Gamifying and carbon scoring urban mobility for a greener Singapore

We are testing the hypothesis that... (withheld because the intervention starts next week)

We are testing the hypothesis that a mix of pro-environmental messaging, carbon scoring, and gamification-based rewards encourages public transit (Salvo and Lee, 2023).

There are three randomized groups of subjects (Control, Treatment EM, Treatment EM+I) and five phases (A to E). See Experimental Design below.

See "Research on Urban Mobility Briefing Deck" document for details, as presented to briefing attendees, regarding research motivation; timeline; and participants' tasks, time demands, and compensation; and data anonymization. A participant's tasks were twofold: (i) sharing their daily trips on a personalized Google Sheet ("daily travel diary") until week 11 of the research study, and (ii) periodically sharing their transaction records on transportation apps and websites, available for both public transport and private car hailing ("transaction reports"), over the past 180 days--to form a baseline--all the way to week 16 of the research study, when the study would end (specifically, May 6, 2024).

During the recruitment briefings, we made no mention of pro-social or pro-environmental behavior in urban mobility (e.g., public transport vs. private cars). We mentioned that Singapore's well-functioning urban transport system needed to continue evolving, that understanding the urban mobility patterns of Singaporeans was important, and thus NUS researchers were partnering with the Land Transport Authority (LTA). We mentioned that until all data were anonymized, by the time the last payment to participants were completed in May 2024, the data would remain with the research team at NUS and not be available to the LTA. We emphasized that the research was not about them individually, but about people like them, and that we were not in a position to judge their preferences and choices; we only sought to learn about these.

Eight one-hour briefing sessions were held in lecture theaters at the Faculty of Arts and Social Sciences on January 11-18, 2024. Across the eight briefing sessions, 340 attendees stayed to the end, sharing their 180-day history of rides on public transport and private car hailing since August 2023 (available to them as users through a variety of digital app and internet platforms; see briefing deck).

When they had completed the briefing tasks, attendees collected their first payment, signing an acknowledgment form and consenting to participate in the research study. See Docs & Materials for a description of how we advertised a research participation opportunity to all NUS undergraduate students. Over 1000 students expressed interest in joining, via a Qualtrics form (see Docs & Materials), and we shortlisted three "clusters" (categories or types) of students:

("Cluster/category 1") Students whose household does not own a car (this is the case for the majority of Singapore's households); who stated taking public transport on at least one day in a typical week during the school term (this is the case for the vast majority of Singaporeans); and who stated taking a taxi or ride-hailed car ride on at least one day in a typical week during the school term.

("Cluster/category 2") Students whose household owns a car; who stated taking public transport on at least one day in a typical week during the school term; and who stated taking a taxi or ride-hailed car ride on at least one day in a typical week during the school term.

("Cluster/category 3") Students whose household does not own a car; who stated taking public transport on at least one day in a typical week during the school term; and who stated taking zero taxi or ride-hailed car ride on at least one day in a typical week during the school term.

2024-02-05

2024-05-06

Modal share of a subject's trips across all elements of their choice set {private car, public transport, walking, bicycling}
Modal share of a subject's distance traveled across all elements of their choice set {private car, public transport, walking, bicycling}
Modal share of a subject's trips across motorized elements of their choice set {private car, public transport}
Modal share of a subject's distance traveled across motorized elements of their choice set {private car, public transport}
Subject's average carbon score (see explanation)
The above are outcomes of interest, for each phase A to E of the experiment.
The intervention starts in phase B on February 5, 2024, after this trial was registered.

Explanation of a subject's average carbon score, as communicated to subjects (where applicable; see Experimental Design):
We calculate your average score based on your motorized mobility choices:
• [High] 200 grams of CO2 emissions for EVERY km traveled in a private car with no other traveler except you (size of travel party = 1)
• [Medium] 100 grams of CO2 emissions for EVERY km traveled in a private car if you are car-pooling with at least one more traveler (size of travel party > 1)
• [Low] 40 grams of CO2 emissions for EVERY km traveled in public transport (bus or train)
For example, if tomorrow I ride 10 km in a car just myself and the Grab driver, 10 km in a car with a friend who is traveling with me, and 10 km on the MRT, my Average Carbon Score will show (10*200 + 10*100 + 10*40) / (10+10+10) = 113 grams of CO2 per motor km.
These numbers are approximate emissions for the different travel modes while keeping things simple. They are good rules of thumb!

Subject's total number of trips across all elements of their choice set {private car, public transport, walking, bicycling}
Subject's total distance traveled across all elements of their choice set {private car, public transport, walking, bicycling}
Subject's trips across motorized elements of their choice set {private car, public transport}
Subject's distance traveled across motorized elements of their choice set {private car, public transport}
Car-pooling share of subject's private car trips (size of traveling party > 1, as explained in the briefing)

The experimental design: Three groups (one control, two treatment groups) and five phases (including a pre-intervention phase and a post-intervention phase). See Docs & Materials for a chart illustrating the experimental design.

The experimental design: Three groups (one control, two treatment groups) and five phases (including a pre-intervention phase and a post-intervention phase). See Docs & Materials for a chart illustrating the experimental design.

