Underreporting

When answering questions on sensitive behaviors, many respondents show a specifie variant of response error: They tend to report fewer instances of undesired behaviors compared to what they have actually experienced. This is called underreporting. It is assumed that respondents avoid reporting unfavorable conduct because they do not want to admit socially undesirable behaviors, which in turn leads to this type of misre-porting. Similar effects are known for responses to survey questions about unpopular attitudes.

Currently, it is not known whether underreporting is the result of a deliberate manipulation of the true answer or whether it occurs subconsciously. Nevertheless, for the most part it is assumed to be a response error that is associated to the cognitive editing stage of the question-answer process. Misreporting due to forgetting or other memory restrictions is considered to be a minor source of underreporting.

Several validation studies use records of the true answers to provide evidence for underreporting. For example, studies on abortion or illegal drug use show that fewer respondents admit those behaviors, or if they admit them, they report fewer instances of the presumed undesired behavior.

Since underreporting is related to social desirability bias it occurs more often in interviewer administered settings compared to self-administered surveys. Accordingly, audio, audiovisual, and telephone audio computer-assisted self-interviewing methods have been used extensively when collecting responses on socially undesirable behaviors or attitudes. These modes reduce the response burden of answering sensitive questions, which in turn is assumed to reduce underreporting of undesirable behaviors or attitudes.

Also, several question techniques have been developed to reduce underreporting. Besides forgiving question wording, the “everybody approach,” and indirect questioning, the sealed ballot method has proven to effectively reduce underreporting. In addition, several variants of the randomized response technique are available. However, because the instructions necessary for those question types are rather complicated and not always understood by the respondent, they are practically seen as less effective in reducing underreporting.