This section presents the travel simulator (section 5.3.1) that was designed to collect the type of data needed to estimate the choice model presented in the previous section. Also, this section briefly discusses participants’ background characteristics (section 5.3.2). Subsequently, participants’ evaluation of the experiment is presented (section 5.3.3).
5.3.1 The travel simulator experiment
In the experiment110, participants were asked to perform 20 trips that were randomly selected as a subsample from all the possible combinations of the given choice scenarios and the different designed attribute levels (to be elaborated later). Figure 5-2 provides a screen shot of an example of a trip that was presented to experiment participant. The top portion of the screen presents the choice situation to the participant, in terms of (i) the preferred arrival time and the importance of being on time for that given day; (ii) the availability of a telework option; and (iii) the availability of fully reliable travel time information for a subset of routes. The left portion in the middle of the screen presents an abstract road ‘network’, consisting of three routes from home to work. Two routes are highways, and there is a secondary route. Travel times (communicated to participants as being door-to-door) for these routes during peak hour are represented on the right hand side, in the middle portion of the screen. Riskiness of travel times is communicated to participants in terms of a best guess travel time,
110
Extensive discussions of the usefulness and limitations of travel simulators for data collection can be found in Bonsall (2004), Mahmassani (2006), Chorus et al. (2007b), and Chorus et al. (2013).
and a range of possible travel times, for each route. The lower portion of the screen represents the travel information service111 (which refers to the two highway routes only).
Figure 5-2: A screen shot of one randomly selected choice situation of the simulator112 After extensive deliberation, and a round of testing using synthetic data, the following attribute-levels were chosen: highway 1’s best guess travel time is varied between 55 minutes and 60 minutes, and its range varies between ±10 minutes and ±15 minutes. Highway 2’s best guess travel time is varied between 45 minutes and 50 minutes, and its range varies between ±20 minutes and ±25 minutes. That is, highway option 1 is the slower (on average) but less risky route, whereas highway option 2 is the faster (on average) but more risky route. The secondary route’s best guess travel time is fixed at 65 minutes, with a fixed range of 5 minutes (which makes it the slowest (on average) but almost riskless route). Concerning the choice- context, five different scenarios were available, which were systematically varied across trips:
111 The simulator, after having presented the traveler with the full set of choice options (routes,
telework option, and the travel information acquisition option), actually first presents a binary choice regarding whether or not to acquire the available travel information. At first sight this may seem to be different from how the choice probability for travel information is formulated in the choice model (being framed in a multinomial context). However, the full choice set was visible for participants during their decision whether or not to acquire the information, and they were explicitly told that after they had acquired information (if they would), they would be asked to choose from the set of travel route and teleworking options given the received information. This implies that mathematically as well as behaviorally speaking the choice to acquire information or not can be cast as being made in the context of a multinomial choice set. See also footnote 108.
112
The screen shot is adapted from the original one where Dutch instead of English is used in the simulator.
the commuter is told that (1) she has no meetings at the office, and she has no preferred arrival time; (2) she has no meetings at the office, but she would like to arrive at the office at 9AM; (3) she has an informal meeting with one colleague at 9AM, and being late would be slightly annoying but not a big deal; (4) she has a multilateral meeting at 9AM for which she would rather not be late; (5) she has a very important meeting at 9AM, and being late would be very embarrassing. The departure time for each choice round is given and is varied among 7:50AM, 7:55AM and 8:00AM. The commuter, therefore, may run the risk of arriving later than the preferred arrival time mentioned in the given scenarios. Finally, trips vary in terms of the availability of the telework option, and of the travel information service. Teleworking, if available, does not cost anything; it was communicated to participants that teleworking would imply having a conference-call from home, in case the scenario featured a meeting. Participants were asked to imagine the situation where they had all the right ICT-tools available for working from home. The cost of information, if available, varies across trips, and can take on the values €0.5, €1, €1.5, and €2.
