Small Nudges, Big Data: Evaluating the Impact of MIT Commuting Benefits Program

TitleSmall Nudges, Big Data: Evaluating the Impact of MIT Commuting Benefits Program
Publication TypeConference Paper
Year of Publication2017
AuthorsAdam Rosenfield, Jinhua Zhao, John Attanucci
Conference NameInternational Conference on Transport Survey Methods
Conference LocationEstérel, Canada
Abstract

Employer-based transportation demand management (TDM) programs have been shown to have a powerful effect on the travel behaviour of commuters. At the Massachusetts Institute of Technology (MIT), a series of TDM initiatives have been established in an effort to manage parking demand on campus while reducing the carbon footprint of its commuting population. In September 2016, MIT launched a new commuter benefits program known as AccessMIT, aimed at providing flexible, affordable and low carbon travel options to its staff, students, faculty and visitors. 

The AccessMITprogram includes a series of benefits that encourage commuters to switch from single-occupancy vehicle (SOV) commuting in favour of transit, carpooling and active modes of transportation. The flagship element of the program is the introduction of zero-cost local transit passes to all benefits-eligible employees—approximately 11,000 individuals living across the Greater Boston Area—making it among the largest employers in Massachusetts to offer such a benefit. In addition, most annual parking permits have been replaced with daily pay-as-you-park pricing in order to remove the sunk cost of yearly permits and provide the opportunity for commuters to save money on days when they choose not to drive. Finally, a series of secondary incentives include an increase to commuter rail subsidy, discounted Massachusetts Bay Transportation Authority (MBTA) station parking fees, and the integration and promotion of other existing commuter programs. An online commuter dashboard called AccessMyCommutehelps employees track their commuting patterns and provides incentives for sustainable travel through gamification and financial rewards.

As MIT rolls out its new commuter benefits program, the Institute is seeking to achieve a goal of ten percent reduction in parking demand over the next two years. The introduction of several simultaneous initiatives will likely increase the chances of achieving a sustained decrease in SOV commuting and parking demand, but may cause difficulty in identifying which interventions are responsible for which changes in behaviour. In order to better understand the relative impacts of each commuting benefit, and to inform lessons learned for MIT and other employers, researchers at the MIT Transit Lab are employing a series of stated and revealed preference survey methods to evaluate the extent to which commuting benefits influence travel behaviour and employee satisfaction.

Through partnership with the MIT Parking & Transportation Office, a multi-faceted data framework has been established to collect near-real-time revealed behaviour of commuters. Given most employees now have a free transit chip built into their employee ID card, MBTA trips can be tracked using automated fare collection technology. Additionally, using the origin, destination and interchange inference (“ODX”) algorithm, the research team can build detailed datasets of passengers’ journeys through the subway and bus system. To measure driving trips, MIT’s network of gated parking lots provides tap-in and tap-out data on when and where employees are parking. When drivers tap their pass to enter the lot, their parking history is stored in a central database and can be linked to employee attributes such as transit usage and demographics. Walking and cycling trips are tracked for employees who opt-in to using a mobile app called Moves, which privately relays daily individual trajectories to the AccessMyCommutedashboard. All other trips, including carpools, taxis, new mobility services, and those parking in un-gated lots, are electively tracked through a self-reporting tool on the online dashboard. 

Supplementing the automated database of real-time commuting patterns is a series of stated preference questionnaires, most importantly the biennial MIT Transportation Survey. Mandated by state and local environmental regulations, this survey has been conducted every other year since 2004, providing intermittent snapshots of student and employee commuting behaviour and perceptions of existing transportation benefits. In its autumn 2016 implementation, the survey has been supplemented to explore the impacts of the AccessMIT benefits program. The new survey includes additional binary, Likert scale and open-response questions to gauge the comparative magnitude and direction of the effect of each incentive on travel behaviour and commute satisfaction. In addition, a supplemental survey published on the AccessMyCommute dashboard elicits responses from those using the online tool to track their commutes, and provides an opportunity for feedback beyond the campus-wide survey.  

While the Institute-wide implementation of universal commuting benefits results in a lack of control group for the purposes of designing a traditional randomized control trial (RCT) experiment, several proxy control groups are identified for comparative analysis. For example, some parkers continue to purchase annual permits (due to their use of ungated lots), and can be compared with those forced to switch to daily pricing. Additionally, some staff are targeted for outreach and marketing, allowing researchers to compare their behaviour to those without these interventions.

Given that the surveys are not anonymized, responses are joined with an employee’s historical commuting and demographic data. The information on past commuting behaviour provides a powerful tool to track temporal changes amongst employees, and allows the researchers to generate three sets of cross-sectional clusters across employees. First, geospatial clusters are identified based on neighborhood of residence, distance to campus and proximity to transit service. Second, behavioural clusters are identified based on benefits used (e.g. MBTA station parking subsidies, pre-tax bicycling benefits, etc.), modes chosen and changes in behaviour over time. Finally, attitudinal clusters are identified based on stated proclivity to take transit or active modes, support for benefits based on environmentalist tendencies, and personality attributes such as price sensitivity. 

This paper will address the challenges and opportunities associated with implementing an employer-based TDM program and evaluating its impacts using a hybrid dataset approach. This research advances the state of combined revealed and stated preference methods using ICT-based data collection through an online dashboard and mobile app, and provides lessons learned for the application and data-driven evaluation of TDM programs. This research will evaluate the impacts of Massachusetts’s largest implementation of universal transit benefits in concert with other incentives to reduce SOV commuting, and provide a case study for other employers seeking to attain similar goals of parking reduction, employee satisfaction and a reduced carbon footprint.