The Limitations of Using Travel Diaries to Understand and Model Travel Behaviour – Are We Missing Something Important?
Siamak Khorgami, AECOM, Peter Jones, UCL
The limitations of using travel diaries to understand and model travel behaviour is discussed in this paper. Analysis was carried out on activity diary data and this is compared versus travel diary data.
For at least the last half century, travel diaries have shaped our understanding of travel behaviour and underpinned the development of a range of transport demand models, both aggregate and disaggregate.
Over the last decade, several efforts have been made to understand travel behaviour and responses to a range of policy measures through the development of operational models of full-day activity-travel patterns (Bowman and Ben Akiva, 2001; Arentze and Timmermans, 2000; Hensher and Ton, 2002; Pendyala, et al, 2004; Bhat, et al, 2006). Given the lack of local activity-based surveys, most activity-travel demand models are developed using conventional travel diaries (Miller and Roorda, 2003).
These models use the trip-based survey data as a proxy for out-of-home activity data. This involves estimating out-of-home activity time from differences in arrival and departure times at each destination, and assuming that the ‘trip purpose’ category adequately describes the activity (ies) carried out at that destination. In particular, travel surveys usually assume that the traveller only takes part in one activity per non-home destination, but this assumption has not been questioned in the literature.
The analysis reported in this paper first uses data from the 2001 UK National Time Use Survey (TUS) to evaluate the validity of this assumption. Respondents in the TUS record their ‘primary’ and ‘secondary’ activities for each 10 minute interval over a 24 hour period of time, with no limit on the number of successive primary activities that can be recorded at the same location. The UK National Travel Survey (which collects a conventional travel diary, but with a much larger sample than the TUS) is then used to explore this further, by using an indirect measure of non-home activity participation.
The paper starts by using the TUS to analyse the type and frequency of activities that individuals take part in at out-of-home locations; results show that a significant number of individuals engage in more than one activity per out-of-home location. Across the whole sample, the average number of primary activities per non-home stop is 1.26; if ‘secondary’ activities are included, this increases to 1.60 activities per non-home destination.
The analysis then looks in more detail at cases where there are multiple primary activities at the same non-home location. It defines a ‘main’ activity at each location, to mirror the reporting of the ‘main’ trip purpose in a travel diary. Since we do not know on what basis a respondent selects their ‘main’ trip purpose, we use four alternative definitions: (i) the activity with the longest duration, (ii) the first activity carried out at each location, (iii) a hierarchical selection, based on some notion of activity importance, and (iv) a random selection of ‘main’ activity. It is possible to compare the activity distribution frequencies resulting from each definition, and to see which one most closely approximates the trip purpose frequencies from the British National Travel Survey (NTS).
Next, ‘primary’ activity combinations at one destination are examined, by estimating the conditional probability of taking part in a set of non-main activities, given participation in a specific main activity. The paper shows the percentage of time allocated to main and non-main activities at different locations, using the four definitions of main activity.
This lack of full reporting of non-home activities is important, for two reasons:
1. In modelling terms, since activity-based models are not fully representing non-home activity participation.
2. In policy terms, since we are not able to examine changing patterns of activity participation over time.
There is evidence in the UK of trip rates declining over time, which might be interpreted as a decline in welfare and a loss of well-being. However, it is possible that part of this reduction is due to activity consolidation at fewer non-home locations (e.g. shop and eat at the same location), in which case this would be interpreted as an increase in efficiency of travel behaviour. The paper sets out to indirectly test this hypothesis using NTS data, to see if the reduction in trip rates is associated with an increase in total activity time spent at each non-home location [analysis in progress].
The paper concludes by discussing some analytical and policy implications of the non-reporting in a conventional travel diary of multiple activities at non-home locations, resulting in a general underestimation of out-of-home activity participation, and what might be done to address this problem.
Association for European Transport