a prostate cancer [Page 143]screening examination. Other times, questions are asked only of those that engage in a certain activity or hold a certain opinion about an issue.

A question that determines if a contingency question is asked is called a “filter,” “skip,” or “branching” question. In the research literature, the terms filter question and contingency question are sometimes used synonymously. However, in practice, the latter is dependent, or contingent, on the response to the former. Filter questions help route respondents through the questionnaire by skipping them over questions that are not relevant. Questionnaire “pathing” can be simple, as when one filter question determines receipt of one contingency question. Complexity is increased when responses to a series of filter questions are used to determine if a respondent gets one or more contingency questions.

Filter and contingency questions can be deployed in any data collection mode. In certain modes (Web, computer-assisted telephone interview [CATTJ, computer-assisted personal interviewing [CAPI], or computer-assisted self-interviewing [CASI]), the determination of who receives a contingency question can be programmed electronically. Once respondent characteristics are pre-loaded, survey programs will automatically skip contingency questions that would otherwise have required asking one or more filter questions. For example, respondents who are known to be male would automatically skip questions contingent on being female without first being asked a filter question about gender.

Contingency questions are not required on survey instruments; however, their use, in conjunction with filter questions, can reduce overall burden by asking respondents only those questions that are relevant. In the absence of filter questions, “Not Applicable” should be added as a response category for items relevant to only a subset of respondents. In the absence of an explicit Not Applicable option, respondents for whom inapplicable questions are asked may respond with a “Don't Know” or “Refuse.” This could be interpreted erroneously as missing data.

Survey researchers should be cognizant of the fact that some respondents may purposely answer filter questions in a way that will result in skipping contingency questions. This may occur when respondents lose interest in the survey, whether it is due to fatigue, boredom, or lack of topic saliency, and when they can too easily anticipate how a particular answer to a filter question will skip them out of another question or series of questions. This can lower data quality, as the result would be undetected missing data on items for which a respondent was actually eligible.