Nonresponse analysis in a longitudinal smartphone-based travel study

Abstract

Currently, travel diary surveys are the standard method for measuring mobility in official statistics. Nonresponse and measurement are problematic in travel surveys, due to the fact that respondents have to recall all their travels over the course of one or multiple days and have to derive distances for all these travels. To overcome these issues, new methods that rely on passive tracking of locations over time have emerged. The aim of this paper is to assess nonresponse in an experimental travel study carried out in the Netherlands. A smartphone application was developed that passively collects GPS coordinates and automatically populates a travel diary. Participants are then asked for additional information in the diary, such as travel mode. In the experiment, respondents from a random sample of the Dutch population participated in a 7-day study that varied how respondents were recruited into the study, as well as the level and timing of a monetary incentive. We study at what stage of the study respondents choose to participate and dropout, and study nonresponse bias across 13 variables from the Dutch population register. We find that respondents receiving lower incentives, respondents of higher age and respondents with lower levels of education are strongly underrepresented. The overall representativity of the study, as expressed in R-indicators and Coefficients of Variation are rather low because of this nonresponse. We found a similar bias in nonresponse for age going in opposite direction when we computed R-indicators for an earlier web-based travel-diary study. This implies that in the future, diary studies should focus on methods to successfully combine smartphone apps and diaries through the web or on paper if the goal is to limit nonresponse successfully.