METHODS A random sample of members of the general public in four UK cities (Exeter, Sheffield, Aberdeen and Glasgow) was obtained from the UK Electoral Register and invited to training sessions in each city. 106 health state descriptions were posted at intervals during 2004-6 and utility values obtained. Recruitment and the characteristics of the panel are described. Participation is defined as having provided any values and compliance as the proportion of potential values provided by participants. The determinants of participation and compliance are explored using univariate and multivariate (where appropriate) methods.

RESULTS Only 5.4% of those approached expressed willingness to participate in the project. The need for training reduced recruitment further, giving an initial panel membership of 112. Recruitment was more successful in Exeter than the other cities, as was reporting of reasons for declining the initial invitation. 83 (74%) people participated in the panel. Socioeconomic and marital status were significantly associated with participation. Mean compliance was 45%. None of the potential explanatory variables were associated with compliance.

CONCLUSIONS It is feasible to establish an internet panel for preference measurement but recruitment rates are very low. Over-sampling according to ethnicity and socioeconomic status are required to ensure a demograpically balanced participative panel.