A sample of 2663 people of voting age from across the country was interviewed after the 2002 election. Households were chosen at random and then a respondent was selected randomly from within each household. The sample design, interviews and initial coding of responses were all carried out by the ESRI. Respondents were drawn for all of the constituencies for 2002 election.
The 2002 questionnaire was fielded after the election in May 2002. There are two versions (with a short split-half) and a drop-off. The main questionnaire typically took 60 minutes to administer, and the drop-off would typically have taken a further 15 minutes. The response rate for the main questionnaire was 60 percent, with over 85 percent of those completing the supplementary drop-off questionnaire (containing, amongst other things, a module of questions from the Comparative Election Systems project (CSES).
Following the 2002 study, respondents were sent mail questionnaires on several more occasions: in November/December 2003, in the summer of 2004 after the June 2004 local and EP elections, and in the first months of 2006. Response rates – measured against the 2002 baseline – were 45 percent, 41 percent and 40 percent respectively. In 2007, after the general election, a final wave of the study was carried out face-to-face as in 2002. This last wave was topped up by a short mail questionnaire sent out late in 2007 to panel respondents who had not been interviewed successfully in 2007, and there were also interviews with a further 220 respondents to provide a more representative sample for 2007. The response rate for the panel element was 38 percent against a 2002 baseline, with a further 4 percent via the mail questionnaire. However, this was actually 54 percent of sought interviews completed, once deaths and changes of address without notification are taken into account. In all 518 respondents completed all five waves of the survey.
The largest wave in terms of items is the 2002 post-electoral survey; the other four waves have broadly overlapping contents. Some of the questions have been included in two, three, four or all five of the interviews, thus offering opportunities for longitudinal comparison of voter behaviour and orientations. Longitudinal comparisons can be made for aggregates as well as for individual respondents. The data is stored in both a wide and a long version to facilitate this.