On New Measures for Detection of Data Quality Risks in Mobility Panel Surveys

Abstract

Multiday and multiperiod panel surveys are state-of-the-art methods to assess changes in individual travel behavior. Though important for transport planners, these surveys are rather time-consuming for participants and therefore might lead to erroneous and biased mobility data. Variability in the data quality significantly affects statistical analyses of mobility figures as well as common microscopic travel demand models that use the mobility data as the basis for generating activity plans. Supplementary to the well-known approach of weighting biases in key figures of mobility, this paper focuses on methods for detecting data quality differences between individual travel diaries. These quality measures address aspects of motivation loss at different stages of the survey. A classification approach based on these new quality measures helps to detect erroneous data and possible dropouts. The results might help reduce dropouts in general by addressing the potential dropouts individually in advance and boosting their motivation. Quality measures are tested with recent data from the German Mobility Panel. For participants older than 60 years of age, the quality measures show good classification results in regard to accuracy, but for participants younger than 35 years of age the quality measures are not effectual in identifying dropouts. Such an individual approach combined with the partial inspection and correction of travel diaries may be useful for microscopic travel demand modeling based on external activity chains.