Measurement of non-random attrition effects on mobility rates using trip diaries data

Lissy Cesarina La Paix Puello, Marie-José Theresia Olde Kalter, Karst Teunis Geurs

    Research output: Contribution to journalArticleAcademicpeer-review

    8 Citations (Scopus)


    This paper examines the influence of panel attrition on the intrapersonal dynamics in self-reported trip rates, using the data from the 2013, 2014 and 2015 waves of the Netherlands Mobility Panel, a large scale household panel. A hybrid choice model (HCM) was developed to simultaneously model the effect of socioeconomic, infrastructure and land use variables, life events and non-random attrition on trip rates, whereby the latent variable (LV) model is composed of panel attrition and survey completeness. The discrete choice model (DCM) includes four trip rate categories, including zero trips. The probability of each trip rate category was estimated for both the HCM and the DCM models; with and without the LV model. The first main conclusion from this paper is that the largest bias due to panel attrition occurs in the probability of reporting no trips per day, and 1–2 trips per day. Also, the HCM models show a correlation between the probability of reporting no trips per day and the tendency to drop out altogether. The second main conclusion is that the results show that the latent variables (attrition and completeness) are statistically significant in estimating mobility. Also, socioeconomic variables (gender, driving license, household type and size), mode preferences, spatial infrastructure and life events determine mobility rates and remain significant after adding attrition/completeness variables. Thirdly, the results proved that attrition effects significantly vary across waves.
    Original languageEnglish
    Pages (from-to)51-64
    JournalTransportation research. Part A: Policy and practice
    Publication statusPublished - 2017


    • METIS-318651


    Dive into the research topics of 'Measurement of non-random attrition effects on mobility rates using trip diaries data'. Together they form a unique fingerprint.

    Cite this