Panel Paper: Contextualizing School Commutes – Multilevel Modeling for Evaluation of Safe Routes to School Effectiveness

Friday, November 3, 2017
Horner (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Timothy G. Reardon, Metropolitan Area Planning Council

Thousands of communities across America have adopted programs to encourage students and their parents to choose walking, biking and other non-auto modes for commutes to and from school. Under the banner of Safe Routes to School (SRTS), these initiatives seek to achieve a variety of objectives, including health benefits for students; air quality improvements; reduced congestion; reduced public school transportation costs; and reduced greenhouse gas emissions. Of course, the ultimate potential of these programs at any given school depends in large part on how many students live close enough to school and have safe and pleasant walking or biking connections. Unfortunately, these factors are poorly captured by conventional (and largely aspatial) survey tools which generally produce aggregate mode share estimates (e.g., percent of students walking to school.) As a result, comparisons across schools/districts or over time may reflect differences in land use or student proximity more than the influence of programmatic interventions.

To help address this gap, the Metropolitan Area Planning Council (Metro Boston's regional planning agency) developed a new survey tool and evaluation method which have been adopted by MassRIDES, the state's SRTS agency. A spatially detailed six-question survey tool has been used to collect responses from over 25,000 students and their parents in over 125 Massachusetts schools since 2014. An online web interface provides school administrators and program coordinators with the ability to manage survey activities and produce automated reports with key results. MAPC also developed a multilevel model to investigate the effects of route, neighborhood, and school characteristics on walking to school. As expected, the model results indicate that the age of students, the distance to school, and the characteristics of the built environment affect the odds of walking to school. More importantly, the multilevel model provides a framework for examining between-school differences in walk-to-school rates while controlling for student body characteristics, assignment policies, and built environment factors. As a result, we can produce valid comparisons of mode share across heterogeneous schools and to contextualize schools' baseline walk share, set appropriate and measurable mode shift goals, and track progress over time. Repeat surveys at multiple schools can be used to evaluate what intervention strategies yield the greatest mode shift and to focus resources accordingly.