FINAL REPORT

UCTC Grant # DTRS99-G-0009

GPS-Based Tracking of Daily Activities

by

Reginald G. Golledge

and

Jianyu Zhou

Department of Geography and

Research Unit on Spatial Cognition and Choice

UCSB

September 2001

Acknowledgement: We acknowledge the assistance of DOT
who made the Lexington CD available to us.

Analysis of Variability of Weekly Travel Behavior
Using GPS-Recorded Data

Preface: This project was initially conceived to investigate how the GPS-based recording of weekly travel behavior in Lexington (KY) could be integrated with the latest disaggregate travel behavior simulation models (SMASH & ALBATROSS) to determine if such recordings of data would improve model prediction capability. Funding limitations required drastic downsizing of the project. The revised project focused only on the GPS data for Lexington, and it was analyzed for patterns of spatial and temporal characteristics that may shed further light on individual and household travel. This Final Report summarizes the results of that work.

Executive Summary

Statement of Purpose

While characteristics of daily travel behavior have been determined from analyses of the reconstructed household travel behavior recorded in travel diaries, such reconstructions are subject to criticisms. Respondents in a survey may lie or falsely recall information about destinations, times of travel, trip purpose, trip destination and other critical characteristics, such as under-reporting of short trips and the number of stops in a trip chain (Brog, et al., 1982; Purvis, 1990). In 1997 the Department of Transportation carried out a one-week study in Lexington, Kentucky in which the cars of 100 households were equipped with GPS and in-car computers. Every stop was logged by the GPS receiver, and the purpose of the stop was recorded in real time on an in-car computer. The final report of the study gave descriptions of travel behavior but performed little analysis on the data so collected. Although the new GPS-involved data collection methodology is not expected to replace the traditional data collection method in behavioral science within a short period of time, it does provide a more robust alternative for defining personal travel than the current methods. After being provided with a CD data record of all the transactions by DOT, a variety of analytical techniques and methods were used on the GPS-collected survey data. Questions that drove these analyses included:

1.      How does the calendar periodicity—the week period –influence different types of people's travel activities in a general way?

2.      What are the periodical variations of people’s travel choices on different days of the week in terms of trip frequency, trip purposes, trip duration, trip distance, and trip direction?

3.      To what extent is the automatic-device-collected data (involving GPS) more accurate than the data collected using traditional methods? Is there any possible inaccuracy involved that may potentially limit our research conclusions?

Answers to these questions are helpful for us to understand the relationship between daily travel activity patterns over a week period (see also Hirsh, et al., 1986; Kitamura, 1987; and Kitamura & Van der Hoorn, 1987), as well as the rationale that incorporates a time factor as a determinant of travel behavior.

A second objective is the evaluation of a new travel data collection method—GPS-integrated data collection. The data set that was generated from such a collection method is for the first time used in analyzing temporal characteristics of weekly travel activities. By adding the effects of real-time on-site data collection using GPS and in-car computer data entry, we aimed to complement and evaluate knowledge gained from traditional diary and survey methods.   

Background and Procedures

An understanding of variability is central to the modeling of travel behavior (Jones & Clarke, 1988). Until the late 80’s, comparatively little attention had been paid to the question of day-to-day variability in behavior. Past attempts usually handled the problem by surveying travel on a common weekday or by collecting and analyzing one-day data from different weeks to obtain a picture of typical travel pattern averaged across individuals and days of the week. Comparison and contrast of day-to-day differences in travel behavior were rarely practiced. For the past two decades, developments in travel behavior analysis and transport policy have led to a greater awareness of the need to examine day-to-day variability in travel behavior.  It is hypothesized that a thorough understanding of variability would provide the chance to make traffic management systems (such as urban traffic control) more efficient or help ensure that road network design is more closely matched to the profile of travel demands. Improved knowledge of variability should also contribute to more efficient sampling procedures for short period traffic counts.

To capture the variability of travel behavior over a longer period, multi-day data is essential for research input. Specifically, with weekly data it becomes possible to answer a series of questions about the distribution of frequency of participation, rather than mean participation rates. The growing interest in longitudinal analysis, using repeat cross-sectional or panel data designs, has also highlighted the need to look at multi-day data sets.

