Analysing interviewer call record data by using a multilevel discrete time event history modelling approach

In recent years, survey agencies have started to collect detailed call record data, including information on the timing and outcome of each interviewer call to a household. In interview-based household surveys, such information may be used to inform effective interviewer calling behaviours, which are critical in achieving co-operation and reducing the likelihood of refusal. However, call record data can be complex and it is not always clear how best to model such data. We present a general framework for the analysis of call record data by using multilevel event history modelling. A multilevel multinomial logistic regression approach is proposed in which the different possible outcomes at each call are modelled jointly, accounting for the clustering of calls within households and interviewers. Of particular interest are the influences of time varying characteristics on the outcome of a call. The analysis of interviewer call record data is illustrated by using paradata from several face-to-face household surveys with the aim of modelling non-response.