Improving the Accuracy of Recall Data: A Test of Two Procedures

Questions have been raised about the validity of data obtained from retrospective survey designs, since these are heavily dependent on the accuracy of respondent recall. One of the most serious problems is over-reporting, typically attributed to "forward telescoping", where respondents report events that occurred outside of the time period under consideration, thereby inflating the results. This study tested two procedures for reducing forward telescoping: the provision of a landmark event to clearly mark the beginning of the recall period; and a double question bounded recall procedure whereby respondents were first asked to recall events for a longer time period (previous eight weeks), then for the time period of interest (previous four weeks). This paper presents the results of this research and discusses the implications for studies dependent upon recall data