For addressing the problem of variability of travel behavior, various methods for measurement exist. Total trip rates or a vector of descriptive attributes (number of journeys, number of stops, travel mode used, duration of journeys, etc) have been used to compare activity patterns by Koppelman and Pas (1984) and Hanson and Huff (1982) respectively. However, their research mainly focused on inter- or intra-person variation of travel patterns rather than on day to day, or one day-of-week to another day-of-week patterns. In this project, measures of activity-travel behavior and its variability among the seven different days-of-week were used. As part of this study, the physical measurements of trips (time duration, distance, frequency, and direction) were analyzed using K-group MANOVA (multiple analysis of variance) procedures to examine how people’s activities show different patterns among days in a week in terms of types of activities pursued. This was followed by a discriminant analysis (Huberty, 1994) of travel patterns for 13 specified trip purposes for each of the seven work days, and the structure matrix of the discriminant coefficients (Zhou & Golledge, 1999). Finally, post-hoc procedures (Hotelling’s T² and univariate t-tests (see Olson, 1974) were used to determine which variables accounted for the associations found in the discriminant analysis). Then time series analysis (Zhou, 2000) was used to further prove the conclusion derived in the former analysis. Circular statistics (Mardia, 1976; Johnson &Wehrly, 1977) were used to examine travel directions. The advantages and disadvantages of using GPS-integrated devices as a reliable means of collecting travel activity data were then analyzed. Finally, suggestions about how to improve the design of experiments involving GPS-integrated data-collecting devices were elaborated.

Discussion of Results

As expected, there exist considerable differences in travel behavior between weekdays and weekends. However, past researchers have not empahsized the difference between Saturday and Sunday. Sunday is characterized with most (relative to other day-of-the-week) reduced travel-activity intensity. But Saturday is not.

As opposed to the research conclusion of Pas (1988), even “regular” travel-activity behavior was found dependent on day-of-the-week in this study. Among weekdays, when people's activities seem pretty routinized because of work or study constraints, differences of travel-activity patterns still exist. In Pas’s paper (1988), only the trip generation rate was considered as a comparison measure, which represents only one perspective of the people’s activity patterns. Our study examined the variation of travel behavior across the week not only from the aspect of trip frequency, but also from the aspects of trip duration and trip distance. Results of the comparisons of activities on week days using variables of trip duration, distance, frequency, and direction give the following results.

Pairwise multivariate tests (T²’s) and univariate t tests are used to determine which pair of groups differ significantly on the set of trip variables. We used a loose criterion (set at the .05 level for univariate t tests) to determine which kind of trips contributed to the significant pairwise difference of people’s travel patterns on different paired days of the week.  Some interesting results are shown as follows:

·        Trips on Thursday - Friday: Not significantly different overall except for a significant difference on purp1 (pickup passenger).

·        Trips on Tuesday – Thursday: Not significant.

·        Trips on Monday – Thursday: Not significant except for purp4 (work-related business).

·        Trips on Thursday – Saturday were significantly different; the difference was emphasized most in: purp1 (pickup passenger), purp3 (work place), purp4 (work-related business) purp5 (school, college, university), and purp11 (return home).

·        Trips on Saturday - Sunday were significantly different, the emphasis mostly being on: purp7 (other errands) and purpl2 (religious activities).

• Trips on Wednesday - Sunday were significantly different, the main contribution to difference deriving from purp3 (work place), purp10 (medical or dental), and purpl2 (religious activities).
• Trips on Tuesday – Saturday were significantly different, the main contributors to difference being purp3 (work place).

·        Trips on Monday - Tuesday were not significantly different.

• Trips on Wednesday – Saturday overall were not significantly different except for purp3 (work place) and purp10 (medical or dental).
• Trips on Friday – Saturday were significantly different, with major contributions from purp3 (work place), purp7 (other errands), and purpl2 (religious activity).

This analysis shows that the variability of activity patterns on weekdays mainly comes from the flexibility associated with the noon, early afternoon or the evening time slots. Activities performed in these time slots may be eat-out, shopping, social or recreational activities-- these activities are typically less obligatory.

With regard to trip intensity, we found an asymmetrical bell curve exists in the plots of trip counts across the week (Young & Willmott, 1973). It peaks on Thursday and falls to the lowest point on Sunday. This indicates people's activities are indeed influenced and shaped by institutional constraints (like the duration of a week). As many social institutional rules are made based on weekly periods, it affects people's decision-making with regard to the allocation of time and effort. Furthermore, the travel behavior of one day may affect that of the next day-of-week (e.g., by delaying trips when problems such as congestion do not allow completion of the original scheduling).

