The increased need for understanding variability in travel behaviour and the parallel advances in travel demand modelling techniques have recently brought in a re-discovery of panel data as a useful replacement for the more typical but limited cross-sectional data. Panel data can be classified into two categories: long survey panels (the most common in the literature), that consist of repeating the same survey at "separate" intervals (e.g. once or twice a year for a certain number of years, or before-and-after an important event); and short survey panels, that consist of multi-day data where repeated measurements on the same sample units are gathered over a "continuous" period of time (e.g. seven or more successive days), but the survey is not necessarily repeated in subsequent years. Recent examples of panel data over a continuous period of time are the two-day time-use diary for the US National Panel Study of Income Dynamics (2002) and the six-week travel and activity diary data panels collected in Germany (Mobidrive, 1999) and Switzerland (Thurgau, 2003). The Santiago Panel (2006-2008), is instead an example of a panel that combines short and long survey.

Long survey panels are useful to study dynamic effects over waves (Smart, 1984; Sunkanapalli et al., 2000), such as habit formation, learning (Arentze and Timmermans, 2003) and the reaction to important policies (Yañez and Ortúzar, 2008). Short survey panels, instead, have been used to detect effects such as rhythms of daily life (Axhausen et al., 2002), to explain current behaviour on the basis of the individuals' history and experience (Cirillo and Axahusen, 2006) and to account for interpersonal variability (Cherchi and Cirillo, 2008).

One of the key issues in models estimated on panel data is that of accounting correctly for the correlation over the various observations provided by the same individual. In this context the use of short survey panels raise new interesting issues, because it might be possible to consider several dimensions of correlation across responses. In fact, different correlation can appear over trips/tours made in the same day, the same week, the day-of-week and/or over individuals and households. Properly accounting for these dimensions of correlation is crucial to better understand the long-term structure of individual choices, to correctly estimate the models and to properly use them in forecasting.

In this paper, we use the six weeks panel data from the Mobidrive survey to estimate a mode choice model that accounts for correlation across individual over three time periods: the single day, the single week, the days of week (all Mondays in the wave, and so on), and also for correlation across individuals over the six weeks. Cherchi and Cirillo (2008) analysed these effects, estimating different models for each type of correlation and drawing conclusions on the basis of the overall test-of-fit obtained with each single model estimated. In this paper we will try to disentangle the relative effect of each type of correlation. A mixed logit model that jointly accounts for the different dimensions of correlation is estimated, allowing to directly comparing their relative effects.

References:

1.       Arentze, T. and Timmermans, H. (2003) Modeling learning and adaptation processes in activity-travel choice. Transportation 30, 37-62.

2.       Cherchi, E. and Cirillo, C. (2008) A modal mixed logit choice model on panel data: accounting for systematic and random heterogeneity in preferences and tastes. 86^th Seminar of the Transportation Research Board. Washington D.C., USA.

3.       Smart, H.E. (1984) The dynamics of change -applications of the panel technique to transportation surveys in Tyne and Wear. Traffic Engineering and Control 25, 595-598.

4.       Sunkanapalli, S. Pendyala, R. and Kuppam, A.R. (2000) Dynamic analysis of traveller attitudes and perceptions using panel data. Transportation Research Record 1718, 52-60.

5.       Yañez, M.F. and Ortúzar, J. de D. (2008) A panel data model to forecast the effect of a radical public transport innovation. 4^th International Symposium on TDM2008. Vienna, Austria.