Survey nonresponse remains a problem despite the development of methods that aim to reduce nonresponse. Statistical techniques help to get most out of the data that is collected from the responses. However, they do not sufficiently address the problem of resource allocation in survey designs in which the focus is on the quality of the survey results given that there will be nonresponse. Therefore, we propose a method in which the optimal allocation of resources in mixed-mode surveys can be determined. The method builds on the idea that population units can be grouped together based on common characteristics that are available as a-priori knowledge. For each group, historical data can be used to produce estimates of contact probabilities and response probabilities for both uni-mode and mixed-mode surveys. Moreover, the total expected costs for any given survey can be separated into group-dependent and group-independent costs, each of these also depending on attempt outcomes (contact and cooperation, contact and no cooperation, non-contact). Our scheduling algorithm assumes two steps. In the first step, we compute per group an optimal schedule of contact attempts for a set of all feasible combinations of survey modes. This is done by maximizing the group quality indicator (e.g., the response rate), given that the corresponding group costs are below a predefined threshold. This step of the algorithm utilizes the pre-estimated group contact probabilities and response probabilities per survey mode. In the second step, we determine the staff assignment such that, for the optimal schedule found in the previous step, constraints on the maximum workload and staff availability are met. The aim of this procedure is to obtain a schedule which is used during the data collection fieldwork. The algorithm is expected to lead to an integrated approach of the scheduling algorithm within adaptive survey designs.
Keywords: Nonresponse; Optimization; Adaptive design
Biography: I am currently carrying out my PhD research on Optimal Resource Allocation for Adaptive Survey Designs at VU University Amsterdam. My main research interests fall in the broad category of optimization, stochastic modelling and online optimization.