In our MANOVA analysis, Friday shows its importance in terms of people's allocation of time for travel. This phenomenon has not been explored in prior research. Combined with the research results derived from time series analysis, we tend to attribute the increase of time spent on travel to the increase of time spent on shopping and social or recreational activities.

The time series analysis revealed the periodicity associated with each type of trip. A rhythmic pattern (Shapcott & Steadman, 1978) for different types of trips was established. Most types of trips are performed on a routine basis across the week period-- at least once a day-- except religious and dental or medical trips. Some trips tend to be performed more frequently during a day (such as work trips, social-recreational trips, or eat-out trips) than other types of trips (such as return-home trips and shopping trips). To establish a reliable spectral signature for different trip purposes, however, we would need data from a longer time period (minimum of about 6 months).

As opposed to what we have assumed, the trip-direction-circular histograms that we produced were not exactly symmetrical about any arbitrary diameter of the base circle (see Zhou, 2000). There exists a mean vector to indicate the possible direction in which multi-purpose trips may have been undertaken. However, a sensible explanation for this effect needs to relate the direction preference to the local activity site distribution and socio-demographic features (as would be done if the data were to be used in the SMASH or ALBATROSS simulations). We hypothesized that direction data can be readily and accurately used in travel behavior and transportation research with the help of GPS. The attempt made in this study shows that the application of circular statistics to travel data is feasible. But more work needs to be done to make it a worthy tool of analysis in activity-related research.

Defects in GPS Collected Data

In the process of compiling and using data collected with the GPS-integrated device, we found the data collection procedure and technique as yet not sufficiently well developed to meet the up-to-date needs of behavior and transportation research.

In the Lexington travel survey, travel data were collected in a general way in terms of the sampling method used. Sampled households are spread evenly throughout the study area. And the sampled people in the households are stratified by age groups.  But, for some reason, the socio-demographic information associated with each sampled driver was not completely recorded, which impedes efforts to relate the revealed travel pattern to various socio-demographic factors (as would have been needed if SMASH and ALBATROSS had been used).

In addition, the collected travel records are restricted to travel made with motorized vehicles. Short trips made by bicycle or on foot are ignored and not recorded. This is due to the fact that the size and weight of the GPS-integrated device made it difficult for individuals to travel with them when biking or walking. Power supply is also a problem. Another problem was that, at times, the respondent simply forgot to turn on the recording device and enter the trip data into it. That caused some loss of trip information, similar to forgetting to enter trips into a diary. In cases where the trip is relatively short, the GPS module may not gain enough time to get a positional fix for the record. What is recorded is comparatively useless information and has to be discarded during the map-matching phase.

We were able to match about 97 percent of the GPS collected trips with digital maps. 87 percent of the corrupted trips have distances less than 0.16 kilometers. This means that probably some short trips were missed. Therefore, using a GPS-integrated device for collecting travel data fixed the problem of "human memory malfunction", but introduced "machine memory malfunction". Furthermore, although more than 1800 trips were traced and recorded, when breaking trips into different trip types, the trip counts were not large enough for statistical analysis.

Another problem with the data set was the classification schema for activities. The classification schema is easy to use for survey purposes (since it is a general classification), but it is not necessarily good for research purposes. For example, a more detailed classification scheme is needed for researchers who are interested in time-budget studies. It will be a blessing if future travel surveys could adopt some standardized schemas and fit the trips made by the respondents into more detailed classes. This is essential for making daily comparative studies possible and will improve researchers’ ability to examine the data set in a more comprehensive way.

Conclusions

In the attempts to reveal the determinants of variability of activity patterns, the earlier works by Hanson (1982) and Pas (1984) have identified their association with social-demographic and spatial or environment variables. The list of explanatory variables that potentially have influences on travel activities includes age, marital status, gender, employment status, education level, presence of young children, income, auto-ownership, residential density and the location of individual or household to potential destinations, such as shopping centers, churches, gas stations, etc. However, none of this literature addressed the role of the weekly period in the daily travel behavior. Kunert (1994) noticed how various life-cycle groups show distinctively different profiles of trip-making over the course of a week. But he considered the social constraint – the week period - only as a time frame for examining travel patterns instead of as an explanatory variable for them.

