Correcting survey measurement with road sensor data
Time-based diary surveys collect data over specified time intervals and impose a heavy response burden. To reduce the effort of reporting, respondents may omit spells or may not respond at all. Correspondingly, those surveys may suffer from underreporting and item-nonresponse. Accordingly, survey estimates might be biased downwards. In this study, the Dutch Road Freight Transport Survey (RFTS) is complemented with road sensor data and capture-recapture methods (CRC) are applied to estimate survey underreporting. The heterogeneity of the vehicles concerning capture and recapture probabilities is modeled through logistic regression and log-linear models. Six different estimators are discussed and compared. The obtained CRC estimates suggest underreporting in the RFTS, although other explanations are possible. The levels are not surprising compared with findings of other validation studies on underreporting in transport, mobility, and travel surveys. All estimators applied yield larger point estimates than the RFTS, although the estimated amount of underreporting varies between the estimators. Linking sensor data to surveys and applying capture-recapture techniques is a promising method to estimate underreporting in surveys.