|
|
|
|
Measuring
Day-to-Day Variability in
Travel Behavior Using GPS Data
LITERATURE REVIEW ON DAY-TO-DAY VARIABILITY IN TRAVEL BEHAVIOR
Introduction
The previous section provided a detailed discussion of the reasons why multiday travel survey data is useful for travel demand modeling. However, what is the amount of day-to-day variability in travel behavior found in data sets derived from multiday surveys? This section reviews the body of knowledge on the day-to-day variability in travel behavior in an attempt to answer this question.
There are two sources of day-to-day variability in travel behavior. First, day-to-day variability occurs because people's needs and desires vary from day-to-day. Thus, for example, the number of trips one takes varies from day-to-day because one need not do grocery shopping each day. Second, behavior varies from day-to-day because of feedback from the transportation system. Thus, one might choose a different route and/or departure time for the work trip today if one encountered severe congestion yesterday on the usual route.
Literature Review
Hanson and Huff have conducted a series of studies examining multiday data from the 1971 Uppsala (Sweden) household travel survey [Hanson and Huff, 1982, 1986, 1988a, 1988b; Huff and Hanson, 1986, 1990]. The Uppsala survey obtained information on all out-of-home travel - activity behavior using self-administered travel diaries for a 35-day period. Hanson and Huff used a representative sample of 149 individuals who completed diaries for the entire 35-day period to examine the day-to-day and week-to-week variability in travel behavior.
Hanson and Huff [1986] and Huff and Hanson [1986] present detailed discussions regarding the habitual and variable behavior of individuals over time. They note that some behaviors, when examined in a disjointed framework (say, a work trip examined in isolation of the overall daily activity-travel pattern), are repeated on a day-to-day or week-to-week basis. However, when the overall daily activity-travel pattern is examined in its entirety, they find that a one-day pattern is not representative of a persons routine travel.
Huff and Hanson [1986] examine for the existence of a typical travel day. They identified the most representative travel day for each person and computed the similarity of that day to all other days in that persons record, and they found a very low sample-wide similarity index. More interestingly, even when they considered five most representative travel days and computed similarity indices (based on sets of days that are most similar to each of the five representative days), the sample-wide mean similarity improved only marginally. They concluded that considerable variability in individuals travel-activity patterns is not explained even when each persons five most representative patterns are considered. Hanson and Huff [1988] and Huff and Hanson [1990] present several further findings that challenge the existence of a typical travel day and the ability of a one-day travel diary survey to capture the variability in travel behavior.
Another interesting finding reported by Huff and Hanson [1986] is that the no-travel day is a significant pattern for several individuals and that modeling efforts should incorporate this aspect of trip-making (in this case, non-trip making) behavior of persons. To do so, they argue, would require the collection of multiday data to capture the travel and no-travel day behavior of individuals. When examining the systematic (nonrandom) portion of intra-individual (within person) variability in travel behavior, Hanson and Huff [1988] find that relatively few stops occurred at regular intervals in the individuals longitudinal record. Of the stops that were non-randomly distributed throughout the record, more were distributed in a clustered fashion than were distributed uniformly. They conclude that little of the day-to-day variability present in the individuals travel-activity pattern could be said to be regularly systematic.
Much of the work reported by Pas and his colleagues in a series of papers [Pas, 1986, 1987, 1988; Koppelman and Pas, 1984; Pas and Koppelman, 1987] utilizes seven-day activity data collected in 1973 in Reading, England. More recently, Pas and Sundar [1995] extended the work done previously by examining day-to-day variability for different travel indicators and across household members using three-day travel diary data collected in 1989 in Seattle. Koppelman and Pas [1984] and Pas [1986] examine the bias and precision levels of parameter estimates obtained when using multiday data to develop regression models of trip generation. They showed that greater levels of precision can be obtained from multiday data than from a one-day data set of equal cost, or the same levels of precision can be obtained more economically using multiday data, as discussed in the previous chapter of this report.
Pas [1987] formulated a paradigm for the representation of variability in behavior. The paradigm distinguishes between interpersonal variability and intrapersonal variability, each of which can be split into two components. Interpersonal variability consists of explained variability (that which may be systematically related to differences in personal characteristics) and unexplained variability. Intrapersonal variability also consists of two components, namely, systematic day-of-week variability and random or unexplained variability.
Pas [1987] examined day-to-day variability in daily trip rates using the Reading data set. In this work, the variance in an individuals daily trip rate about his or her daily average was used as a measure of intrapersonal variability. Using this approach, Pas [1987] found that about 50 percent of the total variability in trip-making in the data set could be attributed to intrapersonal day-to-day variability in trip generation. Pas and Koppelman [1987], in an extension of this work, found that the level of intrapersonal variability varies significantly across demographic segments. For example, they found that females exhibit higher levels of intrapersonal variability than males, possibly due to the roles traditionally played by females in households.
Pas and Sundar [1995] examine day-to-day variability in urban travel using a three-day travel data set collected in 1989 from Seattle, Washington in the U.S.A. In this research effort, the authors consider day-to-day variations in trip chaining and daily travel time in addition to examining the variability in trip generation. The paper finds considerable day-to-day variability in the trip frequency, trip chaining, and daily travel time of the sampled persons. Interestingly, they find that the level of variability in trip generation is very similar to that found previously in the 1973 Reading activity data set. In the Seattle data set, they find that about 38 percent of the total variability in the daily trip rate is due to the intrapersonal or day-to-day variability in travel behavior, compared to 50 percent in the Reading data set. They point out that the higher variability in the Reading data set may be attributable to the longer reporting period (five weekdays) adopted in the Reading activity survey.
Several other studies have examined the variability in travel characteristics on a day-to-day basis. Kitamura and van der Hoorn [1987] examine the timing with which an individual repeats a certain behavior. They found that about 70 percent of the male workers and 59 percent of the female workers in a Dutch National Mobility panel data set had identical daily patterns of shopping participation on five or more of the days of each of two weeks (six months part). A study by Hirsh, et al. [1986] regarding time allocation indicated that daily and weekly variation accounted for most of the variation in time allocation for home and travel activities, while seasonal variations contributed little.
A series of papers by Mahmassani and his colleagues [Mahmassani, 1997; Mahmassani and Chang, 1985, 1986; Mahmassani and Stephan, 1988; Mahmassani and Herman, 1990] study the day-to-day departure time choice dynamics under congested network conditions. Through a series of controlled laboratory experiments and a survey of commuters in Austin, Texas, they study the dynamics of departure time choice, trip chaining, and route choice on the journey-to-work. They find that departure time switching is more frequent than route switching in both the laboratory and field data and develop models to explain the daily variation in departure time and route choice. They do note, however, that their results are dependent on the measures of variability used and the criterion used to identify a switch in behavior. Day-to-day variability in departure time choice and route choice have also been examined and found to be significant by Mannering [1989].
In summary, a vast body of knowledge and evidence supports the notion that there is considerable day-to-day variability in travel behavior. The evidence challenges the existence of a typical travel day representative of the daily or weekly activity-travel patterns exhibited by individuals. While some variations in the extent of day-to-day variability have been found depending on the specific measures of variability used, travel characteristics examined, and data sets employed, it is clear that day-to-day variability in travel behavior exists and is substantial.
United States Department of Transportation - Federal Highway Administration