If information is acquired, the travel times for the two highway-routes are shown, and the traveler is subsequently asked to make a choice between the three routes (two of which are not risky anymore) and, if available, the telework option. If no information is acquired, the traveler is asked to make a choice based on risky travel times for all three routes. In case a route is chosen (as opposed to the telework option), the travel time for the chosen route is shown on screen, as is the arrival time at the office.
5.3.2 The experiment participants
Participants were asked to perform 20 trips, and the average duration of the experiment equaled roughly 30 minutes. Participants were offered two ways to participate: they could do the experiment from home (online), or they could come to one of our university’s computer rooms. In the former situation, the introduction was presented on screen, while in the latter situation, the introduction was presented by one of the graduate students (as well as being available on screen). Participants received a fixed fee for participating. Participants who came to our university’s computer room to do the experiment received €25, and those that did the experiment from home (through the internet) received €10. These participants were asked to identify with the choice situations as if these were real, and to make choices accordingly based on the particular travel time distributions, information costs, and contextual scenarios. Participants were recruited by means of ads in local newspapers. It was made clear, although no actual screening was implemented, that participants were expected to have experience with commuting to work or university by car. In total, 137 participants were recruited for the stated preference experiment – Table 5-1 shows their socio-demographic characteristics of the participants to the stated preference experiment, and distinguishes between the online versus computer room setting.
Table 5-1: Characteristics of participants
Variable
Stated preference treatment (N=137) Computer room (N=35) Online (N=102)
Frequency % Frequency %
Age
15-24 8 22.9 20 19.6
35-44 9 25.7 29 28.4 45-54 3 8.6 12 11.8 55-64 2 5.7 4 3.9 65 or older 0 0.0 0 0.0 Occupation Employee 23 65.7 76 74.5 Volunteer 1 2.9 4 3.9 Student 10 28.6 20 19.6 Others 1 2.9 2 2.0 Gender Male 20 57.1 57 55.9 Female 15 42.9 45 44.1
Highest completed education113
Elementary school 0 0.0 0 0.0
Secondary school 4 11.4 16 15.7
Vocational education 3 8.6 19 18.6
Higher education 28 80.0 65 63.7
Others 0 0.0 2 2.0
Net monthly household income (€) 114
No income 1 2.9 2 2.0 <1000 7 20.0 16 15.7 1000-2000 7 20.0 18 17.6 2000-3000 5 14.3 19 18.6 3000-4000 6 17.1 13 12.7 4000-5000 5 14.3 9 8.8 >5000 3 8.6 7 6.9
I do not want to answer or I do not know 1 2.9 18 17.6
Main commuting mode
Car (as a driver) 20 57.1 65 63.7
Car (as a passenger) 2 5.7 2 2.0
113 The categories of education are classified based on Dutch education system, which were used to
ask participants in the study: a) Elementary school includes Basisonderwijs, b) Secondary school includes VMBO, MAVO, HAVO, VWO/Gymnasium, and VAVO, c) Vocational education includes MBO, and d) Higher education includes HBO and WO.
Public Transport 6 17.1 21 20.6
Motorcycle/Moped 0 0.0 2 2.0
Bicycle 6 17.1 10 9.8
By foot 0 0.0 0 0.0
Others 1 2.9 2 2.0
5.3.3 The experiment evaluation
Participants were asked five survey questions to evaluate the experiment after they finished the experiment. These questions took the form of propositions (e.g., “I found it difficult to concentrate during the experiment”), with associated Likert-scale answer categories (1 = totally disagree; 5 = totally agree). Results (Table 5-2) suggest that participants of the experiment in general enjoyed the experiment, and did not find it hard to concentrate or identify with the choice situations and contextual scenarios presented to them. Mean scores quite strongly resemble those obtained in a previous simulator study where the same questions were asked to participants (Chorus et al., 2007b).
Table 5-2: Evaluation of the experiment (stated-preference)
Questions Mean Score
1. Difficult to concentrate during the experiment 1.88
2. Easy to understand the experiment 4.11
3. Difficult to empathize the choice situations 2.09
4. Nice to do the experiment 4.28
5. Problematic to stay motivated during the experiment 1.74