This study advanced our understanding of the unusual determinant (the weekly period) of activity pattern by a breakdown and analysis of its influence on travel distance, frequency, purpose, direction, and the type and temporal characteristics of activity pursued at an aggregate level. Kunert’s finding (1994) that there is tendency of increasing trip making from Monday through Friday and relatively low trip making at the weekend was confirmed. As a bonus, Thursday was revealed with the highest trip intensity compared to other days of the week.

Another outcome of this study would be the evaluation of a new travel data collection method.  The Lexington study was the first attempt to utilize a travel data collection based on GPS. The GPS-involved travel data collection method has its advantages in accuracy and reliability. In addition, the overall data post-processing procedure can be automated. Up to the time that LTS data was collected, the commercially available GPS-integrated device was still bulky and difficult to carry individually. This led to the fact that only travel by automobiles was recorded in the LTS data set, which potentially modifies any conclusion based on this data set. However, without losing generality, the research conclusions derived in this study are valid if we limit its applicability to motorized travel only. But we also recognize that hardware developments over the past 2-3 years have made GPS easily portable. This should lead to increased use of GPS to collect spatially accurate and complete data that obviates the need for recall from long-term memory and difficulties in remembering to complete surveys or diaries.

Selected References

Brog, W., Erl, E., MeyBurg, A. H., & Wermuth, M. J. (1982). Problems of non-reported trips in surveys of nonhome activity patterns. Transportation Research Record, 891, 1-5.

Golledge, R. G., & Stimson, R. J. (1997). Spatial Behavior: A Geographic Perspective. New York: Guilford Press.

GPS City.com. (2001). Differential GPS (DGPS) Explained. http://gpscity.securesites.com/products/dgps_explained.html.

Hanson, S. (1980). The importance of the multi-purpose journey to work in urban travel behavior. Transportation, 9, 229-248.

Hanson, S. (1982). The determinants of daily travel activity patterns: Relative location and socio-demographic factors. Urban Geography, 3(3), 179-202.

Hanson, S., & Hanson, P. (1993). The geography of everyday life. In T. Gärling & R. G. Golledge (Eds.), Behavior and Environment (pp. 249-269). Amsterdam: Elsevier Science Publishers.

Hanson, S., & Huff, J. O. (1982). Assessing day-to-day variability in complex travel patterns. Transportation Research Record, 891, 18-24.

Hanson, S., & Huff, J. O. (1988). Repetition and day-to-day variability in individual travel patterns: Implications for classification. In R. G. Golledge & H. Timmermans (Eds.), Behavioral Modelling in Geography and Planning (pp. 368-398). London: Croom Helm.

Hirsh, M., Prashker, J. N., & Ben-Akiva, M. (1986). Dynamic model of weekly activity pattern. Transportation Science, 20(1), 24-36.

Huberty, C. J. (1994). Applied Discriminant Analysis. New York: Wiley.

Kunert, U. (1994). Weekly mobility of life cycle groups. Transportation, 21, 271-288.Marble, D. F., Hanson, P., & Hanson, S. (1972). Household Travel Behavior Study 1: Field Operations and Questionnaires. Evanston, IL: The Transportation Center at Northwestern University.

Johnson, R. A., & Wehrly, T. E. (1977). Measures and models for angular correlation and angular-linear correlation. Journal of the Royal Statistical Society, Series B, 39, 222-229.

Jones, P. M., & Clarke, M. (1988). The significance and measurement of variability in travel behavior. Transportation, 15, 65-87.

Kitamura, R., & van der Hoorn, T. (1987). Regularity and irreversibility of weekly travel behavior. Transportation, 15, 227-251.

Kitamura, R. (1987). An analysis of weekly activity patterns and travel expenditure. In R. G. Golledge & H. Timmermans (Eds.), Behavioral Modelling Approaches in Geography and Planning (pp. 399-423). New York: Croom Helms.

Koppelman, F. S., & Pas, E. I. (1984). Estimation of disaggregate regression models of person trip generation with multi-day data. In J. Volmuller & R. Hamerslag (Eds.), Proceedings of the Ninth International Symposium on Transportation and Traffic Theory (pp. 513-531). Utrecht, The Netherlands: VNU Science Press.

Mardia, K. V. (1976). Linear-circular correlation coefficients and rhythmometry. Biometrika, 63, 403-405.

Olson, C. L. (1974). Comparative robustness of six tests in multivariate analysis of variance. Journal of the American Statistical Association, 69, 894-908.

