Conditioning effects on a (relatively) high-frequency random probability web-CATI panel
Jul 26, 11:00
Conditioning effects are a concern for longitudinal research ‰ÛÒ to what extent does participation in research studies affect the participant‰Ûªs behaviour and attitudes? This paper addresses those questions in the particular context of NatCen‰Ûªs web-CATI general access panel. Non-probability panels have proliferated in recent years in response to demand for cheaper and quicker research findings in market research and the opportunities opened up by web-based data collection methods. The NatCen Panel was developed two years ago in response to similar pressures within social research, but with a random probability approach at its heart. It is the first general access random probability web-CATI panel in Britain.
The Panel uses a ‰Û÷piggy back‰Ûª design whereby those interviewed as part of the face-to-face British Social Attitudes survey are asked to be included in the Panel at the end of that interview. Those who agree are then invited to a series of surveys using online and CATI data collection modes (the latter mode ensuring coverage of those who do routinely access the internet).
The Panel was primarily conceived of as a cross-sectional research resource, but high levels of participation from one survey to the next (c.80%) have enabled some limited longitudinal analysis, for instance before and after the General Election this year. We are considering the potential for other studies to include a more ambitious longitudinal component, but given a primary aim to provide good quality cross-sectional estimates, it is important for us to understand more about the potential impact of conditioning on some key measures.
We implemented an experiment in the Panel‰Ûªs November/December 2017 survey to help us understand conditioning effects. This takes advantage of the recent recruitment to the panel of around 3,000 new members from the BSA‰Ûªs 2017 fieldwork. The newly recruited sample will be included alongside panel members recruited from the 2016 BSA who have been included in up to seven previous panel surveys. Items were repeated in the recent survey from surveys conducted previously with the 2016 recruits. These reflected some recurrent themes of the last year, notably Brexit and interest in politics. They also include items relating to life satisfaction and awareness of and trust in a public body. The paper will share analysis of the comparisons of the two samples on these measures and discuss the specific context of conditioning effects in a relatively high frequency online data collection vehicle.