Conferences

Establishing contact is an important part of the response process and effective interviewer calling behaviours are critical in achieving contact and subsequent possible cooperation. Recent developments in the survey data collection process have led to the collection of so-called process data or paradata, which greatly extend the basic information on interviewer calls. Research is needed to establish how best to use such data to inform nonresponse processes, as well as further methodological development in the specification of response propensity models.
This paper aims to build and improve response propensity models based on paradata to predict the likelihood of contact at each call, conditioning on household and interviewer characteristics. We explore the best times of contact for different types of households controlling for influences of the interviewer on establishing contact.

An advantage of this study is that we have access to rich paradata, including detailed call-record data, i.e. information recorded by the interviewer during the data collection stage at each call to the household even if contact was not made, interviewer observation data about the sample unit also recorded by the interviewer and rich information about the interviewers themselves. A key strength is that individual and household characteristics are linked to these data. The resulting data have a multilevel structure with individuals nested within households, which are nested within a cross-classification of interviewers and areas.

To model the propensity of contact at each call, we use multilevel discrete-time event history analysis. The model conditions on all information available to the interviewer, such as from administrative data and prior calls, and include call record data as time-varying covariates.
The paper aims to provide guidance to academic researchers and survey practitioners on how to model and use such paradata for the design of effective and efficient interviewer calling strategies. It is anticipated that this research will inform the improvements of responsive survey designs and the design of call-backs and follow-ups of nonrespondents, with implications for survey agencies for the allocation of time and staff resources.