Pas, E. I. (1984). The effects of selected socio-demographic characteristics on daily travel activity behavior. Environment and Planning A, 16, 571-581.

Pas, E. I., & Koppelman, F. (1984). Intrapersonal variability in weekday urban travel and related behavior: Formulation and examination of some hypotheses. Paper presented at the 63rd Annual Meeting of the Transportation Research Board, Washington, DC.

Pas, E. I. (1988). Weekly travel-activity behavior. Transportation Research B, 15, 89-109.

Pas, E. I., & Koppelman, F. (1985, January). Analysis of multi-day travel-activity patterns. Paper presented at the 64th Annual Meeting of the Transportation Research Board, Washington, DC.

Purvis, C. L. (1990). Survey of travel surveys II. Transportation Research Record, 1271, 23-32.

Shapcott, M., & Steadman, P. (1978). Rhythms of urban activity. In D. Parkes & N. Thrift (Eds.), Human Activity and Time Geography (pp. 49-74). London: Edward Arnold.

Upton, G., & Fingleton, B. (1985). Spatial Data Analysis by Example: Volume 2, Categorical and Directional Data. New York: John Wiley & Sons.

Young, M., & Willmott, P. (1973). The Symmetrical Family. London: Routledge and Kegan Paul.

Zhou, J. (2000). Analysis of variability of weekly travel behavior using GPS-recorded data. Unpublished M.A., University of California, Santa Barbara.

Zhou, J., & Golledge, R. G. (1999, August). A GPS-based Analysis of Household Travel Behavior. Paper presented at the Western Regional Science Association Annual Meeting, Kauai, HI.

Zhou, J., & Golledge, R. G. (2000, February 26-March 1). An analysis of household travel behavior based on GPS. Paper presented at the International Association of Travel Behavior Researchers Annual Meeting, Gold Coast, Australia.

Bio-Sketch

PROFESSOR REGINALD G. GOLLEDGE

Professor Reginald G. Golledge is a Professor of Geography at UCSB.  He is a senior professor with interests in behavioral geography, including disaggregate transportation modelling, spatial cognition, cognitive mapping, individual decision-making, household activity patterns, and the acquisition and use of spatial knowledge across the life-span.  His research has included work on adults, children, teenagers, mentally retarded persons, and adventitious and congenitally blind people.  He has had more than two decades of experience in designing survey instruments and collecting data in both field and laboratory situations.  He has published extensively in the literature of several fields including geography, regional science, and psychology.  He has considerable experience supervising post-doctoral researchers, Ph.D., and Master students, in managing large grants and contracts, and has been chair of his department.

Professor Golledge has written or edited fourteen books, sixty chapters in books, more than 100 papers published in refereed academic journals, and about 150 miscellaneous publications including technical reports, book reviews, published research notes, and so on.  He has presented more than two hundred papers at local, regional, national, and international conferences in geography, regional science, planning, psychology, and statistics.  He received an Association of American Geographers Academic Honors Award in 1981 and the Institute of Australian Geographers International gold medal in 1999.  He was a Guggenheim Fellow in 1987-88.  He is an Honorary Life-Time Member of the Institute of Australian Geographers and a Fellow of the American Association for the Advancement of Science.  He has been Associate Editor and Editor of the journal Geographical Analysis and a Founding Editor of the journal Urban Geography.  He has served on Editorial Boards of the Annals of the Association of American Geographers, The Professional Geographer, Tijdschrift Voor Economische en Sociale Geografie, Environment and Behavior, and The Journal of Spatial Cognition and Computation, and he has been a reviewer for many different journals, as well as for the National Science Foundation, the National Institute of Justice, the National Institute of Health, the Canada Council, the Australian Research Grants Council, and the European Science Foundation.  Professionally, he has served on national research grants committees, on the AAG Honors Committee, and three times on the AAG Program Committee—acting as chair of that committee for the San Diego Meetings.  In 1998-99 he was elected Vice President of the Association of American Geographers, and he was elected President for 1999-2000.

Professor Golledge became legally blind in 1984.  He now works almost exclusively in a multi-researcher environment.  Over the years he has worked with statisticians, educational psychologists, cognitive psychologists, cognitive scientists, developmental psychologists, economists, planners, and regional scientists, as well as transportation engineers.  His collaborative research has taken place within the United States and in Canada, England, Sweden, The Netherlands, New Zealand, and Australia.  He has also interacted extensively with the private sector, including state governments and transportation planning agencies in the states of New South Wales, Victoria, and West Australia in Australia; the National Government of The Netherlands; and with private firms in Japan, Canada, and the United States.

