U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
SUMMARY REPORT |
This summary report is an archived publication and may contain dated technical, contact, and link information |
Publication Number: FHWA-HRT-15-015 Date: June 2015 |
Publication Number: FHWA-HRT-15-015 Date: June 2015 |
The research team used three analytical approaches in the study: a mail and telephone survey; activity-based modeling of respondents’ choice of travel modes, based on the survey results; and agent-based modeling of commuters and their preferences regarding cost, travel time, and safety considerations. The agent-based model, calibrated with data from the four neighborhoods, was developed by first using parameters from the travel behavior literature and then incorporating parameters from the study’s activity-based model. Commuters in the model choose transportation options based on their preferences, characteristics of the environment, and the travel modes available to them.
The research team conducted a survey of at least 150 respondents in each of the four neighborhoods. Respondents lived within 1.5 mi (2.4 km) and worked within 3 mi (4.8 km) of a CTA rail station. The research team designed the household survey to collect information about the household, the respondent’s current travel behavior, mode preferences under the potential transportation and design improvements, and perceptions of mode cost, convenience, time, and safety.
Staff from the Survey Research Laboratory at the University of Illinois at Chicago used a multimode design for the household survey. The staff used address-based sampling from the Delivery Sequence File constructed by the U.S. Post Office to obtain an initial 7,700 addresses within the A and B focus areas of the four neighborhoods. The staff divided sampling equally among the neighborhoods, and carried out recruitment in eight repetitions of the same experimental conditions during March–April 2012. Because of the low responses in Cicero, Pilsen, and Skokie, the survey staff added an additional 8,998 addresses in those neighborhoods in two additional samplings, with recruitment occurring during May–June and December 2012. Thus, 16,698 households received the initial recruitment contact.
The team conducted the survey in three steps:
1. Recruitment mailing to random sample:
2. Study packet mailing to eligible respondents:
3. Telephone interview (30 minutes):
The recruitment mailing asked for participation in the study and offered a cash incentive to eligible respondents who completed the written survey and the followup telephone interview. Anyone living in the household and over the age of 18 who spoke either English or Spanish (interviews and survey materials were available in either language) was eligible to return the response form, which asked whether the respondent works or goes to school outside the home, his or her work or school address, and how many times during the week the person travels to his or her nearest CTA rail station or within a half-mile (0.8 km) of that station.
The survey team used a screening process to eliminate respondents who did not work or go to school outside the home, did not travel at least once a week to their CTA rail station area, or did not work or attend school within 3 mi (4.8 km) of the CTA system. The team sent the survey study packet containing the travel diary and booklet only to respondents deemed eligible. At least 1 day after the date assigned to the respondent to complete the travel diary, the survey team telephoned to conduct the followup interview.
The overall returns to the recruitment mailings were 10.1 percent, and the final participation rate—respondents who completed the telephone interview—was 3.8 percent and ranged from 2.0 percent in Cicero to 9.2 percent in Evanston.
Travel Diary
The travel diary centered on the respondent’s travels on a randomly selected weekday. The questionnaire devoted three pages to the main tour of the day (e.g., the series of trip segments between home and a destination or from a destination to home), with up to three segments, and contained space to add additional tours. Questions involved destination, travel times, travel modes, length of trip, and number of stops. The travel mode options offered were private vehicle as driver, private vehicle as passenger, bus, train, walking, cycling, and car-sharing. Questions also included any costs for parking or highway tolls.
Survey Booklet and Telephone Interview
The telephone interview had four parts:
The survey booklet that was mailed as part of the study packet contained six worksheets covering the transit station approach with which the respondent’s address was associated. Each worksheet was illustrated with one of six “improved” scenarios (a sample is shown in figure 4). These worksheets were completed one by one during the telephone interviews, which were conducted by using CASES, the Computer-Assisted Survey Execution System Program, version 5.4.(7) The interviewer supplied travel times and costs for each travel mode based on the times, starting point, and destination data for the main trip in the respondent’s travel diary. Survey software generated the values for the interviewer. The interviewer then asked the respondent to reflect on the worksheet information and image and select the mode of transportation that they would prefer to use for the trip recorded in the diary.
Limitations of the Survey
Although the survey team sent the initial recruitment mailing to a random sample within the defined geographic areas, the survey respondents were a self-selected sample, and their participation likely reflects a greater interest in transit, cycling, and walking as alternatives to driving than that of the general population. As a result, the survey results cannot be used to make population inferences. The assumed positive bias, however, is in line with the mission of the study, which was to estimate a best-case scenario for the community shuttle.
© University of Michigan
Figure 4. Sample worksheet with illustration of transportation improvements at the West Cullerton Street and South Damen Avenue approach to Pilsen's Hoyne/Damen CTA Station.
Data from the neighborhood surveys were the basis of the activity-based model. The survey incorporated both the respondents’ current travel patterns and their stated preferences after potential implementation of the improvements, as described and illustrated in the survey materials. The stated preference portion of the data was used to estimate an activity-based model of mode choice. Utilities from this model were then used to revise initial parameters used in the agent-based model. In this way, the empirical estimation of the activity-based model was combined with the capacity of the agent-based model to simulate changes over time.
In the stated-preference survey, respondents chose from five travel modes:
Information presented about the travel options included travel time to the CTA station by shuttle, time between shuttles, driving time to the destination, parking cost at the destination, CTA train frequency, station-to-station travel time, and fare. Some variables changed according to the residential location of the respondent but did not vary among the hypothetical scenarios. These variables included: walk time to the CTA station, cycle time to the CTA station, cycle time to the destination, CTA travel time from station to station, and CTA train frequency. The respondents’ actual choices were compared against their stated preferences.
The research team built a prototype agent-based model to represent commuters living within a 1.5-mi (2.4-km) radius of a transit station in their neighborhood and work or study in the downtown area, making transit a feasible commuting choice. As part of a larger metropolitan area, commuters have a choice of various transportation modes, and they base their decisions on their preference for travel time, transportation cost, and perceived safety. The neighborhoods, commuting modes, and parameters were modeled after the four Chicago neighborhoods in the study. The agent-based model incorporated parameters from the activity-based model, and the simulations were run on this integrated model.
A permanent link to the agent-based last mile commuter behavioral model is available here. The model is built on NetLogo 5.0.2 and works on multiple operating systems.