Publication type
Conference Paper
Series
Royal Statistical Society (Social Statistics Section) Seminar
Author
Publication date
December 17, 2002
Abstract:
This paper will analyse data from a large-scale experiment carried out on the 2001 British Crime Survey, following the recommendations of Lynn and Elliot (2000). The experiment involved two alternative methodologies - the 'old BCS design', where respondents are asked about events in the previous calendar year (Jan-Dec) and most interviews take place between Jan and May, and the 'new BCS design,' where interviewing is continuous (equal sample size every month) and respondents are asked about the 12 month period up to and including the most recently completed month. Under the new design, the recall period is more recent on average. It can be hypothesised that this should improve recall of less salient or otherwise memorable events. However, the recall period is no longer (for 11/12 of respondents) bounded by the natural bounds of the beginning and end of calendar years. It can be hypothesised that this might lead to respondents having greater difficulty placing events within the correct time frame and that this might cause 'telescoping' for some types of events. These hypotheses will be explored and the implications for BCS time series will be drawn out.
#517560