There is a need for practical advice to inform decisions about when to mix modes and how, since survey designers are making these decisions in an ad hoc manner, driven by considerations of costs and response rates, but often ignoring the potential impact on data comparability. Some aspects of mixed modes are well researched, but others have not been examined. What is needed is a theoretical framework, based on existing and new complementary findings, about the causes and consequences of mixing modes.

A three-year project to address this issue was launched in Great Britain in October 2007, as part of the ESRC Survey Design and Measurement Initiative.

This project consists of three sub-projects: (1) a review of the literature and development of a theoretical framework, (2) quantitative analysis of existing datasets and new experimental data, and (3) qualitative research using cognitive interviewing to identify causes of mode effects.

This paper will present some initial results from sub-project (2), including some findings from a mixed modes experiment carried out in March-May 2009.

A sample of NatCen Omnibus respondents with access to the internet were randomly assigned to one of three modes: face-to-face, telephone and web. The questionnaire included about 60 questions that had been designed to test a set of hypotheses about the causes and consequences of mixed mode effects. The questions were classified according to type of question (e.g. attitude, behaviour, other factual), question format (e.g. closed/open, scale, # response categories), task difficulty and sensitivity of the question. The mode comparisons were dissected according to interviewer presence (face-to-face, telephone, none), delivery of question (visual, aural), response list (visual, aural) and recording of responses (oral, written). We will present the initial results of the experiment and highlight those research questions which will be taken forward in sub-project (3).