Intoduction: Panel web recruited online are the solution to the lack of frame population in the case of CAWI survey.
From January 2011 we started to build a panel online, recruited offline, in particular while gathering data via CATI interviews.
Research question: The main complaint of the subscribed is the screen-out. With this study we have tried to understand how the screen-out can produce the online panel's dropout and if this increases the panel's bias
Methods & Data: To carry out this analysis the metadata coming from the panel online Opinione.net have been analysed, in particular the information about the panelists (gender, age, education level, occupation) and the outcome of the different invitations which was available until 16th November 2012.
Furthermore, a survey on panelists satisfaction has been made to try to understand which are the critical aspects.
Results: We divide the panelists in 'actives', for those who have responded to at least one questionnaire in the last two months, and the 'inactives'. The gender-age distribution and the level of education distribution are significantly different (chi squared test) between the active panelists and the inactive panelists.
The survival analysis applied to the panel (duration measured as distance between subscription and the last answer) considering the number of screen-out on the total of 'clicks' (screen-out+complete) shows how the number of failures influences the permanence of the panelists. Analysing the joint distribution of the outcomes to the last request and the outcomes to the previous one, we note how in the majority of the cases a screen-out is followed 'unanswered'. The average of failures in the active panelists and in the inactives ones is high in both groups: ranging from 60% in the active, to 75% in the inactive that have responded more than twice. The variance analysis and the post-hoc test highlights a significant difference between the average of the actives and the average of the two groups of inactive (ANOVA: F=143,6 p=0,00).
The customer satisfaction that we have conducted highlighted that the complaint increases with the increase of screen-out suffered.