The “Recall Question”

The so-called “recall question” is only a very limited indicator of actual voting behavior four years prior to the survey interview, since panel studies have demonstrated that respondents’ recall of their earlier voting behavior is highly influenced by how they intend to vote at the time of the interview. In other words, many respondents project their current opinion onto the past. In the case at hand, however, this phenomenon is irrelevant. Despite the well-known distortion in responses to the recall question, the clear difference found between voting intentions at the time of the interview and respondents’ recall of their voting behavior in 1998 does, in fact, reflect the difference in actual voting behavior, and the familiar distortion in voter recall tends to make the shift in voting behavior appear smaller in this case than it actually was. The existence of the distortion effect, therefore, does not detract from the validity of the findings presented in Table 1; on the contrary, they are, in fact, strengthened by it. For a more detailed discussion of the problems involved with the recall question, see Donsbach and Petersen 1999, 255-56.