Richard Boreham, NatCen Social Research
multi-mode response web CAWI
Understanding Society is moving to a multi-mode design from Wave 7 onwards, where a proportion of households will be initially issued to web, and then transferred to face-to-face if necessary. This clearly has potential cost savings, although these will only be realised if the whole household is productive. Introducing mixed mode may also impact the overall response rate – at IP5 the mixed mode experimental group where cases were initially issued to web and subsequently transferred to face-to-face had a household response rate that was 4 points lower than the sample issued directly to face-to-face. Is it possible to identify households where we can make cost savings by getting full household response in web, and not damage the overall response rate for those that do not fully respond in web? This paper sets out an approach to determine which households should be issued to web. The technique is to use logistic regression to estimate the probability of individuals taking part in web, and then from this derive a probability of everyone in a household taking part and then select those households which have the highest probability. Initial analysis suggests that it is possible to select 25% of households to issue to web and to double the web full household response from 23% for the whole sample to 46% for this targeted group of households.