- The pre-intervention, Phase A, started on Monday January 15, 2024 for participants briefed on January 11, 12, or 15, 2024, and on Monday January 22, 2024 for participants briefed on January 16, 17, or 18, 2024. We randomized participants briefed by January 15 across the three groups. We later randomized participants briefed after January 15 across the three groups. In the pre-intervention, subjects became familiarized with the daily travel diary (see slide 7 of "Research on Urban Mobility Briefing Deck," Docs & Materials).
- The intervention consists of three phases, Phase B to D, each phase three weeks long. The intervention will start on February 5, 2024, after this trial was registered. See below for the different treatments by group over time.
- The post-intervention, Phase E, will start on Monday April 8, 2024. There will be no more logging of urban trips in the daily travel diary, no more environmental messaging, and no more rewards for low carbon scores. At the end of Phase E on May 6, 2024, also marking the completion of the RCT, subjects will submit their final set of transaction reports downloaded from transportation apps and websites.

- Control group: During all Phases A to D, subjects will log their daily urban mobility choices in a personalized travel diary on Google Sheet. Slide 7 of "Research on Urban Mobility Briefing Deck" (see Docs & Materials) shows the daily travel diary used in the pre-intervention Phase A and in the intervention Phases B to D.

- Environmental Messaging (EM) Treatment Group: In Phase B (weeks 3 to 5), we will introduce pro-environmental messages. These will change every Tuesday, Thursday, and Sunday and will be conveyed at the top of the daily travel diary and in the Telegram reminders for subjects to log their daily travel (a minority of subjects do not use Telegram and we use email reminders instead). The first Telegram reminder to Group EM, at the start of Phase B, states: "From now on, your travel diary will display informative messages on the benefits of taking fewer private car trips, to you and to others." See "Environmental message repository (sample)" in Docs & Materials. The messages are classified into different categories: (1) climate benefit, (2) air quality co-benefit, (3) accident externality, (4) land use and wear and tear, (5) motivation and social influence, (6) value of time, (7) pro-social behavior and sustainability, and (8) government policy. We will vary the environmental messages, drawing from the different categories over time. For example, the first environmental message of Phase B, sent to subjects in both treatment groups EM and EM+I, speaks to climate benefits: "When you make your travel choices, remember that no one likes melted ice cream! Likewise, we don’t want our glacial ice to melt. Ditch the car ride – it's the cooler thing to do."

- Environmental Messaging (EM) Treatment Group, CONTINUED: In Phases C (weeks 6 to 8) and D (week 9 to 11), on top of environmental messaging as explained above, we will provide subjects with their 3-week Average Carbon Scores from their motorized travel choices. We will communicate these as follows:
"We are adding an exciting feature to your travel diary: Your Average Carbon Score. Shown at the top of your travel diary is the average emissions from your motorized trips (public transport and private car rides). It will be your average score for the next three weeks. We calculate your average score based on your motorized mobility choices:
• [High] 200 grams of CO2 emissions for EVERY km traveled in a private car with no other traveler except you (size of travel party = 1)
• [Medium] 100 grams of CO2 emissions for EVERY km traveled in a private car if you are car-pooling with at least one more traveler (size of travel party > 1)
• [Low] 40 grams of CO2 emissions for EVERY km traveled in public transport (bus or train)
For example, if tomorrow I ride 10 km in a car just myself and the Grab driver, 10 km in a car with a friend who is traveling with me, and 10 km on the MRT, my Average Carbon Score will show (10*200 + 10*100 + 10*40) / (10+10+10) = 113 grams of CO2 per motor km.
These numbers are approximate emissions for the different travel modes while keeping things simple. They are good rules of thumb!
Refer to the image for more details!"
See Docs & Materials for the image we will share with subjects to explain the addition of Average Carbon Score to the travel diary. Note that after Phase C, we reset 3-week Average Carbon Scores for Phase D, giving subjects the opportunity to challenge themselves again.

- Environmental Messaging with Financial Incentives (EM+I) Treatment Group: In Phase B (weeks 3 to 5), we will introduce both pro-environmental messages (via Telegram messages and daily travel diaries) and Average Carbon Scores. See the description for Treatment Group EM above (except that Treatment Group EM will receive environmental messaging from Phase B and Average Carbon Scores from Phase C).

- Environmental Messaging with Financial Incentives (EM+I) Treatment Group, CONTINUED: In Phases C (weeks 6 to 8) and D (week 9 to 11), on top of environmental messaging and 3-week average carbon scores as explained above, we will provide subjects with low-carbon rewards. Like 3-week Average Carbon Scores, low-carbon rewards reset at each phase. These will be:
• Average carbon score at or below 70 grams of CO2 per km of motorized travel, and at least 100 km of motorized travel = $5
Plus
• Additional bonus: Average carbon score at or below 50 grams of CO2 per km of motorized travel, and at least 300 km of motorized travel = an additional $5, totaling $10.

Randomization done in office computer.

Individual. We randomized within each of the three "clusters" (categories or types) of students. See Intervention above. Within each category, one-third of students were assigned to Control, one-third of students were assigned to treatment Environmental Messaging (EM), and one-third of students were assigned to treatment Environmental Messaging with financial Incentives (EM+I)

No

340 students were briefed and joined the experiment.

340 students were briefed and joined the experiment.

114 students (who were briefed and consented to participate) were randomly assigned to Control, 112 students were randomly assigned to treatment Environmental Messaging (EM), and 114 students were randomly assigned to treatment Environmental Messaging with financial Incentives (EM-I).

Total sample size = 246; effect size f = .1; alpha error probability = .05; 1-beta error probability = .8; number of groups = 3; see page 12 of the application to the Institute for Public Understanding of Risk Seed Grant, submitted December 15, 2023 (timestamped and available on demand).