Behavioral Question

Behavioral questions are survey questions that ask about respondents' factual circumstances. They contrast with attitude questions, which ask about respondents' opinions. Typical behavioral questions target the respondent's household composition, sources of income, purchases, crime victimizations, hospitalizations, and many other autobiographical details. The Current Population Survey (CPS), for example, asks:

Similarly, the National Crime Survey (NCS) includes the following behavioral item:

Although these examples call for a simple “Yes” or “No” response, other behavioral items require dates (When was the last time you…?), frequencies (How many times during the last month did you…?), amounts (How much did you pay for …?), and other data. The CPS and NCS examples concern the respondents' behavior in a loose sense, but other questions are less about behavior than about existing or past states of affairs. For example, the following question, from the National Health Interview Survey (NHIS), is more difficult to peg as a behavioral matter:

For questions such as this, “factual question” may be a better label than “behavioral question.”

Because behavioral questions often probe incidents in the respondents' pasts, such as jobs and burglaries, they place a premium on the respondents' memory of these incidents. Inability to recall relevant information is thus one factor that affects the accuracy of responses to such questions. Questions about events that took place long ago, that are unremarkable, or [Page 52]that can be confused with irrelevant ones are all subject to inaccuracy because of the burden they place on memory.

People's difficulty in recalling events, however, can lead them to adopt other strategies for answering behavioral questions. In deciding when an event happened, for example, respondents may estimate the time of occurrence using the date of a better-remembered neighboring event (“The burglary happened just after Thanksgiving; so it occurred about December 1”). In deciding how frequently a type of event happened, respondents may base their answer on generic information (“I usually go grocery shopping five times a month”), or they may remember a few incidents and extrapolate to the rest (“I went grocery shopping twice last week, so I probably went eight times last month”). These strategies can potentially compensate for recall problems, but they can also introduce error. In general, the accuracy of an answer to a behavioral question will depend jointly, and in potentially complex ways, on both recall and estimation.

Answers to behavioral questions, like those to attitude questions, can depend on details of question wording. Linguistic factors, including choice of words, grammatical complexity, and pragmatics, can affect respondents' understanding of the question and, in turn, the accuracy of their answers. Because behavioral questions sometimes probe frequencies or amounts, they can depend on the respondents' interpretation of adverbs of quantification, such as usually, normally, or typically (How often do you usually/normally/typically go grocery shopping each month?) or quantifiers of amounts, such as a lot, some, or a little (as in the NHIS example). Similarly, answers to these questions are a function of respondents' interpretation of the response alternatives. Respondents may assume, for example, that the response options reflect features of the population under study and base their response choice on this assumption.

...