Measuring Day-to-Day Variability in
Travel Behavior Using GPS Data

DEGREE OF REPETITION IN BEHAVIOR

Introduction

This section focuses on an analysis of the degree of repetition in travel characteristics across multiple days. Table 5 provides the analysis for various multiday samples without controlling for day-of-week while Table 6 provides a similar analysis for multiday samples while controlling for day-of-week (by including weekday samples only). The tables consider various multiday samples separately to account for the fact that some individuals provided fewer days of data than others. For each sample, the percent of individuals who exhibited the same value of a characteristic over all days, all but one day, and all but two days is provided. For travel time and distance measures, the percent of individuals who exhibited values of characteristics within ±20% of their median value is identified. For example, suppose an individual's median travel time over a five day period is 60 minutes. If the person had daily travel times that all lie between 48 minutes and 72 minutes, then the person is considered for inclusion in the first column (% same on all days). The median was used instead of the mean to eliminate the effect of outlier values. With respect to departure times and arrival times, the time of day was converted to a continuous clock time (e.g., 6 AM is equal to 360 min starting at 12 midnight) for ease of calculation of median and the ±20% range.

Summary of Results

In general, both Tables 5 and 6 show that there is considerable variability in travel behavior across multiple days. The percent of individuals in each sample who exhibit the same characteristic across all days (the first column in each table) is extremely small, with many entries showing zero. Both tables show that the extent of variability increases as the period of observation increases. This finding is consistent with expectations; as one observes an individual for a longer duration, greater differences will surface. In general, the tables show that the greatest stability occurs in regard to the last three variables that represent temporal events - first departure from home, final arrival at home, and final departure from work. The figures in these columns are a function of the range of time considered (here, ±20% of the median is considered) and should be interpreted with caution. When ±10% of the median value was chosen as the defining range, it was found that the percent of individuals who exhibited times within that range over multiple days dropped dramatically.

In comparing Table 5 against Table 6, it can be seen that controlling for day-of-week effects does not greatly influence the extent of variability and repetition seen in travel characteristics. Table 6 includes samples that have at least three weekdays of observations.

Table 5. Analysis of Repetition of Behavior in Multiday Samples

Table 6. Analysis of Repetition of Behavior in Weekday Samples

Note: For the 3-weekday sample, the last column is not applicable (no repetition).

In general, the percentages here are rather comparable to those seen in corresponding samples in Table 5. This finding is not really contrary to expectations considering that the sample is not a pure commuter sample. One might expect considerable differences when controlling for day-of-week if the sample includes only commuters. In comparing across characteristics and using "±20% of the median" as the defining range for time and distance variables, it appears that mid-day non-work trips and time of arrival at home show the greatest level of similarity across multiple days.

The descriptive analysis presented in Tables 5 and 6 provided a clear indication that, even though sample-wide variability in average travel characteristics across days (as seen in Tables 2, 3, and 4) does not appear significant, there is considerable day-to-day variability in travel characteristics for individual travelers when viewed at a disaggregate level. It was therefore considered appropriate to measure the extent of intra-personal variability present in the data set, determine its sensitivity to the period of observation, and compare it against values reported in the literature. This is presented in the next section.

Back to Main Page

FHWA Home | OHPI Home | Feedback

United States Department of Transportation - Federal Highway Administration