Bio-Sketch

Jianyu (Jack) Zhou

Personal Data

Date of Birth: 03/06/74; Email: zhou@geog.ucsb.edu Homepage: www.geog.ucsb.edu 

Address: Department of Geography, UCSB, Santa Barbara, CA 93106-4060

Education

Ph.D. Student in Geography, UCSB (GPA 3.93) (2000-present)

M.S.  Student in Computer Science, UCSB (GPA 3.93) (2000-present)

M.A., Department of Geography, UCSB (GPA 3.93) (1997-2000).

B.S., Beijing University (China) Economic Geography, Department of Geography (GPA 3.7).

Employment and Research Experience

Graduate Researcher, “Personal Navigation System for Blind People,” a joint project of Department of Geography and Department of Psychology, UCSB

Graduate Researcher, NCGIA, UCSB: (1999.1---1999.8)

Graduate Researcher, Department of Geography, UCSB (1998.9 ---present)

Graduate Researcher, NCGIA, UCSB: (1998.6---1998.9)

Programmer/Analyst, Department of Geography, UCSB

Teaching Assistant, Department of Geography, UCSB: 1997-1999

Undergraduate Student Researcher, Department of Geography, Peking University

Computer Skills

Familiar with UNIX, DOS, Window95/98, and Window NT. Experience with ArcView/Avenue, ArcInfo/AML , CorelDraw, FoxPro, Visual Café. Programming experience in C/C++, VC, HTML, Java and JDBC, SQL, Fortran, Pascal.

Honors

Department Block Grant Fellowship, 2000-2001, 1999-2000, 1998-1999, 1997-1998 UCSB.

Outstanding Student Awards, 1994-1995, Peking University.

Publications and Presentations:

Goodchild, M. F., & Zhou, J. Finding geographic information: Collection-level metadata. Geoinformatica, Submitted.

Zhou, J. (2000). Analysis of variability of weekly travel behavior using GPS-recorded data. Unpublished M.A., University of California, Santa Barbara.

Zhou, J., & Golledge, R. G. (1999, August). A GPS-based Analysis of Household Travel Behavior. Paper presented at the Western Regional Science Association Annual Meeting, Kauai, HI.

Zhou, J., & Golledge, R. G. (2000, February 26-March 1). An analysis of household travel behavior based on GPS. Paper presented at the International Association of Travel Behavior Researchers Annual Meeting, Gold Coast, Australia.

Zhou, J., & Golledge, R. G. (2000, April). An analysis of variability of travel behavior within one-week period based on GPS. Paper presented at the 1st IGERT Graduate Student Research Conference, UC Davis, Davis, CA.

Papers Presented

Zhou, J. (2000). Analysis of variability of weekly travel behavior using GPS-recorded data. Unpublished M.A., University of California, Santa Barbara.

Zhou, J., & Golledge, R. G. (1999, August). A GPS-based Analysis of Household Travel Behavior. Paper presented at the Western Regional Science Association (WRSA) Annual Meeting, Kauai, HI.

Zhou, J., & Golledge, R. G. (2000, February 26-March 1). An analysis of household travel behavior based on GPS. Paper presented at the International Association of Travel Behavior Researchers (IATBR) Annual Meeting, Gold Coast, Australia.

Zhou, J., & Golledge, R. G. (2000, April). An analysis of variability of travel behavior within one-week period based on GPS. Paper presented at the 1st Integrative Graduate Education and Research Traineeship Program (IGERT) Graduate Student Research Conference, UC Davis, Davis, CA.

2001 UCTC Final Report Recipients

Sent

Elizabeth Deakin (5)

Jack Zhou (5)

Jean Wolfe (DOT) (1)

Elaine Murakami (FHWA DOT) (3)

Tommy Gärling (1)

John Rayner (Ohio State) (1)

Mike McNally (Irvine) (1)

Will Recker (Irvine) (1)

Tom Golob (Irvine) (1)

Harvey Miller (Utah) (1)

Don McFadden (Econ; UCB) (1)

H. Mahmassani (Texas) (1)

Ben-Akiva (MIT) (1)

Scott Ramming (MIT) (1)

14 copies left in office