Building a Rich Activities-travel Database from an OD Survey

Building a Rich Activities-travel Database from an OD Survey


M Munizaga, S Jara-Díaz; J Olguín, Universidad de Chile, CL


We explore the possibility of obtaining from an OD survey all information required to calibrate a time use-mode choice model. We analyze time use and activity patterns, and explore the aggregation of observations from different individuals.


Although an OD survey is devoted to describe in detail the travel pattern in a particular period, the information collected allows recovering, with some level of aggregation, a description of time use of the individuals observed. This allows the possibility to calibrate the time use ? mode choice model system proposed by Jara-Diaz and Guerra (2003) with a very rich and trustable database. We explore that possibility for the Santiago 2001 OD survey (Ortúzar et al, 2003) that has more than 12,000 households sampled randomly from the population of Santiago, each observed during a whole day. As a first stage of this process, we analyze the available information and identify the possible ways to build adequate descriptions of the individuals? time assignment to different activities. Also, we analyze the sociodemographic characteristics of the sample, which could be used as segmentation variables and to link observations. The objective of this work is to obtain from the OD survey, a database similar to one obtained specifically with the purpose of calibrating the time use ? mode choice model system, which is much smaller in size.

We identify the information required to calibrate the time use ? mode choice model system, and how it can be obtained from the Santiago 2001 OD survey. From this information, we obtain the time allocation to different activities of a representative sample of the inhabitants of Santiago (the OD sample), and analyze the individuals? characteristics that influence these time allocations. There are two basic ways to describe the time assignment of a group of individuals: trough the average time assignment and through the activity pattern, that represents the proportion of individuals who are conducting each particular activity at any instant through a period of time, for example a day.

The Santiago OD survey was conducted from July 2001 until April 2002. 12,346 households were surveyed during the normal season, and 3,191 during the summer season. For each surveyed household, all the characteristics of the household and those of each of person in the household are available. The modelling framework we want to apply models the decisions of allocating time to work, leisure activities, and constrained activities. Therefore we will work only with the workers observed in the sample during the normal season.

Because the OD survey is a travel survey and not a time use survey, the activities can only be obtained through trip purposes, and the degree of aggregation is constrained to the trip purposes included in the OD survey. The detail of what happens inside each of these aggregate activities is not known. The activities that can be deduced from the Santiago OD survey are: stay at home, work (out of home), study (out of home), recreation (out of home), shopping and errands, travel, other activity out of home.

As each individual was observed only during one day, the activity patterns of the different days of the week come from different individuals. Although, we can make comparisons among days, it would be interesting to build weekly observations through linking observations of very similar individual observed in different days of the week. We explore the possibility of finding ?twins? or ?triplets? to aggregate the information of different individuals observed in different days to a single weekly observation.

Given the size of the database, we can also compare activity patterns for different types of individuals, defined according to their socioeconomic characteristics. This type of analysis has great value for future aggregation and segmentation work. Our preliminary analysis show that the most important variables to explain differences in time assignment are: gender, income, age and household location.

The analysis will be centred in the possibility of aggregating information and generating databases rich in variance, that allow to calibrate the time use ? mode choice model system, already applied to smaller databases. If we prove it feasible to obtain all the information required from such a large and rich database, we will then count with a very powerful approach to the better understanding of individuals? behaviour.


Jara-Díaz, S.R. and Guerra, R. (2003) Modeling activity duration and travel choice from a common microeconomic framework. 10th International Conference on Travel Behaviour Research Conference, Lucerne, August 2003.

Ortúzar, J. de D., Ivelic, A.M., Malbrán, H. and Thomas, A. (2003), ?The 1991 Great Santiago Origin-Destination Survey: Methodological Design and Main Results?, Traffic Engineering and Control, Volume 34, pp. 362-368


Association for European Transport